PAIN AND PAIN MANAGEMENT IN A MEDICAID WAIVER PROGRAM
By
Elizabeth Annette Byma

A DISSERTATION
Submitted to
Michigan State University
In partial requirements
For the degree of
DOCTOR OF PHILOSOPHY
Nursing
2010

ABSTRACT
PAIN AND PAIN MANANAGEMENT IN A MEDICAID WAIVER PROGRAM
By
Elizabeth Annette Byma
Background/Purpose: Poor older adults are a vulnerable population at increased risk for cancer,
pain, poor pain management and pain management outcomes. There is no known research that
has examined the differences in pain, pain management and pain management outcomes or the
transfer of HCBWP participants to nursing homes between persons with and without a diagnosis
of cancer among Medicaid-enrolled, older adults participating in Home and Community-Based
Waiver programs (HCBWP). Research examining pain, pain management and pain management
outcomes among older adults has been cross-sectional. Longitudinal research would be better
able to examine the associations among pain, pain management and pain management outcomes
overtime and how pain, pain management and pain management outcomes associate with the
admission of older adult HCBWP participants to nursing homes.
Conceptual Model: Based on the Symptom Management Theory, the pain experience, pain
management strategies and pain management outcomes are associated with the domains of
person and health and illness. The pain experience, pain management strategies, pain
management outcomes and domains of person and health and illness are conceptualized as
impacting the admission of older adult HCBWP participants to nursing homes.
Methods: The study was a secondary analysis of data from the Minimum Data Set Home Care
(MDS-HC), the State of Michigan Cancer Surveillance Data and Michigan Medicaid paid claims
files and was of a longitudinal design. The sample was comprised of 4054 older adult HCBWP
participants Generalized Estimating Equations, logistic regression and survival analysis methods
were used.

Results: Older adult HCBWP participants who were female, had a higher comorbid conditions
score or higher score for behaviors indicative of depression were more likely to experience daily
pain over time. Older adult HCBWP participants who were African American, older age, or
cognitively impaired were less likely to report daily pain over time. Cancer was not associated
with daily pain over assessment time points. Older adult HCBWP participants with daily pain
and in the initial phase of diagnosis of cancer were less likely to be prescribed non-opioid pain
medications and adjuvant pain medications than older adult HCBWP participants without daily
pain. Pain, age, race, sex, cognitive functioning, behaviors indicative of depression and comorbid
conditions were associated over time with prescribed pain medications and physical functions.
Behaviors indicative of depression was significantly associated with perceived pain control, such
that as the measure of behaviors indicative of depression increased, the likelihood of pain control
occurring decreased. Finally, diagnosis of cancer was not significantly associated with admission
to a nursing home. Older adult HCBWP participants with daily pain were less likely to be
admitted to a nursing home. Older adult HCBWP participants who had a higher measure of
comorbid conditions, were cognitively impaired or were white had an increased hazard of being
admitted to a nursing home.
Conclusion: Diagnosis of cancer was not significantly associated with the pain experience, pain
management outcomes and admission to a nursing home among older adult HCBWP participants
and a limited association with prescribed pain medications. The prescription of pain medication
was most consistently associated with the measure of pain. Pain had a negative association with
the admission of older adult HCBWP participants to a nursing home and a positive association
with physical functioning over time.

Copyright by
Elizabeth Annette Byma
2010

Dedication
This dissertation is dedicated to my family:
My dear husband Gary and our two precious children Maggie and Aaron, my parents, my sisters
and their families: I could not have completed this dissertation or any of the doctoral program
without your constant love, encouragement and support.
A special dedication in memory of my mother-in-law Ann Lillian Werkema Byma, R.N. Her
prayers and pride in my accomplishments meant the world to me. She was so proud to be a
registered nurse, and was still interested in all matters nursing long after she gave up her career
to be a pastor’s wife and mother of five.
I wish she was still here to see me complete this dissertation and graduate.

v

ACKNOWLEDGEMENTS

My appreciation to the following organizations for their support for my doctoral program
training and of this dissertation project:


National Institutes of Health, National Institute of Nursing Research, Individual Ruth L.
Kirschstein National Research Service Award, Grant 1F31NR011523-01A1. Project
Title: Pain and Pain Management in a Medicaid Waiver Program. Awarded 8/16/2010 to
present.



Mary Margaret Walther Program of the Walther Cancer Institute, Mary Margaret Walther
Pre-doctoral Fellowship, Awarded 8/2007-8/2009



Michigan State University Graduate School and Michigan State College of Nursing,
University Enrichment Fellowship, Awarded 8/2006-present.

My sincere thanks to the following:


Dr. Joseph Cesario and Dr. Manfred Stommel for serving on my guidance committee.



The Institute for Health Care Studies at Michigan State University and State of Michigan
Department of Community Health for their work in developing the dataset used for this
study.



Dr. Barbara Given and Dr. Charles Given for your encouragement, support, leadership
and wisdom in guiding me through this study and throughout my doctoral program.



Dr. Bill Corser, for your expertise in comorbidities. Dr. Siran Koroukian, for your
knowledge of working with large datasets and healthcare system issues. Dr. Deb Schutte
for your understanding of the health care needs of older adults and those with dementia.

vi

ACKNOWLEDGEMENTS (Continued)


Ms. Mei You for your statistical expertise and unending patience with my creative SAS
coding. I could not have learned SAS in such a short amount of time or completed the
analyses without your help.



My cohort, Ms. Denise Soltow. A special thank you for your friendship and support. I
couldn’t have made it through without you.

vii

TABLE OF CONTENTS
LIST OF TABLES…………………………………………………………..xi
LIST OF FIGURES…………………………………………………………xiv
CHAPTER 1………………………………………………………………...1
Pain among Older Adults………………………………….………..5
Pain Management among Older Adults…………………………….9
Pain Management Outcomes among Older Adults………………...12
Significance of Study………………………………………............13
Research Questions…………………………………………...……15
Purpose of Study…….…………………………………………......16
CHAPTER 2…………………………………………………………..…...18
Domain of Person………………………………………………......21
Domain of Health and Illness………………………………….…...24
The Pain Experience………………………………………………..27
Pain Management Strategies……………………………………….29
Pain Management Outcomes………………………………..….......31
Admission to a Nursing Home……………………………….….....33
Conclusion of Chapter 2……………………………………...........35
CHAPTER 3………………………………………………………………...36
The Home and Community Based Waiver Program………………..36
Domains of Person and Health and Illness………………………….41
The Pain Experience……………………………………….……......60
Pain Management Strategies……………………………………......67
Pain Management Outcomes……………………………………......75
Time and Pain…………………………………………………….....79
Admission of HCBWP Participants to Nursing Homes……….……81
Conclusion of Chapter 3……………………………….……………84
CHAPTER 4…………………………………………………………..….....86
Purpose and Research Questions……………………………..…..…86
Sample………………………………………………………....….…87
Setting………………………………………………….……...…….89
Data Sources…………………………………………….…..….……93
Operational Definitions and Measures……………………..…...…...95
Dependent Variables………………………………………..….…....96
Covariates………………………………………………………........97
Independent Variables/Covariates………………………….……....104
Dependent/Independent Variables……………………………...…..105
Data Preparation and Management………………………….…..….115
Analysis of Data……………………………………………..….….117
Study Limitations……………………………………………...…...122
Study Strengths………………………………………………….....123
viii

Protection of Human Participants…………………………………………...123
Data Security………………………………………………………..124
Women and Minority Inclusion in Clinical Research………………124
Conclusion of Chapter 4…………………………………...….…….125
CHAPTER 5…………………………………………………….………….126
Sample…………………………………………………….……..….131
Measures……………………………………………….………...….134
Results………………………………………………….…………...164
Research Question 1……………………………...……........165
Research Question 2…………………………………...........170
Pain Management Strategies………………….….…173
Pain Management Outcomes …………….……...….189
Physical Function………………………....…189
Pain Control…………………………....……194
Conclusion for Results of
Research Questions 2 Analyses…………….……….198
Research Question 3……………………………………...…198
Nursing Home Admission…………….…….…..…..200
Time to Nursing Home Admission…………...…….202
Conclusion of Research Question 3……………......205
Discussion of Study Results……………………..……… …205
The Pain Experience…………………………..… …206
Pain Management Strategies………………..………209
Pain Management Outcomes……………………......217
Admission to a Nursing Home and
Time to Nursing Home Admission…………..……..222
Study Limitations…………………………………………...…..….226
Conclusion of Chapter 5……………………………….………..….229
CHAPTER 6
Contributions to Science…………………………………………….231
Implications…………………………………………………….……232
Implications for Clinical Practice……………………………232
Implications for Research……………………………….…...236
Implications for Policy………………………………………242
Conclusion of Dissertation…………………………………..………243

ix

APPENDIX A……………………………………………………………….247
APPENDIX B……………………………………………………………….248
APPENDIX C……………………………………………………………….250
APPENDIX D……………………………………………………………….251
APPENDIX E…………………………………………………...…………..252
REFERENCES……………………………………………………………...255

x

LIST OF TABLES
Table 1: Items from the Adapted and Original Depression Rating Scale….........100
Table 2: Study Variables by Place in Conceptual Model, Data Source, Item,
Coding and Variable Type ……………………………………….........110
Table 3: Reported Diagnosis of Cancer at Each Assessment Time Point………130
Table 4: Revised Measure of Reported Diagnosis of Cancer at
Each Assessment Time Point ..………………………………...……....130
Table 5: Significance of Differences in Diagnosis of Cancer by Socio-demographic
Characteristics at Time 1 among Study Subjects (n=4054)………...….132
Table 6: Significance of Differences in Cognitive Functioning by
Diagnosis of Cancer at Each Assessment Time Point………………....135
Table 7: Significance of Differences in Mean Behaviors Indicative of Depression
By Diagnosis of Cancer At Each Assessment Time Point……..……. 138
.
Table 8: Model Fit Statistics for Multinomial Logistic Regression Models
for the Association between Pain Scale and
Different Measures of Comorbid Conditions………….……………...139
Table 9: Multinomial Logistic Regression Parameter Estimates for
Final Comorbid Condition Model:
Pain Scale (0-3) =Comorbid Conditions ……………..……………….140
Table 10: Significance of Differences in the Mean Measure of
Comorbid Conditions by Diagnosis of Cancer at
Each Assessment Time Point………………….………………..........141
Table 11: Number and Percentage of Cancer Site at each Assessment
Time Point among Subjects with Cancer …………………………….143
Table 12: Number and Percentage of Stage of Cancer at Each Assessment
Time Point among Subjects with Cancer …………………………….144
Table 13: Significance of Differences in Pain Frequency by
Diagnosis of Cancer at Each Assessment Time Point (n=4054)……..145
Table 14: Significance of Differences in Pain Intensity by
Diagnosis of Cancer at Each Assessment Time Point
for Subjects With Daily Pain ………………………………..………146

xi

Table 15: Significance of Differences in Pain Scale by Diagnosis of Cancer at
Each Assessment Time Point (n=4054)………………………..…………..146
..
Table 16: Number and Percentage of Prescribed Pain Medication at
Each Assessment Time Point (n=4054)………………………………..…..148
Table 17: Significance of Differences in Prescribed Pain Medication by
Diagnosis of Cancer at Each Assessment Time Point (n=4054)…………..149
Table 18: Number and Percentage of Prescribed Pain Medication by
Measure of Pain at each Assessment Time Point (n=4054)…………..…..151
Table 19: Significance of Differences in Hospice Services by
Diagnosis of Cancer at Each Assessment Time Point…………………..…153
Table 20: Significance of Difference in Pain Control by
Diagnosis of Cancer at Each Assessment Time Point (n=4054)………….156
Table 21: Number and Percentage for the Measure of Pain Control by
Measure of Pain at Each Assessment Time Point…………………..……157
Table 22: Number and Percentage of Pain Control by Measure of
Prescribed Pain Medication at each Assessment Time Point…………….159
Table 23: Significance of Differences in Mean Physical Function
by Diagnosis of Cancer at Each Assessment Time Point……………..…161
Table 24: Percentage of Older Adult HCBWP Participants
Admitted to Nursing Home Prior to 12/31/2005…..…………….………163
Table 25: Differences in Admission to a Nursing Home
Prior to 12/31/2005 by Diagnosis of Cancer……………..…………..…..163
Table 26: Mean Time to Nursing Home Admission
by Diagnosis of Cancer……………………………….………….………164
Table 27: Parameter Estimates, Odds Ratios and Significance Values of
Parsimonious Model for Generalized Estimating Equation
Fitted to the Dependent Variable of No Pain/Less than
Daily Pain vs. Daily Pain Using a Binomial Distribution (n=4054)…….167

xii

Table 28: Parameter Estimates, Odds Ratios and Significance Values of
Parsimonious Model for Generalized Estimating Equation Fitted to the Dependent
Variable of Daily Pain, Not Unusually Intense vs. Daily Pain, Unusually
Intense Using a Binomial Distribution (n=2243)..……................................….169
Table 29: Parameter Estimates, Odds Ratios and Significance Values of Parsimonious Model
for Generalized Estimating Equation Fitted to the Dependent Variable of Non-Opioid
Prescribed Pain Medications vs. No Prescribed Pain Medications
Using a Binomial Distribution (n=2243)……………………………………….174
Table 30: Parameter Estimates, Odds Ratios and Significance Values of Parsimonious Model
for Generalized Estimating Equation Fitted to the Dependent Variable of
Opioid Prescribed Pain Medications vs. No Prescribed Pain Medications Using a
Binomial Distribution (n=2243)………………………………………………..177
Table 31: Parameter Estimates, Odds Ratios and Significance Values of Parsimonious Model for
Generalized Estimating Equation Fitted to the Dependent Variable of Adjuvant
Prescribed Pain Medications vs. No Prescribed Pain Medications
Using a Binomial Distribution (n=2243)……………….…………….………..182
Table 32: Parameter Estimates, Odds Ratios and Significance Values of Parsimonious Model for
Generalized Estimating Equation Fitted to the Dependent Variable of Opioid
Prescribed Pain Medications vs. Non-Opioid Prescribed Pain Medications Using a
Binomial Distribution (n=2243)………………………………………………...187
Table 33: Parameter Estimates, Odds Ratios and Significance Values of Parsimonious Model for
Generalized Estimating Equation Fitted to the Dependent Variable of Physical
Function Using a Poisson Distribution (n=2243)……………………….………193
Table 34: Parameter Estimates, Odds Ratios and Significance Values of Parsimonious Model
for Generalized Estimating Equation Fitted to the Dependent Variable of Pain Control
Using a Binomial Distribution (n=2243)………………………….………….…195
Table 35: Coefficient, Standard Error, Odds Ratios and Confidence Interval from Logistic
Regression Model for Admission to Nursing Home…………………………….201
Table 36: Coefficient, Standard Error, Hazard Ratios and Confidence Intervals from Cox
Proportional Hazard Model for Time to Nursing Home Admission…………….203
Table 37: Items of the Adapted Cognitive Performance Scale Utilizing MDS-HC Items…248

xiii

LIST OF FIGURES
Figure 1: Symptom Management Theoretical Model……………..……….……………………19
.
Figure 2: Conceptual Model for Study: Pain Management among Older Adults
Within a Home and Community-Based Waiver Program …………………...….…...20
Figure 3: Proposed Sample Selection Flow Chart….…………………………….…...….……..89
Figure 4: Final Exclusion Criteria and Flow Chart for Determination of Sample……….……..133
Figure 5: Mean Reported Behaviors Indicative of Depression by
Diagnosis of Cancer at Each Assessment Time Point…………………….…………137
Figure 6: Mean Measure of Documented Comorbid Conditions by
Diagnosis of Cancer at Each Assessment Time Point………………..…….………..141
Figure 7: Mean Physical Function
by Diagnosis of Cancer at Each Assessment Time Point………………….………...162
Figure 8: Diagram Depicting Scoring Rubric for Cognitive Performance Scale………...……250
Figure 9: Diagram Depicting the Scoring of Pain Scale……………...……………………….251

xiv

Chapter 1
The proportion of the United States population that is 65 and older will increase from
12.4% in 2007 to 20% in 2030 (Administration on Aging, 2007). Older adults are at increased
risk of developing cancer when compared to younger populations as more than 60% of cancers
diagnosed and 70% of cancer-related deaths occur in older adults (Bourbonniere & Van Cleave,
2006; Yancik & Ries, 2000). Therefore, the impending increase in the size of the older adult
population will likely bring a marked increase in the number of persons diagnosed with cancer as
well as cancer survivors (Erikson, Salsberg, Forte, Bruinooge, & Goldstein, 2007; Smith, Smith,
Hurria, Hortobagyi, & Buchholz, 2009).
An increased number of older adults with cancer or surviving cancer will result in an
increase in older adults experiencing pain related to cancer or its treatment. Research has found
that older adults with cancer are more likely to experience pain when compared to older adults
without cancer (Buchanan, Barkley, Wang, & Kim, 2005; Reyes-Gibby, Aday, Todd, Cleeland,
& Anderson, 2007; Rodin, 2008). The presentation of pain among older adults, however, is much
more complicated than simply if one does or does not have cancer. Pain occurs in older adults
not only because of cancer and its treatment, but also because of the presence of comorbid
conditions that are commonly associated with increased age and pain such as arthritis, diabetes
and peripheral vascular disease (Bruckenthal & D'Arcy, 2007; Davis & Srivastava, 2003;
Freedman, 2002).
The presence of comorbid conditions not only causes pain among older adults but also
hinders the resolution of pain due to provider reluctance to manage pain aggressively due to
fears of opioid side effects and drug interactions with medications prescribed for comorbid
conditions (Duncan, Forbes-Thompson, & Bott, 2008; Goldstein & Morrison, 2005; McNeill,

1

Reynolds, & Ney, 2007). Pain management among older adults may be hampered by the
reluctance of older adults to report pain as well as cognitive impairment which inhibits the
production of and assessment of verbal reports of pain (Delgado-Guay & Bruera, 2008;
Goldstein & Morrison, 2005). In summary, older adults are at risk of pain due to the increased
presence of cancer and other diseases that are associated with pain (Bruckenthal & D'Arcy, 2007;
Davis & Srivastava, 2003; Freedman, 2002). Older adults who experience pain are at risk for
perceiving that their pain is not assessed and managed appropriately for a variety of patient and
system-related reasons and are therefore likely to experience poor pain management outcomes
(American Geriatrics Society Panel on Persistent Pain in Older Persons, 2002).
Poor pain management outcomes among older adults include anxiety, depression,
decreased social interaction, sleep disturbances, impaired physical function, agitation, delirium,
decreased appetite, delayed healing, lower quality of life and higher health care utilization and
costs (American Geriatrics Society Panel on Persistent Pain in Older Persons, 2002). Pain is a
frequent deterrent to quality of life among older adults (Deane & Smith, 2008) and therefore,
pain, pain management and pain management outcomes among older adults are important for
health care providers and researchers to address.
Previous research studies that have examined the issues of pain, pain management and
pain management outcomes among older adults have several limitations that the study will
address. First, the research has been primarily focused on community-dwelling and nursing home
settings and has not fully examined pain, pain management and pain management outcomes in
other older adult populations (Bryant, Grigsby, Swenson, Scarbro, & Baxter, 2007; Shega,
Hougham, Stocking, Cox-Hayley, & Sachs, 2006; Walke, Byers, McCorkle, & T.R., 2006; Won
et al., 2004; Zyczkowska, Szczerbinska, Jantzi, & Hirdes, 2007). An alternative to nursing home

2

placement for older adults is the Medicaid Home and Community Based Waiver Program
(HCBWP). The HCBWP allows Medicaid-eligible older adults to receive care services in their
homes instead of being admitted to a nursing home for similar care services. Services covered
under waiver programs may include homemaker services, respite care, adult day care,
environmental modifications, transportation, medical supplies, personal emergency response
system, private duty nurse, counseling, home delivered meals, physical and occupational therapy
and personal care supervision (Shugarman, Fries, & James, 1999). Despite prevalence of pain
among HCBWP participants of 53-69% (Fries, James, & Aliaga, 2004; L. Li & Conwell, 2007)
very little else is known about pain, pain management and pain management outcomes among
older adult HCBWP participants or differences between pain, pain management and pain
management outcomes among older adult HCBWP participants with and without a diagnosis of
cancer.
The issues of pain, pain management and pain management outcomes among older adult
HCBWP participants and differences in pain, pain management and pain management outcomes
between older adult HCBWP participants with and without a diagnosis of cancer are of concern
to nursing. Because of their Medicaid eligibility, HCBWP participants are assumed to be
impoverished and poverty is associated with an increased risk of developing cancer as well as
experiencing poor pain management (Green et al., 2003; Ward et al., 2004b). Therefore, not only
are older adult HCBWP participants at risk of pain, poor pain management and cancer because of
their increased age, they are also considered at risk of pain and cancer due to their poverty. As
such, older adult HCBWP participants are at risk for pain and poor pain management and poor
pain management outcomes “above and beyond” other older adults. Research is needed to
examine pain, pain management strategies and pain management outcomes in this older adult

3

population as well as differences in pain, pain management strategies and pain management
outcomes between HCBWP participants with and without a diagnosis of cancer.
A second limitation in previous research that has examined pain, pain management and
pain management outcomes among older adults is that previous research has been primarily
cross-sectional (Jakobsson, Klevsgard, Westergren, & I.R., 2003; Walke, et al., 2006; Won, et
al., 2004; Zyczkowska, et al., 2007). Pain, pain management and pain management outcomes
have strong temporal (time-related) components and occur as a series of events (Henly, Kallas,
Klatt, & Swenson, 2003). Temporal aspects of pain consist of variation of pain over time, how
frequently pain occurs, as well as the duration of pain (Jensen, 2003). Pain management can
include either short or long-acting pain medications, treating pain that may be either sporadic or
continuous across time (American Pain Society, 2005; NCI, 2010b). Pain management
outcomes are the end results of care (Patrick, 1997) and are assessed after the provision of pain
management interventions (Humphreys & et al., 2008). The assessment of pain management
outcomes is timed to occur after an intervention has had an effect. For example, pain
management guidelines recommend that the level of pain is reassessed after a dose of pain
medication has had time to act, based on pharmacokinetics (American Pain Society, 2005;
NCCN, 2010; NCI, 2010b). In summary, pain, pain management and pain management
outcomes occur over time and therefore, longitudinal research may be a better choice than crosssectional research to examine relationships among pain, pain management and pain management
outcomes over time among older adults with and without a diagnosis of cancer.
Healthcare services for older adults can be viewed as a continuum, moving from
community to institutionalization (L. Li & Zullo, 2003; Williams, 2001). Longitudinal research
will allow for analyses of how pain, pain management and pain management outcomes may

4

influence the admission of older adult HCBWP participants to nursing homes. As the goal of the
HCBWP is to prevent or delay nursing home admission (Fries, Shugarman, Morris, Simon, &
James, 2002), knowledge gained about the impact of pain, pain management and pain
management outcomes on the admission of older adults to nursing homes and differences
between participants with and without a diagnosis of cancer would be beneficial for guiding the
development of care strategies for assisting older adults in staying in the community and
avoiding institutionalization.
In this study, associations between pain, pain management and pain management
outcomes among older adult HCBWP participants with and without cancer over time in the
HCBWP were systematically explored. Additionally, an analysis of the influence of pain, pain
management and pain management outcomes on the admission of older adult HCBWP
participants to a nursing home was completed.
The purpose of Chapter 1 was to provide a brief overview of the pain, pain management
and pain management outcomes experienced by older adult, HCBWP participants aged 65 and
older as an introduction to the study. Differences in pain, pain management, and pain
management outcomes among older adults with and without a diagnosis of cancer will be
described. An overview of the predictors of admission to a nursing home will be presented.
Research questions developed from this overview and a synopsis of the proposal will be
presented at the conclusion of this chapter.
Pain among Older Adults
Pain is “…an unpleasant sensory and emotional experience associated with actual or
potential tissue damage or described in terms of such damage” (International Association for the
Study of Pain, 2009, para. 36). Pain is highly subjective, being “…whatever the experiencing

5

person says it is, existing whenever he/she says it does” (McCaffery & Beebe, 1989, p. 7)
Individual-related characteristics contribute to the multidimensional nature of pain (American
Pain Society, 2005; Armstrong, 2003; Dodd et al., 2001; Humphreys & et al., 2008) and are
associated with the experience of pain among older adults.
Individual characteristics such as sex and race assist in explaining differences in the pain
among older adults. Female older adults are more likely to experience pain than males (ReyesGibby, Aday, et al., 2007; Soldato et al., 2007; Thomas, Peat, Harris, Wilkie, & Croft, 2004).
Regarding race, while Horgas, Yoon, Nichols and Marsiske (2008) found no difference in pain
presence, intensity, locations and durations between black and white older, community dwelling
older adults, Teno, Kabumoto, Wetle, Roy and Mor (2004) reported that African American
nursing home residents were less likely to experience daily, excruciating pain than white nursing
home residents. In comparison, Reyes-Gibby, Aday and Cleeland (2002) found that among
community-dwelling older adults, more Hispanics than Whites, Blacks and American Indians
were often bothered by pain. Additional research is needed to clarify associations between race
and pain among older adults.
Differences in pain among older adults may be explained in part by age-related changes
in nerve physiology and function, resulting in diminished endogenous opioid mechanisms and
pain modulation (R. R. Edwards, Fillingim, & Ness, 2003; Washington, Gibson, & Helme,
2000), reduced tolerance of and increased response to higher intensity experimental pain stimuli
(Gibson & Helme, 2001) and slower resolution of post-injury hyperalgesia (increased sensitivity
to pain) (Zheng, Gibson, & Helme, 2000). Therefore, the neuro-physiological changes older
adults experience may lead to an increased and prolonged response to injury and pain stimuli,
thereby placing older adults at risk of pain that persists or in need of extended pain management.

6

Another health-related characteristic of older adults associated with pain is the presence
of conditions associated with pain. Depression is comprised of both emotional and physical
symptoms, with pain as a physical symptom (Delgado, 2004). Therefore, depression may be a
possible source of pain among older adults and associations between pain and depression must
be accounted for in research examining pain among older adults. In addition to depression being
a possible source of pain, depression and pain are associated with one another such that an
increased number of depressive symptoms has been shown to predict worsening levels of pain
among community dwelling older adults (Rosso, Gallagher, Lubrosky, & Mossey, 2008). Older
adults with pain are more likely to have recognized depression (L. Li & Conwell, 2007), have a
higher rate of developing a new onset of depression and have a slower resolution of depression
than older adults without pain (Geerlings, Twisk, Beekman, Deeg, & van Tilburg, 2002).
Therefore, pain is associated with depression among older adults, with the presence of one
increasing the likelihood of occurrence and worsening the prognosis of the other (Geerlings, et
al., 2002; L. Li & Conwell, 2007).
Physical conditions associated with pain, such as arthritis, diabetes and cancer are more
likely to occur in older adults than in younger populations and place older adults at risk for pain
(Bruckenthal & D'Arcy, 2007; Davis & Srivastava, 2003; Freedman, 2002). For example,
diabetes may contribute to demyelization of peripheral nerves from decreased blood supply
which may result in ectopic and spontaneous nerve discharges (Backonja, 2003; Pasero, 2004).
Arthritis is common among older adults, affecting about 50% of community-dwelling older
adults (Blyth et al., 2008; Zyczkowska, et al., 2007) with over half of older adults with arthritis
reporting pain (Reyes-Gibby, et al., 2002).The effect of comorbid conditions on pain is not only
singular, i.e. the specific effect of a disease, such as diabetes on pain, but also one of the

7

combined effect of multiple comorbid conditions, with a higher number of comorbid conditions
associated with an increased presence of pain (Mao et al., 2007; Reyes-Gibby, Aday, et al.,
2007). The additional effect of comorbid conditions on cancer is a specific concern in regards to
older adults, as a diagnosis of cancer is likely to be made in the context of pre-existing health
conditions (Yancik, Ganz, Varricchio, & Conley, 2001).
Unfortunately, older adults bear a disproportionate burden of cancer in the United States
(Yancik, et al., 2001). Older adults comprise 60% of the population of persons diagnosed with
cancer (Yancik & Ries, 2000) and are more likely to have pain at the time of diagnosis than
younger populations (Freedman, 2002; McNeill, et al., 2007). Pain has been noted to be one of
the most distressing symptoms in patients with cancer (Laird, Colvin, & Fallon, 2008; Valeberg
et al., 2008). Pain may occur due to the cancer itself or to its treatment (Chang, Janjan, Jain, &
Chau, 2006). For example, a tumor may compress a nerve and cause pain, or chemotherapy may
alter nerve function, resulting in pain. Older adults with cancer are more likely to experience
pain when compared to older adults without a history of cancer in both community and nursing
home settings (Buchanan, et al., 2005; Reyes-Gibby, Aday, et al., 2007; Rodin, 2008). In
summary, cancer, depression and the presence of other comorbid conditions increase the risk of
pain among older adults.
While comorbid conditions common to older adults place older adults at risk of pain, the
presence of cognitive impairment lessens the likelihood that the presence of pain among older
adults is recognized (Procter & Hirdes, 2001; Reynolds, Hanson, DeVellis, Henderson, &
Steinhauser, 2008; Sengupta, Bercovitz, & Harris-Kojetin, 2010). The presence of pain is
acknowledged by a patient’s verbal report of pain or pain-related behaviors made by the patient
in response to stimuli perceived to be painful (NANDA, 2005; Sykes, Fallon, & Patt, 2003).

8

Cognitive impairment may diminish the ability of patient to verbally report pain (Bruckenthal,
2008; Helme & Gibson, 2001) and may therefore result in less pain being reported among older
adults. Pain prevalence among cognitively impaired older adults is routinely less than pain
prevalence among cognitively intact older adults, after controlling for diseases known to likely
produce pain (Procter & Hirdes, 2001; Reynolds, et al., 2008; Sengupta, et al., 2010). In
summary, older adults with cognitive impairments are less likely than cognitively intact older
adults to be able to make verbal reports of pain and therefore, less pain is attributed to
cognitively impaired older adults than cognitively intact older adults.
In conclusion, individual characteristics such as sex, race, age and presence of
depression, cancer, comorbid conditions and cognitive impairment are associated with pain
among older adults and influence the production and perception of pain as well as the ability of
the patient to report pain. The patient report of pain or pain behaviors are then used for the pain
assessment. As the goal of pain assessment is to direct pain management, an inadequate pain
assessment will contribute to poor pain management among older adults.
Pain Management among Older Adults
Pain management begins with pain assessment (Dodd, et al., 2001). Because the
assessment of pain is primarily based on a person’s verbal self-report of pain, characteristics
unique to older adults may affect the self-report of pain, resulting in under-reporting of pain
among older adults (Bruckenthal, 2008; Helme & Gibson, 2001) resulting in inadequate pain
management. Older adults may expect pain to be part of aging and disease processes and
therefore, may underreport their pain (Delgado-Guay & Bruera, 2008; Goldstein & Morrison,
2005). Age and disease-related changes in the production and sensation of pain may influence
the older adult’s perception of painful stimuli and alter the older adult’s ability to accurately

9

convey reports of pain. (Bruckenthal, 2008; Gibson & Helme, 2001). Pain may be challenging
for healthcare providers to accurately assess in older adults with cognitive changes or
communication difficulties if pain-related behaviors must be assessed in place of verbal reports
of pain (Delgado-Guay & Bruera, 2008; Goldstein & Morrison, 2005). Assessments of pain
prevalence among cognitively impaired older adults are routinely less than pain prevalence
among cognitively intact older adults, after controlling for diseases known to likely produce pain
(Procter & Hirdes, 2001; Reynolds, et al., 2008). In summary, characteristics of older adults
may result in underreported pain and a poorer pain assessment. Pain management is based on
pain assessment and if the pain assessment in poor, pain management may therefore be poor as
well.
Older adults are at risk for inadequate pain management (American Geriatrics Society
Panel on Persistent Pain in Older Persons, 2002; Landi et al., 2001). Older adults often have
complex clinical presentations of symptoms, including pain, as they may have multiple, cooccurring chronic diseases (Duncan, et al., 2008; Goldstein & Morrison, 2005; McNeill, et al.,
2007). Health care providers may have concerns about pain medication interactions with other
medications prescribed for chronic diseases and may limit their pain medication prescribing
patterns, negatively affect a patient’s pain management (Duncan, et al., 2008; Goldstein &
Morrison, 2005; McNeill, et al., 2007). Providers may be more concerned about opioid sideeffects in older adults and may therefore not prescribe opioids to older adults (Duncan, et al.,
2008; Goldstein & Morrison, 2005; McNeill, et al., 2007). As noted in the previous section,
cognitively impaired older adults report less pain than cognitively intact older adults (Procter &
Hirdes, 2001; Reynolds, et al., 2008). Cognitive impairment impacts not only the reporting and
therefore the assessment of pain, but also pain medication use among older adults with cognitive

10

impairment. Older adult nursing home residents with cognitive impairment and pain receive less
pain medications when compared to cognitively intact older adult nursing home residents with
pain (Reynolds, et al., 2008).
In addition to the prescribing of medication by clinicians for pain management, hospice
services are also used among older adults for pain management and palliative care. Hospice, as a
health care service provider, provides palliative rather than curative care and coordinates the
provision of medical, emotional and spiritual care for terminally ill patients (life expectancy of
less than 6 months) and their families (Hospice Association of America, 2006). Although
hospice services are often associated with cancer care, older adults with other terminal diseases
(heart failure, chronic obstructive pulmonary disease (COPD) and dementia and Alzheimer
disease) may also receive hospice services for symptom management (Locher, Kilgore,
Morrisey, & Ritchie, 2006; Rodin, 2008).
The provision of hospice services can take place in the home, nursing homes, hospitals or
hospice centers (Hospice Association of America, 2006). Only a small percentage of older adult
nursing home residents who would benefit from hospice services receive them (Buchanan, et al.,
2005; Duncan, et al., 2008). Despite its limited use in nursing homes, hospice services has been
shown to provide high quality end of life care and result in positive outcomes, such as reduced
hospitalizations and improved pain management (Stevenson & Bramson, 2009). There is no
known research which has examined the provision of hospice services within a HCBWP or
associations between the pain experience and hospice services. As HCBWP are an alternative to
nursing home care, knowledge is needed regarding the use of hospice services among older
adults and to make comparisons regarding hospice use between HCBWP participants with and
without cancer.

11

In conclusion, older adults are at high risk for poor pain management and limited hospice
service use. Pain management among older adults is influenced by both characteristics unique to
older adults that affect patient reporting of pain and assessment of pain and therefore the ability
of healthcare providers to manage pain and complex clinical presentation of older adults with
comorbid conditions. Pain that is poorly managed may result in poor pain management
outcomes.
Pain Management Outcomes among Older Adults
Pain management outcomes consist of the end results of pain management. In research
literature, pain control is an outcome of interventions directed toward pain and equated with a
decrease in or improvement of pain levels, functional impairments and frequency due to pain
interventions (Christine Miaskowski et al., 2002; Oliver, Kravitz, Kaplan, & Meyers, 2001;
Shvartzman et al., 2003). Research examining pain management among older adults in both
community and nursing home settings has noted that pain is poorly controlled due to under
assessment and treatment (Landi, et al., 2001; Teno, et al., 2004; Won, et al., 2004). In summary,
pain control is an outcome of pain management and equated with a decrease in or improved
values of pain severity levels, functional impairments and frequency due to pain interventions
(Christine Miaskowski, et al., 2002; Oliver, et al., 2001; Shvartzman, et al., 2003) and among
older adults, pain is poorly controlled (Teno, et al., 2004; Won, et al., 2004).
Outcomes of pain management include the measure of physical function, as a goal of
pain management is improvement of physical functioning (Turk et al., 2003). Physical
functioning is the physical ability of a person to engage in various activities, ranging from simple
mobility to complex activities requiring adaptation to the environment (Bennett, Winters-Stone,
& Nail, 2006). Pain has been shown to have a significant, negative effect on physical

12

functioning and activities of daily living among older adults (Bryant, et al., 2007; Helme &
Gibson, 2001; Onder et al., 2006; Reyes-Gibby, et al., 2002; Soldato, et al., 2007). Among older
adults with cancer, pain, fatigue and insomnia were significant and independent predictors of a
change in physical functioning from pre-diagnosis physical functioning levels to 8 weeks after
initial treatment (B. Given, Given, Azzouz, & Stommel, 2001). Adult long-term cancer survivors
55 and older, with long-term survivorship defined as diagnosed more than 4 years prior, were
more likely than adults 55 and older without a diagnosis of cancer to experience pain and
diminished mobility and activity of daily living limitations (Keating, Norredam, Landrum,
Huskamp, & Meara, 2005). In summary, pain is a predictor of change in physical functioning
among older adults in general and older adults with cancer receiving cancer treatment. Aging
adults with cancer are more likely than aging adults without a diagnosis of cancer to have
diminished physical function. Thus, pain and differences between older adults with and without a
diagnosis of cancer is important to address and manage as pain management may preserve and/or
improve the physical function of older adults.
In conclusion, older adults are at risk of poor pain management outcomes. Pain
management outcomes include pain control and physical function. Older adults with pain are at
risk for poor pain control and diminished physical functioning.
Significance of Study
Older adults, aged 65 and above are at risk for experiencing pain, poor pain management
and poor pain management outcomes (American Geriatrics Society Panel on Persistent Pain in
Older Persons, 2002; Landi, et al., 2001). For the purpose of this study, older adults are adults 65
and older, as 65 is the age at which adults are eligible for Medicare services (Centers for
Medicare & Medicaid Services, 2008). Older adults are at risk for experiencing pain as diseases

13

often associated with pain, such as arthritis, diabetes and cancer which occur at higher rates in
older populations (Bruckenthal & D'Arcy, 2007; Davis & Srivastava, 2003; Freedman, 2002).
Older adults with cancer are more likely to experience pain when compared to older adults
without a history of cancer in both community and nursing home settings (Buchanan, et al.,
2005; Reyes-Gibby, Aday, et al., 2007; Rodin, 2008). Older adults are at risk for poor pain
management due to inadequate recognition and treatment of pain caused by diminished patient
cognitive performance, patient beliefs in pain being part of aging and provider fear of opioid
side-effects and medication interactions (Bruckenthal, 2008; Delgado-Guay & Bruera, 2008;
Duncan, et al., 2008). Pain and poor pain management are frequent and powerful deterrents to
quality of life in older adults (American Geriatrics Society Panel on Persistent Pain in Older
Persons, 2002; Deane & Smith, 2008).
The importance of research addressing pain, pain management and pain management
outcomes has been noted by nursing leadership and U.S. government research agencies
(American Nurses Association, 2010; National Institute of Nursing Research, 2006a, 2006b).
Historically, nurses have been responsible for providing comfort and alleviating suffering from
pain. The nursing profession’s focus on the alleviation of pain continues to be at the forefront of
practice and research directives. Nurses are to address issues such as physical comfort,
discomfort and pain via nursing interventions (American Nurses Association, 2010). Research
examining pain had been prioritized by the National Institute of Nursing Research (NINR) in
order to improve patient quality of life (National Institute of Nursing Research, 2006a, 2006b).
The NINR has emphasized research addressing the elimination of disparities to ensure that all
persons benefit from health care strategies (National Institute of Nursing Research, 2006b) such
as pain management. Older adult HCBW participants experience disparities in regards to pain

14

due to their poverty and therefore, an increased risk of cancer, pain, poor pain management and
poor pain management outcomes (Green, et al., 2003; Ward et al., 2004a). Finally, the Institute
of Medicine has noted that addressing the physical comfort needs of patients is an important
component of patient centered care, which should be a primary focus of health care providers
(Institute of Medicine, 2001).
In summary, the importance of this research which addresses longitudinal associations of
pain, pain management and pain management outcomes among older adult HCBWP participants
and differences in pain, pain management and pain management outcomes between older adults
HCBWP participants with and without a diagnosis of cancer, is supported. The research also
examines the impact of pain, pain management and pain management outcomes on the admission
of older adult HCBWP participants to a nursing home.
The results of this descriptive research will contribute to science by providing
information regarding the pain experience, pain management strategies and pain management
outcomes of a vulnerable and at-risk population that has not been thoroughly examined: poor
older adults with and without cancer. The results of this research can then serve to inform
policies regarding the assessment of pain, the provision and reimbursement of pain management
strategies as well as the quality of care received in HCBWPs in order to prevent or delay nursing
home placement among older adult HCBWP participants.
Research Questions
The following research questions were developed to be addressed in this longitudinal
research study.
1) How does the pain experience differ between older adult HCBWP participants in regards
to diagnosis of cancer over time? How is the relationship between the pain experience

15

and diagnosis of cancer affected by sex, age, race, comorbid conditions, depression and
cognitive functioning over time?
2) How does the pain experience of older adult, HCBWP participants relate to pain
management strategies and pain management outcomes and how does this relationship
differ in regards to diagnosis of cancer over time? How is the relationship between the
pain experience, pain management strategies, pain management outcomes and diagnosis
of cancer affected by sex, age, race, comorbid conditions, depression and cognitive
functioning over time?
3) How do the pain experience, pain management strategies and pain management outcomes
of older adult, HCBWP participants predict the admission and time to admission of older
adult HCBWP participants to a nursing home over time and how does this relationship
differ in regards to diagnosis of cancer while accounting for sex, age, race, comorbid
conditions, depression and cognitive functioning?
Purpose of Study
The primary purpose of this study is to examine longitudinal differences in the pain
experience, pain management strategies and pain management outcomes among older HCBWP
participants with respect to diagnosis of cancer while participating in the HCBWP. The
secondary purpose of this study is to determine what differences exist in how the pain
experience, pain management strategies and pain management outcomes among older adult,
HCBWP participants associates with the admission of older adult HCBWP participants to a
nursing home, with respect to diagnosis of cancer, over the course of time while participating in
the HCBWP. This study will not determine if pain experienced by HCBWP participants is acute,

16

chronic, or cancer-related as this is beyond the scope of the project and of the dataset that will be
used. Instead, pain is viewed in general terms, encompassing all types of pain.
Following the previous overview, in Chapter 2, the revised symptom management model
(Dodd, et al., 2001; Humphreys & et al., 2008) will be discussed and conceptual definitions for
each variable provided. In Chapter 3, a review of the literature describing pain, pain
management, pain management outcomes and predictors of nursing home admission, as well as
the influence of cancer, among older adults will be presented. Study methods and planned
analysis will be described in Chapter 4. Research findings are presented in Chapter 5.
Implications for clinical practice, study limitations and future direction for nursing research
based on this research are discussed in Chapter 6.

17

Chapter 2
The purpose of Chapter 2 is to present the conceptual model that was developed for this
study as well as provide definitions for the concepts within the model. The symptom
management theory (SMT) which was initially developed by the School of Nursing at the
University of California in San Francisco (Larson et al., 1994) and revised by Dodd and
colleagues (2001) and Humphreys an colleagues (2008) (See Figure 1) was selected as a guide
for the development of the conceptual model for the study (Figure 2).
The SMT was created to be utilized with any symptom experienced by patients. Dodd
and colleagues (2001) defined a symptom as a “…subjective experience reflecting changes in the
biopsychosocial functioning, sensations or cognition of the individual” (p 669). Within the SMT
model, the domains of person, health and illness and environment are shown to be in continuous
interaction with each other as well as with the dimensions of symptom experience, symptom
management strategies and pain management outcomes.
For the purposes of this study, the SMT model (Dodd, et al., 2001; Humphreys & et al.,
2008) was adapted to focus on the differences in the pain experience, pain management
strategies, pain management outcomes and admission of older adult HCBWP participants to a
nursing home in regards to diagnosis of cancer (See Figure 2). Pain occurs due to actual or
potential tissue damage and is comprised of both sensory and emotional components
(International Association for the Study of Pain, 2009). Pain is shaped by the contribution of
personal and social characteristics (American Pain Society, 2005) and is highly subjective, being
“…whatever he experiencing person says it is, existing whenever he/she says it does”
(McCaffery & Beebe, 1989, p. 7).

18

Person
Demographic, Psychological, Sociological,
Physiological, Developmental

Symptom
Experience
Perception of
Evaluation
Symptoms
of Symptoms

Components of
Symptom Management
Strategies
Who? (Delivers)
What?
When?
Where?

Response to
Symptoms

How?
To Whom?
How Much?
Why?

Environment
Physical
Social
Cultural

Outcomes
Emotional
Status

Symptom
Status

Functional
Status

Mortality
Self-Care
Morbidity &
Co-morbidity

Costs

Adherence

Health & Illness
Risk Factors
Health Status
Disease &
Injury

Quality of Life

Figure 1. Symptom Management Theoretical Model. Reprinted from “Advancing the Science of Symptom Management” by M.
Dodd, S.Janson, N. Facione, J.Faucett, E. S. Froelicher, J. Humphreys...D. Taylor, 2001 , Journal of Advanced Nursing, 33, p.670.
Reproduced with permission.

19

Domain of
Person
-Age
-Sex
-Race/Ethnicity
-Cognitive
Functioning
-Behaviors
Indicative of
Depression

Adherence

Pain Management
Strategies
-Prescribed Pain
Medication
-Hospice
Services

Pain Experience
-Pain

ADMISSION TO
NURSING HOME

Domain of
Health & Illness
-Comorbid
Conditions
-Diagnosis of
Cancer
-Cancer stage
-Cancer Site

Pain
Management
Outcomes
-Pain Control
-Physical Function

TIME IN HCBWP
TIME TO NURSING HOME
ADMISSION

Figure 2. Conceptual Model for Study: Pain Management Among Older Adults within a Home and Community-Based Waiver
Program Note. Adapted from “Advancing the Science of Symptom Management” by M. Dodd, S.Janson, N. Facione, J.Faucett, E. S.
Froelicher, J. Humphreys...D. Taylor, 2001 , Journal of Advanced Nursing, 33, p.670.

20

Not all older adult HCBWP participants however, are able to make verbal reports of pain.
Dodd and colleagues (2001) then allows for interpretation of the patient’s pain experience by
those who care for the patient for the purpose of intervening. Within the study, proxy reports of
the pain experience, as provided by either healthcare providers or family members, were
recorded if the HCBWP participant was unable to make verbal reports of pain. The American
Geriatric Panel on Persistent Pain in Older Persons recommends the inclusion of information
regarding an older adult’s pain from caregivers and family members in the pain assessment
(2002). The pain experience, pain management strategies, pain management outcomes of pain
control and physical functioning of HCBWP participants, and admission of older adult HCBWP
participants to a nursing home, as affected by the domains of person, health and illness, are
depicted within the conceptual model for this study (See Figure 2). The following sections of this
chapter will describe the conceptual model for this study and define concepts, beginning with the
domain of person.
Domain of Person
Pain is subjective and shaped by person-related characteristics (American Pain Society,
2005). As depicted in Figure 2, the domain of person includes factors that may affect the manner
in which the HCBWP participant views and responds to the pain experience, pain management
strategies and pain management outcomes. For this study, the domain of person is
conceptualized as including the age, sex, race/ethnicity, behavior s indicative of depression and
cognitive functioning of the HCBWP participant.
Age. Age is conceptualized as having an effect on the pain experience, pain management
and pain management outcomes of older adult HCBWP participants despite inconclusive results
from previous research examining associations between pain and age. While some researchers

21

have noted that pain is more prevalent as one ages (Jakobsson, et al., 2003; Tsang et al., 2008),
others have found a negative association between pain prevalence and age (Sawyer, Lillis,
Bodner, & Allman, 2006; Thomas, et al., 2004) or even no association between age and pain
(Reyes-Gibby, Aday, et al., 2007). Differences in these findings regarding the association
between age and pain may be attributed to differences in pain measures used as well differences
in the time period where pain was measured (i.e. daily, weekly etc.).
For this study, age was conceptually defined as one’s chronological age at the time of
assessment and all subjects will be 65 years and older. The age of 65 is generally used in olderadult related research to define older adult samples as the federal government has previously
used the age of 65 as a marker for full social security (Social Security, 2009) and Medicare
benefits (Centers for Medicare & Medicaid Services, 2008), as well as reporting statistics
describing older adult populations (Administration on Aging, 2007).
Sex. Sex was conceptually defined as the distinction between male and female and is
conceptualized as impacting the pain experience, pain management strategies and pain
management outcomes of older adult HCBWP participants. Among older adults, females more
significantly likely to have pain reported than males in NH (Reyes-Gibby, Aday, et al., 2007;
Sawyer, et al., 2006).
Race/Ethnicity. Race/ethnicity is conceptually defined as the racial or ethnic group the
HCBWP participant perceives him or herself as belonging to. Edwards, Fillingim and Keefe
(2001) noted that “race distinguishes major groups of people according to their ancestry and
more or less distinctive combinations of physical characteristics” (p. 134). In comparison,
ethnicity was described as a social concept, providing a group designation, such as a shred
nationality, tribal affiliation, religious faith, shared language or cultural and traditional origins

22

and background (Ezenwa, Ameringer, Ward, & Serlin, 2006; Huff & Kline, 1999; J. L. Riley et
al., 2002). The National Institutes of Health (NIH) provides ethnic categories of “Hispanic of
Latino” and “not Hispanic or Latino” separate from race. For the purposes of the study, the
variable defining race/ethnicity will combine aspects of both race and ethnicity and describe
whether the HCBWP participant is Caucasian, Black, American Indian, Other (includes Asian
and Pacific Islander), Unknown and Hispanic by self or proxy report.
Past research among older adult populations has suggested that differences exist in the
pain experience between racial and ethnic groups among older adults (Im et al., 2007; ReyesGibby, Aday, et al., 2007; Sawyer, et al., 2006; Teno, et al., 2004). Race and ethnicity influences
pain management as well. African American and Hispanic persons are more likely to have their
pain underestimated and undertreated and less likely to receive opioid pain medications when
compared to whites (Cintron & Morrison, 2006; Sengupta, et al., 2010). If pain is under treated,
then poor pain management outcomes may occur (American Geriatrics Society Panel on
Persistent Pain in Older Persons, 2002).
Behaviors Indicative of Depression. Behaviors Indicative of Depression is
conceptualized as the HCBWP participant exhibiting feelings of sadness, persistent anger,
repetitive anxious complaints, sad facial expressions, recurrent crying and withdrawal from
activities of interest, as per the Depression Rating Scale (Burrows, Morris, Simon, Hirdes, &
Phillips, 2000). Additional information regarding the Depression Rating Scale (Burrows, et al.,
2000) will be presented in Chapter 4. Pain is associated with depression among older adults, such
that the presence of one increases the likelihood of the occurrence and worsening the prognosis
of the other (Geerlings, et al., 2002; L. Li & Conwell, 2007). For this study, depression will be

23

conceptualized as influencing the pain experience, pain management strategies and pain
management outcomes and will not be measured as an outcome of pain management.
Cognitive Functioning. Cognitive functioning is conceptualized as the overall mental
performance of an individual regarding memory, decision making capability, awareness, and
communication abilities, as per the Cognitive Performance Scale (Morris et al., 1994). The
Cognitive Performance Scale examines one’s short term memory, decision making capability,
communication abilities and self-performance in eating (Morris, et al., 1994). Additional
information regarding the Cognitive Performance Scale (Morris, et al., 1994) will be presented in
Chapter 4.
Cognitive functioning is conceptualized as influencing the ability of the older adult
HCBWP participant to convey their pain experience (Cohen-Mansfield, 2004; Fisher et al., 2002;
Procter & Hirdes, 2001). Pain management strategies based on a poorly conveyed pain report by
older adults with cognitive impairment may fail to truly meet the pain management strategy
needs of the patient, as previous research has suggested that cognitively impaired older adult
nursing home residents receive less pain medications than cognitively intact older adult nursing
home residents (Reynolds, et al., 2008).
In summary, the domain of person includes age, sex, race, behaviors indicative of
depression and cognitive functioning and is conceptualized (See Figure 2) as including factors
which influence the way an individual views and responds to the pain experience, pain
management strategies and pain management outcomes.
Domain of Health and Illness
The domain of health and illness is comprised of variables unique to the individuals
health and illness state (Dodd, et al., 2001; Humphreys & et al., 2008) and will include variables

24

defining the diagnosis of cancer and comorbidity. Comparisons regarding the pain experience,
pain management strategies, pain management outcomes of (pain control, pain medication costs
and physical function) and the admission of HCBWP participants with and without a diagnosis
of cancer to a nursing home are a purpose of this study.
Diagnosis of cancer, cancer site and cancer stage. Diagnosis of cancer is
conceptualized as the identification of the presence of a malignant cells or neoplasm within the
body and the resulting phases of health care that occur after the identification of malignant cells.
Over the course of time while participating the HCBWP, participants with a diagnosis of cancer
moved through the phases of cancer care as defined by Yabroff and colleagues (Yabroff, Warren,
Knopf, Davis, & Brown, 2005; Yabroff et al., 2009): The initial phase, continuation phase and
terminal phase. These phases are in relation to the time between diagnosis of cancer and death.
In this research, cancer diagnosis was represented as no diagnosis of cancer, initial phase,
continuation phase and terminal phase. The development of the measure of diagnosis of cancer
will be described in detail in Chapter 4. Older adults with cancer are more likely to experience
pain when compared to older adults without a history of cancer in both community and nursing
home settings (Buchanan, et al., 2005; Reyes-Gibby, Aday, et al., 2007; Rodin, 2008).
Diagnosis of cancer was conceptualized as impacting the pain experience, pain management
strategies, pain management outcomes of older adult HCBWP participants and admission of
older adult HCBWP participants to a nursing home.
Diagnosis of cancer can be further characterized by cancer site and stage. Cancer site is
conceptualized as the anatomical location or locations within the body where the cancer is
present and will be described in terms of the primary or initial anatomical location of the cancer.
Cancer stage is conceptualized as the method of defining how progressed or metastasized the

25

cancer is. Cancer stage will be presented in line with Surveillance, Epidemiology and End
Results (SEER) summary stage: in situ, local, regional, distant, un-staged and invasive but site
unknown.
Comorbid conditions. Comorbid conditions are conceptualized as the “…the cooccurrence of health conditions or diseases in reference to an index disease” (Yancik et al., 2007,
p. 276) and as impacting the pain experience, pain management strategies and pain management
outcomes. For this study, cancer is the index disease as comparisons will be made between older
adults in regards to diagnosis of cancer.
Comorbid conditions such as arthritis, diabetes, peripheral vascular disease occur at
higher rates in older adults populations and are conceptualized as affecting the pain experience,
as a higher number of comorbidities is associated with an increased presence of pain (Mao, et al.,
2007; Reyes-Gibby, Aday, et al., 2007). The presence of comorbid conditions may influence
pain management strategies used by healthcare providers, as there may be concerns about drug
interactions between pain medications and medications used to treat comorbid conditions
(Duncan, et al., 2008; Goldstein & Morrison, 2005; McNeill, et al., 2007). Pain management
strategies that are inappropriate for the patient’s pain experience may be utilized, thereby
negatively influencing outcomes. Comorbid conditions were represented by a weighted index
score which reflected both the summed number of comorbidities and the association between
each comorbid condition and pain experienced by older adult HCBWP participants. Additional
information describing the procedure for the development of the measure of comorbid conditions
will be presented in Chapter 4.
In summary, variables defining diagnosis of cancer and comorbid conditions are
conceptualized within the domain of health and illness. The domain of health and illness

26

contributes to the unique pain experience, pain management strategies and pain management
outcomes (Buchanan, et al., 2005; Duncan, et al., 2008; Mao, et al., 2007; Reyes-Gibby, Aday, et
al., 2007; Rodin, 2008).
The Pain Experience
The pain experience is the complex process of perceiving, evaluating, and responding to
a sensory input (Dodd, et al., 2001; Humphreys & et al., 2008; Kandle, Schwartz, & Jessell,
2000; Turk & Okifuji, 1999) and for the purposes of the study, the pain experience is the verbal
report of or behavioral responses to pain defining both pain frequency and pain intensity which
will be examined over the time of participation in the HCBWP.
The Pain Experience is conceptualized as being affected by the domains of person and
health and illness. For example, aging and disease processes may alter nerve function,
predisposing one to pain production (Backonja, 2003; Gibson & Helme, 2001; Pasero, 2004;
Zheng, et al., 2000). The pain experience is initiated with sensory inputs, which are sensed by
nociceptors and transmitted to the dorsal horn in the spinal cord and to the brain where the
sensory inputs are then interpreted as painful or not painful (Basbaum, Bushnell, & Devor, 2005;
Kandle, et al., 2000). Once a sensory input is interpreted, the painful stimulus is evaluated by the
patient and factors such as intensity, location, temporal nature, frequency and the affective
impact of pain are used by the patient to appraise and characterize the pain (Dodd, et al., 2001;
Humphreys & et al., 2008). Whether or not a stimulus is perceived as pain depends not only on
the nature of the stimulus, but also on the context within which the stimulus was experienced, as
well as memories and emotions in relation to pain (Basbaum, et al., 2005).
The patient’s evaluation of pain is influenced by pain-related beliefs developed over
one’s lifetime (Turk & Okifuji, 2002). Once a patient has evaluated and characterized the pain,

27

the patient responds to the pain. Response to pain may include behaviors that communicate pain
to others such as verbal reports, posturing, grimacing, moaning or crying out when moved
(Goldstein & Morrison, 2005; NANDA, 2005; Turk & Okifuji, 1999). Because the sample for
the study include older adult HCBWP participants who are cognitively impaired and therefore,
not able to provide verbal reports of pain, pain behaviors such as posturing, grimacing, moaning
or crying out when moved are used by assessors to define pain frequency and pain intensity.
Pain. For this study, pain is both the frequency and the intensity of the pain experienced
by the HCBWP participant, as measured by the MDS Pain Scale (Fries, Simon, Morris,
Flodstrom, & Bookstein, 2001). Pain frequency is the varying time dimension which
characterizes a temporal aspect of pain: the frequency with which pain occurs in a certain period
of time (Jensen, 2003; Lenz, Pugh, Milligan, Gift, & Suppe, 1997). Pain intensity is associated
with the sensory dimension of pain and is the level of severity, strength or overall magnitude of
the pain (C. S. Cleeland & Ryan, 1994; R. H. Dworkin et al., 2005; Jensen, 2003; Lenz, et al.,
1997). Pain frequency and intensity are both assessed via patient or proxy report. Pain frequency
and intensity are capable of changing over time in response to person and health and illness
characteristics as well as pain management interventions (Harris et al., 2005; Henly, et al., 2003;
Humphreys & et al., 2008).
In summary, the dimension of pain experience addresses the patient perceiving,
evaluating and responding to pain. The pain experience dimension includes the concept of pain
to characterize the patient’s pain experience. Pain will be measured over time, to examine pain
frequency and pain intensity over the course of care within the HCBWP. Within the conceptual
model (See Figure 2), the pain experience dimension is depicted as influencing and being

28

influenced by the pain management strategies dimension (Bruckenthal & D'Arcy, 2007; Dodd, et
al., 2001; Humphreys & et al., 2008).
Pain Management Strategies
The pain management strategies dimension included what is done for and by the older
adult HCBWP participant to manage pain (Dodd, et al., 2001; Humphreys & et al., 2008). Pain
management strategies are based on the assessment of the older adult HCBWP participant’s pain
experience (Bruckenthal & D'Arcy, 2007; Dodd, et al., 2001). Pain management strategies can
be pharmacological and non-pharmacological (Deane & Smith, 2008; JCAHO, 2000) and within
this study, pain management strategies were conceptualized as including prescribed pain
medication and the provision of hospice services.
Prescribed Pain Medications. Prescribed pain medications are conceptually defined as
medications that are prescribed by health care providers for pain management. In this study,
pain medications were characterized as to whether they are non-opioid, opioid or adjuvant as
per the World Health Organization Analgesic Ladder (NCI, 2006; WHO, 1996). Whether
opioid, non-opioid or adjuvant pain medications were prescribed depends on the assessment of
the pain experience (Humphreys & et al., 2008; NCCN, 2010; NCI, 2010b; WHO, 1996). For
example, for mild pain, acetaminophen alone may have been prescribed, whereas for severe pain,
an opioid such as morphine may have been prescribed (Humphreys & et al., 2008; NCCN, 2010;
NCI, 2010b; WHO, 1996).
In addition to being directed by the pain assessment, prescribed pain medications were
conceptualized as being influenced by patient and health and illness characteristics. For
example, non-steroidal anti-inflammatory medications should be used with caution in older
adults with renal impairment (Hanlon, Guay, & Ives, 2005) and older adults are at greater risk

29

for opioid side effects such as sedation, nausea, vomiting , respiratory depression and
constipation (Weiner & Hanlon, 2001). Thus, providers may be concerned about medication side
effects among older adults and may therefore limit what they do prescribe based on person and
health and illness-related factors (Duncan, et al., 2008; Goldstein & Morrison, 2005; McNeill, et
al., 2007).
Hospice Services. Hospice services were conceptually defined as the provision of
palliative care including medical, emotional and spiritual care for terminally ill patients (life
expectancy of less than 6 months) and their families (Hospice Association of America, 2006).
Hospice care serves to identify goals for care, provide effective prevention and management of
end of life symptoms such as pain and dyspnea and provide attention to psychosocial and
spiritual issues and prevention of suffering, and assist with completion of developmental tasks
related to the dying process and successful negotiation of the grief and bereavement processes
(Ersek & Wilson, 2003). The provision of hospice services can take place in the home, nursing
homes, hospitals or hospice centers (Hospice Association of America).
In summary, for this study, the pain management strategies dimension focused on
prescribed pain medications and hospice services. Among older adults, pain management
strategies were conceptualized as being influenced by the pain experience and person and health
and illness characteristics (Bruckenthal & D'Arcy, 2007; Dodd, et al., 2001; Duncan, et al., 2008;
Hanlon, et al., 2005; Humphreys & et al., 2008; Weiner & Hanlon, 2001). In this study,
relationships between prescribed pain medications and hospice services, the pain experience, and
the domains of person and health and illness were assessed over time. In addition to being
influenced by the pain experience dimension, the pain management strategies dimension affects
the pain management outcomes dimension.

30

Pain Management Outcomes
Outcomes are the end results of care (Patrick, 1997). The pain management outcomes
dimension, as depicted in Figure 2, includes the end results of pain management strategies, is
influenced by pain management strategies and has an effect on the pain experience (Dodd, et al.,
2001; Humphreys & et al., 2008). For this study, pain management outcomes included pain
control and physical function and were examined over the period of time that the older adult
participates in the HCBWP. The outcomes of pain control and physical function are described in
further detail in the following sections.
Pain Control. Past definitions of pain control are aligned with its role as an outcome of
pain management strategies. Pain control is the result of pain assessment and analgesic
treatments (Allard, Maunsell, Labbe, & Dorval, 2001; Shvartzman, et al., 2003). Pain control
was conceptually defined as an outcome of pain management whereby pain is perceived by the
patient or proxy as limited or decreased from a previous pain level and is an indicator of the
effectiveness of pain management strategies (Allard, et al., 2001; Christine Miaskowski, et al.,
2002; Oliver, et al., 2001; Shvartzman, et al., 2003). Pain control is therefore associated with
perceived medication effectiveness.
For this study, pain control was represented by the HCBWP participant’s or proxy’s
response to the statement “pain controlled by medication”. The representation of pain control by
the HCBWP participant’s or proxy’s response to the statement “pain controlled by medication”
is a limitation of this study, as pain control was limited to medication only. The researcher
acknowledges that pain control is an outcome of any pain management strategy, including pain
medications, non-pharmacological strategies such as heat, cold, massage etc, or procedures such
as nerve injection.

31

As a pain management outcome, pain control was influenced by pain management
strategies and impacts the pain experience. For example, pain management strategies are
implemented and reassessment of the pain experience takes place after pain management
(NCCN, 2006). If pain management does not result in pain control, then pain management
strategies should be altered and the pain experience reevaluated to determine if pain control has
occurred (NCCN, 2006). This inter-relationship between the pain experience, pain management
strategies and pain management outcomes is depicted in Figure 2.
Physical Function. Physical function is an outcome of pain and pain management
strategies as previous research has shown that pain has a significant, negative effect on physical
function (Onder, et al., 2006; Reyes-Gibby, et al., 2002; Soldato, et al., 2007). The concept of
physical function has been noted to be within the overarching concept of functional status
(Ferrans, Zerwic, Wilbur, & Larson, 2005) as depicted in the Symptom Management Theory
conceptual model (See Figure 1) (Dodd, et al., 2001; Humphreys & et al., 2008).
Physical functioning has been noted to be “…a broad concept that includes physical
abilities that range from simple mobility to engagement in complex activities that require
adaptation to an environment” (Bennett, et al., 2006, p. 41). The Initiative on Methods,
Measurement and Pain Assessment in Clinical Trials (IMMPACT) supports the use of activities
of daily living as a fundamental measure of physical functioning to determine the effect of pain
and pain management strategies on physical function (R. H. Dworkin, et al., 2005; Turk, et al.,
2003). For this study, physical functioning was conceptualized as the dependency of the older
adult HCBWP participant in activities of daily living (ADL).
This study examined predictors of the admission of older adult, HCBWP participants to a
nursing home. Dependencies in ADL are a significant predictor of nursing home admission

32

(Gaugler, Duval, Anderson, & Kane, 2007). Therefore, physical function was conceptualized in
Figure 2 as affecting the admission of HCBWP participants to a nursing home. Physical function
is influenced by the pain experience and pain management strategies. Physical functioning was
represented by a count of the following ADLs that the HCBWP participant is dependent in:
dressing, personal hygiene, toilet use, bathing and eating. Additional information describing the
summed ADL dependency scale will be presented in Chapter 4.
In summary, the pain management outcomes dimension includes the end results of pain
management and consists of pain control and physical function. This study examined
relationships among the pain experience, pain management strategies, pain management
outcomes and completed an exploration of the associations among pain experience, pain
management strategies, pain management outcomes and admission to the nursing home.
Admission to a Nursing Home
For this study, the domains of person and health and illness, the pain experience, pain
management strategies and pain management outcomes were conceptualized as impacting the
admission of older adult HCBWP participants to a nursing home.
Admission to a Nursing Home. Admission to a Nursing Home, as conceptualized in
Figure 2, is conceptually defined as the movement of older adult, HCBWP participants to a
nursing home facility, without returning to the HCBWP. Admission to a Nursing Home is part of
the healthcare services continuum of older adults, where older adults move from community to
institutionalization (L. Li & Zullo, 2003; Williams, 2001). A goal of the HCBWP is to prevent
or delay admission to a nursing home (Fries, et al., 2002).
The concept of time was not included in the original SMT (Dodd, et al., 2001;
Humphreys & et al., 2008) but is addressed in the developed conceptual model (Figure 2). Time

33

was included in the adapted model to account for the temporal aspects of pain, pain management
and pain management outcomes as well as to conceptualize the long-term care continuum that
exists for older adults (L. Li & Zullo, 2003; Williams, 2001) . Within the conceptual model
depicted in Figure 2, there are two time-related variables: Time in the HCBWP and Time to
Nursing Home Admission. These two time-related concepts are clarified in the following
paragraphs.
Time in HCBWP. Time in the HCBWP is conceptually defined as the period of time in
months that the older adult is a HCBWP participant among the pain experience, pain
management strategies, pain management outcomes which are influenced by the domains of
person and health and illness (Dodd, et al., 2001; Humphreys & et al., 2008). Even though Time
in the HCBWP is not specifically included in the research questions as a study variable, it will be
included in the analysis as a covariate in order to account for differences between subjects in
time spent in the HCBWP. The statistical reason for the inclusion of Time in the HCBWP as a
covariate will be described in more detail in Chapter 4.
Time to Nursing Home Admission. The timeline in the conceptual model also depicts
time for Time to Nursing Home Admission, which is conceptualized as the time in months that a
HCBWP participant spends in the HCBWP prior to admission to a nursing home, without
returning to the HCBWP. As the goal of the HCBWP is to prevent or delay admission to a
nursing home (Fries, et al., 2002), the concept of time to nursing home admission is included in
the model to account for the influence of the pain experience, pain management strategies and
pain management outcomes has on the above goal of the HCBWP.
In summary, the concept of time was included in the adapted model (Figure2) to
conceptualize the longitudinal design of this study in order to examine the relationships between

34

dimensions and domains and predictors of time before admission of older adult HCBWP
participants to a nursing home within the long-term care continuum.
Conclusion of Chapter 2
The conceptual model for this study accounts for the interactive nature of the pain
experience, pain management strategies, pain management outcomes dimensions and admission
to a nursing home along with the domains of person, and health and illness among older adult
HCBWP participants. In this study, relationships between variables in the pain experience, pain
management strategies, pain management outcomes dimensions and the concept of admission to
a nursing home will be examined over time while accounting for differences among HCBWP
participants according to variables with the domains of person and health and illness.
Chapter 3 will provide a literature review as to what is known regarding the issue
pain, pain management and pain management outcomes among older adults as well as
the need for and significance of research addressing the pain management needs of older
adult HCBWP participants.

35

Chapter 3
The purpose of Chapter 3 was to provide an overview and synthesis of the literature
regarding the Medicaid Home and Community Based Waiver Program (HCBWP) and the pain
experience, pain management strategies, pain management outcomes and admission to a nursing
home of older adult HCBWP participants. The developed conceptual model (See Figure 2)
serves as a guide for the following literature review. The model domains (Person and Health and
Illness), dimensions (Pain Experience, Pain Management Strategies, and Pain Management
Outcomes) and admission will serve as section headings. Because the model was conceptualized
to be within the HCBWP, a review of the literature regarding the HCBWP will be provided first.
The review of the literature describing the HCBWP will present the overall history of the
HCBWP as well as describe HCBWP policies, HCBWP admission criteria and HCBWP client
characteristics.
The Home and Community Based Waiver Program
The Medicaid Home and Community Based Waiver Program (HCBWP) was established
by the Omnibus Budget Reconciliation Act of 1981 (OBRA-81) and was incorporated into the
Social Security Act at Section 1915(c) (Duckett & Guy, 2000). Medicaid is a joint federal and
state program that provides health insurance coverage to provide services for certain categories
of low-income individuals, including children, pregnant women, parents of eligible children,
persons with disabilities and older adults (Centers for Medicare & Medicaid Services, 2005).
Prior to the OBRA -81, Medicaid coverage of home and community-based services was limited
and favored institutional-based care for long-term care needs (Duckett & Guy, 2000; Marek et
al., 2005). The OBRA-81 supported the expansion of Medicaid coverage of home care benefits

36

provided by individual states via “Medicaid waivers” (L. Li & Zullo, 2003; Shugarman, et al.,
1999).
Medicaid waivers “waive” certain regulatory requirements regarding individual state’s
Medicaid plans thereby easing the expansion of Medicaid-covered home and community-based
services (Kitchener & Harrington, 2003; Kitchener, Ng, & Harrington, 2004; Shugarman, et al.,
1999). HCBWP services may be provided to individuals who are elderly and disabled,
physically disabled, developmentally disabled, mentally retarded, mentally ill or suffer from
chronic debilitating diseases, such as Acquired Immune Deficiency Syndrome caused by the
Human Immunodeficiency Virus (Duckett & Guy, 2000; N. A. Miller, Elder, Kitchener, Kang,
& Harrington, 2008). Services covered under the HCBWP includes homemaker services,
respite care, adult day care, home modifications, transportation, medical supplies, personal
emergency response system, private duty nurse, counseling, home delivered meals, physical and
occupational therapy, and personal care supervision (Kitchener, et al., 2004; Shugarman, et al.,
1999). Services can be provided in the participant’s home or in residential-care or assisted living
facilities (Kitchener, Hernandez, Ng, & Harrington, 2006). The overall purpose of the HCBWP
is to provide Medicaid covered home care services to those 18 and older who are at risk for
nursing home placement because of the need for assistance with activities of daily living or
medical services, in order to delay or avoid the high costs of institutional long-term care
(D'Souza, James, Szafara, & Fries, 2009; Sands et al., 2008; Shugarman, et al., 1999)
HCBWP costs are controlled by individual states through the use of four mechanisms,
which allow the states a great deal of flexibility in the designs of HCBWPs. Variations in the
use of these four cost-containment mechanisms exist at both the inter and intra-state level
(Kitchener, et al., 2004; Kitchener, Ng, & Harrington, 2007). First, states must demonstrate

37

HCBWP participant Medicaid costs are no greater than the Medicaid costs of comparable
nursing home level care (Kitchener, Carrillo, & Harrington, 2003; Kitchener, et al., 2007).
Second, states may limit the number of available HCBWP participant slots in order to limit total
costs (Kitchener, et al., 2007). Third, states at their discretion, may set HCBWP participant
medical and financial eligibility criteria and spending caps on provided services (Kitchener, et
al., 2007). Fourth, states may limit waiver programs to specific geographical areas (i.e. a county
with high need for long-term care services) and population groups (i.e. the elderly, persons with
HIV/AIDS, persons with mental illness) (Duckett & Guy, 2000; Kitchener, et al., 2007; N. A.
Miller, et al., 2008). Therefore, states have at their disposal methods to keep HCBWP costs
under control. However, concerns have been raised regarding the effect of cost-containment
mechanisms on access to HCBWP services and other HCBWP participant outcomes (Tonner &
Harrington, 2003).
HCBWP cost-containment mechanisms may restrict individual choice of healthcare
services and result in unnecessary institutionalization. Kitchener et al. (2004) performed a
nationwide survey of cost-containment strategies utilized by 229 different waiver programs.
There are inter and intra-state variations in financial eligibility criteria, which varied from
between 300% of Supplemental Security Income (SSI) to 100% of SSI as well as other financial
eligibility criteria based on other poverty indicators (Kitchener, et al., 2004). Kitchener et al.
noted that 33% of the states were using more stringent financial criteria for their HCBWPs than
for nursing home eligibility, a policy which would clearly favor nursing home admissions.
D’Souza et al. (2009) explored variation in HCBWP participant outcomes in response to major
state-level budget restrictions in Medicaid funding as well as policies which increased medicaleligibility threshold for HCBWP admission. D’Souza et al. examined 112,182 HCBWP

38

assessment records collected over a four-year time period. During time periods of worsening
state budgets, HCBWP participants experienced decreases in the amount of care hours provided
with resulting increases in emergency room use by HCBPW participants and increased caregiver
burden when compared to time periods with the least state-level budget restrictions (D'Souza, et
al., 2009). In summary, HCBWPs must strive to work within budgetary constraints while also
meeting participant needs, as HCBWP participant outcomes appear to suffer with worsening
budget situations (D'Souza, et al., 2009).
Despite difficulties with meeting both budget restrictions and participant needs, the use of
HCBWP services results in improved participant outcomes compared to populations not
receiving HCBWP services. Medicaid-eligible older adults with disabilities in activities of daily
living (ADL) receiving home care services experienced a significantly reduced risk of death
when compared to Medicaid-eligible older adults with disabilities in ADL not receiving home
care services (Albert, Simone, Brassard, Stern, & Mayeux, 2005). Marek et al. (2005) compared
clinical outcomes (cognitive decline, functional status decline, depression, pressure ulcers and
incontinence) of HCBWP participants aged 50 and older to nursing home residents of a similar
case mix over a 2 year time period. HCBWP participants had significantly better outcomes of
functional status, cognition, depression and incontinence over time when compared to the
matched nursing home residents (Marek, et al.). In summary, HCBWP services benefit HCBWP
participants when compared to persons not receiving HCBWP services.
Although there are known benefits from receiving HCBWP services among older adults,
little is known about any benefits of pain management and pain management outcomes
experienced by older adult HCBWP participants with and without a diagnosis of cancer. Pain is
reported by 53%-75% of HCBWP participants (Fries, James, & Aliaga, 2004; L. Li & Conwell,

39

2007; Spoelstra, Given, von Eye, & Given, 2010). Cancer prevalence rates among older adult
HCBWP participants range from 13% (Spoelstra, et al., 2010) to 15% (L. Li, 2005) and this rate
is slightly higher than the reported 11% of nursing home residents who have a diagnosis of
cancer at admission (Buchanan, et al., 2005; Rodin, 2008). Therefore, over half of older adult
HCBWP participants experience pain and cancer occurs at a higher rate than among older adult
nursing home residents.
HCBWP participants must meet financial eligibility below 300% of the federal poverty
level (L. Li & Zullo, 2003) and are therefore considered impoverished. Poverty is associated
with increased pain and poor pain management and pain management outcomes. Fuentes, HartJohnson and Green (2007) found that as neighborhood socioeconomic status decreased (became
poorer), pain and disability increased among older black and white older adults. Medicaid and
Medicare recipients are at risk for poor pain management due to healthcare providers not
prioritizing pain management as well as restrictive drug coverage benefits (Jost, 2000).
Therefore, older adult HCBWP participants are at higher risk for pain and poor pain management
and pain management outcomes when compared to wealthier older adults (Fuentes, et al., 2007;
Jost, 2000).
In summary, the HCBWP was developed to provide Medicaid covered home care
services to those 18 and older who are at risk for nursing home placement due to the need for
medical care or assistance in activities of daily living in order to delay or avoid the high costs of
institutional long-term care (D'Souza, et al., 2009; Shugarman, et al., 1999). Home and
Community-Based Waiver Programs utilize various strategies to control the costs of health care
services which result in decreased access to health care services and poor health outcomes
(D'Souza, et al., 2009; Dodd, et al., 2001; Kitchener, et al., 2004; Tonner & Harrington, 2003).

40

Healthcare services directed toward pain and pain management among older adults are
both frequent and costly occurrences in the healthcare system. Pain is the most common reason
individuals seek medical attention (JCAHO, 2000) and older adults who experience pain utilize
more healthcare services than older adults without pain (Mossey & Gallagher, 2004). The
financial costs to the health care system related to pain management have been reported to
exceed $4000.00 per year for persons with chronic pain (Turk, 2002) and average $891.00 per
month for persons diagnosed with cancer (Fortner et al., 2003). While cost-containment is a key
HCBWP strategy, very little is known regarding pain, pain management and pain management
outcomes in HCBWP. Because of the high costs of pain and pain management, research is
needed to describe differences in the pain experience, pain management and pain management
outcomes of HCBWP participants with and without a diagnosis of cancer.
Domains of Person and Health and Illness
Within this research, participant characteristics are represented by the domain of person
and domain of health and illness (Figure 2). The domain of person includes variables defining
age, sex, race/ethnicity, participant cognitive function, depression. The domain of health and
illness is represented by variables defining comorbid conditions and cancer. These variables are
frequently associated with pain in research examining the pain experience, pain management and
pain management outcomes among older adults. The following section of the literature review
will examine each variable listed above as it relates to the pain experience, pain management and
pain management outcomes among older adults in general.
Domain of Person. Variables within the domain of person comprise personal factors
that may affect the manner in which the HCBWP participant views and responds to the pain
experience, pain management strategies, pain management outcomes (Dodd, et al., 2001;

41

Humphreys & et al., 2008). Differences among persons in response to the pain experience, pain
management strategies and pain management outcomes occur at the genetic level, causing
differences in pain processing as well as drug metabolizing enzyme production (Webster, 2008)
and opioid efficacy (Reyes-Gibby et al., 2007). For this study, variables within the domain of
person included age, sex, race/ethnicity, cognitive functioning and depression. These variables
are frequently examined in research examining the pain experience, pain management and pain
management outcomes among older adults.
Age. The following review of the association of age to the pain experience, pain
management and pain management outcomes presents a synthesis of findings from research
examining the pain experience, pain management strategies and pain management outcomes
among community dwelling older adults and older adult nursing home residents. The age of
older adult HCBWP participants tends to be younger than older adult nursing home residents
with 24% of HCBWP participants under 65, vs. 11% of nursing home residents, 36% of HCBWP
participants are aged 65-79% vs. 27% of nursing home residents and 40% of HCBWP
participants are aged 80 and older, compared to 62% of nursing home residents (Fries, James, &
Aliaga, 2004). The present research focuses on HCBWP participants 65 and older. Research
examining the association of age and the pain experience, pain management strategies and pain
management outcomes among older adults has compared older adults to younger adults as well
as developed group comparisons among older adults (i.e. comparing 65-74 year olds to 85 and
older).
Aging contributes to widespread changes in the cellular and neuro-chemical substrates of
the nociceptive system (Gibson & Farrell, 2004; Helme & Gibson, 2001). In a normally
functioning central nervous system, pain stimuli causes the release of endogenous opioids,

42

which then attach to opioid receptors in the periaqueductal grey and dorsal horn and modulate
pain transmission, decreasing pain perception (Basbaum, et al., 2005; Heinricher, 2005). Aging
leads to diminished endogenous opioid mechanisms and differences in modulation to painful
cold stimuli in healthy adults aged 51 and older (n=48) when compared to populations aged 1825 (n=45) (R. R. Edwards, et al., 2003). Older adults aged 73 and above (n=10) experienced
slower resolution of post-injury pain and tenderness from a capsaicin patch than younger adults
aged 23-36 (n=10) due to age-related changes in the central nervous system (Zheng, et al., 2000).
In a review of human and animal research examining associations between aging and pain,
Gagliese and Farrell (2005) noted that although data from the reviewed research was inconsistent
and not robust, older adults experience impairment of nerve restoration after nerve injury and
that older adults are at greater risk of persistent pain following injury or disease. In summary,
aging leads to physiological changes in the nervous system that influence the perception and
production of pain such that older adults experience an increased and prolonged response to
injury and pain stimuli (Basbaum, et al., 2005; R. R. Edwards, et al., 2003; Gagliese & Farrell,
2005; Gibson & Farrell, 2004; Heinricher, 2005; Helme & Gibson, 2001; Zheng, et al., 2000),
thereby placing older adults at risk of pain that persists or in need of extended pain management.
The results of research examining the association between pain prevalence and age have
been inconclusive, with some researchers finding that pain is more prevalent as one ages
(Jakobsson, et al., 2003; Tsang, et al., 2008), while others have noted a negative association
between pain prevalence and age (Sawyer, et al., 2006; Soldato, et al., 2007) or no association
between age and presence of pain (C. Given, Given, Azzouz, Kozachik, & Stommel, 2001).
Jakobsson et al. examined the prevalence of pain among 4, 093 community dwelling and
institutionalized older adults aged 75 and older and compared those who reported pain (n=1,654)

43

with those who did not report pain (n=2,439). The prevalence of pain was significantly higher
within the older age groups. Among 75-79 year olds, 34.1% reported pain, while in the 90 and
older group, 50.1% reported pain (Jakobsson, et al., 2003). Tsang and colleagues (2008)
examined associations between age and prevalence of pain among 42,249 adults from 17
different countries. Tsang et al. did not clarify if any of the participants were community
dwelling or institutionalized. Mean age varied from 35 to 51. The prevalence of pain was
positively associated with age, with 21%-35% of participants aged 18-35 reporting pain while
47%-73% of participants aged 66 and older reported pain, after controlling for comorbid
conditions (Tsang, et al.).
As noted in the previous paragraph, Given et al. (2001) noted no association between age
and presence of pain while Sawyer et al. (2006) and Soldato et al. (2007) each reported a
negative association between pain prevalence and age. Given et al. examined predictors of pain
and fatigue among 841 patients aged 65 and older who had been diagnosed with breast, colon,
lung or prostate cancer. There was no significant association between age and presence of pain.
Sawyer et al. reported that lower odds of substantial daily pain were associated with older age
among 27,628 nursing home residents aged 65 and above. Soldato et al. utilized a longitudinal
analysis of data from 1, 520 randomly selected elderly European home care patients to assess
associations between pain and disability over a one year time period and found that subjects with
daily pain were significantly more likely to be younger (2007). While Thomas, Peat, Harris,
Wilkie and Croft (2004) also noted a negative association between pain prevalence and age, the
authors did report a positive association between age and pain interference with daily activities
among . Utilizing cross-sectional data from 11, 230 adults aged 50 and older, Thomas et al.
found a significant increase in pain interference in activities of daily living with increased age.

44

Research results reporting the association between pain severity and age are consistent.
Zyczkowska et al. explored pain among 193,158 institutionalized and community-dwelling older
adults (age 65 and older). Based on 5-year age groups beginning at age 65, the mean reported
pain severity score was lower with each age increment in age for both men and women
(Zyczkowska, et al.). Likewise, Baker and Green (2005) presented findings noting a significant
difference in pain severity between Black and White American adults less than 50 vs. Black and
White American adults 50 and older who were seeking care for pain at a pain management
clinic. Overall, research examining associations between pain and age among both community
dwelling and institutionalized older adults found that while it is not clear whether pain
prevalence increases or decreases with age (Jakobsson, et al., 2003; Sawyer, et al., 2006;
Soldato, et al., 2007; Tsang, et al., 2008), interference from pain may increase and pain severity
diminish as one ages (Thomas, et al., 2004).
In addition to being associated with pain prevalence, pain severity and pain interference,
age is also associated with pain management. Among community-dwelling older adults with
dementia, insufficient analgesia was three times more likely for each additional year of age
(Shega, et al., 2006). Won et al. (2004) noted that older adult nursing home residents 85 and
older were significantly less likely to have received analgesics when compared to younger older
adult groups. Among community-living older adults, Landi and colleagues (2001) found that
older adults 85 and older were significantly less likely than younger older adults to receive pain
medication. Zyckowska (2007) reported that among older adults who were receiving home care
or nursing home care and reporting severe pain, older age groups were less likely than younger
older adults (aged 65-69) to receive stronger opioid pain medications (codeine, morphine) as
odds ratios for receiving opioid pain medications dropped from 0.76 for older adults aged 70-74

45

to 0.42 for older adults aged 100-115 (Zyczkowska, et al., 2007). In summary, older adults are at
increased risk of not receiving adequate pain management when compared to younger adults
(Landi, et al., 2001; Shega, et al., 2006; Won, et al., 2004; Zyczkowska, et al., 2007).
In summary, research that has examined the associations between age, the pain
experience and pain management has taken place among community-dwelling older adults and
older adult, nursing home residents. Aging may cause physiological changes in one’s perception
and resolution of pain (Basbaum, et al., 2005; Gagliese & Farrell, 2005; Gibson & Farrell, 2004;
Heinricher, 2005; Helme & Gibson, 2001). Age is positively associated with pain interference
and negatively associated with pain severity (Baker & Green, 2005; Thomas, et al., 2004;
Zyczkowska, et al., 2007). Results describing associations between age and pain prevalence are
inconsistent (B. Given, et al., 2001; Jakobsson, et al., 2003; Sawyer, et al., 2006; Soldato, et al.,
2007; Tsang, et al., 2008).
Sex. The following review of the association of sex to the pain experience, pain
management and pain management outcomes will present findings from research examining the
pain experience, pain management and pain management outcomes among community-dwelling
older adults and older adult nursing home residents. Differences between sexes must be
considered research examining pain, pain management and pain management outcomes.
Females experience an overall higher prevalence of pain (Chou & Chi, 2007; C. Given, et
al., 2001; Jakobsson, et al., 2003; Reyes-Gibby, Aday, et al., 2007; Soldato, et al., 2007), more
frequent pain (Soldato, et al., 2007), more severe pain (Reyes-Gibby, Aday, et al., 2007;
Zyczkowska, et al., 2007) and worsening pain over time (Rosso, et al., 2008) when compared to
males. Female, community-dwelling and nursing home resident older adults consistently report a

46

higher prevalence of pain as well as experience more frequent and severe pain when compared to
male, community-dwelling and nursing home resident older adults.
Research results regarding differences in pain management between older adult males
and females, however, are inconsistent. Won et al. (2004) presented results of lower use of
opioid pain medications by men (32% vs. 40%) when compared among older adult nursing home
residents with and without persistent pain (n=21,380). These results seem logical after
considering that female older adults report a higher prevalence of pain and more severe pain, as
noted in the previous section. However, Soldato et al. (2007) found no difference in pain
medication use between community-dwelling male and female older adults who reported daily
pain. More evidence regarding the association between sex and pain management among older
adults is needed. In summary, it is unclear from previous research if differences in prescribed
pain medications exist between male and female older adult nursing home residents and
community-dwelling older adults.
There is limited research examining the association between sex and pain management
outcomes among older adults. In regards to physical function, Soldato et al.(2007) noted that the
association between daily pain and disability was stronger among female community-dwelling
older adults when compared to male community-dwelling older adults. Deimling, Bowman and
Wagner (2007) reported sex as a significant predictor of the effects of pain and fatigue on
physical functioning among long-term (5 years) cancer survivors aged 60 and older, after
controlling for other personal, age-related and cancer-related characteristics as well as weakness,
pain, and fatigue. Results indicate that being female increases difficulty with physical
functioning, after controlling for other personal, age-related and cancer-related characteristics as
well as weakness, pain, and fatigue. In summary, limited research results find that among

47

community-dwelling older adults, females are more likely to experience pain-related diminished
physical functioning.
Differences between the sexes in regards to pain, pain management and pain management
outcomes have been explained by the increased likelihood of females to experience diseases
which result in chronic pain as well as psychosocial differences between males and females
(Greenspan et al., 2007; Wiesenfeld-Hallin, 2005). Chronic exposure to pain results in central
sensitization, where neurons in the central nervous system (CNS) become hyper-excitable
(Samad, 2004). Prolonged central sensitization leads to permanent structural changes in the CNS
that includes death of inhibitory neurons, replacement with new excitatory neurons, and creation
of aberrant excitatory synaptic connections (Samad, 2004). These changes collectively lead to a
state of fixed sensitization that is resistant to most analgesic medications (Samad, 2004). In
addition, central sensitization increases the size of the field in which dorsal field neurons respond
to stimuli, thereby extending pain sensitivity well past the site of injury (Miaskowki, 2004).
Psychosocial factors specific to women may account for their increased pain sensitivity: hypervigilance toward threatening situations, greater bodily monitoring and greater prevalence of
anxiety and depression when compared to men (Rollman, Abdel-Shaheed, Gillespie, & Jones,
2004). Pain is a stressful experience and differences between the sexes in stress responses,
including the inflammatory response due to stressors, contribute to sex differences in pain
(Greenspan, et al., 2007).
In summary, while older adult females report more prevalent, severe and worsening pain
over time than older adult males (C. Given, et al., 2001; Jakobsson, et al., 2003; Reyes-Gibby,
Aday, et al., 2007; Rosso, et al., 2008; Zyczkowska, et al., 2007), there are inconsistent results
regarding differences between community dwelling and nursing home resident older adult males

48

and females regarding pain management (Soldato, et al., 2007; Won, et al., 2004) and the
association between sex and pain on physical functioning (Deimling, et al., 2007; Soldato, et al.,
2007). Biological and psychosocial differences between males and females account for
differences in the pain experience, pain management and pain management outcomes of older
adult HCBWP participants.
Race/Ethnicity. Race/ethnicity was utilized as a variable which allowed for comparison
between racial/ethnic groups in terms of the pain experience, pain management and pain
management outcomes. Racial breakdown of HCBWP participants has been noted to be
predominantly “white” (77%), with 23% “black” and 1% “other” (Fries, James, & Aliaga, 2004).
The following is a review of the association between race/ethnicity and the pain experience, pain
management and pain management outcomes among older adults residing in the community or
nursing homes. As noted by Ezenwa, Ameringer, Ward and Serlin (2006) there is misuse and
confusion regarding the use of race and ethnicity throughout research literature. The use of the
terms of race and ethnicity, as well as group-names in the review of literature in the following
review is as the original authors reported.
Regarding association between the pain experience and race, while Horgas, Yoon,
Nichols and Marsiske (2008) found no difference in pain presence, intensity, locations and
durations between black and white older, community dwelling older adults, Teno, Kabumoto,
Wetle, Roy and Mor (2004) reported that African American nursing home residents were less
likely to experience daily, excruciating pain than white nursing home residents. Won et al.
(2004) as well noted that among elderly nursing home residents (n=21,380) African Americans
experienced a much lower rate (36%) of persistent pain than whites (49.6). Additionally, Sawyer
et al. (2006) also noted that white nursing home residents were more likely to experience pain

49

than African American nursing home residents (n=27, 628). In comparison, Reyes-Gibby, Aday
and Cleeland (2002) found that among community-dwelling older adults, more Hispanics than
Whites, Blacks and American Indians were often bothered by pain although these results
diminished when education and insurance status were controlled for. Results of a literature
review found that African Americans perceived greater pain intensity than whites (Cleland,
Palmer, & Venzke, 2005). Among community-dwelling older adults aged 51 and older, ReyesGibby et al. (2007) noted that Non-Hispanic blacks and Hispanics had a higher risk of severe
pain compared with Non-Hispanic whites. Variation exists in the pain experience in older adults
residing in the community or nursing homes among racial and ethnic groups.
Research addressing differences between racial/ethnic groups and pain management and
pain management outcomes among older adults is limited. The body of literature examining
racial/ethnic differences in pain management and pain management outcomes is primarily
focused on a general adult population, without developing a specific focus on older adults.
Overall, white adults younger than 65 are significantly more likely to receive appropriate pain
management than adults from other racial groups and Hispanics with similar pain levels (Cintron
& Morrison, 2006; J. A. Cleeland, Palmer, & Venzke, 2005; Green, et al., 2003; Rodin, 2008).
Specific to older adults residing in nursing homes, Won et al. (2004) found a lower use of opioid
pain medications by African American nursing home residents when compared to white nursing
home residents. Regarding the association between race and ethnicity and pain management
outcomes, Cleeland and colleagues reported that there is a clear trend that adult African
Americans experience higher pain-related disability compared to white adults. In summary, these
results from research primarily conducted with a general adult population found that compared to

50

white adults, minority adults are less likely to receive appropriate pain management and African
Americans are more likely to experience disability from pain.
In summary, Fries et al. (2004) reported that a HCBWP sample was 77% white, still
leaving a significant portion of that HCBWP sample to be counted as from a minority group.
There is significant variation in the pain experience of older adults residing in the community or
nursing homes among racial and ethnic groups. Research examining differences in pain and pain
management outcomes by race or ethnicity in general adult populations indicate that compared to
white adults, minority adults are less likely to receive appropriate pain management (Cintron &
Morrison, 2006; J. A. Cleeland, et al., 2005; Green, et al., 2003; Rodin, 2008; Won, et al., 2004)
and African Americans are more likely to experience disability from pain (J. A. Cleeland, et al.).
There is limited research specific to older adults examining differences in pain management and
pain management outcomes by race and/or ethnicity. Additional research is needed to clarify the
effect of race/ethnicity on the pain experience, pain management and pain management
outcomes in older adults as well as older adult HCBWP participants.
Cognitive functioning. Cognitive functioning is vitally important to address in any painrelated research among older adult populations. Poor cognitive functioning, or cognitive
impairment occurs in 37% to 41% of HCBWP participants, compared to 80% of nursing home
residents (Fries, James, & Aliaga, 2004; L. Li & Conwell, 2007). Impaired cognitive functioning
may influence the older adult’s ability to report pain as well as the health care providers’ ability
to assess the patient’s pain (Delgado-Guay & Bruera, 2008; Goldstein & Morrison, 2005). It is
through pain-related behaviors (i.e. verbal reports of pain, grimacing, guarding etc) that the pain
experience is conveyed by the patient to a health care provider (Sawyer, et al., 2006). The
subjective report of pain is the gold standard for pain assessment, as pain is a subjective

51

experience and is best conveyed by the one who is experiencing the pain (American Pain
Society, 2005; Dodd, et al., 2001; McCaffery, 1972). If changes in cognitive functioning alter the
ability of the older adult to convey pain, then the older adult’s pain experience may not be
accurately conveyed. If the pain experience is not accurately conveyed, pain management (which
is based on the pain experience) may be inadequate, thereby resulting in poor pain management
outcomes.
Older adults with poor cognitive function consistently report less pain than cognitively
intact older adults such that as the level of cognitive performance deteriorates pain prevalence
decreases (Chu, Schnelle, & Osterwell, 2004; Duncan, et al., 2008; Procter & Hirdes, 2001;
Sawyer, et al., 2006; Zyczkowska, et al., 2007) and this effect was noted in both nursing home
and community-dwelling older adult populations. Reynolds, Hanson, DeVellis, Henderson and
Steinhauser (2008) utilized a cross-sectional method to examine disparities in pain management
between cognitively intact and cognitively impaired nursing home residents (n=551).
Cognitively impaired older adults were more likely than cognitively intact nursing home
residents to have their pain recognized as occurring less than daily and of mild severity.
Reynolds et al. also noted that among nursing home residents, as the degree of cognitive
impairment increased, the less likely the nursing home resident was to receive treatment for pain.
Eighty-percent of cognitively intact nursing home residents received pain medications, while
only 56% of those with cognitive impairment received pain medications (Reynolds, et al.).
In summary, about 37% to 41% of HCBWP participants have some degree of cognitive
impairment (Fries, James, & Aliaga, 2004; L. Li & Conwell, 2007). For this study, cognitive
functioning was treated as a covariate in analytic models in order to account for the negative
effects of poor cognitive functioning on the pain experience, pain management and pain

52

management outcomes among older adult HCBWP participants. Cognitively impaired
community dwelling older adults and older adults who are nursing home residents have their
pain consistently assessed as less prevalent and severe as well as receive less analgesic
medication than cognitively intact older adults (Chu, et al., 2004; Duncan, et al., 2008; Procter &
Hirdes, 2001; Reynolds, et al., 2008; Sawyer, et al., 2006; Zyczkowska, et al., 2007).
Behaviors indicative of depression. Depressive symptoms are commonly associated with
pain in research literature. Depressive symptoms and pain are associated with one another such
that an increased number of depressive symptoms has been shown to predict worsening levels of
pain among community dwelling older adults (n=241) (Rosso, et al., 2008). Sawyer et al. (2006)
found that a presence of a sad/depressed mood was a significant predictor of substantial daily
pain among nursing home residents, after controlling for other personal and facility-related
factors. Community-dwelling older adults with reported pain are more likely to have depressive
symptoms (Jakobsson, et al., 2003; L. Li & Conwell, 2007; Reyes-Gibby, et al., 2002), have a
higher rate of developing a new onset of depressive symptoms, and have a slower resolution of
depressive symptoms than older adults without pain (n=652) (Geerlings, et al., 2002). Therefore,
pain is associated with depressive symptoms among older adults, with the presence of one
increasing the likelihood of occurrence and worsening the prognosis of the other (Geerlings, et
al., 2002; L. Li & Conwell, 2007; Reyes-Gibby, et al., 2002). Compared to older adult nursing
home residents, HCBWP participants are half as likely to report “feelings of sadness or being
depressed” (20% vs. 42%) (Fries, James, & Aliaga, 2004).
Researchers have struggled to determine the direction of the relationship between pain
and depression. Chou and Chi (2007) completed a longitudinal study over one year to examine
the relationship between depressive symptoms and pain among 457 elderly Chinese primary care

53

patients living in the community. Participants who reported pain at baseline were significantly
more likely to report depressive symptoms one year later. However, the presence of depressive
symptoms at baseline did not predict pain one year later (Chou & Chi, 2007), supporting that the
direction of the relationship is from pain to depressive symptoms and not depressive symptoms
to pain. In another longitudinal study, Geerlings and colleagues (2002) noted that the prognosis
of comorbid pain and depressive symptoms was much worse than having depressive symptoms
alone over a three year period among community-based older adults aged 55-85. The percentage
of subjects developing depressive symptoms was more than three-times higher in those with pain
at baseline when compared to subjects without pain at baseline (Geerlings, et al.). Additionally,
Geerlings et al. found that pain and depressive symptoms had a reciprocal association where one
predicted the other over the three year time span of the study.
In summary, pain and depressive symptoms are strongly associated with each other
among older adults in both community and nursing home settings. Depressive symptoms are
positively associated with worsening pain (Rosso, et al., 2008). The presence of pain increases
the risk of developing depressive symptoms and of experiencing a slower resolution of
depressive symptoms than older adults without pain (Geerlings, et al., 2002). Additional
longitudinal research is needed to clarify the direction of the relationship between pain and
depressive symptoms. HCBWP participants are half as likely to report feeling of sadness or
depressed mood when compared to nursing home residents (Fries, James, & Aliaga, 2004).
Although this result seems to potentially indicate less of a problem of depressive symptoms
among HCBWP, the association between depressive symptoms and the pain experience, pain
management and pain management outcomes of older adult, HCBWP participants has not been
examined. Research is needed to examine behaviors indicative of depression among HCBWP

54

participants and its association with the pain experience, pain management and pain management
outcomes.
Domain of Health and Illness. The domain of health and illness was comprised of
variables unique to the HCBWP participant’s state of health (Dodd, et al., 2001) and included
variables defining comorbidity and cancer. These variables are frequently examined in research
examining the pain experience, pain management and pain management outcomes among older
adults in community and nursing home settings. The following is a review of the association
between comorbidities and the pain experience, pain management and pain management
outcomes among older adults residing in the community and in nursing homes. A review of the
association between cancer and the pain experience, pain management and pain management
outcomes among older adults residing in the community or in nursing homes will follow.
Comorbid conditions. Pain is known primarily as a symptom of injury or disease
(Reyes-Gibby, Aday, et al., 2007). While the researcher acknowledges that chronic pain (pain
lasting greater than six months) (Basbaum, et al., 2005; Greener, 2009) can be considered a
chronic condition and therefore capable of occurring along with other conditions (i.e. comorbid),
(Staud, Price, Robinson, Mauderli, & Vierck, 2004) this research did not distinguish between
chronic and acute pain. Therefore, for the purposes of this research pain was conceptualized as a
symptom influenced by the factor of comorbid conditions.
Comorbid conditions are noted to be the co-occurrence of conditions in relation to an
index disease (Yancik, et al., 2007). For this research, cancer was conceptualized as the index
disease. The researcher acknowledges that cancer may not in fact be the actual index disease
experience by each subject and therefore the effect of the diagnosis of cancer may be influenced
or superseded by the effect of an alternative index condition. The researcher also acknowledges

55

that the comorbid conditions may resolve over time and therefore, the effect of each comorbid
condition on the pain experience, pain management strategies and pain management outcomes
would change.
Older adult populations are at increased risk of experiencing conditions that are
commonly associated with pain such as arthritis, diabetes and peripheral vascular disease
(Bruckenthal & D'Arcy, 2007; Davis & Srivastava, 2003; Freedman, 2002; Yancik, et al., 2007).
Prevalence rates of hip fracture, osteoporosis and diabetes among older adult HCBWP
participants is 5%, 22% and 38% respectively (Fries, James, & Aliaga, 2004). The burden placed
on the body from multiple, co-occurring conditions is referred to as pathologic load, which is an
overriding factor contributing to the increased pain complaint among older adults (Helme &
Gibson, 2001; Reyes-Gibby, Aday, et al., 2007). Among older adults, the risk and behavior of
cancer is strongly associated with the presence of comorbid conditions and their related
treatment (Extermann, 2007)Therefore, associations between comorbid conditions and the pain
experience, pain management and pain management outcomes must be considered when
examining pain among older adult populations.
Research results have confirmed the effect of pathological load on pain, concluding that
increases in the number of comorbid conditions has an effect on pain prevalence, pain severity
and pain interference among older adults. In research examining pain, frailty and comorbidity
among 1705 community dwelling older men (age 70 and older), Blythe et al (2008) noted that
having greater than one comorbid condition significantly and incrementally increased the risk of
pain interference. The risk of presence of pain and pain severity is positively associated with the
presence of comorbid conditions among community-dwelling and institutionalized older adults
(Soldato, et al., 2007; Zyczkowska, et al., 2007). In summary, the presence of comorbid

56

conditions increases the risk of older adults experiencing pain, pain severity and pain
interference (Blyth, et al., 2008; Soldato, et al., 2007; Zyczkowska, et al., 2007).
Specific comorbid conditions have been noted to cause more pain than others. Proctor
and Hirdes (2001) utilized descriptive research to examine relationships among pain and
cognitive status among Canadian nursing home residents. Among cognitively intact nursing
home residents, those with arthritis, osteoporosis and pressure ulcers were more likely to
experience pain than those with other conditions (Procter & Hirdes). In comparison, ReyesGibby et al. (2002) noted a higher prevalence of pain among community dwelling older adults
with lung disease, heart disease and stroke. Sawyer et al. (2006) found that among older adult
nursing home residents the presence of musculoskeletal disease, anemia and cancer were
associated with a significantly greater likelihood of substantial daily pain. In summary, specific
conditions have been found to be more highly associated with pain among older adults than other
conditions: lung and heart disease, musculoskeletal disease, anemia, stroke and cancer (Procter
& Hirdes, 2001; Reyes-Gibby, et al., 2002; Sawyer, et al., 2006).
The presence of comorbid conditions influences the provision of pain management
among older adult HCBWP participants. For example, certain conditions, such as liver or kidney
disease warrant caution in prescribing pain medications as these conditions affect drug
metabolism (Goldstein & Morrison, 2005). Older adults often have complex clinical patient
presentations, as they may have multiple, co-occuring chronic diseases (Duncan, et al., 2008;
Goldstein & Morrison, 2005; McNeill, et al., 2007). As a result, health care providers may have
concerns about drug interactions with other medications prescribed for chronic diseases which
may affect pain management by limiting pain medication prescribing patterns (Duncan, et al.,
2008; Goldstein & Morrison, 2005; McNeill, et al., 2007). Providers may be concerned about

57

opioid side effects with older adults and may therefore not prescribe opioid pain medications to
older adults (Duncan, et al., 2008; Goldstein & Morrison, 2005; McNeill, et al., 2007).
In summary, older adults from community and nursing home settings are at increased risk
of developing comorbid conditions and comorbid conditions increase the likelihood of older
adults experiencing pain (Bruckenthal & D'Arcy, 2007; Davis & Srivastava, 2003; Freedman,
2002; Yancik, et al., 2007). The presence of comorbid conditions may act as a barrier to pain
management, as clinicians may be more cautious in prescribing pain management due to side
effects of medications and interactions with medications prescribed for comorbid conditions
(Duncan, et al., 2008; Goldstein & Morrison, 2005; McNeill, et al., 2007).
Diagnosis of cancer. More than 60% of cancers diagnosed and 70% of cancer-related
deaths occur in older adults (Yancik & Ries, 2000). Older adults are more likely to have pain at
the time of diagnosis (Freedman, 2002; McNeill, et al., 2007). Pain has been noted to be one of
the most distressing symptoms in patients with cancer (Laird, et al., 2008; Valeberg, et al., 2008).
With cancer, pain may occur due to the cancer itself or to its treatment (Chang, et al., 2006).
Older adults with cancer are more likely to experience pain when compared to older adults
without a history of cancer in both community and nursing home settings (Buchanan, et al.,
2005; Reyes-Gibby, Aday, et al., 2007; Rodin, 2008).

Therefore, the presence of cancer may

further increase the risk of pain in older adult HCBWP participants.
Past research has noted that 11% of nursing home residents have a diagnosis of cancer at
admission and residents with a diagnosis of cancer were more likely to have pain when
compared to residents without cancer at admission (Buchanan, et al., 2005; Rodin, 2008). Fries
et al. (2004) noted that 11% of older adult HCBWP participants reported having been diagnosed
with cancer (other than skin cancer) in comparison to 7% of nursing home residents. Other

58

research found that 20% of nursing home residents who reported daily, excruciating pain also
have a diagnosis of cancer (Teno, et al., 2004). Because of the high prevalence of pain among
older adult nursing home residents with cancer, Buchanan et al. emphasized the need for research
examining the pain management needs of older adults with cancer and developing research
comparing pain, pain management and pain management outcomes between older adult nursing
home residents with and without cancer.
A diagnosis of cancer has an impact on the outcomes of physical function and pain
control. Buchanan et al. (2005), in an analysis of nursing home residents with and without cancer
(n=61,980) noted that a larger proportion of those with cancer were totally dependent and
required extensive assistance with activities of daily living when compared to nursing home
residents without cancer. Therefore, older adult nursing home residents with cancer have more
physical function impairments than their cancer-free counterparts (Buchanan, et al., 2005).
Regarding pain control, older adults with cancer who are community-dwelling or residing in a
nursing home experience higher pain prevalence as well as higher pain severity than their cancerfree counterparts (Buchanan, et al., 2005; Duncan, et al., 2008; Reyes-Gibby, Aday, et al., 2007;
Rodin, 2008). More frequent and severe pain may require more aggressive pain management
than less frequent or milder pain. In summary, based on the above results among communitydwelling and nursing home-dwelling older adults, HCBWP participants with cancer may be
more at risk for pain-related physical function impairment as well as potentially having more
difficulty controlling pain.
In summary, older adults are more likely to be diagnosed with cancer when compared to
younger populations. Older adults living in the community or in nursing homes and diagnosed
with cancer report more pain than older adults without a diagnosis of cancer. Therefore, older

59

adults with cancer are at risk for experiencing pain above and beyond older adults without
cancer. Older adult diagnosed with cancer and experiencing pain are more likely to experience
physical function impairments as well as potentially needing more aggressive pain management.
Additional research is needed to examine differences in the pain experience, pain management
strategies and pain management outcomes of older adult HCBWP participants with and without
cancer.
Summary
The above review of research presents the overall impact that age, sex, race/ethnicity,
depression, cognitive functioning, comorbidities and cancer have on the pain experience, pain
management strategies and pain management outcomes among older adults. The above review
also presents the limitations in research that has examined associations among person and health
and illness-related factors and pain, pain management and pain management outcomes. The
results of this study will expand what is known regarding associations among age, sex,
race/ethnicity, depression, cognitive functioning, comorbidities, cancer, the pain experience, pain
management strategies and pain management outcomes among older adults. Most importantly,
however, this study will add to the science by examining associations among person and health
and illness-related factors and pain, pain management and pain management outcomes in an
infrequently studied vulnerable population who is at risk for both pain and cancer (Green, et al.,
2003; Ward, et al., 2004b).
The Pain Experience
Research Question 1 of this research determined what differences exist in the pain
experience of older adult HCBWP participants over time in regards to diagnosis of cancer, while

60

controlling for age, sex, race/ethnicity, cognitive functioning, depression and comorbid
conditions.
Pain is an unpleasant sensory and emotional experience (International Association for the
Study of Pain, 2009) that is influenced by one’s social history, cultural expectations, as well as
individual differences in physiological, developmental and psychological makeup (Dodd, et al.,
2001; Hollenack, Cranmer, Zarowitz, & O'Shea, 2006). The very foundation of physiological
differences among persons, genetic makeup, is associated with differences in nerve functioning
and sensitivity to painful stimuli (Diatchenko et al., 2006; Diatchenko et al., 2005; C.
Miaskowski, 2009). The pain experience is the culmination of perceiving, evaluating and
responding to pain (Dodd, et al., 2001). The human pain experience involves complex
interactions of sensory, cognitive and behavioral factors which aging may influence (American
Geriatrics Society Panel on Persistent Pain in Older Persons, 2002; Hollenack, et al., 2006). It is
logical, therefore, to surmise that age-related changes in sensation and cognition would influence
the perception of pain as well as the ability to evaluate and respond to pain and thereby impact
the pain experience of older adult, HCBWP participants.
Age-related changes in sensation were reviewed and presented earlier in Chapter 3. To
summarize: aging may lead to physiological changes in the nervous system that influence the
perception and production of pain such that older adults may experience an increased and
prolonged response to injury and pain stimuli, thereby placing older adults at risk of pain that
persists or in need of extended pain management (Basbaum, et al., 2005; R. R. Edwards, et al.,
2003; Gagliese & Farrell, 2005; Gibson & Farrell, 2004; Heinricher, 2005; Helme & Gibson,
2001; Zheng, et al., 2000). In addition to age-related changes in sensation, the presence of
comorbidities such as cancer, cancer and arthritis can alter the sensation of painful stimuli

61

(Backonja, 2003; Chang, et al., 2006; Pasero, 2004; SAS System for Windows (Version 9.3.1),"
2002). Age-related increases in diseases associated with pain such as diabetes, cancer and
arthritis are more likely to occur in older adults than in younger populations and place older
adults at risk for pain (Bruckenthal & D'Arcy, 2007; Davis & Srivastava, 2003; Freedman,
2002).
Diabetes and chemotherapeutic treatments for cancer may both alter nerve function,
leading to increased nerve sensitivity as well as ectopic and spontaneous nerve discharges
(Backonja, 2003; R. Dworkin et al., 2003; Pasero, 2004). Cancer-related tumors can press on
nerves or internal organs, causing pain. The added burden of cancer increases pain among older
adults as nursing home residents with a diagnosis of cancer at admission were more likely to
have pain when compared to residents without cancer at admission (Buchanan, et al., 2005;
Rodin, 2008). In nursing homes, 20% residents who reported daily, excruciating pain also had a
diagnosis of cancer (Teno, et al., 2004).
Arthritis is a common, chronic condition among older adults, affecting about 50% of
community-dwelling older adults (Blyth, et al., 2008; Zyczkowska, et al., 2007). Chronic
exposure to pain may lead to central sensitization where changes occur in the dorsal horn in the
spinal cord that cause the threshold for nerve activation to be lowered, the response to stimuli
increased, the size of the receptive field expanded (Beydoun & Backjona, 2003; Pasero, 2004),
and inhibitory pathways are lost or suppressed (Pasero, 2004). Therefore, central sensitization
leads to less stimuli being needed to produce pain as well as an alteration of the central nervous
system’s ability to inhibit incoming painful stimuli. In summary, age-related and disease and
treatment-related changes in the nervous system may alter the production of and sensation of
painful stimuli and therefore, the pain experience of older adults.

62

Based on the above evidence that older adults are at increased risk of experiencing pain
due to age-related physiological changes and disease-related effects on the nervous system, one
would expect that older adults would experience a greater prevalence of pain and more severe
pain than younger adults. A higher pain prevalence and more severe pain is not clearly shown in
research examining pain in older adults, as research findings of comparisons of pain prevalence
between older and younger adults have been inconsistent (C. Given, et al., 2001; Jakobsson, et
al., 2003; Sawyer, et al., 2006; Soldato, et al., 2007; Tsang, et al., 2008) and older adults have
reported less severe pain than younger adults (Baker & Green, 2005; Zyczkowska, et al., 2007).
This discrepancy between older adults being more at risk for pain than younger adults yet not
having a consistent higher prevalence of pain or more severe pain than younger adults may be
explained by characteristics of older adults that influence the evaluation and reporting of pain.
A clinician’s awareness of a patient experiencing pain is dependent on the response to
pain via the verbal report or through pain behaviors, which is preceded by the patient evaluating
the pain (Dodd, et al., 2001; Humphreys & et al., 2008). Pain is evaluated by the patient in
relation to its location, frequency as well as by its sensory and affective dimensions (Dodd, et al.,
2001). Pain is multidimensional, having both a sensory dimension (level of severity, strength or
overall magnitude of the pain) (C. S. Cleeland & Ryan, 1994; R. H. Dworkin, et al., 2005;
Jensen, 2003) and an affective dimension (the extent to which pain affects or impacts day-to-day
functioning) (C. S. Cleeland & Ryan, 1994; Jensen, 2003; Thomas, et al., 2004). Once pain is
evaluated by the patient, the patient then responds by either ignoring the pain or reporting the
pain via a verbal report and/or pain-related behaviors.
The resulting verbal report and/or pain behaviors can then be assessed. For this study the
pain experience was assessed through the Pain Scale, a measure of both pain frequency and pain

63

intensity (Fries, et al., 2001). Pain is multidimensional, including sensory, affective and temporal
dimensions (C. S. Cleeland & Ryan, 1994; Jensen, 2003). Therefore, the pain experience can be
assessed by using measures of the different dimensions of pain. For example, pain intensity
(sensory dimension) can be measured through the use of the numeric rating scale, the visual
rating scale (C. S. Cleeland & Ryan, 1994; Jensen, 2003). The affective dimension of pain can be
measured through a pain interference scale, which assesses how pain interferes with activities (C.
S. Cleeland & Ryan, 1994; Jensen, 2003). Temporal aspects of pain can be measured by pain
frequency or how often pain occurs i.e. daily, less than daily (Jensen, 2003).
In addition to the pain experience, pain control is assessed in the present research. Cancer
pain guidelines (American Pain Society, 2005; National Comprehensive Cancer Network, 2006;
WHO, 1996) recommend assessing changes in pain levels, impairments and frequency to
determine the effectiveness of pain interventions. Allard, Maunsell, Labbe, and Dorval (2001)
defined pain control as a result of “ …routine assessment of pain characteristics and intensity,
and analgesic treatment based on the regular intake of agonist opioids combined with additional
doses of relief of breakthrough pain” (p. 192) and equated pain control with changes in pain
levels. Shvartzman (2003) assessed cancer pain control with appropriateness of pain medication,
as based on worst pain intensity levels, and whether or not pain levels were affected by
medications and noted that pain control is related to adequate and correct treatment of pain.
Miaskowski and colleagues (2002) used significant differences in pain intensity scores between
around-the-clock pain medication dosing and as needed dosing to determine better pain control.
Overall, cancer pain control was shown to be an outcome and equated with a decrease in or
better values of pain levels, impairments and frequency due to pain interventions. Farrar, Berlin
and Strom (Farrar, Portenoy, Berlin, Kinman, & Strom, 2000) noted that a decrease in pain

64

intensity scores of 33% overtime was equated with clinically significant difference in pain pre
and post intervention. Due to data limitations, the researcher did not have knowledge as to the
timing of the assessment of the pain experience in relation to the provision of pain management
strategies. Additionally, because pain was assessed with the Pain Scale and not a numeric scale
the researcher could not determine if changes in the pain experience were clinically significant.
Response to pain is shaped by the contribution of personal and social characteristics and
is therefore, individualistic (American Pain Society, 2005). Awareness of these characteristics
and their effect on response to pain assists in explaining “…discrepancies between expected
pain, pain behaviors, and patient’s self reports of pain” (American Pain Society, 2005, p. 22).
Older adults may expect pain to be part of aging and disease processes and therefore, may
underreport or be reluctant to report their pain (Delgado-Guay & Bruera, 2008; Goldstein &
Morrison, 2005; Yong, Gibson, de L. Horne, & Helme, 2001). Among older adults, the ability to
evaluate and to report pain may be compromised by the presence of cognitive impairment
(Bruckenthal, 2008; Helme & Gibson, 2001). Cognitively impaired older adults consistently
report less pain than cognitively intact older adults such that as the level of cognitive
performance deteriorates pain prevalence decreases and this effect was noted in both nursing
home and community-dwelling older adult populations. (Chu, et al., 2004; Duncan, et al., 2008;
Procter & Hirdes, 2001; Sawyer, et al., 2006; Zyczkowska, et al., 2007). Proctor and Hirdes
found that there was no difference between cognitively intact and impaired nursing home
residents in the prevalence of diseases likely to cause pain, but the cognitively impaired nursing
home residents had a significantly lower prevalence of pain than the cognitively intact nursing
home residents. Cognitively impaired older adults report less pain not because they experience
less pain-causing diseases than cognitively intact older adults, but because they are not able to

65

convey an accurate, understandable verbal report of pain (Chu, et al., 2004; Duncan, et al., 2008;
Procter & Hirdes, 2001; Sawyer, et al., 2006; Zyczkowska, et al., 2007).
Because of the difficulties with pain assessment among cognitively impaired older adults,
some researchers examining pain among older adults have chosen to exclude subjects with
cognitive impairment (Soldato, et al., 2007). This however, would reduce generalizability of
research results to older adults, as cognitive impairment may occur in 29% -82% of nursing
home residents (Levin et al., 2007; Procter & Hirdes, 2001) and 37% to 41% in older adult,
HCBWP participants (Fries, James, & Aliaga, 2004; L. Li & Conwell, 2007). Therefore, older
adults with cognitive impairment represent a significant portion of nursing home residents and
HCBWP participants and they should be included in research examining pain among older
adults.
Summary. In summary, this research examined if differences exist in the pain experience
of older adult HCBWP participants with and without a diagnosis of cancer over time. Despite
being at high risk for developing diseases that are associated with pain as well as experiencing
age-related changes in the nervous system, research results have not confirmed that older adults
have a higher prevalence of pain compared to younger adults. This discrepancy may be
explained in part by reluctance of older adults in reporting pain and the detrimental effect
cognitive impairment has on the patient self-report of pain.
Even though the prevalence of pain among older adults is not higher than younger adults,
pain remains prevalent among older adult populations. Pain prevalence among HCBWP
participants has been reported to be 53-69% (Fries, James, & Aliaga, 2004; L. Li & Conwell,
2007). The prevalence of pain has been reported to be between 28% and 72% in communitydwelling older adults (Landi, et al., 2001; Reyes-Gibby, Aday, et al., 2007; Thomas, et al., 2004).

66

In comparison, pain prevalence has been estimated to be between 41% to 80% for older adult
residing in nursing homes (American Geriatrics Society Panel on Persistent Pain in Older
Persons, 2002; Teno, Weitzen, Wetle, & Mor, 2001; Won, et al., 2004). Pain is a prevalent
problem among older adults. The study will fill a vital gap in the knowledge of the pain
experience of older adult HCBWP participants.
The pain experience and the reporting of the pain experience is what pain management is
based on. Accurate assessment of the pain experience is a necessary precursor to pain
management as pain management would be difficult and dangerous to in the absence of basic
knowledge of the patient’s pain experience (Deane & Smith, 2008). In the following section,
literature examining pain management strategies among older adults will be presented.
Pain Management Strategies
The first part of Research Question 2 of the study examined how the pain experience of
older adult, HCBWP participants related to pain management strategies and how did this
relationship differ in regards to diagnosis of cancer, while controlling for age, sex, race/ethnicity,
cognitive functioning, depression and comorbid conditions. For the purposes of this research,
pain management strategies will include both prescribed pain medications and the provision of
hospice services.
Prescribed pain medications are categorized as non-opioid, opioid and adjuvant pain
medications. Non-opioid medications include acetaminophen and non-steroidal antiinflammatory (NSAID) medications such as ibuprofen or naproxen sodium and are useful for
mild to moderate pain and in conjunction with opioid medications for more intense pain
(American Pain Society, 2005). The mechanism of action for acetaminophen is still unknown,
but it is postulated that it has a central nervous system mechanism, because of its pain and fever

67

reducing effects (Schug, 2005). In comparison, the mechanism of action of NSAIDs is well
known. NSAIDs inhibit cyclooxygenase, an enzyme that catalyzes the production of
prostaglandins, which are key instigators of the inflammatory process (American Pain Society,
2005). Because of this mechanism, NSAIDs are especially useful in treating inflammatory pain,
as they prevent the very process that causes it (Samad, 2004).
Opioid pain medications are the medications most frequently used for moderate to severe
pain because of their effectiveness, ease of titration, and favorable risk-to-benefit ratio
(American Pain Society, 2005). Opioid medications include morphine, hydromorphone,
methadone, codeine, oxycodone, hydrocodone, levorphanol, and fentanyl (American Pain
Society, 2005). Opioid pain medications may be a combination of narcotic pain medications and
acetaminophen or non-steroidal anti-inflammatory medications. Opioid medications act on
opioid receptors which are found both peripherally and centrally in nerve tissue, in
gastrointestinal, respiratory, and cardiovascular organs, and the bladder (Lipman & Gautier,
1997). One particularly opioid receptor-rich area in the central nervous system is the
periaqueductal gray, which is a key area in the modulation or control of pain (Heinricher,
2005). When an opioid binds to the opioid receptor, an excitatory or inhibitory response occurs,
which inhibits the transmission of pain impulses in the brain and spinal cord (Sweeney & Bruera,
2003).
The term adjuvant analgesics describes “…non-opioid medications that have painrelieving effects in certain conditions, but whose primary or initial indication was not for the
treatment of pain” (American Pain Society, 2005, p. 73). Medications that have been used as
adjuvant pain medications include anticonvulsants and antidepressants (American Pain Society,
2005). Adjuvant medications diminish pain by altering nerve function. Anticonvulsants, such as

68

phenytoin and carbamazepine work by blockading the sodium channels and stabilizing the nerve
membrane (Kalso, 2005). Antidepressants, such as amitriptyline increase the availability of
neurotransmitters, block sodium channels, and block receptors (Kalso, 2005). If sodium
channels are blocked, then nerve depolarization and stimulation will be affected, and nerve
hyper-excitability is diminished (Kalso, 2005).
The type of pain medication prescribed (i.e. non-opioid, opioid, adjuvant) is an important
indicator of pain management quality as pain management guidelines recommend specific types
of medication in response to different reports of pain (American Pain Society, 2005; NCCN,
2006; NCI, 2006; WHO, 1996). While the researcher acknowledges that the management of
pain can take place via medications and non-pharmacological or procedural interventions this
study was limited to the use of prescribed pain medication data in Medicaid paid claim files.
Since the purpose of the study is to examine differences in the pain experience, pain management
and pain management outcomes among older adult HCBWP participants between those with and
without a diagnosis of cancer the provision of hospice services for palliative care will be
examined as well.
Literature describing and examining the issue of pain management among older adults is
primarily framed by a discussion of barriers to pain management among older adults. The
majority of older adult pain management literature is in clinical guideline format. Pain
management barriers are described in terms of where the barrier originates: patient, provider or
health care system (American Pain Society, 2005; NCI, 2006). Pain management barriers to the
treatment of pain among older adults have been noted to take place at both during the pain
assessment phase as well as during the provision of prescribed pain medication. The overarching
goal of pain assessment in older adults is to provide successful pain management. Therefore,

69

pain management must be preceded by a successful and comprehensive pain assessment
(Bruckenthal, 2008). This study was limited to a measure of the pain experience (pain frequency
and pain intensity) taken approximately every three months and the data did not include
information detailing when the assessment occurred in relation to the reception of pain
management strategies. In the following section, literature describing pain management among
community-dwelling and nursing home-residing older adults will be presented and discussed in
accordance with where the barrier originates.
Patient Barriers to Pain Management among Older Adults. Patient-related barriers to
pain management among older adults are primarily focused on older-adult factors that may
prevent a valid pain assessment from taking place. An inadequate assessment would impact any
resulting pain management. Older adults may expect pain to be part of aging and disease
processes and therefore, may underreport their pain (Delgado-Guay & Bruera, 2008; Goldstein &
Morrison, 2005). Older adults may fear being a nuisance to health care providers by reporting
pain or may be concerned about the financial costs of extensive testing to determine the cause
and treatment of pain (American Geriatrics Society Panel on Persistent Pain in Older Persons,
2002; Bruckenthal & D'Arcy, 2007). As noted earlier in this chapter, older adult nursing home
residents who are cognitively impaired are less likely than a cognitively intact older adult to
receive treatment for pain (Reynolds, et al., 2008). Reynolds et al. noted that 88% of cognitively
intact nursing home residents received pain medications, while only 56% of those with cognitive
impairment received pain medications.
Other patient-related barriers to pain management are not necessarily isolated to just
older adults but may be common to all persons with pain and include issues of race/ethnicity and
fears of addiction. White adults are significantly more likely to receive appropriate pain

70

management than adults from other racial groups and Hispanics with similar pain levels (Cintron
& Morrison, 2006; Cleland, et al., 2005; Green, et al., 2003; Rodin, 2008; Won, et al., 2004).
Fears of addiction are a well known barrier to the use of opioid pain medications by persons with
pain (American Pain Society, 2005; NCI, 2006).
In summary, patient-related barriers to pain management among older adults may include
cognitive impairment, reluctance to report pain, race/ethnicity, and fears of addiction. These
barriers inhibit a valid pain assessment from taking place and from the patient receiving adequate
pain management strategies.
Provider-Related Barriers to Pain Management Among Older Adults. Professional
barriers are behaviors that include inadequate knowledge of opioid medications and proper
assessment and management of pain and concerns about controlled substance regulation and
patient addiction (APS, 2005; Chang, Hwang, & Kasimis, 2002; JCAHO, 2000; Lin & Mathew,
2005; Sun et al., 2007). Healthcare providers assessing pain among older adults with cognitive
impairment may be unfamiliar with assessing pain behaviors such as grimacing, moaning and
agitation (Barkin, Barkin, & Barkin, 2005). Pain may be challenging for healthcare providers to
accurately assess in older adults with cognitive changes or communication difficulties if painrelated behaviors must be assessed instead of verbal self- reports of pain (Delgado-Guay &
Bruera, 2008; Goldstein & Morrison, 2005). However, pain behaviors may be a reliable indicator
of response to pain management. In results from a pilot study, Elliot and Horgas (2009) noted
that among community-dwelling older adults with dementia, pain behaviors decreased in
frequency in response to scheduled acetaminophen dosing. For the purposes of the current study,
the pain item is based on two pain measures which use patient verbal report of pain and verbal
report by a proxy that the patient frequently complains of pain or shows evidence of pain via

71

pain behaviors and that the pain is unusually intense. The pain item therefore relies not only on
the patient verbal report, but on the interpretation and report of patient behaviors by the proxy.
The pain measure used in the study was the Pain Scale (Fries, et al., 2001) which is based on
measures of pain frequency and intensity and is described in more detail in Chapter 4.
Health professionals may be hesitant to treat pain aggressively in older adults due to
concerns about opioid side-effects, complex clinical presentations, as well as drug interactions
with other medications prescribed for comorbid conditions (Duncan, et al., 2008; Goldstein &
Morrison, 2005; McNeill, et al., 2007). Health professions may have inadequate knowledge of
the treatment of pain in older adults with comorbid conditions and the best pain medication
choices for those with organ impairment (Bruckenthal & D'Arcy, 2007). Older adults may
experience pain that is persistent or chronic in nature due to the lingering conditions they are at
increased risk for experiencing, such as osteoarthritis and diabetes. Chronic pain, unfortunately,
is often negatively associated with psychiatric problems, drug seeking and futility in treatment by
healthcare providers and may prevent the health care provider from treating an older adult’s pain
aggressively (American Geriatrics Society Panel on Persistent Pain in Older Persons, 2002).
In summary, provider-related pain management barriers among older adults include
inappropriate pain assessment skills of pain in older adults, lack of knowledge of pain
management among older adults and fears of addiction. These barriers act to prevent the older
adult from receiving adequate pain management.
System-Related Barriers to Pain Management. Older adults may receive health care
services in a number of settings. In the community they may receive pain management via home
care and/or primary care in clinics; they may reside in nursing homes and receive pain

72

management there; and finally, in relation to the study, receive pain management services,
including hospice services, while participating in the HCBWP.
The healthcare system has an obligation to provide comfort and pain management for
older patients (American Geriatrics Society Panel on Persistent Pain in Older Persons, 2002).
Both JCAHO (2001) and the American Pain Society (2005) have noted that there may be poor
reimbursement of the most appropriate pain treatments and therefore, these effective treatments
may be too costly for patients. The provision of pain management and hospice services is of key
importance in the population of interest for the study, as the HCBWP provides services for poor
older adults who are dual eligible for Medicaid and Medicare services and pain management
coverage and hospice services may be limited.
In addition to limited insurance coverage, older adults may receive poor pain
management due to where they receive pain management services. Pain experienced by older
adult nursing home residents is poorly assessed and under treated (Fisher, et al., 2002; Reynolds,
et al., 2008; Sawyer, et al., 2006; Won, et al., 2004). Weissman, Griffie, Muchla and Matson
(2001) noted that the organization of nursing home care presents an obstacle to effective pain
management. Nursing home-related obstacles to effective pain management include that
physicians are rarely on site, that nursing assistants are the primary care providers and are
untrained in pain assessment and there is a reluctance to use narcotic pain medications due to
fear of scrutiny by state and federal surveyors (Weissman, et al.). Older adults receiving home
care services at home are also at risk of poor pain management due to patient, provider and
system-related barriers to pain management (Delgado-Guay & Bruera, 2008; Goldstein &
Morrison, 2005; Lin & Mathew, 2005; Sun, et al., 2007). Soldato et al. (2007) noted that among
1520 older adult home care recipients 46% had daily pain, but only 38% were taking pain

73

medications. In summary, older adults are at risk for poor pain management in both home care
and nursing home health care settings.
In addition to prescribed pain medication, the study examined the use of hospice services
as a pain management strategy. Hospice, as a health care service provider, provides palliative
rather than curative care and coordinates the provision of medical, emotional and spiritual care
for terminally ill patients (life expectancy of less than 6 months) and their families (Hospice
Association of America, 2006). While hospice services are primarily equated with care of
persons with cancer, patients with a terminal prognosis due to other diseases may receive hospice
services as well such as heart failure, chronic obstructive pulmonary disease (COPD) and
dementia and Alzheimer’s disease (Locher, et al., 2006; Rodin, 2008). Ersek and Wilson (2003)
noted that as the U.S. population continues to age, end of life and hospice services will become
an increasingly important issue to address.
Palliative and hospice care serves to identify goals for care, provide effective prevention
and management of end of life symptoms such as dyspnea and pain, provide attention to
psychosocial and spiritual issues and prevention of suffering, and assist with completion of
developmental tasks related to the dying process and successful negotiation of the grief and
bereavement processes (Ersek & Wilson). The provision of hospice services can take place in
the home, nursing homes, hospitals or hospice centers (Hospice Association of America). The
provision of hospice services in nursing homes has been examined and found to be lacking.
Among older adult nursing home residents with cancer (n=61, 890), 67% had a terminal
diagnosis, but only 19% were receiving hospice services (Buchanan, et al., 2005). Similarly,
Duncan, Bott, Thompson and Gajewski (2009) found that among nursing home residents with
cancer, 30% had clinically deteriorated within three months of admission and needed improved

74

symptom management (pain, dyspnea and weight loss). However, only 11.9% were receiving
hospice services (Duncan, et al.).
Despite its limited use in nursing homes, hospice services has been shown to provide
high quality end of life care and result in positive outcomes, such as reduced hospitalizations and
improved pain management (Stevenson & Bramson, 2009). As HCBWP are an alternative to
nursing home care, knowledge is needed regarding the use of hospice services among older
adults and to make comparisons regarding hospice use between HCBWP participants with and
without cancer.
Summary. In summary, pain management among other older adults is often inadequate,
as patient, provider and health care system-related barriers prevent adequate pain management
from taking place (Duncan, et al., 2008; Goldstein & Morrison, 2005; McNeill, et al., 2007). In
the study, pain management strategies will include prescribed pain medications and the provision
of hospice services. Hospice services are not only be provided for those with cancer, but also
older adults with other chronic, terminal illnesses such as heart failure, COPD, dementia and
Alzheimer’s (Locher, et al., 2006; Rodin, 2008). In nursing homes, hospice services are not
provided as frequently as they perhaps should be, based on reported symptoms and terminal
prognosis (Buchanan, et al., 2005; Duncan, et al., 2009). Because HCBWP are designed to be an
alternative to nursing home care, knowledge is needed regarding the use of hospice services in
HCBWPs. Associations between the pain experience and pain management strategies and pain
management outcomes can then be examined.
Pain Management Outcomes
Outcomes are the end results of care (Patrick, 1997). In the Symptom Management
Theory (Figure 1), outcomes are conceptualized as including symptom status, functional status,

75

emotional status, self-care, costs, quality of life, morbidity/co-morbidity and mortality (Dodd, et
al., 2001; Humphreys & et al., 2008). While the researcher acknowledges the importance of all
of the above outcomes pain management outcomes were limited to physical functioning and pain
control for the present research. Physical functioning and participant ratings of improvement
and/or satisfaction with treatment have been noted to be core outcomes for pain clinical trials (R.
H. Dworkin, et al., 2005). Pain control is conceptually defined as an outcome of pain
management whereby pain is perceived by the patient or proxy as limited or decreased from a
previous pain level and is an indicator of the effectiveness of pain management strategies
(Allard, et al., 2001; Christine Miaskowski, et al., 2002; Oliver, et al., 2001; Shvartzman, et al.,
2003). For this study, pain control was represented by the HCBWP participant’s or proxy’s
response to the statement “pain controlled by medication”. The Initiative on Methods,
Measurement and Pain Assessment in Clinical Trials (IMMPACT) supports the use of activities
of daily living as a fundamental measure of physical functioning to determine the effect of pain
and pain management strategies on physical function (R. H. Dworkin, et al., 2005; Turk, et al.,
2003). For this study, physical functioning was conceptualized as the dependency of the older
adult waiver participant in activities of daily living (ADL). The second part of Research
Question 2 of the study examined how the pain experience of older adult, HCBWP participants
related to pain management strategies and pain management outcomes of physical functioning
and pain control and how these relationships differed in regards to diagnosis of cancer, while
controlling for age, sex, race/ethnicity, cognitive functioning, behaviors indicative of depression
and comorbid conditions.
Physical Functioning. Decreased physical functioning has been noted to be a serious
consequence of pain as it fosters learned helplessness, social isolation and greater healthcare

76

costs because of more dependencies in activities of daily living and therefore, more nursing care
needs (Weiner & Hanlon, 2001). Associations between pain and physical functioning have been
found to occur among both community-dwelling older adults as well as older adult nursing home
residents (Jakobsson, et al., 2003; Soldato, et al., 2007; Teno, et al., 2004). Soldato and
colleagues (2007) assessed the association between pain and risk of developing a need for
assistance with the following activities of daily living (ADL): eating among 1,520 older adults
receiving home care services: dressing, transferring, mobility in bed, personal hygiene or
bathing. When compared to older adults with no pain, older adults with daily pain had a
increased hazard of developing new onset disability even after controlling for age, gender, flareup of comorbid conditions, number of comorbid conditions and level of physical activity (Odds
Ratio=1.36, CI=1.05-1.78) (Soldato, et al.). Compared to older adults with no painful sites, older
adults with multiple pain sites had an increased hazard of developing a need for assistance with
ADLs (Odds Ratio=1.56, CI=1.13-2.15) (Soldato, et al.).
Jakobsson, Klevsgard, Westergren and Hallburg (2003) identified variables associated
with pain prevalence among 4, 093 older adults aged 75 and older residing in nursing homes and
in the community. Jakobsson and colleagues utilized a measure called functional health status
which was comprised of items which assessed walking, mobility and activities of daily living
including personal hygiene, food intake and dressing. Pain was reported by 40.4% of subjects.
When compared to older adults without pain, older adults with pain were significantly more
likely to experience walking problems, mobility problems, as well as require assistance with
activities of daily living (Jakobsson, et al., 2003). Teno, Kabumoto, Wetle, Roy and Mor (2004)
completed a cross-sectional analysis of data from 2,138,442 assessments of nursing home
residents to examine associations between excruciating daily pain outcomes among nursing

77

home residents and demographic characteristics, functioning ( cognitive functioning, activities of
daily living and change in self-sufficiency) and measures of disease burden. Activities of daily
living was measured by a Activities of Daily Living Score which ranged from 0=minimal
oversight for ADL to 5=high dependence. When compared to older adults who did not have
daily, excruciating pain older adults with daily, excruciating pain were more likely to require
greater than or equal to extensive assistance with ADLs (69% vs. 62%). In summary, pain is
associated with physical functioning among older adults in both community and nursing home
settings.
Pain Control. Pain control is a key outcome in pain management as the goal of pain
management is to reduce or control pain. Allard, Maunsell, Labbe, and Dorval (2001) describe
pain control as a result of “ …routine assessment of pain characteristics and intensity, and
analgesic treatment …” (p. 192). Among younger adults and in pain management guidelines pain
control is equated with decreased or diminished pain level, impairments or frequency in response
to a pain management intervention (American Pain Society, 2005; Carpenter, Hastie, Morris,
Fries, & Ankri, 2006; Christine Miaskowski, et al., 2002; NCCN, 2006; NCI, 2006; Oliver, et al.,
2001; Shvartzman, et al., 2003). For the purposes of the present study, pain control was
conceptually defined as an outcome of pain management whereby pain is perceived by the
patient or proxy as limited or decreased from a previous pain level and is an indicator of the
effectiveness of pain management strategies (Allard, et al., 2001; Christine Miaskowski, et al.,
2002; Oliver, et al., 2001; Shvartzman, et al., 2003).
Research examining pain and pain management in older adults in either nursing homes or
community settings has been primarily cross-sectional (Reynolds, et al., 2008; Won, et al., 2004)
and not longitudinal, where response to pain management strategies could be measured overtime.

78

Elliott and Horgas (2009) measured both frequency and duration of pain behavior in a
longitudinal pilot study, noting a decline in pain frequency and duration in response to scheduled
acetaminophen dosing among community-dwelling older adults with Alzheimer’s disease. The
present study used longitudinal methods to assess pain control over time among HCBWP
participants in response to pain management strategies. Pain control was operationalized as a yes
or no response via patient or proxy to the question “Pain controlled by medication?”
Conclusion of Pain Management Outcomes. In conclusion, physical functioning is
associated with pain among older adults residing in nursing homes and in the community
(Jakobsson, et al., 2003; Rosso, et al., 2008; Soldato, et al., 2007; Teno, et al., 2004). The
outcome of pain control determines if the patient or a proxy noted a limited or decreased pain
level through the use of pain management strategies and is an indicator of the effectiveness of
pain management strategies (Allard, et al., 2001; Christine Miaskowski, et al., 2002; Oliver, et
al., 2001; Shvartzman, et al., 2003).
Time and Pain
This research examined differences in pain, pain management strategies and pain
management outcomes among those with and without cancer over time in the HCBWP. Pain,
pain management strategies and pain management outcomes have strong temporal (time-related)
components and occur as a series of events (Henly, et al., 2003). Temporal aspects of pain
consist of variation of pain over time, how frequently pain occurs, as well as the duration of pain
(Jensen, 2003). The perception, evaluation and response to pain, as presented in the Symptom
Management Theory (Humphreys & et al., 2008) not only occur over time, but are affected by
the past events over time. For an older adult, the perception, evaluation and response to pain are
influenced by a lifetime of personal, environmental and health and illness events. For example,

79

one’s perception, evaluation and response to pain are affected by beliefs and culture (Cleland,
Palmer, & Venzke, 2005). Chronic exposure to pain leads to permanent structural changes in the
CNS that includes death of inhibitory neurons, replacement with new excitatory neurons, and
creation of aberrant excitatory synaptic connections (Samad, 2004). These changes result in an
increased area which dorsal field neurons respond to, thereby extending pain sensitivity well past
the site of injury (Miaskowki, 2004). Patients can experience pain that is persistent (greater than
12 hours per day) and or transient exacerbations of significant or severe pain or “breakthrough”
pain (Mishra, Bhatnagar, Chaudhary, Pratap, & Rana, 2009).
Pain management strategies are associated with time through the scheduling or timing of
pain management strategies and pharmacokinetics. Pharmacokinetics is activity of drugs in the
body over a period of time, including the processes by which drugs are absorbed, distributed in
the body, localized in the tissues, and excreted (NCI, 2010a). Pain medications can include either
short or long-acting pain medications, treating pain that may be either sporadic (as needed) or
continuous (scheduled) across time (American Pain Society, 2005; NCI, 2010b).
Pain management outcomes are the end results of care (Patrick, 1997) and should be
assessed after the provision of pain management interventions (Humphreys & et al., 2008). The
assessment of pain management outcomes should be timed to occur after an intervention has had
an effect. For example, pain management guidelines recommend that the level of pain is
reassessed after a dose of pain medication has had time to act, based on pharmacokinetics
(American Pain Society, 2005; NCCN, 2010; NCI, 2010b). A limitation in this research was
that the timing of pain management strategies and assessment of outcomes was not known from
the study data. Instead, the pain management outcomes of physical function and pain control and
the pain management strategy of hospice services were assessed approximately every 90 days,

80

with the measure of prescribed pain medications including all prescribed pain medications in the
30 days prior to assessment (described in more detail in Chapter 4).
In summary, pain, pain management and pain management outcomes occur over time and
therefore, longitudinal research may be a better choice than cross-sectional research to examine
relationships among pain, pain management and pain management outcomes over time among
older adults with and without a diagnosis of cancer. Optimally, pain management outcomes
should be assessed within specific time periods after the provision of pain management
strategies. This research was limited by the lack of information regarding the timing of pain
management strategies and assessment of outcomes in the study data, i.e. it could not be known
how much time transpired between the taking of pain medications or reception of hospice
services and the assessment of pain management outcomes.
Admission of HCBWP Participants to Nursing Homes
The overall purpose of the HCBWP is to provide Medicaid covered home care services to those
18 and older who are at risk for nursing home placement because of the need for assistance with
activities of daily living or medical services, in order to delay or avoid the high costs of
institutional long-term care (D'Souza, et al., 2009; Sands, et al., 2008; Shugarman, et al., 1999).
As the goal of the HCBWP is to prevent or delay nursing home admission (Fries, et al., 2002),
knowledge gained about the impact of the pain experience, pain management strategies and pain
management outcomes on admission of HCBWP participants to nursing homes and differences
between HCBWP participants with and without a diagnosis of cancer would be beneficial for
guiding the development of care strategies for assisting older adults in staying in the community
and avoiding institutionalization. Research Question 3 of the study guides an examination of the
effect of the pain experience, pain management strategies and pain management outcomes on the

81

admission of HCBWP participants to nursing homes. The review that follows will examine
predictors of admission of older adults to nursing homes. Research examining predictors of
nursing home placement among older adults addresses two populations: a general older adult
placement and older adults with dementia and/or diagnosis of Alzheimer’s disease.
Predictors of Nursing Home Admission among a General Older Adult Population.
In a meta-analysis of 77 reports that included community-based samples and longitudinal designs
Gaugler, Duval, Anderson and Kane (2007) noted that the strongest predictors of nursing home
admission were three or more activities of daily living (ADL) dependencies, cognitive
impairment and prior nursing home use. In a 14-year longitudinal study of 2,805 Australian
elders there was a 9% nursing home rate (McCallum, Simons, Simons, & Friedlander, 2005).
Forty-four percent of these placements were due to dementia, with dementia being listed as a
secondary diagnosis on admission in an additional 20% of the admissions. Urinary incontinence,
impaired peak flow, physical disability and depression were other significant predictors of
nursing home admission in this sample (McCallum, et al.).
In summary, the strongest predictors of nursing home placement among a general
population of community-dwelling older adults were ADL/physical deficiencies, cognitive
impairment, prior nursing home use, urinary incontinence, depression and impaired peak flow.
Predictors of Nursing Home Admission Among Older Adults with Dementia.
Separate research has examined predictors of admission of older adults with cognitive
impairment or dementia to nursing homes. Results of a systematic review of 80 studies by
Gaugler et al. (2009) noted that among older adults with dementia the most consistent predictors
of a nursing home admission were severity of cognitive impairment, diagnosis of Alzheimer’s
disease, basic activities of daily living (ADL) dependencies, behavioral symptoms and

82

depression. Caregiver characteristics that predicted nursing home admission included indication
of greater emotional distress, desire to institutionalize the care recipient and feelings of being
“trapped” in care responsibilities (Gaugler, et al., 2009).
Yaffe et al. (2002) performed an analysis of data from the Medicare Alzheimer’s Disease
Demonstration and Evaluation study to develop and validate a predictive model of predictors of
nursing home placement of 5, 788 community-living persons with advanced dementia. Patient
characteristics that were predictive of nursing home placement were black race or Hispanic
ethnicity, living alone, one or more ADL dependencies, high cognitive impairment and one or
more difficult behaviors. Caregiver characteristics predictive of care recipient admission to a
nursing home included caregiver age of 65 or older and high care giver burden (2002). Gaugler
et al. (2000) noted that caregivers of cognitively impaired older adults who received assistance
from family members in overnight monitoring and ADL performance were significantly less
likely to admit the care recipient to a nursing home than care givers who did not receive
assistance.
In summary, the predictors of older adults with dementia being institutionalized were
similar to older adults from a general population and included level of cognitive impairment,
physical impairment and behavior issues. Caregiver characteristics that predicted
institutionalization of older adults with dementia included older age, high care giver burden and
not receiving assistance in ADL and night supervision.
Of interest in this study is the association of the pain experience, pain management
strategies and pain management outcomes to the admission of older adults to a nursing home.
Pain is highly associated with impairment in physical function and depression and impaired ADL
performance or physical functioning and depression is associated with admission to a nursing

83

home. Therefore, the pain experience, pain management strategies and pain management
outcomes may have a role in the admission of HCBWP participants to a nursing home.
Time to Admission to Nursing Home. For Research Question 3, which examines
predictors of nursing home admission, time to admission will also be explored. Because a goal of
the HCBWP is to prevent or delay nursing home admission (Fries, et al., 2002), the amount of
time spent in the HCBWP program prior to admission to a nursing home is important
consideration. If the pain experienced, pain management strategies received and assessed pain
management outcomes are predictive of time to nursing home admission then information gained
from this study can guide the development of pain assessment and pain management strategies
for HCBWP participants. Additionally, if there are differences in predictors of admission to a
nursing home and time to admission between older adult HCBWP participants with and without
a diagnosis of cancer, focused pain assessment and pain management strategies can be developed
for older adult HCBWP participants at higher risk for nursing home admission. These
associations will be examined via the third research question of the study.
Conclusion of Chapter 3
The primary purpose of this study was to examine longitudinal differences in the pain
experience, pain management strategies and pain management outcomes among older HCBWP
participants with respect to diagnosis of cancer while participating in the HCBWP. The
secondary purpose of this study was to determine what differences exist in how the pain
experience, pain management strategies and pain management outcomes among older adult,
HCBWP participants associates with the admission of older adult HCBWP participants to a
nursing home, with respect to diagnosis of cancer, over the course of time while participating in
the HCBWP.

84

The above literature review presented findings from research examining the pain
experience, pain management and pain management outcomes of community-dwelling older
adults and older adult nursing home residents and predictors of nursing home admission.
Community-dwelling older adults and nursing home residents are at high risk for pain, poor pain
management and poor pain management outcomes and which older adults with cancer are more
likely to experience pain than older adults without cancer.
This study addressed a vital gap in the literature addressing the needs of frail, poor, older
adults, as there is no known research examining differences in the pain experience, pain
management strategies, pain management outcomes or admission to a nursing home between
older adult, HCBWP participants with and without a diagnosis of cancer overtime. Chapter 4
follows and will present a thorough description of the research design and methods which were
utilized to address the research questions of the study.

85

Chapter 4
The purpose of Chapter 4 is to present design and methodology of the study. First, the
design, sample and setting of the study are introduced. Next, instruments, operational definitions
and measurement of experimental variables are presented. Last, the proposed data analysis plan,
data management and protection of human subjects are described.
The study is a secondary analysis of a large dataset comprised of data from Michigan
Medicaid paid claims files and eligibility data; the Minimum Data Set Home Care (MDS-HC);
and Michigan Cancer Registry and death certificate information from the Michigan Division for
Vital Records and Health Statistics. The study is a longitudinal, descriptive design that utilized
data obtained over the time period that subjects participated in the state of Michigan Home and
Community-Based Waiver Program (HCBWP), known as MIChoice.
Purpose and Research Questions
The primary purpose of this study is to examine longitudinal differences in the pain
experience, pain management strategies and pain management outcomes among older HCBWP
participants with respect to diagnosis of cancer while participating in the HCBWP. The
secondary purpose of this study is to determine what differences exist in how the pain
experience, pain management strategies and pain management outcomes among older adult,
HCBWP participants associates with the admission of adult HCBWP participants 65 and older to
a nursing home, with respect to diagnosis of cancer, over the course of time while participating
in the HCBWP. The study answered the following research questions:
Among HCBWP participants 65 and older and with a minimum of five Minimum Data
Set-Home Care assessments:

86

1) How does the pain experience differ between older adult HCBWP participants in
regards to diagnosis of cancer over time? How is the relationship between the pain
experience and diagnosis of cancer affected by sex, age, race, comorbid conditions,
behaviors indicative of depression and cognitive functioning over time?
2) How does the pain experience of older adult, HCBWP participants relate to pain
management strategies and pain management outcomes and how does this
relationship differ in regards to diagnosis of cancer over time? How is the relationship
between the pain experience, pain management strategies, pain management
outcomes and diagnosis of cancer affected by sex, age, race, comorbid conditions,
behaviors indicative of depression and cognitive functioning over time?
3) How do the pain experience, pain management strategies and pain management
outcomes of older adult, HCBWP participants predict the admission and time to
admission of older adult HCBWP participants to a nursing home over time and how
does this relationship differ in regards to diagnosis of cancer while accounting for
sex, age, race, comorbid conditions, behaviors indicative of depression and cognitive
functioning?
The above research questions follow a natural progression from the assessment and
management of pain in a HCBWP to whether pain, pain management or pain management
outcomes among HCBWP participants act as predictors of admission to a nursing home from a
HCBWP in regards to diagnosis of cancer.
Sample
The target sample for the study were older adults aged 65 and above who participated in
the Michigan Medicaid HCBWP, known as MIChoice and who had a minimum of five

87

assessments. Assessments were completed using the Minimum Data Set Home Care assessment
tool on admission and quarterly thereafter. The Minimum Data Set Home Care (MDS-HC) will
be described in detail later in this chapter.
The sample consists of older adults aged 65 and above who were continuously Medicaideligible between 1/1/2002-12/31/2005 and who were enrolled in the MIChoice HCBWP between
1/1/2003-12/31/2005 (See Figure 1 for Sample Selection Flow Chart). Continuous Medicaid
eligibility was required as Medicaid eligibility can vary month to month depending on assets and
thus, data from subjects who lose eligibility, are discharged from the HCBWP and then re-enter
at a later date could be sporadic. Requiring continuous Medicaid eligibility ensures multiple
assessment measures over time for each subject in order to complete longitudinal analysis.
The date restrictions were selected because Medicare Plan D, which provides prescription
drug coverage for Medicare beneficiaries, began 1/1/2006. Older adult HCBWP participants
aged 65 and older may be eligible for both Medicare and Medicaid coverage. The proposed data
set does not include Medicare data, as the researcher did not have access to Medicare paid claim
files. The study did however have access to Medicaid paid claim files. In order for Medicaid to
be the sole payer for pain medications so that prescribed pain medications could be ascertained,
the dataset was restricted to the time period before Medicare Part D began.
Additional inclusion criteria required a minimum of five MDS-HC assessments and that the five
MDS-HC assessments were completed by 12/31/2005. As the MDS-HC is completed at
admission and quarterly, a minimum of five assessments would allow for a minimum of
approximately one year’s worth of data for analysis. Analysis of data will begin at the second
assessment in order to allow for 30 days prior to the second assessment to examine the Medicaid

88

paid claim files for the billing of pain medications. The second assessment will be referred to as
“Time 1” beginning in Chapter 5.

-Continuously Medicaid-Eligible, 65 and
older between 1/1/02 and 12/31/07.
-Enrolled in Home and Community-Based
Waiver Program between 1/1/02 and
12/31/07.
- Begin assessing subjects at second MDS
Assessment to allow time for Medicaid claim
files prior to assessment

Exclude subjects who:
-Did not have 5
consecutive MDS
assessments completed
before 12/31/05
-Left waiver program for
reasons other than death
or transition to nursing
home

Final Sample

Figure3. Proposed Sample Selection Flow Chart
Setting
The data for the study were collected from MIChoice participants in private homes,
assisted living and group homes and nursing homes throughout the state of Michigan as part of

89

the assessment of participants in the MIChoice program. A brief history and description of the
HCBWP and MIChoice follows.
The Medicaid Home and Community Based Waiver Program (HCBWP) was established
by the Omnibus Budget Reconciliation Act of 1981 (OBRA-81) and was incorporated into the
Social Security Act at Section 1915(c) (Duckett & Guy, 2000). Medicaid is a joint federal and
state program that provides health insurance coverage to certain categories of low-income
individuals, including children, pregnant women, parents of eligible children, persons with
disabilities and older adults (Centers for Medicare & Medicaid Services, 2005). Prior to the
OBRA -81, Medicaid coverage of home and community-based services was limited and favored
institutional-based care for long-term care needs (Duckett & Guy, 2000; Marek, et al., 2005).
The OBRA-81 supported the expansion of Medicaid coverage of home care benefits
provided by individual states via “Medicaid waivers” (Shugarman, et al., 1999). Medicaid
waivers “waive” certain regulatory requirements regarding individual state’s Medicaid plans
thereby easing the expansion of Medicaid-covered home and community-based services
(Kitchener & Harrington, 2003; Kitchener, et al., 2004; Shugarman, et al., 1999). The overall
purpose of the waiver program is to provide Medicaid-covered home care services to those 18
and older who are at risk for nursing home placement in order to delay or avoid more expensive
institutional long-term care (Shugarman, et al., 1999). Services covered under waiver programs
may include homemaker services, respite care, adult day care, environmental modifications,
transportation, medical supplies, personal emergency response system, private duty nurse,
counseling, home delivered meals, physical and occupational therapy and personal care
supervision (Shugarman, et al., 1999).

90

The state of Michigan’s HCBWP was initiated in 1992 and expanded statewide in 1998
and became part of MIChoice (L. Li & Zullo, 2003; Shugarman, et al., 1999). The MIChoice
waiver program is managed by the Michigan Department of Community Health, which contracts
with home health agencies, Area Agencies on Aging, community mental health boards and
private nonprofit organizations across the state to provide the program at a regional level (L. Li
& Zullo, 2003; Tilly & Kasten, 2001). MIChoice waiver participants must meet financial and
medical eligibility criteria for Medicaid-funded nursing home care in order to receive waiver
services (Fries, James, Hammer, Shugarman, & Morris, 2004; Shugarman, et al., 1999).
Financially eligible persons include those currently receiving supplemental social security
income or those with income at or below 300% of the SSI level (Fries, et al., 2002). Medical
eligibility is determined via a two-step process. First, the potential participant who desires home
and community based services or a representative calls a waiver agent program for a 15-20
minute telephone-screening by the waiver agent (Fries, James, Hammer, et al., 2004). The
screening process determines level of care (nursing home, home care, intermittent personal care,
homemaker services and information and referral with no formal services) needed by the
potential participant based on care needs (Fries, James, Hammer, et al., 2004).
MIChoice participants designated by the telephone screening as requiring nursing home
level of care are then prioritized by the waiver agency to receive a second, more thorough inperson assessment by team of a registered nurse and social worker to determine if the potential
participant meets criteria for nursing home care (Fries, James, Hammer, et al., 2004; Fries, et al.,
2002; L. Li & Zullo, 2003). Criteria for nursing home level of care is defined as having
functional limitations in activities of daily living, complex and unstable medical needs, failing
social supports, a recent history of hospitalization, and/or requiring care by a trained aid (Fries,

91

James, Hammer, et al., 2004; Shugarman, et al., 1999). If the individual is determined to need
nursing-home level care, they are admitted to the MIChoice waiver program.
At or shortly after admission to the MiChoice waiver program and quarterly thereafter, an
assessment takes place in the participant’s home, nursing home or hospital room utilizing the
Minimum Data Set Home Care (MDS-HC) instrument. The MDS-HC is a comprehensive,
standardized questionnaire for evaluating the care needs, strengths and preferences of clients of
home care agencies (Landi et al., 2000) and has been validated for use in United States and
international populations (Kwan, Chi, Lam, & Chou, 2000; Landi, et al., 2000; Morris, Fries,
Steel, et al., 1997). The MDS-HC is comprised of 223 items that assess home care clients’
sociodemographic characteristics, functional and cognitive status, social support, psychosocial
well-being, clinical diagnoses and nursing needs relevant to care planning (Landi et al., 2005;
Morris, Fries, Steel, et al., 1997; Shugarman, et al., 1999). Additional information regarding the
MDS-HC will be presented in the section in this chapter detailing Instruments and Measures.
Instructions detailing the process of administering the MDS-HC are presented in the
MDS-HC instruction manual (Morris, Fries, Bernabei, et al., 1997) for the version of the MDSHC used by MIChoice during the inclusion dates for the proposed research. The MDS-HC items
consists of questions which are primarily asked of the waiver participant if possible or the
caregiver or family member of the waiver participant if the waiver participant is not capable of
responding (Morris, Fries, Bernabei, et al., 1997). Observations of the waiver participant are
also to be made by the assessor (Morris, Fries, Bernabei, et al., 1997). The potential for
information for key study variables by proxy is included in the operational definitions and
measures.

92

Data Sources
Data for this study were obtained for the time period when subjects were participating in
the MIChoice HCBWP. Sources of data include Michigan Medicaid paid claims files, the
Michigan Cancer Registry Data and death certificate data from the Michigan Division for Vital
Records and Health Statistics and the Minimum Data Set Home Care. However, the primary
source of data for the study is the Minimum Data Set Home Care (MDS-HC), version 1
instrument (Morris, Fries, Steel, et al., 1997). Development of and access to this combined
dataset was overseen by the Institute of Health Care Studies (IHCS) at MSU. Additional
information regarding the creation of the dataset will be presented in the Procedures section.
Each data source will be described in more detail in the following.
Michigan Medicaid Paid Claim Files. Medicaid is a health insurance program for
persons with low incomes and resources that is jointly funded by the federal and state
governments (Centers for Medicare & Medicaid Services, 2009b). Medicaid recipients may
include low income children, pregnant women, the blind, aged and disabled and persons who are
eligible to receive federal income assistance (Bradley et al., 2007; Centers for Medicare &
Medicaid Services, 2009b). Medicaid paid claim files contain claims for inpatient, outpatient and
health care provider services and nursing home services and prior to 1/1/2006, pharmacy claims
(Bradley, et al., 2007). The use of Medicaid pharmacy claims for estimating medication use has
been found to be accurate. McKenzie, Semradek, McFarland, Mullooly and McCamant (2000)
compared Medicaid pharmacy claims to nursing home resident chart information and found an
85% agreement. For this study, the researcher examined pharmacy Medicaid paid claim files
data to determine which pain medications were billed to Medicaid as billed medications will

93

indicate medications that were ordered by a clinician and received at a pharmacy for the
MIChoice participant.
Michigan Cancer Registry and Death Certificate Data. The Michigan Cancer
Registry and death certificate data are collected by the Michigan Division for Vital Records and
Health Statistics through reporting of diagnoses of cancer and deaths by physicians. The
Michigan Division for Vital Records and Health Statistics is supervised by the Michigan
Department of Community Health for the purpose of monitoring the health of Michigan citizens
(Michigan Department of Community Health, 2009). The Michigan Cancer Registry contains
data, reported by physicians, regarding patient’s date of diagnosis, cancer site, histology, cell
behavior, summary stage and related morphological descriptors. Death certificates must be filed
for every known death. Death certificate data includes: data defining date of death, underlying
and related causes of death and categorical place of death code.
Minimum Data Set Home Care. The Minimum Data Set Home Care (MDS-HC)
(Morris, Fries, Steel, et al., 1997) is completed around the time of admission to the MIChoice
waiver program by a nurse or social worker case manager trained in using the MDS-HC.
Additional MDS-HC assessments are then performed quarterly or sooner if a major change
occurs in the waiver participant’s health or a major event, such as a fall. This study utilized
MDS-HC data from multiple points in time over the course of participation in the MIChoice
waiver program.
The MDS-HC was developed from the Minimum Data Set (MDS) (Morris, Hawes, &
Fries, 1990). The MDS is used in Medicaid and Medicare certified nursing homes as part of the
federally mandated process for clinical assessment of all residents (Centers for Medicare &
Medicaid Services, 2009a). The MDS-HC is a comprehensive, standardized questionnaire for

94

evaluating the care needs, strengths and preferences of clients of home care agencies (Landi, et
al., 2000) and has been validated for use in United States and international community-based
populations (Kwan, et al., 2000; Landi, et al., 2000; Morris, Fries, Steel, et al., 1997) with the
weighted kappa score for the MDS-HC averaging 0.72 (Morris, Fries, Steel, et al., 1997).
The MDS-HC is comprised of 223 items that assess home care clients’ sociodemographic
characteristics, functional and cognitive status, social support, psychosocial well-being, clinical
diagnoses and nursing needs relevant to care planning (Landi, et al., 2005; Morris, Fries, Steel, et
al., 1997; Shugarman, et al., 1999). While the MDS-HC contains many items directly from the
MDS, home care-specific items were also developed and included in the MDS-HC to address
areas unique to the home environment, such as role of informal supports and indicators of abuse
(Morris, Fries, Steel, et al., 1997).
MDS items within specific domains were combined to create internal scales that include
the Cognitive Performance Scale (CPS) (Morris, et al., 1994), the MDS Depression Rating Scale
(DRS) (Burrows, et al., 2000) and the MDS Activities of Daily Living scale (ADLS) (Morris,
Fries, & Morris, 1999). While these internal scales were originally developed from MDS items,
researchers have since altered the scales to incorporate items from the MDS-HC, as some MDSHC items differ from their MDS item counterparts in order to better assess the home-care client
instead of the nursing home client (Morris, Fries, Steel, et al., 1997). The proposed research will
use the CPS, the DRS and the ADLS. These scales will be further described in following
Operational Definitions and Measures section.
Operational Definitions & Measures
Operational definitions and measurement of the key study variables are described below
and are organized according to their function within the proposed analyses: as dependent

95

variables, covariates, as both independent variables and covariates and as both independent and
dependent variables. Cancer, cancer site and cancer stage will serve as both independent
variables and covariates, depending on the research question. Variables defining pain experience,
pain management strategies and pain management outcomes will serve as both independent and
dependent variables within the proposed research, depending on the research question. At the
completion of this section, Table 2 is presented as a summary of study variables and includes
place in the conceptual model, data source, item information, coding and variable type.
Dependent Variables
Admission to a Nursing Home. Admission to a Nursing Home is the movement of a
HCBWP participant from the HCBWP to a nursing home as detected by a change in the
Medicaid paid claim files level of care coding from “22” (MIChoice) to “2” (nursing home).
The measure of transition described if during over the period of four MDS-HC assessments and
prior to 12/31/2005, the subject is admitted to a nursing home. Admission to a Nursing Home
was rated as “remains in MIChoice”=0 vs. “to nursing home”=1. Admission to a Nursing
Home was used as a dichotomous variable.
Time to Nursing Home Admission. Time to Nursing Home Admission is the number of
months from admission to the HCBWP until the admission of HCBWP participant to a nursing
home. Time to nursing home admission was measured by a variable developed from Medicaid
paid claim files and will indicate the time, in total number of months from HCBWP admission
to admission to nursing home. The admission of a HCBWP participant to a nursing home was
detected in the Medicaid paid claim files by a change in the level of care coding from a “22”
(MIChoice) to “2” (nursing home). Subjects who stayed in the HCBWP prior to 12/31/2005

96

were censured, i.e. their time to nursing home admission will be from the date of HCBWP
admission to 12/31/2005. Time to nursing home admission was treated as a continuous
variable.
Covariates
Covariates for the proposed analyses were comprised of variables defining the domains
of person and health and illness and included age, sex, race/ethnicity, cognitive functioning,
depression, comorbidities and time in HCBWP.
Age. Age is the chronological age of the HCBWP participant, as measured in years and is
determined from the HCBWP participant or proxy reported date of birth of the HCBWP
participant as recorded on the MDS-HC face sheet. Age was used as a continuous variable in the
analytic models (Appendix A).
Sex. Sex is the sex - male or female of the HCBWP participant, as reported by the
participant or proxy and recorded on the MDS-HC face sheet. Sex was treated as a
dichotomous variable in analytical models (Appendix A).
Race/Ethnicity. Race/Ethnicity is the race and/or ethnicity of the HCBWP participant, as
identified and reported by the participant or proxy and recorded on the MDS-HC face sheet:
Caucasian, Black, American Indian, Other (includes Asian and Pacific Islander), Unknown and
Hispanic. Race/ethnicity was a categorical variable (Appendix A).
Cognitive functioning. Cognitive functioning is represented by the dichotomized
hierarchical scale score of the Cognitive Performance Scale (CPS) (Morris, et al., 1994),
adapted for use with items from the MDS-HC (Fries, James, & Aliaga, 2004). The original CPS
is a measure of cognitive impairment and has been validated against (Morris, et al., 1994;
Paquay et al., 2007) and correlated (r=-0.65) with the Mini-Mental Health Examination

97

(Gruber-Baldini, Zimmerman, Mortimore, & Magaziner, 2000). The adapted CPS has been
correlated against the Mini-Mental Health Examination (r=0.81) (Landi, et al., 2000). While
the adapted CPS has been used in research examining home care populations, additional
validity and reliability measures of the adapted CPS were not reported (L. Li & Conwell, 2007;
Soldato, et al., 2007).
The adapted CPS is comprised of four MDS-HC items which includes the HCBWP
participant’s memory recall after 5 minutes, ability to make decisions, make self understood and
eating dependency (Fries, James, & Aliaga, 2004; Morris, et al., 1994). The nurse or social work
assessor asks the items of the caregiver and/or family member to determine cognitive
performance. A scoring algorithm was then completed using the categorical responses to each of
the four CPS items to develop the hierarchical scaled CPS score (Appendices A & B). Scoring of
the CPS was as follows: 0=cognitively intact, 1=borderline intact, 2=mild impairment,
3=moderate impairment, 4=moderately severe impairment, 5=severe impairment, 6= very severe
cognitive impairment, with “2” or greater indicative of cognitive impairment (Morris, et al.,
1994). For the analyses the CPS score was dichotomized: 0=cognitively intact (score 0-1) and
1=cognitively impaired (> 2). Instead of using the full CPS scale (0-5) the CPS score was
dichotomized to clearly distinguish which subjects was cognitively intact versus cognitively
impaired.
Behaviors indicative of depression. Behaviors indicative of depression is represented by
the score of the Depression Rating Scale (DRS) (Burrows, et al., 2000), adapted for use with
seven MDS-HC items. The original DRS scale was comprised of seven items from the MDS that
document behaviors that are indicative of depression (Table 1). During development, the original

98

DRS was validated against Hamilton Depression Rating Scale and Cornell Scale for Depression
(Burrows, et al., 2000) and has demonstrated adequate internal consistency reliability
(Cronbach’s alpha 0.71-0.74) (Burrows, et al., 2000; Martin et al., 2008). Additional validity
testing of the DRS in older adult populations has been recommended (Anderson, Buckwalter,
Buchanan, Maas, & Imhof, 2003; Burrows, et al., 2000). The six items in the adapted DRS (See
Table 1) differ from the original DRS as items indicative of depression in the MDS-HC (Fries,
James, & Aliaga, 2004) differ from the MDS. Li and Conwell (L. Li & Conwell, 2007) adapted
the DRS for use with MDS-HC items while examining the mental health status of communitydwelling Michigan older adults with demonstrated reliability (Cronbach’s alpha 0.74).
The adapted DRS was the HCBWP participant or proxy observation of the HCBWP
participant exhibiting in the 30 days prior to assessment feelings of sadness, persistent anger,
repetitive anxious complaints, sad facial expressions, recurrent crying and withdrawal from
activities of interest (Burrows, et al., 2000; L. Li & Conwell, 2007). Response coding for the
adapted DRS included 0=indicator not exhibited in last 30 days; 1=indicator of this type
exhibited daily or almost daily up to 5 days a week; 2=indicator of this type exhibited daily or
almost daily (6-7 days a week). Responses were then summed to create a possible score of 0-12,
with “ 0” meaning no behaviors indicative of depression over the previous 30 days and “12”
meaning feelings of sadness, persistent anger, repetitive anxious complaints, sad facial
expressions, recurrent crying and withdrawal from activities all exhibited daily or almost daily up
to 5 days a week. A DRS score of 3 or above was indicative of depression (Burrows, et al., 2000;
L. Li & Conwell, 2007). For the study, the adapted DRS score was used as a continuous
variable.

99

Table 1.
Items from the Adapted and Original Depression Rating Scale
Adapted MDS-HC
Original MDS-NH
Depression Rating Scale
Depression Rating Scale
Indicators
Indicators
(Li & Conwell, 2007)
(Burrows et al., 2000)
 A feeling of sadness or being  Resident made negative
depressed, that life is not
statements
worth living, that nothing
matters and he or she is of no
use to anyone or would
rather be dead
 Persistent anger with self or  Persistent anger and
others (easily annoyed, anger
irritability with self and
at care received
others
 Repetitive anxious
complaints, concerns

 Expressions of what appear
to be unrealistic fears
 Repetitive health complaints
 Repetitive anxious
complaints/concerns (nonhealth-related)

 Sad, pained, worried facial
expressions

 Sad, pained, worried facial
expressions

 Recurrent crying, tearfulness  Crying, tearfulness
 Withdrawal from activities
of interest

 N/A

Comorbid conditions. Comorbid conditions was represented by a comorbidities measure
indicating the summed, weighted effect of the presence of multiple diseases other than cancer
(diabetes, chronic pulmonary disease, congestive heart failure, cerebrovascular disease,
peripheral vascular disease, paralysis, acute myocardial infarction, old myocardial infarction,
moderate/severe renal disease, diabetes with complications, ulcer disease, rheumatologic
disease and mild liver disease) on the pain experience, pain management strategies, pain
management outcomes and admission to a nursing home of HCBWP participants. The measure
100

of comorbid conditions was developed from MDS-HC data utilizing the same statistical
methods used by Klabunde, Potosky, Legier and Warren (2000) while adapting the Charlson
comorbidity index (CCI) (Charlson, Pompei, Ales, & MacKenzie, 1987) to determine noncancer related mortality.
Klabunde and colleagues’ (2000) adaptation of the CCI was selected to develop the
comorbidity index for the proposed research for the following reasons. First, Klabunde and
colleagues developed the adaptation of the CCI with outpatient ICD-9 coding which is more
similar to the MDS-HC data than inpatient hospital ICD-9 codes used in older CCI adaptations
(Deyo, Cherkin, & Ciol, 1992; Romano, Roos, & Jollis, 1993). Second, Klabunde and colleagues
corrected statistical errors (Harrel, 1996; Romano, et al., 1993) used in the original CCI and
other adaptations. Lastly, Klabunde and colleagues’ method had greater ease of use and equal
predictive power compared to Elixhauser and colleagues’ comorbidity measure with 30 comorbid
conditions (Baldwin, Klabunde, Green, Barlow, & Wright, 2006). The development of the
comorbidity index measure is described in the following sections.
Data to develop the comorbid conditions measure came from the MDS-HC Section I,
questions 7 and 9 and utilized data from the admission, or first assessment of the waiver
participant. Section I in the MDS-HC collects data detailing physicians, hospitalizations and
presence of endocrine, circulatory, cardiovascular, musculoskeletal, neurological, psychiatric,
pulmonary and sensory conditions. Question 7 assesses for presence of specific diseases, with
possible responses as 0=not present, 1= present, not subject to focused treatment, 2=present,
monitored or treated by home care nurse. MDS-HC Section I, question 9 allows for the assessor
to enter in specific ICD-9 codes not addressed or addressed fully in Question 7. The MDS-HC
data was then searched for the presence of the following 13 comorbid conditions: diabetes,

101

chronic pulmonary disease, congestive heart failure, cerebrovascular disease, peripheral vascular
disease, paralysis, acute myocardial infarction, old myocardial infarction, moderate/severe renal
disease, diabetes with complications, ulcer disease, rheumatologic disease and mild liver disease.
Comorbid conditions were indicated as not present=0, present=1 based on their presence in the
MDS-HC section I, questions 7 or 9.
The dichotomized individual comorbidities were then used in multinomial logistic
regression models which were fitted using the Pain Scale score (Fries, et al., 2001) as the
dependent variable. Comorbidity data from the first MDS-HC assessment was used to predict the
Pain Scale score from the first MDS-HC assessment. Using the method described in Klabunde et
al. (2000), the estimated coefficients from the multinomial logistic regression models for each
comorbidity were multiplied by its dichotomous indicator (not present, present) and then
summed over all conditions to create the comorbidity index score for each subject, with a higher
score indicating higher comorbidity. The resulting summed comorbidities index score was used
as a continuous variable.
The measure of comorbid conditions used in this research had several weaknesses. First,
the comorbid condition measure may have underestimated the presence of comorbid conditions
in each subject. Other researchers who have developed comorbid measures have used Medicaid
and Medicare claim files data and chart data to determine the presence of comorbid conditions
(Charlson, et al., 1987; Deyo, et al., 1992; Klabunde, et al., 2000; Romano, et al., 1993). The
present research used data from the MDS-HC, as the researcher did not have access to Medicare
paid claim files and in older adults 65 and older have Medicare as the primary health care service
payor. Although the researcher had access to Medicaid paid claim files, health care providers
may not have billed Medicaid (the secondary payor) for services after Medicare paid and

102

therefore, there would be no ICD-9 coding for conditions related to treatment located in the
Medicaid paid claim files. Second, information regarding the presence of comorbid conditions
was at a single moment in time at each assessment. Whereas researchers utilizing Medicare and
Medicaid data could search the data for specific time periods before or after an event, the MDSHC data was at one moment in time. This may have led to an underestimation of the presence of
comorbid conditions. Third, the data did not allow the researcher to determine if conditions that
were present were severe or mild and therefore, the researcher was not able to weight conditions
according to severity.
Time in HCBWP. Time in the HCBWP is the number of months from the date of
admission to the date of each of the assessments for each subject. Time in HCBWP was
measured by a variable developed from Medicaid paid claim files and the MDS-HC. The date
that each subject began the HCBWP is noted in the Medicaid paid claim files, while the date of
each assessment for each subject is recorded in the MDS-HC. The date of beginning the program
will be subtracted from each assessment to provide the number of days, at the time of
assessment, from the date the subject began the HCBWP.
The incorporation of the Time in HCBWP variable as a covariate is necessary as the
study data is unbalanced- meaning each subject may have data from different points in time from
his or her admission to the HCBWP (Fitzmaurice, Laird, & Ware, 2004). Even though the goal
is to assess each patient quarterly, some assessment may be closer or further apart. Therefore,
time in the HCBWP was included as a covariate (Fitzmaurice, et al., 2004). Time in HCBWP
was a continuous variable.

103

Independent Variables/Covariates
Diagnosis of cancer, cancer site and stage are variables within the domain of health and
illness that were used as both independent variables and covariates in analytic models in order to
make comparisons between HCBWP participants with and without a diagnosis of cancer.
Diagnosis of cancer. Diagnosis of cancer was categorically defined using “phases of
care” definitions based on the date of death in relation to the date of cancer diagnosis (Brown,
Riley, Potosky, & Etzioni, 1999; G. F. Riley, Potosky, Lubitz, & Kessler, 1995; Yabroff, et al.,
2005; Yabroff, et al., 2009). “Phases of care” definitions were originally developed in
accordance with the use of healthcare services within specific time periods after a diagnosis of
cancer and before death (Brown, et al., 1999; G. F. Riley, et al., 1995; Yabroff, et al., 2005;
Yabroff, et al., 2009). Data from the Michigan Cancer Surveillance Data by the Michigan
Division for Vital Records Coding were utilized. Coding was as follows: 0=no cancer, 1=initial
phase, 2=continuing phase, 3=terminal phase.
The measure of diagnosis of cancer was determined from cancer registry data. Subjects
who do not have a diagnosis of cancer recorded in the cancer registry data were placed in the “no
diagnosis of cancer” level of the diagnosis of cancer measure. For subjects who survived at least
24 months after diagnosis of cancer, the initial phase level of the diagnosis of cancer measure
was defined as the first 12 months after diagnosis. The terminal phase of the diagnosis of cancer
measure was defined as the final 12 months preceding death and the continuing phase as all
months between the initial and terminal phases. For subjects surviving less than 24 months, the
final 12 months was the terminal phase and all other months were part of the initial phase. There
were no continuing phase for these subjects (Yabroff, et al., 2005; Yabroff, et al., 2009).
Subjects with cancer were further described according to cancer stage and cancer site.

104

Cancer stage. Cancer stage was classified categorically based on the Surveillance,
Epidemiology and End Results (SEER) summary stage from categorical data from the
Michigan Cancer Surveillance Data by the Michigan Division for Vital Records and Health
Statistics: 1=in situ, 2=local, 3=regional, 4=distant, 05=un-staged, 09=invasive unknown.
Cancer site. Cancer site was classified categorically in terms of the primary or initial
anatomical location of the cancer from the Michigan Cancer Surveillance Data by the Michigan
Division for Vital Records and Health Statistics. Preliminary analysis was completed to
determine distribution and cancers were categorized and coded as follows: 1=Colon, 2=Lung,
3=Lymphoma/Leukemia, 4=Breast, 5=Female Reproductive, 6=Prostrate, 7=other.
Dependent/ Independent Variables
Pain. The variable pain, which defines the pain experience, is a hierarchal
variable- The MDS Pain Scale- developed from two MDS-HC items which describe pain
frequency and pain intensity (Chou & Chi, 2007; Fries, et al., 2001; L. Li & Conwell, 2007).
The measure of pain experience was used as a dependent variable in question 1 and as an
independent variable in question 2 and 3.
To develop the pain variable, MDS-HC Section J items were used. Section J of the MDSHC was designated as the “Health Conditions and Preventive Measures” section. This section
included items that inquired about preventative health (vaccinations, prostate check,
mammogram, pap smear), pain, fall risk, changes in health status and hygiene and abuse
screening. Two items in Section J, item 8a and b, were used for the MDS Pain Scale: 8a) Over
the past 7 days, the HCBWP participant or proxy reported frequently complains or shows
evidence of pain (no pain, pain less than daily, pain daily) and 8b) HCBWP participant or proxy
reported pain is unusually intense (0=no, 1=yes) were used. The MDS-HC Pain Scale scoring is

105

as follows: “0”= no pain; “1”=mild pain (less than daily); “2”= moderate pain (daily, not
unusually intense); 3= intense pain (daily, unusually intense). The hierarchical scoring is
presented by diagram in Appendix D.
Pain management strategies. Variables defining the pain management strategies
dimension include prescribed pain medications and hospice services. Prescribed pain
medications and hospice services will be used as dependent variables for Question 2 analyses
and independent variables in Question 3 analyses.
Prescribed pain medications. Prescribed pain medications is the type of pain medications
prescribed by a health care provider for the HCBWP participant and billed to Medicaid in the 30
days period before each assessment. The measure of prescribed pain medication was developed
from Medicaid paid claim files pharmacy data. Within the pharmacy data are three drug codes:
Generic Therapeutic Class, Therapeutic Drug Class and Specific Therapeutic Drug Class. These
three drug codes were used to find medications used to treat pain. The process of categorizing
prescribed pain medication from the Medicaid pain claim files pharmacy data was a very time
intensive process due to inconsistencies in the data and this process is described further in the
following section.
Descriptive analyses was used initially examine the pharmacy paid claim files and
Generic Therapeutic Drug Class (GTDC) was found to most stable and basic variable to begin
the process of categorizing prescribed pain medications. The GTDC codes were then examined
to determine which medications were clinically consistent with pain management in accordance
with the literature and the researcher’s clinical experience. These included: 02=Analgesics,
03=Analgesics and Antihistamine combinations, 05=Anesthetics, 11=Anti-arthritics, 44=CNS
Drugs and 80=Psychotherapeutic drugs.

106

The GTDC codes were then cross-tabulated with the Standard Therapeutic Drug Class
(STDC) to determine which STDC codes existed within each GTDC. Within Analgesics, there
was a STDC code for “emetics”, but on further examination with the 3rd code (HIC3 Specific
Therapeutic Drug Class (HIC3)) these “emetics” were found to be aspirin, Tylenol, Oxycontin
and Vicodin and not emetics. Furthermore, the GTDC code for anti-arthritic medications was
the same as the STDC code for psycho-stimulants. On further examination, medications noted to
be psycho-stimulants under the STDC coding had HIC3 codes for Vioxx, Motrin, which are not
psycho-stimulants but COX-2 inhibitors, a type of non-steroidal anti-inflammatory medication.
Overall, by carefully using all three medication codes to carry out extensive crosstabulation procedures among the three drug codes, the research was able to create a prescribed
pain medication variable which was coded 0=no prescribed pain medications, 1=non-opioid
prescribed pain medications, 2= opioid prescribed pain medications and 3=adjuvant prescribed
pain medications. Non-opioid, opioid and adjuvant pain medications were described in Chapter
3. To briefly review, non-opioid pain medications include acetaminophen and non-steroidal antiinflammatory medications. Opioid pain medications included medications that were either in part
or fully narcotic. Adjuvant analgesics describes “…non-opioid medications that have pain-relieving
effects in certain conditions, but whose primary or initial indication was not for the treatment of pain”
(American Pain Society, 2005, p. 73). Medications that are used as adjuvant pain medications include
anticonvulsants, antidepressants, and local anesthetics (American Pain Society, 2005). Adjuvant
medications diminish pain by altering nerve function. There was unfortunately no measure within the

data documenting medication administration or subject ingestion of medication. Prescribed pain
medications were used as an independent variable in the analytic models examining the
association between the pain experience, prescribed pain management strategies and pain
management outcomes.
107

Hospice Services. Hospice services was the reporting of scheduling and adherence of
hospice service utilization by the HCBWP participant per the MDS-HC. The measure of hospice
services was developed from Section Q of the MDS-HC titled Service Utilization. Items in
Section Q inquired about recent surgery, formal care that the participant is already receiving,
treatments (for example, drug treatment, chemotherapy, cardiac rehabilitation), therapies
(exercise, physical therapy), programs (day care, hospice care, clinic visit) and special
procedures done in the home (EKG, skin treatment, special diet). For the measure of hospice
services, Item 3y was used and worded as follows, Special treatments, therapies, program
received or schedules during the last 14 days (received in home or outpatient basis) and
adherence to the required schedule. Response to the item was 0=N/A, 1=scheduled, full
adherence as prescribed, 2=scheduled partial adherence, 3=scheduled, not received. For this
research, hospice was dichotomized as follows: 0=Hospice services not received (original code 0
and 3) 1=Hospice services received (original code 1 and 2).
Pain management outcomes. Variables defining the pain management outcomes
dimension include pain control and physical function. Pain control and physical function are
used as dependent variables for Question 2 analyses and independent variables in Question 3
analyses.
Pain control. Pain control is the waiver participant’s or the HCBWP participant’s or
proxy’s response to an MDS-HC item in Section “J” Health Conditions and Preventative Health
Measures. Section “J” included items that inquired about preventative health (vaccinations, prostate
check, mammogram, pap smear), pain, fall risk, changes in health status and hygiene and abuse screening.

The measure of pain control was developed from a single item (J8e) asking “pain controlled by
medication” with possible responses as 0=no pain, 1=medication offered, no control and 2=pain
is partially or fully controlled by medication (Appendix E). Pain control was utilized as a
108

categorical variable.
Physical function. Physical function is represented by a count of the following activities
of daily living (ADLs) that the HCBWP participant is dependent in: dressing, personal hygiene,
toilet use, bathing and eating (Appendix E). Physical function score ranges from 0 to 5, with a
higher score indicating more dependency with ADLs. As pain has been shown to have a
significant, negative effect on physical functioning among older adults (Helme & Gibson, 2001;
Onder, et al., 2006; Reyes-Gibby, et al., 2002; Soldato, et al., 2007), physical function, as
measured by the number of ADLs the HCBWP participant is dependent in, is an outcome of pain
management strategies.
The measure indicating physical function is comprised of the observed ability of the
HCBWP participant in regards to dressing, personal hygiene, toilet use, bathing and eating over
the previous 7 days as measured by MDS-HC items (Section P, Question 2) (C. Given,
Spoelstra, You, Haque, & Given, 2010; Morris, et al., 1999). Responses to each item address the
amount of assistance the MIChoice participant needs with mobility in bed, transferring,
locomotion, dressing, eating, toilet use and personal hygiene over the previous 7 days:
0=independent, 1=supervision, 2=limited assistance, 3=extensive assistance and 4=total
dependence 5=Activity did not occur, regardless of ability . For this study, item scores were then
recoded to : 0 to1=independent “0” and > 2 =dependent “1”. Responses to each of the five items
were then summed together for a total possible score ranging from 0-5, with a higher value
indicating greater dependence with activities of daily living. The continuous variable defining
physical function was used as both a dependent variable for Question 2 analyses and as a
covariate for Question 3 analyses.

109

Table 2.
Study Variables by Place in Conceptual Model, Data Source, Item, Coding and Variable Type
Variable

Age

Place in
Conceptu
al Model
Domain of
Person

Data
Source
MDS-HC
face sheet,
#2

Item
Information/
Question
 Date of birth

Coding
 Continuous
variable
Number of years

Sex

Domain of
Person

MDS-HC  Gender
face sheet,
#3

 Male
 Female

Race/
Ethnicity

Domain of
Person

MDS-HC  Race
face sheet,
#7






Cognitive
Domain of
Functioning Person

MDS-HC
Sections
E 1& 2
F2
P 2e





 Will be
assessed using
the Cognitive
Performance
Scale (CPS), a
subscale created
from 4 items of
the MDS-HC.
The CPS scores
will then be
dichotomized to
indicate

whether a
HCBWP
participant is
cognitively
intact vs.
cognitively
impaired

110

Variable
Type
 Continuous
 Covariate
Questions
1,2,3
 Categorical
 Covariate
Questions 1,
2, 3
 Categorical
 Covariate
Questions
1,2,3

Caucasian
Black
American Indian
Other (includes
Asian and Pacific
Islander
Unknown
Hispanic
The CPS is a 7 Categorical
point hierarchical  Covariate
summary scale
Questions
that rates
1,2,3
cognitive
impairment from
0=intact to
6=very severe.
The CPS is
scored via a
decision tree.
The CPS is then
dichotomized:
0-1=cognitively
intact
> 2=cognitively
impaired

Table 2 (Continued )

Variable

Place in
Conceptu
al Model
Domain of
Person

Data Source

Diagnosis
of Cancer

Domain of
Health &
Illness

Michigan
Cancer
Surveillance
Registry

Cancer site

Domain of
Health &
Illness

Michigan  Recorded via
Cancer
ICD-9 codes
Surveillance
Registry

Cancer
stage

Domain of
Health &
Illness

Michigan
Cancer
Surveillance
Registry

Behaviors
Indicative
of
Depression

Item
Coding
Variable
Information/
Type
Question
MDS-HC  Will be
 The DRS is a  Continuous
Section G 2
assessed using
summary
 Covariate
the Depression
scale 0-12
Questions
Rating Scale,
possible with
1,2,3
a subscale
a higher score
created from 6
indicating
MDS-HC
more
items.
behaviors that
are indicative
of depression
 Diagnosis of
Cancer

 Records stage
according to
Surveillance,
Epidemiolog
y and End
Results
(SEER)
summary
stage.
111

 No diagnosis
of cancer
 Initial Phase
 Continuation
Phase
 Terminal
Phase

 Categorical
 Independent
Questions
1,2,
 Covariate 3

 1=Colon
 Categorical
 2=Lung
 Independent
Question 1,2
 3=Lymphoma
or Leukemia  Covariate 3
 4=Breast
 5=Female
Reproductive
 6=Prostate
 7=Other







1=In situ
2=Local
3=Regional
4=Distant
5=Un-staged
9=Invasive,
unknown

 Categorical
 Independent
Questions
1,2
 Covariate 3

Table 2 (Continued )

Variable

Comorbid
conditions

Pain

Place in
Conceptual
Model
Domain of
Health &
Illness

Pain
Experience
Dimension

Data
Source
MDS-HC
Section I

MDS-HC
Section J8

Item
Coding
Variable
Information/
Type
Question
Responses are  Continuous
 Disease/
infection that categorized as:  Covariate
is indicated
 A summed,
Questions
as present
weighted
1,2, 3
and affects
index was
client’s
developed
status,
utilizing
requires
Klabunde et
treatments or
al’s method
requires
(Klabunde,
symptom
Harlan, &
management.
Warren,
2006;
Klabunde, et
al., 2000).
 Range for
mortality as
outcome: 027.
Note: For
present
research
used pain as
outcome.
 MDS Pain
 0-3, with
 Categorical
Scale
0=no
 Dependent
-Frequently
reported pain
Question 1
complains or
to 3=daily
shows evidence
pain that is
 Independent
of pain in last 7
unusually
Questions 2
days?
intense
&3
-Pain is
unusually
intense?

112

Table 2 (Continued )

Variable

Prescribed
pain
medication

Hospice
Services

Place in
Conceptual
Model
Pain
Management
Strategies
Dimension

Pain
Management
Strategies
Dimension

Data
Source
Medicaid
paid claim
files

MDS-HC
Section Q
3y

Item
Coding
Information/
Question
 Includes all  Medications
pain
known to be
medications
specific for
billed to
pain
Medicaid
treatment
for time
will be
period of 30
categorized
days prior
& coded as
to each
0=no pain
assessment.
medications
1=nonopioid pain
medication
2=opioid
pain
medication
3=adjuvant
pain
medication
 Special
 0= Hospice
treatments,
services not
therapies,
received
programs
 1=Hospice
received or
services
scheduled
received
during the
last 14 days
(received in
the home or
on an
outpatient
basis) and
adherence to
the required
schedule

113

Variable
Type
 Categorical
 Dependent
and
Independent
Question 2
 Independent
Question 3

 Dependent
Question 2
 Independent
Question 3

Table 2 (Continued)

Variable

Physical
Function

Place in
Conceptual
Model
Pain
Management
Outcomes

Data
Source
MDS-HC
Section P 2

Item
Coding
Information/
Question
 Will be
 Summary
assessed
scale from 0using a sum
5, with a
of the
higher value
number of
indicating
ADLs the
more ADL
participant
dependencies
is
.
dependent
in
 Includes
dressing,
eating,
toileting,
personal
hygiene,
and bathing

Variable
Type
 Continuous
 Dependent
Question 2
 Independent
Question 3

Pain
control

Pain
Management
Outcomes

MDS-HC  Pain
 0=Medicatio
Section J 8e
controlled
n offered no
by
control,
medication?  1=Pain is
partially or
fully
controlled

Admission
to Nursing
home

Admission to
Nursing
Home

Medicaid  Variable
 0=Remains
 Categorical
paid claims
will describe in MIChoice  Dependent
files
if subject
(NO)
Question 3
was
 1=To nursing
admitted to
home (YES)
a nursing
home from
the HCBWP
prior to
12/31/05

114

 Categorical
 Dependent
Question 2
 Independent
Question 3

Table 2 (Continued )

Variable

Time to
nursing
home
admission

Place in
Conceptual
Model
Time to
Nursing
Home
Admission

Time in the
HCBWP

Time in
HCBWP

Data
Source
Medicaid
paid claims
files

Medicaid
Paid Claim
Files &
MDS-HC
Section A2

Item
Coding
Information/
Question
 Time from  Total
HCBWP
number of
admission
months from
to
admission
admission
day to
to nursing
HCBWP
home
until
admission to
 Subjects
a nursing
who stay in
home or end
the waiver
of
program
measurement
prior to
time.
12/31/05
will be
censured
 Event=
admission
to nursing
home

Variable
Type
 Continuous
 Dependent
Question 3

 Time from  Total
 Continuous
HCBWP
number of
 Covariate
admission
months from
Questions 1
to each
HCBWP
2&3
MDS-HC
admission
Assessment
day to each
MDS-HC
assessment

Data Preparation and Management
The study was a secondary analysis of a large dataset comprising data from Michigan
Medicaid paid claims files; the home care version of the Minimum Data Set (MDS-HC); and
cancer diagnosis, tumor staging, and death certificate information from the Michigan Cancer
Surveillance Data by the Michigan Division for Vital Records and Health Statistics. The study

115

data set included multiple assessments of the waiver participant, services billed to Medicaid,
diagnoses of cancer.
Development of and access to this combined dataset was overseen by the Institute of
Health Care Studies (IHCS) at Michigan State University. The IHCS has been granted direct
access to the data by the Michigan Department of Community Health (MDCH). IHCS staff first
extracted Medicaid paid claim files from 6/1/2002 to 12/31/2005 and MDS-HC data from
1/1/2003 to 12/31/2005. The MDS-HC data was then matched to the Medicaid data through
social security number, date of birth and first and last name. With this initial dataset constructed,
IHCS staff consulted with the state registrar to obtain date of death, date of cancer diagnosis,
staging and tumor description records from the Michigan Division for Vital Records. The state
registrar matched death and cancer-related records to the Medicaid paid claim files and MDS-HC
data by social security number, date of birth and first and last name. The completed dataset was
then returned to the IHCS where staff removed all patient identification from the data. Subjects
will not be identifiable, directly or through identifiers linked to the subjects in the final study
dataset to which the applicant will have access.
The de-linked dataset was provided in its entirety to a statistician in the College of
Nursing Research Center and is kept on a password protected computer to which only the
statistician has access to. The researcher was not able to identify persons in the data set because
it is de-linked. The researcher consulted with the statistician and obtained a copy of the de-linked
dataset. All data was received in electronic form. The study data was kept on a password
protected laptop computer that only the researcher has access to. A codebook was developed for
the MDS-HC data. Coding guides were also available for the cancer diagnosis, tumor staging,
and death certificate and Medicaid data. Both the MDS-HC and Medicaid data were broken

116

down into sections for ease of use. The statistician utilized both SAS and SPSS to complete the
analyses for this study.
Analysis of Data
The primary purpose of this study was to examine longitudinal differences in the pain
experience, pain management strategies and pain management outcomes among older HCBWP
participants with and without a diagnosis of cancer over the course of time of participating in the
HCBWP. The secondary purpose of this study was to determine differences in how the pain
experience, pain management strategies and pain management outcomes among older adult,
HCBWP participants with and without a diagnosis of cancer associates with the admission of
older adult HCBWP participants to a nursing home over the course of time of participating in the
HCBWP. Data were analyzed using SPSS and SAS. Tests had a 0.05 set level of significance.
Power. Power analysis was conducted using G*power software (Faul, Erdfelder, Lang, &
Buchner, 2007) utilizing a z-test for the difference between two independent proportions. For
the P1 and P2 proportion values the researcher used values for the prevalence of pain in HCBWP
participants with a diagnosis of cancer after admission to the HCBWP and HCBWP participants
with no diagnosis of cancer: 0.77 and 0.72, respectively. The last assessment for the presence of
pain (no pain or less than daily vs. daily pain) was used. For the allocation ratio the researcher
used the percentage of older adult HCBWP participants with a diagnosis of cancer over the
percentage of HCBWP participants without a diagnosis of cancer for an allocation ratio of 5.67.
A two-tailed test with 0.05 error probability and 0.90 power was computed. For an actual power
of .90, the total sample size required was 6189 subjects: 928 in the group with a diagnosis of
cancer and 5261 in the group without a diagnosis of cancer. Effect size was 0.11, indicating a
small effect (Cohen, 1988).

117

For Medicaid-eligible adults, 65 and older and HCBWP participants from 1/1/0312/31/05, the sample size for the proposed research would include approximately 12,750
individuals. With the proposed research being a secondary data analysis of such a large dataset,
the sample adequacy to detect differences between groups should be assured by the large dataset
size.
Analysis.

Initial data analysis served to determine sample size and to assess

distribution of data for each variable in order to make final decisions regarding categorical
coding of variables. Descriptive statistical analysis was completed to profile the study sample in
terms of demographic characteristics and other study variables. Continuous variables were
summarized with the number of observations, mean, standard deviation, range and 95% CI for
the mean. The relationships among the variables were examined.
Research questions and proposed plans for analysis of longitudinal data were as follows:
1) How does the pain experience differ between older adult HCBWP participants in
regards to cancer over time? How is the relationship over time between the pain
experience and diagnosis of cancer affected by sex, age, race, comorbid conditions,
depression and cognitive functioning over time?


Dependent variables: Pain.
o For Research Question 1, the pain measure was dichotomized as follows Pain
Scale 0-1=”0”, Pain Scale 2-3=”1” to provide a dichotomous dependent variable
for Generalized Estimated Equation Model.

 Independent variable: Diagnosis of cancer as: no cancer, initial phase, continuation phase,
terminal phase. This variable was named as “group” in the following statement and the rest
of the analysis section.

118

 The following model addressed research question 1:
Pain= Diagnosis of cancer group + covariates
The main effect was the group (no cancer, initial phase, continuation phase, terminal
phase). Covariates include age, sex, race/ethnicity, depression, cognitive functioning and
comorbidities. Age, depression, comorbidities and Time in HCBWP are continuous variables
while sex, race/ethnicity and cognitive functioning are categorical variables. Repeated measures
analysis was used. SAS procedure GENMOD with Generalized Estimating Equations modeling
(Lipsitz & Kim, 1994) was used for the categorical outcome of pain. To account for the relation
of multiple measurements across time within a patient, auto regressive 1 (ar(1)) covariance was
specified. Categorical outcomes with more than two levels were treated as ordinal response and
the proportional odds model was used. If the proportional odds assumption (M. E. Miller, Davis,
& Landis, 1993) did not stand, then the dependent variables were dichotomized as binaries based
on their frequency.
2. How does the pain experience of older adult, HCBWP participants relate to pain
management strategies and pain management outcomes and how does this relationship
differ in regards to diagnosis of cancer over time? How is the relationship between the pain
experience, pain management strategies, pain management outcomes and diagnosis of
cancer affected by sex, age, race, comorbid conditions, depression and cognitive
functioning over time?
Dependent variable: prescribed pain medications, hospice, pain control and physical
function. Each dependent variable was used once in the model to create four separate GEE
models. Prescribed pain medications was categorical 1=opioids, 2=adjuvants, 3=non-opioid, as
defined by the World Health Organization analgesic ladder (World Health Organization, 2002)

119

as well as 0=no pain medication. For Research Question 2, dummy variables were created to
provide a dichotomous dependent variable for GEE modeling. The dummy variables compared
each prescribed pain medication level (1-3) against prescribed pain medication=”0” no pain
medication. Separate GEE models were then carried out for each dummy variable.
Independent variable: 1) Groups as diagnosis of cancer: no cancer, initial phase,
continuation phase, terminal phase 2) Pain.
The following model addressed research question 2: dependant variable (prescribed pain
medications, hospice, pain control, physical function) across time= group + covariates + Pain .
The main effect was group and pain. Pain was represented by the full pain scale (Fries, et al.,
2001), including values 0 to 3. Covariates include age, sex, race, comorbid conditions,
depression, cognitive functioning and Time in HCBWP.
Repeated measures analyses were implemented. GENMOD with Generalized Estimating
Equations modeling was used for binary, count and categorical outcomes (Lipsitz & Kim, 1994).
For categorical outcomes, proportional odds methods was used. To account for the relation of
multiple methods across time with patients, auto regressive 1 (ar(1)) covariance was specified.
3. How does the pain experience, pain management strategies and pain management outcomes
of older adult, HCBWP participants predict the admission and time to admission of older
adult HCBWP participants to a nursing home over time and how does this relationship
differ in regards to diagnosis of cancer while accounting for sex, age, race, comorbid
conditions, depression and cognitive functioning?


For research question 3, an analysis was performed that included subjects who stayed in
the HCBWP and those who were admitted to a nursing home in order to investigate
whether variables defining the pain experience, pain management strategies and pain

120

outcomes related to this admission and time to admission. Subjects who had died or left
the program for reasons other than admission to a nursing home were excluded from this
analysis.
 Dependent variable:
1) Admission to nursing home: yes/no. Logistic regression was used for this
analysis
2) Time to transition: Cox proportional survival analysis models were used. For those who
transferred to the nursing home, the time in the waiver program was programmed as “event” and
those who remained in the waiver program until 12/31/05 were censored. Proportional hazard
ratio was tested to ensure that the assumption of analysis stood. If the proportional hazard ratio
had not stood then time to nursing home admission would have been treated as a continuous
variable and general linear modeling would have been used to carry out the analysis.
 Independent variable: Groups as diagnosis of cancer, pain, prescribed pain medications,
hospice, pain control, physical function
 Covariates: include age, sex, race, comorbid conditions, depression, cognitive functioning and
Time in HCBWP.
The following models were used to address research question 3:
 Admission to nursing home (yes versus no) =independent variables + covariates
 Time to admission to nursing home= group + independent variables + covariates
 The main effect in both models were pain, prescribed pain medications, hospice, pain
control, physical function. Pain was represented by the full pain scale (levels 0-3).
Prescribed pain medications was represented by 0=no pain medications, 1=non-opioid pain
medications, 2=opioid pain medications, 3=adjuvant pain medications. The covariates

121

included groups as diagnosis of cancer: no cancer, initial phase, continuation phase,
terminal phase.
Study Limitations
This research was a secondary analysis of pre-existing data and therefore, data analyses
was limited to what was already recorded. The study was limited to mostly categorical data and
therefore information regarding pain measures was not as detailed as when numeric or visual
pain measurement scales are used. The MDS-HC was collected by interviewing both the patient
and informal caregivers. If the patient was unable to respond, the data would be limited to what
the caregiver provided. Proxy reporting can be inaccurate and thus the pain assessment measures
provided by the caregiver may not have truly reflected what the patient’s pain experience was.
Generalizability to older adults who are able to self-report measure of pain was therefore limited.
The data does not include information regarding patient need of pain medication and whether a
subject took pain medication, only that pain medication was prescribed and billed to Medicaid.
This research was limited to four assessments, thereby limiting the sample to older adult
HCBWP participants who were in the HCBWP for approximately 15 months. Therefore, older
adult HCBWP participants who left the HCBWP in less than 15 months were excluded, creating
a selection bias.
The researcher acknowledges that pain management strategies include more than
prescribed pain medication and hospice services. Pain management strategies include both
pharmacological and non-pharmacological strategies (JCAHO, 2000) such as cold and heat
therapy, massage, acupuncture, for example. The consequences of poor pain management
among older adults includes clinical symptoms of depression, anxiety, decreased social
interaction, sleep disturbances, impaired physical function, agitation, delirium, decreased

122

appetite, delayed healing, lower quality of life and higher health care utilization and costs
(American Geriatrics Society Panel on Persistent Pain in Older Persons, 2002). Although the
researcher acknowledges the many negative outcomes of pain and poor pain management, this
research was limited to examining the effects of pain and pain management on pain control,
physical function and transfer to a nursing home. Additional outcomes such as the financial costs
of pain management were not possible due to data limitations. Additional pain management
outcomes will be examined in future research. Finally, the proposed research did not seek to
determine if pain was acute, chronic or cancer-related, as this was outside of the scope of the
project.
Study Strengths
The dataset of this study was very large, thereby allowing for more precision in
estimation of population properties than a smaller dataset. The proposed research examined
cancer, pain and pain management in older adult waiver program participants. An extensive
literature search showed no previous research has addressed this issue. Therefore, this research
examined an area that although significant, had not been previously addressed. The proposed
research was longitudinal and allowed for multiple measures over time. This provided a clinical
and management “picture” of each subject as they participated over time in MIChoice. The
proposed research served to provide an initial descriptive assessment of pain in a HCBWP on
which future research can build.
Protection of Human Participants
The study was granted exempt status by the Michigan Department of Community Health
(MDCH) Institutional Review Board and the Michigan State University Institutional Review
Board (IRB# X08-742), as the study was a secondary analysis of de-identified data. For the

123

study, the researcher applied for and received permission to access Medicaid paid claim files,
Michigan Cancer Registry data, death certificate information from Vital Records, and MDS-HC
data from the Michigan Department of Community Health (MDCH) Institutional Review Board.
Data use and non-disclosure agreements were completed by the researcher as required by the
MDCH. As the data are de-linked, there are minimal additional risks to the subject. All data
were kept confidential. Although the subjects of the proposed research will not likely receive
additional benefits from the proposed research, future older adult HCBWP participants may
benefit from the findings of the research.
Data Security
All data for the study were stored on a password protected laptop computer that only the
researcher had access to. All data had been de-linked and the researcher does not have access to
the identification key. The researcher completed mandatory IRB training regarding protection of
human subjects.

Women and Minority Inclusion in Clinical Research
The study was a secondary analysis of existing data and therefore, the inclusion of
women and minorities in the proposed study was as they are already represented in the existing
data set. From 2005 records, the Bureau of Medicaid Policy and Actuarial Services reported the
breakdown by gender among MIChoice participants as 72% female (Bureau of Medicaid Policy
and Actuarial Services of Michigan Department of Community Health, 2004, 2006). Also from
the 2005 records, minorities comprised a small percentage of MIChoice participants: African
Americans 20% and Hispanics 1%. The majority of persons in MIChoice in the 2005 records
were Caucasian, 75% (Bureau of Medicaid Policy and Actuarial Services of Michigan
Department of Community Health, 2004, 2006). A similar breakdown in sex and race in the
124

dataset for the proposed research was expected and descriptive statistics regarding sex and race
are reported in Chapter 5.
Conclusion of Chapter 4
The purpose of this chapter was to present the design and methods that were used for this
study as well as human subject protection and data safety. Chapter 5 will describe the results of
the analyses. Chapter 6 will present contributions to science and implications for policy, clinical
practice and research.

125

Chapter 5
The primary purpose of this study was to examine longitudinal differences in the pain
experience, pain management strategies (prescribed pain medications and hospice services) and
pain management outcomes (physical function and pain control) among older HCBWP
participants with respect to diagnosis of cancer while participating in the HCBWP. The
secondary purpose of this study was to determine what differences exist in how the pain
experience, pain management strategies and pain management outcomes among older adult,
HCBWP participants associates with the admission of older adult HCBWP participants to a
nursing home, with respect to diagnosis of cancer while participating in the HCBWP. This study
addressed the following research questions:
1) How does the pain experience differ between older adult HCBWP
participants in regards to diagnosis of cancer over time? How is the relationship
between the pain experience and diagnosis of cancer affected by sex, age, race,
comorbid conditions, depression and cognitive functioning over time?
2) How does the pain experience of older adult, HCBWP participants relate
to pain management strategies and pain management outcomes and how does this
relationship differ in regards to diagnosis of cancer over time? How is the
relationship between the pain experience, pain management strategies, pain
management outcomes and diagnosis of cancer affected by sex, age, race,
comorbid conditions, depression and cognitive functioning over time?
3) How do the pain experience, pain management strategies and pain
management outcomes of older adult, HCBWP participants predict the admission
and time to admission of older adult HCBWP participants to a nursing home over

126

time and how does this relationship differ in regards to diagnosis of cancer while
accounting for sex, age, race, comorbid conditions, depression and cognitive
functioning?
Generalized estimating equation (GEE) analyses (Liang & Zeger, 1986) were used to
respond to Research Questions 1 and 2. GEE is an extension of generalized linear models that
provides a semi-parametric approach to longitudinal data analysis with univariate outcomes for
which the quasi-likelihood formulation if sensible, i.e. normal, Poisson, binomial and gamma
response variables. This approach allows for the analyses of longitudinal data where there are
multiple measures of a subject characteristic over time (Liu, Dixon, Qiu, Tian, & McCorkle,
2009; Molenberghs & Verbeke, 2005). When a subject has a characteristic assessed at multiple
time points, the assumption of independence cannot be met as subject response at one time point
is very likely to predict the subject’s response at a future time point (Fitzmaurice, et al., 2004;
Singer & Willett, 2003). GEE analysis takes the dependence among multiple measures of a
subject characteristic over time into consideration and allows for examination of changes in the
subject characteristic over time (Fitzmaurice, et al., 2004; Singer & Willett, 2003).
GEE analysis requires the specification of distribution (normal, Poisson, binomial and
gamma response variables) and link functions which connects the expected value of the
dependent variable to the linear combination of the independent variable and covariates (Liu, et
al., 2009). SAS and SPSS can use the GEE approach for the longitudinal analysis of binomial or
multinomial outcomes. Few examples in the literature were found to support the use of
computer language other than SPSS or SAS to perform multinomial longitudinal analysis (Lee,
Kang, Liu, & Seo, 2010). Thus, the technique of multinomial longitudinal data analysis is not

127

well established and for the current research the researcher chose to use the built in approach in
SAS.
Based on the dependent variables of research questions 1 and 2, an ordinal multinomial
GEE approach would have been preferred. However, this approach had two disadvantages. First,
it required that the outcome had an ordinal distribution (i.e. the proportional odds assumption
stands). Second, only an independence covariance matrix could be assumed. To test the ordinal
distribution of the outcomes of pain and prescribed pain medications, the model at time point 1
with covariates was carried out by SAS logistic procedure. The proportional odds assumption
was rejected (p value< 0.001 for both models ). Therefore, the ordinal multinomial approach was
not appropriate for the analyses for Research Question 1 and 2.
For the analyses, Research Question 1 was decomposed into 2 sub-questions: 1) How do
the covariates affect the older adult HCBWP participant who experienced daily pain vs. those
with no pain or less than daily pain and 2) How do the covariates affect the older adult HCBWP
participant with daily unusually intense pain vs. those with daily not unusually intense pain.
For Research Question 1, a binomial link function or distribution was specified. Pain was
dichotomized “0”=Pain Scale 0-1 (no pain, less than daily pain) and 1= (daily pain, both not
unusually intense and unusually intense). Because pain was dichotomized for GEE analyses for
Research Question 1 the results were limited to examining the likelihood of older adult HCBWP
participants with daily pain vs. older adult HCBWP participants with no pain or less than daily
pain instead of the full pain scale (0-3).
The first section of Research Question 2 examined the association between the pain
experience and pain management strategies. The dependent variable, prescribed pain
medications, was originally a nominal variable. Because the GEE analyses were limited to either

128

dichotomous or count dependent variables, dichotomous dependent variables which compared no
prescribed pain medications and non-opioid, opioid and adjuvant pain medications were
developed. For the second part of Research Question 2, which addressed associations between
the pain experience, pain management strategies and physical function among older adult
HCBWP with and without cancer, a Poisson distribution was specified as the dependent variable
was a count of the number of ADL dependencies. The Poisson distribution (also known as loglinear) is used for the analysis of counts of the number of times an event occurs in time or space
(Fitzmaurice, et al., 2004; Liu, et al., 2009). For this research, physical function was represented
by the count of activities of daily living that the subject was dependent in. The use of and results
of the GEE analyses are described further in the Results section later in Chapter 5.
The following section will present sample characteristics followed by descriptive
analyses of study variables. All descriptive analyses of study variables included examining for
differences in regards to diagnosis of cancer. Diagnosis of cancer was represented using the
“phases of care” definitions based on the date of death in relation to the date of cancer diagnosis
((Brown, et al., 1999; G. F. Riley, et al., 1995; Yabroff, et al., 2005; Yabroff, et al., 2009).
Preliminary descriptive analyses of the diagnosis of cancer measure revealed that there were very
few subjects in the “terminal” phase at each assessment, making longitudinal modeling difficult
(Table 3). Therefore, subjects in the continuing phase and terminal phase were consolidated
under a single category “continuing/terminal”. Descriptive statistics for the revised diagnosis of
cancer at each assessment are presented in Table 4. Across four assessments, cancer diagnosis
remained consistent over time with approximately 6-7% of subjects having a diagnosis of cancer
at each assessment.

129

Table 3
Reported Diagnosis of Cancer at Each Assessment Time Point
No
Diagnosis
of Cancer
Time 1
(%)

3812
(94)

Initial
Phase
N
(%)
96
(2)

Continuing
Phase
N
(%)
145
(4)

Terminal
Phase
N
(%)
1
(0)

Time 2
(%)

3797
(94)

92
(2)

160
(4)

5
(0.1)

4054

Time 3
(%)

3786
(93)

80
(2)

177
(4)

11
(0.3)

4054

Time 4
(%)

3774
(93)

68
(2)

193
(5)

19
(0.5)

4054

Total
Response
N
4054

Table 4
Revised Measure of Reported Diagnosis of Cancer at Each Assessment Time Point
No
Cancer
N
(%)
Time 1
(%)

3812
(94)

Initial
Phase
N
(%)
96
(2)

Continuing/Terminal
Phase
N
(%)
146
(4)

Time 2
(%)

3797
(94)

92
(2)

165
(4)

4054

Time 3
(%)

3786
(93)

80
(2)

188
(5)

4054

Time 4
(%)

3774
(93)

68
(2)

212
(5)

4054

130

Total
Response
N
4054

Sample
A total of 4054 subjects met the inclusion criteria for this study. Exclusion criteria were
used to ensure that subjects had completed five MDS assessments prior to 12/31/2005 which was
the date that Medicaid ceased being the sole payer of prescription medications for dually eligible
(Medicaid and Medicare) older adults. Additionally, in order to clearly specify diagnosis of
cancer, cancer survivors (those who had a diagnosis of cancer five years prior to the MDS
assessment) were excluded. Other exclusion criteria and the determination of the final sample
size of 4054 subjects are depicted in Figure 4. Analyses were started at the second MDS
assessment to allow for at least 30 days in the Medicaid paid claim files prior to each MDS
assessment. The second assessment will be referred to as Time 1 from this point on.
Subject socio-demographic characteristics by diagnosis of cancer at Time 1 are presented
in Table 5. Overall, the majority of subjects were female (80%, n=3238) and Caucasian (74%
n=3010). Race level “other” was comprised of American Indian (n=5), Asian and Pacific
Islanders (n=12), Unknown (n=74) and Hispanic (n=37. The overall mean subject age at Time 1
was 77 years (not in Tables). There was no significant association between diagnosis of cancer
and age or race at Time 1 (Table 5) and there was no significant difference in mean age by
diagnosis of cancer at Time 1 (Table 5).

131

Table 5
Significance of Differences in Diagnosis of Cancer by Sociodemographic Characteristics at
Time 1 among Study Subjects (n=4054)
Variable

No
Diagnosis
of Cancer
N
(%)

Initial
Phase of
Cancer
N
(%)

Continuing
/Terminal
Phase of
Cancer
N
(%)

Total
Response
N

Signif.
Testing

760
(93)

23
(3)

33
(4)

816

Chi-Square
p-value=
0.47

3052
(94)

73
(2)

113
(4)

3238

2825
(94)

71
(2)

114
(4)

3010

Black

866
(94)

23
(3)

27
(3)

916

Other

121
(94)

2
(1)

5
(4)

128

Mean Age at
Time 1
(SD)

77.39
(7.91)

76.19
(7.45)

76.97
(8.11)

4054

Sex
(% of row)
Male

Female

Race/Ethnicity
Caucasian

132

Chi-Square
p-value=
0.76

ANOVA
Df=2
F=1.26
Sig=0.28

-Continuously Medicaid-Eligible, 65 and
older between 1/1/02 and 12/31/07.
-Enrolled in Home and Community-Based
Waiver Program between 1/1/02 and
12/31/07.
- Begin assessing subjects at second MDS
Assessment to allow time for Medicaid claim
files prior to assessment
n=12,750

Exclude subjects who:
-Did not have at least 5
consecutive MDS
assessments completed
before 12/31/05
(n=5848)
-Left waiver program for
reasons other than death
or transition to nursing
home (n=2722)
-Those who only had
“benign” staging in
Cancer Registry Data.
(n=1)
-Those who had all
cancer diagnoses > 5
years before 4
consecutive MDS
assessments (n=125)

Final Sample
n= 4054

Figure 4. Final Exclusion Criteria Flow Chart for Determination of Sample

133

Measures
Operational definitions and measurement of study variables were presented in Chapter 4.
The following section presents descriptive analyses results of sample response to study variables.
The purposes of this study are focused on differences in the pain experience, pain management
strategies, pain management outcomes and admission to a nursing home in respect to diagnosis
of cancer. Therefore, each variable was examined in relation to the measure of diagnosis of
cancer.
Cognitive Functioning
Cognitive functioning was represented by the dichotomized Cognitive Performance Scale
(CPS) (Morris, et al., 1994) score. The CPS was comprised of four MDS-HC items: memory
recall after 5 minutes, ability to make decisions, ability to make self understood and eating
dependency. A scoring algorithm (Appendix C) used the scores for memory recall after 5
minutes, ability to make decisions, ability to make self understood and eating dependency to
score the CPS score as follows: 0=cognitively intact, 1=borderline intact, 2=mild impairment,
3=moderate impairment, 4=moderately severe impairment, 5=severe impairment 6=very severe
cognitive impairment.
The CPS score was then dichotomized as 0=cognitively intact (0 to 1) and > 1=
indicative of cognitive impairment (> 2) (Morris, et al., 1994) (Table 6). Overtime (Table 6), the
percentage of subjects who were cognitively impaired remained consistent in those with no
cancer, initial phase or continuing/terminal phase. There was no significant association between
cognitive functioning and diagnosis of cancer at each assessment time point.

134

Table 6
Significance of Differences in Cognitive Functioning by Diagnosis of Cancer at Each Assessment
Time Point (n=4054)

Time
Point

X2 of

Cognitive Functioning
0=Cognitively Intact
1=Cognitively Impaired

Cognitive
Functioning by
Diagnosis of
Cancer at Each
Time Point

0
n
(% of
row)

1

2

1
n
(% of
row)

Total
Response
n

2539
(63)

1501
(37)

4040

2488
(62)

1551
(38)

4039

X2 =1.47
Df =2
Sig = 0.48

X2 = 0.60
Df = 2
Sig = 0.74

3

4

2430
(60)

1610
(40)

2371
(59)

1669
(41)

X2 =0.19

4040

4040

Df =2
Sig = 0.91

X2 =2.31
Df =2
Sig =0.31

Note. Response column does not include missing data

Behaviors Indicative of Depression
The measure of behaviors indicative of depression was represented by the Depression
Rating Scale (DRS) (Burrows, et al., 2000). The DRS is comprised of 6 items from the MDS-HC
that document behaviors indicative of depression: feelings of sadness, persistent anger, repetitive
135

anxious complaints, worried facial expressions, recurrent crying and withdrawing from social
activities (Fries, James, & Aliaga, 2004; L. Li & Conwell, 2007). Item response coding is as
follows for the 30 days before each MDS-HC assessment: 0= indicator not exhibited; 1=indicator
exhibited daily or almost daily up to 5 days a week; 2=indicator exhibited daily or almost daily
(6-7 days per week). Responses were then summed to create the DRS scale, with 0-12 possible
and a higher number indicating more behaviors indicative of depression.
For the present study, the Cronbach’s alpha for the DRS was 0.74, which is consistent
with the DRS Cronbach’s alpha of 0.74 presented by Li and Conwell (2007). Table 7 presents
descriptive statistics for behaviors indicative of depression at each time point. A DRS score of a
“3” or above is considered indicative of depression (Fries, James, & Aliaga, 2004; L. Li &
Conwell, 2007). Using this criterion, 13-15% of subjects had a DRS score > 3 at each time point
(not in Tables).
There were no significant differences in mean behaviors indicative of depression in
regards to diagnosis of cancer at each time point (Table 7). Mean behaviors indicative of
depression across the four time points by diagnosis of cancer are presented in Figure 5. There
was some variation in mean behaviors indicative of depression over the four time points by
diagnosis of cancer. Most noticeably, subjects in the “initial phase” of cancer diagnosis appeared
have a more pronounced increase in mean behaviors indicative of depression at Time 2 than
subjects with no cancer or subjects in the “continuing/terminal phase”. However, analysis via
ANOVA showed no significant differences in mean behaviors indicative of depression in regards
to diagnosis of cancer at Time 2 (Table 7).

136

Figure 5. Mean Reported Behaviors Indicative of Depression by Cancer Diagnosis across Time
Points (n=4054)

137

Table 7
Significance of Differences in Mean Behaviors Indicative of Depression by Diagnosis of Cancer
at Each Assessment Time Point
ANOVA:
Behaviors
Indicative of
Depression
By Diagnosis of
Cancer At Each
Time Point
Df=2
F=0.62
Sig=0.54

Time

Mean
(sd)

Response

Time 1

1.03
(1.72)

4052

Time 2

0.99
(1.69)

4052

Df=2
F=0.65
Sig=0.52

Time 3

0.99
(1.68)

4052

Df=2
F=0.76
Sig=0.47

Time 4

0.99
(1.70)

4051

Df=2
F=1.15
Sig=0.32

Note. Response column does not include missing data

Comorbid Conditions
Comorbid conditions were represented by a comorbid conditions measure indicating the
summed, weighted effect of the presence of multiple diseases other than cancer (diabetes,
chronic pulmonary disease, congestive heart failure, cerebrovascular disease, peripheral vascular
disease, paralysis, acute myocardial infarction, old myocardial infarction, moderate/severe renal
disease, diabetes with complications, ulcer disease, rheumatologic disease and mild liver disease)
(Klabunde, et al., 2000) on pain, as measured by the Pain Scale (Fries, et al., 2001). As part of

138

the development process for the comorbid conditions index multinomial logistic regression
models were used to make comparisons of the association of the pain scale (Fries, et al., 2001)
with different comorbid condition measures: 1) Klabunde’s (Klabunde, et al., 2000) method for
a comorbid index measure; 2) a simple count of comorbid conditions and 3) Charlson’s
comorbidity index method (Charlson, et al., 1987).
Table 8
Model Fit Statistics for Multinomial Logistic Regression Models for the Association between the
Pain Scale (0-3) and Different Measures of Comorbid Conditions
-2 LL

AIC

R2

10860.13

10874.13

0.02

Simple
Comorbidity
Count

10925.51

10933.51

0.002

Via
Charlson’s
method

10931.19

10939.19

0.0006

Comorbidity
Measure
Via
Klabunde’s
method

The separate multinomial logistic regression models based on the different measures of
comorbid conditions used the pain scale from Time 1 (Fries, et al., 2001) as the dependent
variable and each comorbid measure as the independent variable. Results are presented in Table
2

8 . Based on the lower -2LL and AIC scores, as well as the higher R value, the use of the
measure of comorbid conditions developed through Klabunde and colleagues’ method
(Klabunde, et al., 2000) was supported. The final model included diabetes without
complications, chronic pulmonary disease, peripheral vascular disease and rheumatologic disease
which had a significant positive association with pain (Table 9).
139

Table 9
Multinomial Logistic Regression Parameter Estimates for Final Comorbid Condition Model:
Pain Scale (0-3) =Comorbid Conditions
Comorbid
Conditions

DF

Estimate

Standard
Error

Wald
ChiSquare

P-value

Diabetes
Without
Complications

1

0.1497

0.06

6.51

0.01

Chronic
Pulmonary
Disease

1

0.3312

0.06

27.14

< 0.001

Peripheral
Vascular
Disease

1

0.2402

0.07

11.22

0.001

Rheumatologic
Disease

1

1.0660

0.24

20.17

<0.001

The parameter estimates of the significant comorbid conditions were then added together
to create the summed comorbid conditions index score. Parameter estimates for the final model
are presented in Table 9. Bootstrap method was used (1000 repetitions) to determine the 95%
2

confidence interval for the R value (0.0174 -0.0193), which was consistent with the final model

R2 of 0.02.

The above significant comorbid conditions coefficient values (developed from Time

1) were then applied to assessment time points 2, 3 and 4 data as well. For all time periods mean
measure of comorbid conditions was approximately 0.22 (Table 10). There were no significant
differences in the mean measure of comorbid conditions in regards to diagnosis of cancer at each
time point (Table 10). Figure 6 shows there was little variability across time points in the mean
measure of comorbid conditions among subjects in regards to diagnosis of cancer.

140

Figure 6. Mean Measure of Documented Comorbid Conditions by Diagnosis of Cancer at Each
Assessment Time Point
Table 10
Significance of Differences in the Mean Measure of Comorbid Conditions by Diagnosis of
Cancer at Each Assessment Time Point
Time

Mean
(SD)

Response

Time 1

0.22
(0.25)

4054

ANOVA
Mean Comorbid Condition by
Diagnosis of Cancer
At Each Time Point
Df=2
F=0.06
Sig.=0.94

141

Table 10 (Continued)

Time

Mean
(SD)

Response

ANOVA
Mean Comorbid Condition by
Diagnosis of Cancer
At Each Time Point
Df=2
F=0.23
Sig.=0.80

Time 2

0.22
(0.25)

4054

Time 3

0.23
(0.25)

4054

Df=2
F=0.20
Sig.=0.82

Time 4

0.23
(0.25)

4054

Df=2
F=0.42
Sig.=0.66

Cancer Site
Per the diagnosis of cancer measure, each subject was either noted to have “no cancer” or
be in the initial or continuing/terminal phases of diagnosis of cancer. In addition to the diagnosis
of cancer measure, cancer site was used to note the primary or initial anatomical location of the
cancer for those in the initial or continuing/terminal phase of diagnosis of cancer. As presented
in Chapter 4, preliminary analysis of cancer site data was completed and cancer sites with small
numbers of subjects (for example Central Nervous System, Urinary, Upper Gastrointestinal)
were consolidated into the “other” category. Table 11 presents the breakdown by cancer site of
those with cancer at each time point. Overall, the percentage of each cancer remained
approximately consistent over time. The number of those in each cancer site increased as the
total number of those in the initial or continuing/terminal phase of diagnosis of cancer increased
over time. Breast cancer was the most common cancer and female reproductive the least
common at each time point.
142

Table 11
Number and Percentage of Cancer Site at each Assessment Time Point among Subjects with
Cancer
Cancer Site
(Column %)

Time 1
Sample
(%)
n=242
37
(15)

Time 2
Sample
(%)
n=257
41
(16)

Time 3
Sample
(%)
n=268
44
(16)

Time 4
Sample
(%)
n=280
45
(16)

Lung

22
(9)

23
(9)

22
(8)

22
(8)

Lymphoma
/Leukemia

21
(9)

23
(9)

25
(9)

29
10)

Breast

72
(30)

76
(30)

79
(30)

81
(30)

Female
Reproductive

18
(7)

20
(8)

21
(8)

22
(8)

Prostate

25
10)

27
(10)

29
(11)

31
(11)

Other
Diagnosis

47
(19)

47
(18)

48
(18)

49
(17)

Colon

Note. The sample “n” at each time point only includes those with a diagnosis of cancer
at each time point.

Cancer Stage
For those in the initial or continuing/terminal phases of diagnosis of cancer, the measure
of cancer stage was used to denote the stage of cancer at diagnosis. The measure of cancer stage
was developed using the Surveillance, Epidemiology and End Results (SEER) coding. Cancer
stage at the each assessment for those with cancer is presented in Table 12. There was very little

143

change over time in the percentage of the different stages among subjects with cancer. Overall,
the majority of cancers were stage 2-3 (local to regional at each assessment) at each time point.
Table 12
Number and Percentage of Stage of Cancer at Each Assessment Time Point among Subjects with
Cancer
Cancer Stage

Time 1
Sample
(%)
(n=242)
21
(9)

Time 2
Sample
(%)
(n=257)
21
(8)

Time 3
Sample
(%)
(n=268)
22
(8)

Time 4
Sample
(%)
(n=280)
23
(8)

2=Local

128
(53)

137
(53)

139
(52)

143
(51)

3=Regional

49
(20)

53
(25)

54
(20)

55
(26)

4=Distant

20
(8)

21
(8)

23
(9)

25
(9)

5=Unstaged

24
10)

25
(10)

29
(11)

32
(29)

9=Invasive
Unknown

0

0

1
(0.4)

1
(0.4)

1=In-situ

Note. The sample “n” at each time point only includes those with a diagnosis of cancer
Pain
The main variable of interest for this study was pain, which was measured via the Pain
Scale (Fries, et al., 2001). The Pain Scale is a hierarchal variable developed from two MDS-HC
items which describe pain frequency and pain intensity (Chou & Chi, 2007; Fries, et al., 2001; L.
Li & Conwell, 2007). The Pain Scale was coded 0=no pain, 1=less than daily pain, 2=daily, not
unusually intense pain and 3, daily, unusually intense pain. The algorithm for the Pain Scale can
be viewed in Appendix D. The two separate MDS-HC items that comprise the Pain Scale assess
144

pain frequency and pain intensity and responses at each time point are presented in Tables 13 and
14. In the MDS-HC, pain intensity is assessed only if pain frequency is > 1, therefore pain
intensity is limited to subjects with pain frequency=2.
Table 13
Significance of Differences in Pain Frequency by Diagnosis of Cancer at Each Assessment Time
Point (n=4054)
Assessment
Time Point

Pain Frequency
MDS-HC Item J8a
0=No Pain
1=Less
2=Daily
Than
Pain
Daily
Pain

X2 p-value

Time 1

911
(22)

874
(22)

2267
(56)

4052

for
Pain Freq.
by
Diagnosis
of Cancer
0.59

Time 2

915
(23)

890
(22)

2248
(55)

4053

0.81

Time 3

959
(24)

875
(22)

2217
(55)

4050

0.85

Time 4
(%)

940
(23)

873
(22)

2237
(55)

4050

0.52

Response

Note. Response column does not include missing data
There were no significant differences in pain frequency or pain intensity by diagnosis of
cancer at each time point (Tables 13 and 14). Table 15 presents the pain scale results at each time
point. Over half of all subjects experienced daily pain and approximately 30% experienced daily
pain that was unusually intense across the four assessments. The overall percentage breakdown
across different pain levels (none, less than daily, daily not usually intense and daily unusually
intense) remained relatively consistent across all four assessments (Table 15). Cross-tabulation
of pain by diagnosis of cancer at each time point showed there were no significant differences in
pain in regards to diagnosis of cancer at each time point (Table 15).
145

Table 14
Significance of Differences in Pain Intensity by Diagnosis of Cancer at Each Assessment Time
Point for Subjects With Daily Pain
Time
Point

Pain Intensity
MDS-HC Item J8b
0=Pain Not
1=Pain
Unusually Unusually
Intense
Intense

X2
Response

p-value for
Pain
Intensity
by
Diagnosis
of Cancer
0.69

Time 1
Sample
(%)

1617
(54)

1400
(46)

3017

Time 2
Sample
(%)

1624
(54)

1382
(46)

3006

Time3
Sample
(%)

1615
(54)

1358
(46)

2973

0.60

Time 4
Sample
(%)

1629
(54)

1360
(45)

2989

0.73

0.22

Note. Response column includes only those with Daily Pain at Each Time Point
Table 15
Significance of Differences in Pain Scale by Diagnosis of Cancer at
Each Assessment Time Point (n=4054)
Time
Point

Time 1

0=No
Pain

912
(23)

Pain Scale
1=Less
2=Daily, Not
than daily
Unusually
pain
Intense

875
(22)

948
(24)

3=Daily,
Unusually
Intense

1237
(31)

146

X2
Response

3972

p-value
Pain by
Diagnosis
of Cancer
0.70

Table 15 (Continued)

Time
Point

0=No
Pain

Pain Scale
1=Less
2=Daily, Not
than daily
Unusually
pain
Intense

3=Daily,
Unusually
Intense

X2
Response

p-value
Pain by
Diagnosis
of Cancer

Time 2

920
(23)

888
(22)

959
(24)

1212
(31)

3979

0.79

Time 3

962
(24)

875
(22)

946
(24)

1202
(30)

3985

0.90

944
870
969
1199
(24)
(22)
(24)
(30)
Note. Response column does not include missing data

3982

0.67

Time 4

Prescribed Pain Medications
Prescribed pain medications were ascertained from mediations billed to the Medicaid
Paid Claim Files in the 30 days prior to each of the four assessment dates. Drug coding within
the Medicaid paid claim files was used to categorize the medications as 0=no pain medications
prescribed, 1=non-opioid pain medications 2=opioid pain medications, 3=adjuvant pain
medications. The descriptive statistics for prescribed pain medications across at each time point
are presented in Table 16. The percentage of each level of Prescribed Pain Medications
remained consistent over the four time points.

147

Table 16
Number and Percentage of Prescribed Pain Medication at Each Assessment Time Point
(n=4054)
Prescribed Pain Medications
Time

Non-Opioid
Pain
Medications
N
(%)
469
(11)

Opioid Pain
Medications
N
(%)

Time 1
(row %)

No Prescribed
Pain
Medications
N
(%)
2498
(62)

960
(24)

Adjuvant
Pain
Medications
N
(%)
127
(3)

Time 2
(row %)

2450
(60)

443
(11)

1036
(26)

125
(3)

Time 3
(row %)

2461
(61)

437
(11)

1026
(25)

130
(3)

Time 4
(row %)

2427
(60)

453
(11)

1032
(25)

142
(4)

At each assessment, diagnosis of cancer was not significantly associated with prescribed
pain medications (Table 17). Cross-tabulation between Pain Scale and prescribed pain
medication data at each time point is presented in Table 18. Interestingly, on average across the
time points, 87% of those with “no pain” had no prescribed pain medications billed in the 30
days prior to each time point. On average across the time points, 43% of subjects with daily,
unusually intense pain had opioid pain medications prescribed pain medications billed in the 30
days prior to each time point. Of concern is that on average across the time points, 40 % of
subjects with daily, unusually intense pain had no pain medications billed in the 30 days prior to
each time point.

148

Table 17
Significance of Differences in Prescribed Pain Medication by Diagnosis of Cancer at Each
Assessment Time Point (n=4054)
Time
Point

Diagnosis
Of
Cancer

1
N=4054

0=No
Cancer
(n=3812)

2354
(62)

447
(12)

892
(23)

119
(3)

1=Initial
Phase
(n=96)

54
(56)

9
(10)

32
(33)

2=
Continuing
/Terminal
Phase
(n=146)

90
(62)

13
(9)

36
(25)

7
(5)

0=No
Cancer
(n=3797)

2303
(61)

417
(11)

958
(25)

119
(3)

1=Initial
Phase
(n=92)

56
(61)

9
(10)

26
(28)

1
(1)

2=
Continuing
/Terminal
Phase
(n=165)

91
(55)

17
(10)

52
(32)

5
(3)

X2

1
(1)

2
N=4054

Prescribed Pain Medication
0
1
2
3
No
Non-Opioid
Opioid
Adjuvant
Pain
Pain
Pain
Pain
Medication Medication Medication Medication
n
n
n
n
(row %)
(row %)
(row %)
(row %)

149

Prescribed
Pain
Medication
by Diagnosis
of Cancer at
Each Time

X2

= 8.43
Df =6
Sig = 0.21

X2 = 4.91
Df = 6
Sig = 0.55

Table 17 (Continued)

Time
Point

Diagnosis
Of
Cancer

3
N=4054

0=No
Cancer
(n=3786)

2296
(61)

409
(11)

960
(25)

121
(3)

1=Initial
Phase
(n=80)

43
(54)

11
(14)

23
(29)

2=
Continuing
/Terminal
Phase
(n=188)

122
(65)

17
(9)

43
(23)

6
(3)

0=No
Cancer
(n=3374)

2249
(60)

431
(11)

960
(25)

134
(4)

1=Initial
Phase
(n=68)

46
(68)

6
(9)

14
(21)

2
(3)

2=
Continuing
/Terminal
Phase
(n=212)

132
(62)

16
(8)

58
(27)

6
(3)

X2

3
(4)

4
N=4054

Prescribed Pain Medication
0
1
2
3
No
Non-Opioid
Opioid
Adjuvant
Pain
Pain
Pain
Pain
Medication Medication Medication Medication
n
n
n
n
(row %)
(row %)
(row %)
(row %)

150

Prescribed
Pain
Medication
by Diagnosis
of Cancer at
Each Time

X2 =3.26
Df =6
Sig =0.78

X2 =5.27
Df =6
Sig =0.51

Table 18
Number and Percentage of Prescribed Pain Medication by Measure of Pain at each Assessment
Time Point (n=4054)
Time
Point

Pain Scale
Freq
(Row %)

Time 1
(N=3972)

0=No Pain

799
(88)

57
(6)

35
(4)

21
(2)

912

1=Less
that Daily
Pain

601
(69)

123
(14)

124
(14)

27
(3)

875

2=Daily
pain, Not
Unusually
Intense

530
(56)

118
(12)

270
(28)

30
(3)

948

3=Daily
Pain,
Unusually
intense

516
(42)

159
(13)

511
(41)

48
(4)

1237

0=No Pain

796
(86)

63
(7)

45
(5)

16
(2)

920

1=Less
that Daily
Pain

599
(67)

112
(13)

149
(17)

28
(3)

888

2=Daily
pain, Not
Unusually
Intense

528
(55)

110
(11)

289
(30)

32
(3)

959

3=Daily
Pain,
Unusually
intense

477
(39)

152
(13)

536
(44)

47
(4)

1212

Time 2
(n=3979)

0=
No Pain
Medication

Prescribed Pain Medications
1=
2=
3=
Non-opioid
Opioid
Adjuvant
Pain
Pain
Pain
Medication Medication Medication

151

Total

Table 18 (Continued)

Time
Point

Pain Scale
Freq
(Row %)

Time 3
(N=3985)

0=No Pain

832
(87)

57
(6)

53
(5)

20
(2)

962

1=Less
that Daily
Pain

601
(69)

107
(12)

142
(16)

25
(3)

875

2=Daily
pain, Not
Unusually
Intense

489
(52)

120
(13)

301
(32)

36
(3)

946

3=Daily
Pain,
Unusually
intense

495
(41)

143
(12)

515
(43)

49
(4)

1202

0=No Pain

817
(86)

54
(6)

54
(6)

19
(2)

944

1=Less
that Daily
Pain

582
(67)

111
(13)

141
(16)

36
(4)

870

2=Daily
pain, Not
Unusually
Intense

514
(53)

132
(14)

294
(30)

29
(3)

969

3=Daily
Pain,
Unusually
intense

473
(39)

142
(12)

528
(44)

56
(5)

1199

Time 4
(N=3982)

0=
No Pain
Medication

Prescribed Pain Medications
1=
2=
3=
Non-opioid
Opioid
Adjuvant
Pain
Pain
Pain
Medication Medication Medication

152

Total

Hospice Services
The measure of hospice services was the reporting, scheduling and adherence of hospice
services utilization by the HCBWP participant, per the MDS-HC. Hospice services was
dichotomized as 0=hospice services not received, 1=hospice services received. Descriptive
statistics for hospice over the four assessments are presented in Table 19. Less than 1% of
subjects per assessment were receiving hospice services. Further analyses over the four
assessments reveal very little change over time in the overall percentage of subjects who were
receiving hospice services (Table 19).
Table 19
Significance of Differences in Hospice Services by Diagnosis of Cancer at Each Assessment
Time Point
Time
(Response for
Hospice
Services at
Each Time
Point)

Diagnosis
Of Cancer
(Response for
Diagnosis of Cancer
at Each Time Point)

Hospice Service
0
1
(% of
(% of
response for Diagnosis of
Diagnosis of
Cancer at
Cancer at
each Time
each Time
Point)
Point)

1
(n=3980)

0=No Cancer
(3739)

3734
(99.9)

5
(0.1)

1=Initial Phase
(95)

95
(100)

0

2=Continuing/Terminal
Phase
(146)

145
(99.3)

1
(0.7)

153

Fisher’s
Exact Test
Hospice
Services by
Diagnosis
of Cancer
at Each
Time
p=0.31

Table 19 (Continued)

Time
(Response for
Hospice
Services at
Each Time
Point)

Diagnosis
Of Cancer
(Response for
Diagnosis of Cancer
at Each Time Point)

2
(n=4047)

0=No Cancer
(3790)

3783
(99.8)

6
(0.2)

1=Initial Phase
(92)

92
(100)

2=Continuing/Terminal
Phase
(165)

165
(100)

0

0=No Cancer
(3782)

3774
(99.8)

8
(0.2)

1=Initial Phase
(80)

80
(100)

0

2=Continuing/Terminal
Phase
(188)

188
(100)

0

0=No Cancer
(3767)

3761
(99.8)

6
(0.2)

1=Initial Phase
(n=68)

67
(98.5)

Fisher’s
Exact Test
Hospice
Services by
Diagnosis
of Cancer
at Each
Time

0

1
(1.5)

3
(n=4050)

4
(n=4047)

Hospice Service
0
0
(% of
(% of
response for response for
Diagnosis of Diagnosis of
Cancer at
Cancer at
each Time
each Time
Point)
Point)

2=Continuing/Terminal
212
Phase
(100)
(212)
Note. Response does not include missing data

p=1.00

p=1.00

p=0.16

0

The Fisher’s exact test of the association between diagnosis of cancer and hospice
services at each time point was insignificant. The Fisher’s exact test was used instead of the ChiSquared test as there were cross tabulation cells with counts of less than five. Interestingly,
154

almost all of subjects receiving hospice services did not have cancer except for 1 in the diagnosis
of cancer continuing phase. Due to the extremely low response in the hospice services variable,
hospice services was not included in further analyses.
Pain Control
Pain control, a pain management outcome, was measured via a single MDS-HC item,
stating “pain is controlled by medication”, with the response of either 0=no pain, 1=medication
offered, no control, 2=pain is partially or fully controlled by medication. Descriptive statistics for
pain control across assessments is presented in Table 20. Percentages for the levels of pain
control were consistent over time.
Cross-tabulation between pain control and pain and pain control and prescribed pain
medications are presented in Tables 21 and 22, respectively. Pain control was highly sensitive to
the Pain Scale, as approximately 95% of those without pain per the measure of pain were also
noted as having no pain per the pain control measure over the four time points. On average over
the four time points, 45% of those who had medication offered with no control had daily,
unusually intense pain (Table 22).

Of subjects who reported no pain at the four time points, on

average only 13% had evidence of prescribed pain medications in the 30 days prior to each time
point, leading to the question if this sub-group’s lack of pain is due to not having pain instead of
pain being managed well by prescribed pain medication that was not captured in the 30 days
before assessment time period (Table 22). Unfortunately, the Medicaid paid claim files did not
capture the use of over the counter pain medications and perhaps this may be an explanation for
the high percentage of subjects with no pain who also had no pain medications prescribed. Over
the four assessments, diagnosis of cancer was not significantly associated with the measure of
pain control (Table 20).

155

Table 20
Significance of Difference in Pain Control by Diagnosis of Cancer at Each
Assessment Time Point (n=4054)

X2

Time
Point

Pain
Control
0=No Pain

Pain Control
1=Medication
Offered , No
Control

Pain
Control
2=Pain is
partially or
fully
controlled

Time 1

894
(23)

100
(2)

2959
(75)

3953

X2 =5.41

Time 2

913
(23)

102
(3)

2954
(74)

3969

X2 =2.56

951
(24)

107
(3)

2922
(73)

3980

953
(24)

101
(3)

2930
(73)

3984

Time 3

Time 4

Note. Response column does not include missing data

156

Response

Pain
Control
by
Diagnosis
of
Cancer
at Each
Time

Df =4
Sig =
0.25

Df =4
Sig =
0.75

X2 =2.82
Df =4
Sig =
0.59

X2 =2.41
Df =4
Sig =
0.66

Table 21
Number and Percentage for the Measure of Pain Control by Measure of Pain at Each
Assessment Time Point
Time
Point

Pain Control
Freq
(Row %)

Pain Scale

847
(95)
2
(2)

28
(29)

17
(18)

49
(51)

96

2=Pain is
partially or
fully
controlled by
medication
Time 2
Response=
3895

0=No pain

1=Less
than
Daily
Pain
25
(3)

1=Medication
offered, no
control

Time 1
Response=
3875

0=No
Pain

32
(1)

782
(27)

905
(31)

1170
(41)

2889

0=No pain

852
(93)

37
(4)

17
(2)

5
(1)

911

1=Medication
offered, no
control

1
(1)

30
(30)

21
(21)

47
(47)

99

2=Pain is
partially or
fully
controlled by
medication

39
(2)

785
(27)

907
(31)

1154
(40)

2885

157

2=Daily
Pain, not
Unusually
Intense
11
(1)

3=Daily
Pain,
Unusually
Intense
7
(1)

Total

890

Table 21 (Continued)

Time
Point

Pain Control
Freq
(Row %)

Time 3
Response=
3914

0=No pain

896
(94)

33
(3)

14
(2)

5
(1)

948

1=Medication
offered, no
control

1
(1)

33
(31)

26
(25)

45
(43)

105

2=Pain is
partially or
fully
controlled by
medication

39
(1)

780
(27)

895
(31)

1147
(40)

2861

0=No pain

888
(94)

42
(4)

13
(1)

5
(1)

948

1=Medication
offered, no
control

1
(1)

27
(27)

33
(33)

38
(39)

99

2=Pain is
partially or
fully
controlled by
medication

33
(4)

778
(27)

913
(32)

1145
(40)

2869

Time 4
Response=
3916

0=No
Pain

Pain Scale
1=Less
2=Daily
than
Pain, not
Daily
Unusually
Pain
Intense

Note. Response column does not include missing data

158

Total
3=Daily
Pain,
Unusually
Intense

Table 22
Number and Percentage of Pain Control by Measure of Prescribed Pain Medication at each
Assessment Time Point
Time
Point

Pain Control
Freq
Row %

Prescribed Pain Medication

0=No pain

785
(88)

49
(5)

36
(4)

24
(3)

894

60
(60)

9
(9)

27
(27)

4
(4)

100

2=Pain is
partially or
fully
controlled by
medication
Time 2
Response=
3969

1=NonOpioid
Pain
Meds

1=Medication
offered, no
control

Time 1
Response=
3953

0=No Pain
Meds

1569
(53)

404
(14)

890
(30)

96
(3)

2959

0=No pain

795
(87)

54
(6)

47
(5)

17
(2)

913

1=Medication
offered, no
control

59
(58)

9
(9)

32
(31)

2
(2)

102

2=Pain is
partially or
fully
controlled by
medication

1523
(52)

374
(13)

954
(32)

103
(3)

2954

159

2=Opioid 3=Adjuvant Response
Pain Meds Pain Meds

Table 22 (Continued)

Time
Point

Pain Control
Freq
Row %

Time 3
Response=
3980

0=No pain

Prescribed Pain Medication
Response
0=No Pain
1=Non2=Opioid 3=Adjuvant
Meds
Opioid
Pain Meds Pain Meds
Pain
Meds
831
48
53
19
951
(87)
(5)
(6)
(2)

1=Medication
offered, no
control

9
(8)

24
(22)

4
(4)

107

2=Pain is
partially or
fully
controlled by
medication

1496
(51)

379
(13)

943
(32)

104
(4)

2922

0=No pain

822
(86)

56
(6)

54
(6)

21
(2)

953

1=Medication
offered, no
control

61
(60)

11
(11)

25
(25)

4
(4)

101

2=Pain is
partially or
fully
controlled by
medication

Time 4
Response=
3984

70
(65)

1482
(51)

384
(13)

949
(32)

115
(4)

2930

Note. Response column does not include missing data
Physical Function
Physical function was represented by a count of the following activities of daily living
(ADL) that the HCBWP participant is dependent including dressing, personal hygiene, toilet use,
bathing and eating. Mean physical function at each time point is presented in Table 23. There
was no significant difference in mean physical function in regards to diagnosis of cancer at each
160

time point (Table 23). Across the four assessments, there was a mild increase in mean physical
function (indicating an increase in the mean number of ADL dependencies) overtime with some
variability among diagnosis of cancer (Figure 7). However, the ANOVA analysis results noted
there was no significant differences in mean physical function in regards to diagnosis of cancer
across time points (Table 23).
Table 23
Significance of Differences in Mean Physical Function by Diagnosis of Cancer at Each
Assessment Time Point
Time
Point

Mean
(sd)

Total at
Each Time

Time 1

1.81
(1.55)

4053

Time 2

1.85
(1.54)

4053

Time 3

1.89
(1.55)

4053

Time 4

1.96
(1.55)

4053

161

ANOVA for
Physical Function
by Diagnosis of
Cancer at Each
Time
Df=2
F=0.19
Sig=0.82
Df=2
F=0.13
Sig=0.88
Df=2
F=0.17
Sig=0.84
Df=2
F=0.01
Sig=0.99

Figure 7. Mean Measure of Physical Function by Measure of Diagnosis of Cancer At Each
Assessment Time Point
Admission to a Nursing Home
Admission to a Nursing Home measured whether a HCBWP participant is admitted to a
nursing home while participating in the HCBWP and does not return to the HCBWP. Admission
to a Nursing Home is dichotomous, with 0=remains in the HCBWP, 1=to a nursing home.
Descriptive statistics for Admission to a Nursing Home can be found in Table 24. Admission to
nursing Home was not re-measured at each assessment, but was regarded as a single event.
Therefore, descriptive analysis was completed as a single event and not at each time point.

162

Approximately 66% of the subjects stayed in the HCBWP prior to 12/31/05. There was no
significant association between admission to a nursing home and diagnosis of cancer (Table 25).
Table 24
Prior to 12 Percentage of Older Adult HCBWP Participants Admitted to Nursing Home /31/2005
Measure

Coding

Admission
to Nursing
Home

0=Subject stayed in
HCBWP
1=Subject admitted to
Nursing Home

Sample
Response
(%)
0=2667 (66)

Response

4054
1=1387 (34)

Table 25
Significance of Differences in the Measure of Admission to a Nursing Home Prior to 12/31/2005
by Diagnosis of Cancer
Diagnosis
Of Cancer

Admission to a Nursing Home
0
1
Total
Remain in
Admitted
Response
HCBWP
to Nursing
Home

0=
No Cancer
(row %)

2506
(66)

1306
(34)

3812

1=
Initial Phase
(row%)

70
(73)

26
(27)

96

2=
Continuing/Terminal
Phase
(row%)

91
(62)

55
(38)

146

163

X2 Admission
to a Nursing
Home by
Diagnosis of
Cancer

X2 =2.95
Df =2
Sig = 0.23

Time to Nursing Home Admission
Time to Nursing Home Admission was represented by the number of months from
admission to the HCBWP until the admission of HCBWP participant to a nursing home.
Subjects who stayed in the HCBWP past 12/31/2005 were censured, i.e. their time to nursing
home admission was from the date of HCBWP admission to 12/31/2005. Time to nursing home
admission was treated as a continuous variable. Mean time to nursing home admission was 33.48
months (sd=12.75) with a range of 3 months to 47 months. There were no significant differences
in mean time to nursing home admission by diagnosis of cancer (Table 26).
Table 26
Mean Time to Nursing Home Admission by Diagnosis of Cancer
Diagnosis
Of Cancer
(response)

Mean Time to
Nursing Home
Admission
(sd)

0=No Cancer
(3739)

33.53
(12.74)

1=Initial Phase
(95)

30.93
(12.62)

2=Continuing/Terminal
Phase
(146)

ANOVA
Admission to
a Nursing
Home by
Diagnosis of
Cancer
Df =2
F=0.199
Sig = 0.82

33.81
(12.75)

Note. Response does not include missing data
Results
The models examining the relationship among pain, pain management strategies, pain
management outcomes and nursing home admission were analyzed using SAS 9.2 (SAS

164

Institute, 2008) and PASW (2009) statistical software. Results are presented in relation to the
three research questions.
Research Question 1: How does the pain experience differ between older adult HCBWP
participants in regards to diagnosis of cancer over time? How is the relationship between the
pain experience and diagnosis of cancer affected by sex, age, race, comorbid conditions,
depression and cognitive functioning over time?
Generalized estimating equation (GEE) analysis models were used to examine the
longitudinal association between pain and diagnosis of cancer as well as how this association
was affected by sex, age, race, comorbid conditions, depression and cognitive functioning. Since
the SAS and SPAA data analysis programs for GEE uses binary or count dependent variables
(Liu, et al., 2009; Molenberghs & Verbeke, 2005) the dependent variable, pain, was
dichotomized “0”=Pain Scale 0-1 (no pain, less than daily pain) and 1= (daily pain, both not
unusually intense and unusually intense). An autoregressive (AR1) working correlation was
specified as AR1 is a very parsimonious correlation model (two parameters) and accounts for
correlation as a function of time (Fitzmaurice, et al., 2004). Estimate statements were used to
generate odds ratios with 95% confidence intervals.
For this model, the dichotomized pain variable was the dependent variable with diagnosis
of cancer (0=no cancer, 1=initial phase, 2=continuing/terminal) as an independent variable. Age,
sex, race/ethnicity, cognitive functioning, comorbid conditions and behaviors indicative of
depression were added to the model as covariates all at once. In addition to the above covariates,
cancer site and cancer stage were added to the model initially. However, the model could not
carry out successfully due to the small number of subjects with cancer which were further
minimized by cancer site and cancer stage categories. Despite further consolidation of the levels

165

of cancer site and cancer stage, the model continued to not properly estimate the binomial
structure. Therefore, the decision was made to do a sub-analysis with cancer site and cancer
stage if diagnosis of cancer was noted to be significant in any of the study models. Insignificant
variables were then removed one at a time and the model re-ran until the most parsimonious
model was achieved. Time in the HCBWP and diagnosis of cancer were insignificant, indicating
no significant effect from diagnosis of cancer or time in the HCBWP on likelihood of
experiencing daily pain over time. Time in the HCBWP and diagnosis of cancer were left in the
model as they were main variables of interest in this study.
Significant findings are described in the following (Table 27). Females were 1.78 times
more likely to experience daily pain over time than males, after adjusting for other covariates.
African American HCBWP participants were 0.74 times less likely to experience daily pain
overtime than white HCBWP participants, after adjusting for other covariates. Age had a
negative association with pain, such that for each year older a subject was, he or she was 0.99
times less likely to experience daily pain, after adjusting for other covariates. Subjects who were
cognitively impaired were 0.72 times less likely to experience daily pain overtime than those
who were cognitively intact, after adjusting for other covariates. As the measure of comorbid
conditions increased by one unit, a subject was 2.32 time more likely to experience daily pain
over time, after adjusting for other covariates. Finally, the measure of behaviors indicative of
depression was positively associated with pain overtime, such that as behaviors indicative of
depression increased by one, a subject was 1.06 times more likely to experience daily pain, after
adjusting for other covariates.

166

Table 27
Parameter Estimates, Odds Ratios and Significance Values of Parsimonious Model for
Generalized Estimating Equation Fitted to the Dependent Variable of No Pain/Less than Daily
Pain vs. Daily Pain Using a Binomial Distribution (n=4054)
Variable

2

Parameter
Estimate
(95% CI)

Odds Ratio
(95% CI)

Pr>X

Ref.

--------

--------

Initial Phase

0.18
(-0.04 – 0.39)

1.19
(0.96-1.48)

0.11

Continuing/Terminal
Phase

0.06
(-0.17 – 0.28)

1.06
(0.85 – 1.33)

0.61

Sex
Male

Ref.

--------

----------

0.57
(0.43 – 0.72)

1.78
(1.53 – 2.05)

<0.001

Ref.

---------

----------

African American

-0.31
(-0.44 - -0.17)

0.74
(0.64 – 0.84)

<0.001

Other

-0.25
(-0.57 – 0.07)

0.78
(0.57 – 1.08)

0.13

Age

-0.01
(-0.019 - -0.004)

0.99
(0.98 – 1.00)

0.003

Diagnosis of
Cancer
No Cancer

Female

Race
White

167

Table 27 (Continued)

Variable

2

Parameter
Estimate
(95% CI)

Odds Ratio
(95% CI)

Pr>X

Cognitive
Functioning
Cognitively Intact

Ref.

-------

--------

Cognitively
Impaired

-0.33
(-0.42 - -0.24)

0.72
(0.66 – 0.79)

<0.001

Comorbid
Conditions

0.84
(0.59 – 1.09)

2.32
(1.80 – 2.99)

<0.001

Behaviors
Indicative of
Depression

0.06
(0.04-0.08)

1.06
(1.04-1.08)

<0.001

Time in HCBWP

-0.002
(-0.008 -0.004)

1.00
(0.99 -1.00)

0.49

A sub-analysis was then completed to examine what differences existed in the
longitudinal association between daily not unusually intense pain or daily unusually intense pain
and diagnosis of cancer as affected by sex, age, race, cognitive functioning , depression and
comorbid conditions. This sub-sample was limited to subjects who had experienced daily pain at
two or more time points (n=2243) which was then dichotomized into 0=not unusually intense
pain, 1=unusually intense pain. GEE modeling was utilized with an autoregressive (AR1)
working correlation. Insignificant variables were removed one by one and the GEE model reran.
Out of the above significant predictors of daily pain vs. not daily pain/no pain (sex, race,
cognitive functioning, age, comorbid conditions and behaviors indicative of depression) only age
and behaviors indicative of depression were predictors of not unusually intense pain vs.

168

unusually intense pain among subjects with daily pain (Table 28). As age increased by one year,
a subject was 0.97 times less likely to experience unusually intense pain vs. not unusually intense
pain, after adjusting for other covariates. As the measure of behaviors indicative of depression
increased by 1, a subject was 1.04 times more likely to experience unusually intense pain, after
adjusting for other covariates.
Table 28
Parameter Estimates, Odds Ratios and Significance Values of Parsimonious Model for
Generalized Estimating Equation Fitted to the Dependent Variable of Daily Pain, Not Unusually
Intense vs. Daily Pain, Unusually Intense Using a Binomial Distribution (n=2243)
Variable

2

Parameter
Estimate
(95% CI)

Odds Ratio
(95% CI)

Pr>X

No Cancer

Ref.

------

------

Initial Phase

-0.19
(-0.54 – 0.15)

0.94
(0.68 – 1.30)

0.71

Continuing/Terminal

-0.20
(-0.51 – 0.10)

0.87
(0.65 – 1.17)

0.38

Age

-0.03
(-0.04 - -0.02)

0.97
(0.96-0.98)

< 0.001

Behaviors
Indicative of
Depression

0.03
(0.003 – 0.055)

1.04
(1.01– 1.07)

0.01

Time in Waiver
Program

0.001
(-0.005 – 0.007)

1.00
(0.99 – 1.01)

0.31

Diagnosis of
Cancer

169

Conclusion for Results of Research Question 1 Analysis
In conclusion, diagnosis of cancer was not significantly associated with the pain
experience over time among older adult HCBWP participants. The pain experience was not
significantly affected by the amount of time spent in the HCBWP. Females were more likely to
experience daily pain than males. African American older adult HCBWP participants were less
likely to experience daily pain than white older adult HCBWP participants. Cognitive
functioning had a negative association with the pain experience: those with impaired cognitive
functioning were less likely to experience daily pain than those who were cognitively intact. As
the measure of comorbid conditions increased, the likelihood of experiencing daily pain
increased as well over time. As age increased, not only did the likelihood of experiencing daily
pain vs. no pain decrease but the likelihood of experiencing daily unusually intense pain vs. daily
not unusually intense pain decreased as well. Finally, as behaviors indicative of depression
increased, not only did the likelihood of experiencing daily pain vs. no pain increase but the
likelihood of experiencing daily unusually intense pain vs. daily not unusually intense pain
increased as well.
Research Question 2: How does the pain experience of older adult, HCBWP participants
relate to prescribed pain management strategies and pain management outcomes and how does
this relationship differ in regards to diagnosis of cancer over time? How are the relationships
between the pain experience, pain management strategies, pain management outcomes and
diagnosis of cancer affected by sex, age, race, comorbid conditions, depression and cognitive
functioning over time?
Generalized estimating equation (GEE) analysis models were used to examine the
longitudinal association between pain management strategies (prescribed pain medications), pain

170

management outcomes (physical function and pain control) and diagnosis of cancer and how
these associations were affected by pain, sex, age, race, depression, cognitive functioning,
comorbid conditions and time in the HCBWP. As mentioned earlier in this chapter, further
analysis with the pain management strategy of hospice services was not completed due to
extremely low response number of those who received hospice services (Tables 19 and 20).
For the GEE models examining prescribed pain medications and pain control a binomial
distribution was utilized. The nominal variable, prescribed pain medications, was dichotomized
as no prescribed pain medications vs. non-opioid pain medications, opioid pain medications and
adjuvant pain medications as well as non-opioid pain medications vs. opioid pain medications. A
Poisson distribution was used for the GEE model examining physical function, as physical
function was a count variable of the total number of activities of daily living the subject was
dependent in. An autoregressive (AR1) working correlation was specified for all models as AR1
is a very parsimonious correlation model (two parameters) and accounts for correlation as a
function of time (Fitzmaurice, et al., 2004). Estimate statements were used to generate odds
ratios with confidence intervals. Research Question 2 analyses were limited to subjects with
daily pain at two or more time points (n=2243). Of special interest was the examination of
subjects who reported daily pain and had no prescribed pain medication vs. opioid pain
medications prescribed, as approximately half of the subjects with daily pain did not have any
prescribed pain medications across all time points (Table 18).
Dependent variables were developed as follows:
1)

For the pain management strategy of prescribed pain medications,
multiple analyses were performed. The measure of prescribed pain medications
was originally coded as 0=no pain medications, 1=non-opioid pain medication,

171

2=opioid pain medications, 3=adjuvant pain medications. The measure of
prescribed pain medications was developed from the Medicaid paid claim files, a
process which was described in Chapter 4. In order to perform the GEE
modeling, the measure of prescribed pain medications was dichotomized through
the use of dummy variables. “No pain medications” was the reference level and
denoted as “0” and compared to “non-opioid”, “opioid” and “adjuvant” in
separate models. In addition, opioid pain medications were compared to a
reference level of non-opioid.
2)

The pain management outcome of physical function was represented by a
count of the following activities of daily living (ADL) that the older adult
HCBWP participant with daily pain is dependent in: dressing, personal hygiene,
toilet use, bathing and eating. Physical function score ranged from 0 to 5, with a
higher score indicating more dependency with ADLs.

3)

The pain management outcome of pain control was originally
operationalized as the HCBWP participant’s or the HCBWP participant’s or
proxy’s response to the MDS-HC item “pain is controlled by medication” with
possible responses as 0=no pain, 1=medication offered, no control and 2=pain is
partially or fully controlled by medication. For the GEE model, the dependent
variable needed to be dichotomous. Dummy variables were constructed for no
pain vs. medication offered no control and no pain vs. pain is partially or fully
controlled by medication among older adult HCBWP participants with daily pain.

In summary, research question 2 examined differences in the pain management strategies
and pain management outcomes among older adults HCBWP participants in regards to diagnosis

172

of cancer as well as the effects of the pain experience, age, sex, race/ethnicity, cognitive
functioning, behaviors indicative of depression, comorbid conditions on the differences. Results
of the GEE analyses follow.
Research Question 2: Pain Management Strategies
Prescribed Pain Medications. Generalized Estimating Equation (GEE) models were
used to explore for differences in prescribed pain medications among older adult HCBWP
participants with daily pain (n=2243) in regards to diagnosis of cancer and how this association
was affected by age, sex, race/ethnicity, cognitive functioning, behaviors indicative of
depression, comorbid conditions and time in the HCBWP.
Separate models were used for each comparison: no pain medications vs. non-opioid
(Table 29), no pain medications vs. opioid pain medications (Table 30), no pain medications vs.
adjuvant pain medications (Table 31) and non-opioid vs. opioid pain medications (Table 32). All
variables (age, sex, race/ethnicity, cognitive functioning, behaviors indicative of depression,
pain, comorbid conditions and time in the HCBWP) were added to each GEE model.
Insignificant variables were then removed one by one and the model re-ran. Diagnosis of cancer,
time in the HCBWP and pain were left in each model, even if insignificant, as they were key
study variables.
No prescribed pain medications vs. Non-opioid pain medications. Only diagnosis of
cancer and pain were significant predictors of whether non-opioid pain medications were
prescribed vs. if no pain medications were prescribed over time among older adult HCBWP
participants with two or more assessments of daily pain (Table 29). Age, sex, race/ethnicity,
cognitive functioning, behaviors indicative of depression, comorbid conditions and time in the

173

HCBWP were not significantly associated with likelihood of being prescribed non-opioid pain
medications vs. no prescribed pain medications over time.
Table 29
Parameter Estimates, Odds Ratios and Significance Values of Parsimonious Model for
Generalized Estimating Equation Fitted to the Dependent Variable of Non-Opioid Prescribed
Pain Medications vs. No Prescribed Pain Medications Using a Binomial Distribution (n=2243)
Variable

2

Parameter
Estimate
(95% CI)

Odds Ratio
(95% CI)

Pr>X

Ref.

_______

______

Initial Phase

-0.99
(-1.93 – -0.06)

0.37
(0.14 – 0.94)

0.04

Continuing/Terminal

-0.31
(-0.73 – 0.11)

0.73
(0.48 – 1.12)

0.15

Ref

------

-------

0.24
(0.07-0.42)

1.28
(1.07 – 1.52)

0.01

-0.002
(-0.02 – 0.01)

0.99
(0.98 – 1.01)

0.75

Diagnosis of Cancer
No Cancer

Pain
Not Daily unusually intense
Daily unusually intense

Time in Waiver Program

When compared to subjects with no diagnosis of cancer, subjects in the initial phase of
diagnosis of cancer were 0.37 times less likely to have non-opioid pain medications prescribed
vs. no pain medications prescribed, after adjusting for other covariates. When compared to
subjects who did not have daily unusually intense pain, subjects with daily unusually intense

174

pain were 1.28 times more likely to have non-opioid pain medications prescribed than no pain
medications, after adjusting for other covariates.
Because diagnosis of cancer was significantly associated with the prescription of nonopioid pain medications vs. no pain medications, a sub-analysis was performed with cancer site
and cancer stage. Neither cancer site (p=0.07) or cancer stage (p=0.27) were significantly
associated with the prescription of non-opioid pain medications vs. no pain medications among
older adult HCBWP participants.
In summary, pain and diagnosis of cancer were significant predictors of the prescription
of non-opioid pain medications vs. no prescribed pain medications among older adult HCBWP
participants experiencing daily pain over time. Having daily, unusually intense pain increased
the likelihood of an older adult HCBWP participant having non-opioid pain medications
prescribed for them. Older adult HCBWP participants who were in the initial phase of diagnosis
of cancer were less likely than subjects with no diagnosis of cancer of having non-opioid pain
medications prescribed for them. There were no significant associations between cancer site and
cancer stage and the likelihood of being prescribed non-opioid pain medications vs. no
prescribed pain medications among older adult HCBWP participants.
The assessment of predictors of whether an older adult HCBWP participant had nonopioid pain medications prescribed was complicated by the fact that many non-opioid pain
medications are available over the counter and therefore, would not be captured on the Medicaid
paid claim file data. The use of over the counter non-opioid pain medications by older adult
HCBWP participants cannot be known via the data used for this study and therefore this analysis
may have been affected by the under-reporting of non-opioid pain medication use.

175

No prescribed pain medications vs. opioid pain medications. Pain, age, sex,
race/ethnicity, cognitive functioning, comorbid conditions and time in the HCBWP were
significant predictors of the prescription of opioid pain medications vs. no pain medications
(Table 30) among older adult HCBWP participants with two or more assessments of daily pain.
Diagnosis of cancer was not significantly associated with the prescription of opioid pain
medications vs. no prescribed pain medications. Significant results among older adult HCBWP
participants with two or more assessments of daily pain were as follows:
Pain. Pain was positively associated with the prescription of opioid pain medications
when compared with no prescribed pain medications. Subjects with daily unusually intense pain
were 1.72 times more likely to have opioid pain medications prescribed when compared to
subjects without daily unusually intense pain, after adjusting for other covariates.
Age. As subject age increased by one year, subjects was 0.96 times less likely to have
opioid pain medications prescribed vs. no pain medications, after adjusting for other covariates.
Race/Ethnicity. African American HCBWP participants were 0.63 times less likely have
opioid pain medications prescribed vs. no pain medications prescribed over time than white older
adult HCBWP participants, after adjusting for other covariates. There was no significant
difference in the likelihood of older adult HCBWP participants in the “other” race/ethnicity
group having prescribed opioid pain medications vs. no pain medications when compared to
white older adult HCBWP participants.
Sex. When compared to males, females were 1.46 times more likely to have opioid pain
medications prescribed than have no pain medications prescribed, after adjusting for other
covariates.

176

Cognitive Functioning. Cognitive functioning was negatively associated with prescribed
pain medications such that subjects who were cognitively impaired were 0.67 times less likely to
be prescribed opioid pain medications vs. no pain medications over time than subjects who were
cognitively intact over time.
Comorbid Conditions. As the measure of comorbid conditions increased by one unit, a
subject was 1.67 times more likely of being prescribed opioid pain medication vs. no pain
medications over time, after adjusting for other covariates.
Time in the HCBWP.As time in the HCBWP increased by one month, subjects were 1.01
times more likely to be prescribed opioid pain medications vs. no pain medications.
Table 30
Parameter Estimates, Odds Ratios and Significance Values of Parsimonious Model for
Generalized Estimating Equation Fitted to the Dependent Variable of Opioid Prescribed Pain
Medications vs. No Prescribed Pain Medications Using a Binomial Distribution (n=2243)

Variable

2

Parameter
Estimate
(95% CI)

Odds Ratio
(95% CI)

Pr>X

Ref.

--------

--------

-0.12
(-0.50 – 0.25)

0.88
(0.60 – 1.29)

0.52

Initial Phase

-0.08
(-0.40 – 0.23)

0.92
(0.67 – 1.26)

0.61

Continuing/Terminal

Diagnosis of
Cancer
No Cancer

177

Table 30 (Continued)

Variable

2

Parameter
Estimate
(95% CI)

Odds Ratio
(95% CI)

Pr>X

Ref.

------

------

0.54
(0.41 – 0.67)

1.72
(1.51 – 1.96)

<0.001

Ref

------

------

Female

0.38
(0.17 – 0.59)

1.46
(1.19 – 1.81)

<0.001

Age

-0.03
(-0.04 - -0.02)

0.96
(0.95-0.98)

<0.001

White

Reference

--------

-------

African American

-0.45
(-0.64 - -0.27)

0.63
(0.52 – 0.76)

<0.001

Other

-0.29
(-0.73 – 0.15)

0.75
(0.48 – 1.16)

0.20

Cognitively Intact

Ref

------

-------

Cognitively
Impaired

-0.36
(-0.50 - -0.21)

0.70
(0.60 – 0.80)

<0.001

Pain
Not daily, not
unusually intense

Daily, unusually
intense
Sex
Male

Race/Ethnicity

Cognitive
Functioning

178

Table 30 (Continued)

2

Variable

Parameter
Estimate
(95% CI)

Odds Ratio
(95% CI)

Pr>X

Comorbid
Conditions

0.51
(0.24 – 0.79)

1.67
(1.27 – 2.20)

<0.001

Time in the
HCBWP

0.01
(0.003 – 0.024)

1.01
(1.00– 1.02)

0.01

In summary, pain, age, sex, race/ethnicity, cognitive function, comorbid conditions and
time in the HCBWP were all significantly associated with the prescription of opioid pain
medications vs. no pain medications over time among older adult HCBWP participants. When
compared to older adult HCBWP participants not experiencing daily unusually intense pain,
older adult HCBWP participants with daily unusually intense pain were significantly more likely
to have opioid pain medications prescribed for them, vs. no pain medications. As the age of
older adult HCBWP participants increased, the likelihood of being prescribed opioid pain
medications decreased. Female older adult HCBWP participants were more likely than male
older adult HCBWP participants to have opioid pain medications prescribed for them. African
American older adult HCBWP participants were less likely than their white counterparts to have
opioid pain medication prescribed for them, as are cognitively impaired older adult HCBWP
participants when compared to cognitively intact older adult HCBWP participants. Finally, the
likelihood of being prescribed opioid pain medications increased as the measure of comorbid
conditions or time in the HCBWP also increased.
As noted earlier in this section, the examination of subjects who reported daily pain and
had no prescribed pain medication vs. opioid pain medications prescribed was of special interest,
179

as approximately half of the subjects with daily pain did not have any prescribed pain
medications across all time points (Table 18). The results of this analysis suggest that subjects
who reported daily pain and were not prescribed opioid pain medications were more likely to be
experiencing not unusually intense daily pain, be older, be African American, be cognitively
impaired, have less comorbid conditions and had spent less time in the HCBWP than older adult
HCBWP participants who were prescribed opioid pain medications.
No prescribed pain medications vs. adjuvant prescribed pain medications. Adjuvant
prescribed pain medications are “…non-opioid medications that have pain-relieving effects in
certain conditions, but whose primary or initial indication was not for the treatment of pain”
(American Pain Society, 2005, p. 73). Medications that are used as adjuvant pain medications
include anticonvulsants, antidepressants, and local anesthetics (American Pain Society, 2005).
Adjuvant medications diminish pain by altering nerve function (Kalso, 2005).
Diagnosis of Cancer, pain, age, and race/ethnicity were significant predictors of the
prescription of adjuvant pain medications vs. no pain medications (Table 31). Significant results
among older adult HCBWP participants with two or more assessments of daily pain were as
follows:
Diagnosis of Cancer. Subjects in the initial phase of diagnosis of cancer were 0.41 times
less likely than subjects with no diagnosis of cancer to have adjuvant pain medications
prescribed vs. no pain medications prescribed. There was no significant difference in the
likelihood of having adjuvant pain medications prescribed vs. no pain medications prescribed
between subjects in the continuing/terminal diagnosis of cancer and subjects with no diagnosis of
cancer.

180

Because diagnosis of cancer was significantly associated with the prescription of
adjuvant pain medications vs. no pain medications, a sub-analysis was performed with cancer
site and cancer stage. With the addition of cancer site to the model containing the above
significant variables, the model failed to converge and error messages were produced. Cancer
site was removed from the model. Cancer stage, when added separately to the model, did not
significantly associate (p=0.61) with the prescription of adjuvant pain medications vs. no pain
medications among older adult HCBWP participants. Therefore, cancer stage was not associated
with the prescription of adjuvant pain medications vs. no prescribed pain medications overtime,
after adjusting for other covariates.
Pain. Pain was positively associated with the prescription of adjuvant pain medications
when compared with no prescribed pain medications. Subjects with daily unusually intense pain
were 1.41 times more likely to have adjuvant pain medications prescribed than no pain
medications when compared to subjects without daily unusually intense pain, after adjusting for
other covariates. Therefore, daily unusually intense pain was significantly associated with the
prescribing of adjuvant pain medications, such that older adult HCBWP participants with daily
unusually intense pain were more likely to have adjuvant pain medications prescribed to them
compared to older adult HCBWP participants experiencing daily pain that was not unusually
intense.
Age. As subject age increased by one year, subjects were 0.95 times less likely to have
adjuvant pain medications prescribed vs. no pain medications over time, after adjusting for other
covariates.
Race/Ethnicity. In regards to race, African Americans were 0.47 times less likely to be
prescribed adjuvant pain medications vs. no prescribed pain medications over time than white

181

subjects, after adjusting for other covariates. There was no significant difference in the
likelihood of being prescribed adjuvant pain medications vs. no pain medications between white
subjects and subjects in the “other” race/ethnicity category.
Table 31.
Parameter Estimates, Odds Ratios and Significance Values of Parsimonious Model for
Generalized Estimating Equation Fitted to the Dependent Variable of Adjuvant Prescribed Pain
Medications vs. No Prescribed Pain Medications Using a Binomial Distribution (n=2243)

Variable

2

Parameter
Estimate
(95% CI)

Odds Ratio
(95% CI)

Pr>X

No Cancer

Ref.

_______

______

Initial Phase

-0.90
(-1.72 – -0.08)

0.41
(0.18 – 0.92)

0.03

Continuing/Terminal

-0.12
(-0.82 – 0.57)

0.88
(0.45 – 1.78)

0.73

Not daily unusually
intense

Ref.

____

_____

Daily, unusually
intense

0.34
(0.07– 0.61)

1.41
(1.08-1.84)

0.01

Age

-0.05
(-0.08 – - 0.03)

0.95
(0.93 – 0.97)

<0.001

Diagnosis of Cancer

Pain

182

Table 31 (Continued)

Variable

2

Parameter
Estimate
(95% CI)

Odds Ratio
(95% CI)

Pr>X

White

Ref.

_____

____

African American

-0.76
(-1.24 - -0.28)

0.47
(0.29 – 0.75)

0.002

Other

-0.26
(-1.30 – 0.77)

0.77
(0.27 – 2.17)

0.62

Time in Waiver
Program

0.02
(-0.002– 0.035)

1.02
(1.00– 1.03)

0.08

Race

In summary, diagnosis of cancer, pain, age and race/ethnicity were significantly
associated with the prescription of adjuvant pain medications vs. no prescribed pain medications
over time among older adult HCBWP participants. As with the prescription of non-opioid pain
medications, older adult HCBWP participants in the initial phase of diagnosis of cancer were less
likely than older adult HCBWP participants without cancer to have adjuvant pain medications
prescribed for them. Older adult HCBWP participants with daily unusually intense pain were
significantly more likely to have adjuvant pain medications prescribed than older adult HCBWP
participants without daily unusually intense pain. African American older adult HCBWP
participants were less likely than white older adult HCBWP participants to have adjuvant pain
medications prescribed. Finally, as age increased among older adult HCBWP participants, the
likelihood of having adjuvant pain medications prescribed decreased.

183

Opioid pain medications vs. Non-Opioid Pain Medications. Pain, age, sex,
race/ethnicity, cognitive functioning, comorbid conditions and time in the HCBWP were
significant predictors of the prescription of opioid pain medications vs. the prescription of nonopioid pain medications (Table 32) among older adult HCBWP participants with two or more
assessments of daily pain. There was no significant association between diagnosis of cancer and
the likelihood of being prescribed opioid pain medications vs. non-opioid pain medications
among older adult HCBWP participants. Significant results among older adult HCBWP
participants with two or more assessments of daily pain were as follows.
Pain. Pain was positively associated with the prescription of opioid pain medications
when compared with the prescription of non-opioid pain medications. Subjects with daily
unusually intense pain were 1.33 times more likely to have opioid pain medications prescribed
than non-opioid pain medications when compared to subjects without daily unusually intense
pain, after adjusting for other covariates.
Sex. When compared to males, females were 1.36 times more likely to have opioid pain
medications prescribed rather than non-opioid pain medications prescribed, after adjusting for
other covariates.
Age. As subject age increased by one year, subjects were 0.97 times less likely to have
adjuvant pain medications prescribed vs. no pain medications over time, after adjusting for other
covariates.
Race/Ethnicity. In regards to race, African Americans were 0.73 times less likely to be
prescribed opioid pain medications vs. non-opioid pain medications over time than “white”
subjects, after adjusting for other covariates. There was no significant difference in the

184

likelihood of being prescribed opioid pain medications vs. non-opioid pain medications between
“white” subjects and subjects in the “other” race/ethnicity category.
Cognitive functioning. Subjects who were cognitively impaired were 0.70 times less
likely than cognitively intact subjects to be prescribed opioid pain medications vs. non-opioid
pain medications over time, after adjusting for other covariates.
Comorbid conditions. As the measure of comorbid conditions increased by one unit,
subjects were 2.04 times more likely to be prescribed opioid pain medication vs. non-opioid
pain medications over time, after adjusting for other covariates.
Time in the HCBWP.As time in the HCBWP increased by one month, subjects were 1.01
times more likely to be prescribed opioid pain medications than non-opioid pain medications
over time, after adjusting for other covariates.
In summary, pain, age, sex, race/ethnicity, cognitive function, comorbid conditions and
time in the HCBWP were all significantly associated with the prescription of opioid pain
medications vs. non-opioid pain medications over time. When compared to older adult HCBWP
participants not experiencing daily unusually intense pain, older adult HCBWP participants with
daily unusually intense pain were significantly more likely to have opioid pain medications
prescribed for them, vs. non-opioid pain medications. As the age of older adult HCBWP
participants increased, the likelihood of being prescribed opioid pain medications vs. non-opioid
pain medications decreases. Female older adult HCBWP participants are more likely than male
older adult HCBWP participants to have opioid pain medications vs. non-opioid pain
medications prescribed for them. African American older adult HCBWP participants are less
likely than their white counterparts to have opioid pain medication vs. non-opioid pain
medications prescribed for them, as are cognitively impaired older adult HCBWP participants

185

when compared to cognitively intact older adult HCBWP participants. Finally, the increased
likelihood of being prescribed opioid pain medications vs. non-opioid pain medications increased
as the measure of comorbid conditions or time in the HCBWP increased.
The predictors of the likelihood of being prescribed opioid pain medications vs. nonopioid pain medications were the same as the predictors of the likelihood of being prescribed
opioid pain medications vs. no pain medications. As noted earlier in this section, older adult
HCBWP participants may have been taking over the counter non-opioid pain medications that
were not captured in the Medicaid paid claim files and therefore were not included in the
prescription of non-opioid pain medication. The resulting identical predictors for the prescription
of opioid pain medications vs. no prescribed pain medications and opioid pain medications vs.
non-opioid pain medication suggest that there were not many differences between the
prescription of non-opioid pain medications and no prescribed pain medications when comparing
to the prescription of opioid pain medications. Strengthening the argument that use of non-opioid
pain medications may not have been fully accounted for with this dataset.

186

Table 32
Parameter Estimates, Odds Ratios and Significance Values of Parsimonious Model for
Generalized Estimating Equation Fitted to the Dependent Variable of Opioid Prescribed Pain
Medications vs. Non-Opioid Prescribed Pain Medications Using a Binomial Distribution
(n=2243)
Variable

2

Parameter
Estimate
(95% CI)

Odds Ratio
(95% CI)

Pr>X

No Cancer

Ref.

--------

-------

Initial Phase

0.62
(0.04 – 1.20)

1.86
(1.04 – 3.32)

0.04

Continuing/Terminal

0.08
(-0.35 – 0.52)

1.09
(0.70– 1.69)

0.70

Pain

Ref.

-------

--------

0.28
(0.10 – 0.46)

1.33
(1.11 – 1.58)

0.002

Reference

--------

-------

Female

0.31
(0.02 – 0.60)

1.36
(1.02-1.82)

0.03

Age

-0.03
(-0.04 – - 0.02)

0.97
(0.95-0.98)

<0.001

Diagnosis of Cancer

Not Daily unusually
intense
Daily, unusually
intense
Sex
Male

187

Table 32 (Continued)

Variable

2

Parameter
Estimate
(95% CI)

Odds Ratio
(95% CI)

Pr>X

White

Reference

--------

-------

African American

-0.31
(-0.57 - -0.06)

0.73
(0.57 – 0.93)

0.02

Other

-0.42
(-1.00 – 0.16)

0.66
(0.37 – 1.18)

0.16

Cognitively Intact

Ref

------

-------

Cognitively Impaired

-0.35
(-0.54 - -0.16)

0.70
(0.58 – 0.85)

<0.001

Comorbid
Conditions

0.71
(0.30 – 1.13)

2.04
(1.35 – 3.09)

<0.001

Time in Waiver
Program

0.01
(0.002– 0.03)

1.01
(1.00– 1.03)

0.03

Race

Cognitive
Functioning

Summary of Pain Management Strategies: Prescribed Pain Medications. In summary,
results indicated that diagnosis of cancer (initial phase vs. no diagnosis of cancer) was negatively
associated with the prescription of non-opioid and adjuvant pain medication (vs. no pain
medications). These findings should be examined further in future research to determine
differences in the pain experience and personal and health and illness characteristics of older
adult HCBWP participants in the initial phase from those with no diagnosis of cancer or those
further along in their cancer treatment. In the present study, descriptive analysis had not shown a

188

significant association between the measure of diagnosis of cancer and personal and health and
illness factors.
The results of these analyses which found significant associations between the
prescription of opioid pain medications and pain, age, sex, race/ethnicity, cognitive functioning,
comorbid conditions are similar to results of previous research that were presented in Chapter 3
(American Geriatrics Society Panel on Persistent Pain in Older Persons, 2002; Bruckenthal,
2008; Bruckenthal & D'Arcy, 2007; Reynolds, et al., 2008; Won, et al., 2004). The significant
time effect on the prescription of opioid pain medication indicated that the longer the older adult
participated in the HCBWP, the more likely he or she was to be prescribed opioid pain
medications. Whether this significant time effect is specific to the HCBWP or there is an overall
time effect for the likelihood being prescribed opioid pain medications would need to be clarified
in future research. The fact that age and likelihood of being prescribed opioid pain medications
had a negative association over time certainly points to a significant time effect specific to the
HCBWP.
The prescription of non-opioid pain medications vs. no prescribed pain medications was
significantly associated with pain and diagnosis of cancer. As discussed earlier, the analysis of
predictors of the prescription of non-opioid pain medications may have been limited by the use
of over the counter non-opioid pain medications that were not captured in the Medicaid paid
claim files. The consistent use of a write-in medication section in the MDS-HC documenting all
medication would assist in assessing the association between personal, health and illness and
environmental characteristics in future research.
The prescription of adjuvant pain medications vs. no prescribed pain medications was
significantly associated with diagnosis of cancer, pain, age and race/ethnicity. The negative

189

association between age and race/ethnicity may be explained in part by drug-related and health
system-related factors. Adjuvant pain medications are primarily medications which have central
nervous system actions. For example, Elavil is a tricyclic antidepressant while Tegretol has been
used for seizures and both drugs have a sedating effect. Healthcare providers may be reluctant to
prescribe sedating drugs to older adults. Additionally, adjuvant pain medications may be
primarily used by healthcare providers associated with pain clinics, a specialty practice. African
American older adult HCBWP participants may have less access to specialty health services like
pain clinics than their white counterparts.
Research Question 2: Pain Management Outcomes
Generalized Estimating Equation (GEE) models were used to examine the association
between pain management outcomes of physical function and pain control among older adult
HCBWP participants with daily pain and diagnosis of cancer and how this association was
affected by age, sex, race/ethnicity, cognitive functioning, behaviors indicative of depression,
comorbid conditions and time in the HCBWP. Separate models were used for physical function
and pain control. All variables including pain, prescribed pain medication, age, sex,
race/ethnicity, cognitive functioning, behaviors indicative of depression, pain, comorbid
conditions and time in the HCBWP were added to each GEE model. Insignificant variables were
then removed one by one and the model re-ran. Diagnosis of cancer, time in the HCBWP and
pain were left in each model, even if insignificant, as they were key study variables.
Physical Function. A GEE model was used to explore differences in physical function
among older adult HCBWP participants who experienced daily pain in regards to diagnosis of
cancer and how this association was affected by pain, age, sex, race/ethnicity, cognitive
functioning, behaviors indicative of depression, comorbid conditions, prescribed pain

190

medications and time in the HCBWP. To review, the pain management outcome of physical
function is represented by a count of the following activities of daily living (ADL) that the
HCBWP participant is dependent in: dressing, personal hygiene, toilet use, bathing and eating.
Physical function score ranged from 0 to 5, with a higher score indicating more dependency with
ADLs.
Pain, race/ethnicity, cognitive functioning, behaviors indicative of depression and time in
the HCBWP were significantly associated with physical function over time (Table 33) among
older adult HCBWP participants with two or more assessments of daily pain. There was no
significant association between diagnosis of cancer and physical function over time among older
adult HCBWP participants with two or more assessments of daily pain. Significant results among
older adult HCBWP participants with two or more assessments of daily pain were as follows:
Pain. Pain was positively associated with physical function. Subjects with daily
unusually intense pain were 1.05 times more likely to experience physical function dependencies
when compared to subjects without daily unusually intense pain, after adjusting for other
covariates.
Race. In regards to race, African Americans were 1.36 times more likely to experience
physical function dependencies when compared to white subjects, after adjusting for other
covariates. There was no significant difference in the likelihood of experiencing physical
function dependencies between “white” subjects and subjects in the “other” race/ethnicity
category.
Cognitive Functioning. When compared to cognitively intact subjects, subjects with
cognitive impairment were 1.19 times more likely to experience an increase in physical function
dependencies over time, after adjusting for covariates.

191

Behaviors Indicative of Depression. The measure of behaviors indicative of depression
had a positive, significant association with physical function over time, such that subjects were
1.01 times more likely to experience dependencies in physical function as behaviors indicative
of depression increased 1 point, after adjusting for other covariates.
Time in the HCBWP. As time in the HCBWP increased by one month, subjects were
1.01 times more likely to experience physical function dependencies, after adjusting for other
covariates.
In summary, pain, race/ethnicity, cognitive functioning, behaviors indicative of
depression and time in the HCBWP were all significantly associated with the measure of
physical function (the total number of activities of daily living the older adult HCBWP
participant was dependent, or needed assistance with) over time. Older adult HCBWP
participants who experienced daily unusually intense pain were more likely to experience
physical function dependencies than those not experiencing daily unusually intense pain. African
American older adult HCBWP participants were more likely than white older adult HCBWP
participants to experience physical function dependencies. A higher likelihood of having
physical function dependencies was experienced by the cognitively impaired as well as those
with increasing behaviors indicative of depression. There was a significant time effect on
physical function, such that as an older adult spent more time in the HCBWP, they experienced
an increase in physical function dependencies. Diagnosis of cancer was not a significant
predictor of physical function over time among older adult HCBWP participants.

192

Table 33
Parameter Estimates, Odds Ratios and Significance Values of Parsimonious Model for
Generalized Estimating Equation Fitted to the Dependent Variable of Physical Function Using a
Poisson Distribution (n=2243)

Variable

2

Parameter
Estimate
(95% CI)

Odds Ratio
(95% CI)

Pr>X

No Cancer

Ref.

--------

-------

Initial Phase

0.04
(-0.05 – 0.14)

1.04
(0.94 – 1.15)

0.38

Continuing/Terminal

0.11
(-0.004 – 0.22)

1.12
(0.99 – 1.25)

0.06

Not daily, unusually
intense pain

Ref.

------

-------

Daily, unusually
intense

0.05
(0.003-0.094)

1.05
(1.00 – 1.10)

0.04

White

Ref.

-------

------

African American

0.31
(0.24 – 0.38)

1.36
(1.27 – 1.46)

<0.001

Other

0.13
(-0.07 – 0.33)

1.13
(0.92 – 1.38)

0.21

Diagnosis of Cancer

Pain

Race

193

Table 33 (Continued)

Parameter
Estimate
(95% CI)

Odds Ratio
(95% CI)

Pr>X2

Cognitively Intact

Ref

------

--------

Cognitively Impaired

0.17
(0.12 – 0.23)

1.19
(1.13 – 1.2)

<0.001

Behaviors Indicative
of Depression

0.01
(0.005 – 0.024)

1.01
(1.00 – 1.02)

0.003

Time in Waiver
Program

0.008
(0.007 – 0.011)

1.009
(1.007– 1.011)

<0.001

Variable

Cognitive
Functioning

Pain Control. The next GEE model was used to explore for differences in pain control
among older adult HCBWP participants with two or more assessments of daily pain in regards to
diagnosis of cancer and how this association was affected by age, sex, race/ethnicity, cognitive
functioning, behaviors indicative of depression, comorbid conditions, prescribed pain
medications, hospice services and time in the HCBWP. Dummy variables were constructed for
pain control: 0= no pain vs. medication offered, no control and 1= no pain vs. pain is partially or
fully controlled by medication. Initial modeling of 0= no pain vs. medication offered produced
an error statement for the hessian matrix not being positive definite. The decision was then made
to revise the pain control measure again to be coded as 0= medication offered, no control, 1=
pain is partially or fully controlled by medication. The GEE model was re-ran with the revised

194

measure of pain control as the dependent variable. The algorithm converged for this new model
and no error statement was generated.
Statistically insignificant variables were removed one by one and the GEE model re-run.
Results of the final model are presented in Table 34. There was no significant association
between prescribed pain medications and pain control over time. Although statistically
insignificant, pain, time in the HCBWP and diagnosis of cancer were left in the model as they
were variables of interest in the study. Only behaviors indicative of depression had a significant
association with pain control. As behaviors indicative of depression increased by one point,
subjects were 0.92 times less likely to experience partial or full control of pain with medication
over time, after adjusting for other covariates.
Table 34
Parameter Estimates, Odds Ratios and Significance Values of Parsimonious Model for
Generalized Estimating Equation Fitted to the Dependent Variable of Pain Control Using a
Binomial Distribution (n=2243)

Variable

2

Parameter
Estimate
(95% CI)

Odds Ratio
(95% CI)

Pr>X

No Cancer

Ref.

_______

______

Initial Phase

-0.22
(-0.75 – 0.31)

0.80
(0.47 – 1.36)

0.41

Continuing/Terminal

-0.45
(-1.04 – 0.14)

0.64
(0.35-1.15)

0.14

Diagnosis of Cancer

195

Table 34 (Continued)

Variable

Parameter
Estimate
(95% CI)

2

Odds Ratio
(95% CI)

Pr>X

____

_____

Pain

Not daily, unusually
intense

Ref.

Daily, unusually
intense

-0.29
(-0.69 – 0.12)

0.75
(0.50 – 1.12)

0.16

Behaviors Indicative
of Depression

-0.08
(-0.14 - -0.02)

0.92
(0.87 – 0.98)

0.01

Time in Waiver
Program

-0.002
(-0.03 – 0.02)

0.99
(0.97 – 1.02)

0.86

In summary, the measure of pain control was dichotomized as medication offered no
control vs. pain is partially or fully controlled by medication. There was no significant
association between diagnosis of cancer and pain control over time. Surprisingly, there was no
significant association between either pain or prescribed pain medication and pain control over
time as well. The measure of behaviors indicative of depression was significantly associated with
pain control. As the measure of behaviors indicative of depression increased, the likelihood of
full or partial control of pain with medication decreased over time. These results suggest that the
presence of depressive symptoms may cause pain to be refractory to treatment. While evidence
addressing the association between depressive symptoms and pain medication among older
adults is limited, pain-related research has found that pain is associated with depressive
196

symptoms among older adults, with the presence of one increasing the likelihood of occurrence
and worsening the prognosis of the other (Geerlings, et al., 2002; L. Li & Conwell, 2007;
Reyes-Gibby, et al., 2002).
An additional analysis was performed to examine if the presence of behaviors indicative
of depression and prescribed pain medication were associated with pain control (0=medication
offered no control, 1=pain is partially or fully controlled by medication). The measure of
behaviors indicative of depression was the dichotomized Depression Rating Scale, with a score >
3 indicative of depression (Burrows, et al., 2000) (0=not indicative of depression, 1=indicative of
depression). The measure of prescribed pain medication was dichotomized into two variables.
Non-opioid pain medication was categorized as 0= no prescribed pain medication, 1=prescribed
non-opioid pain medication. Opioid pain medication was categorized as 0= no prescribed pain
medication, 1=prescribed opioid pain medication. Non-opioid pain medication was not
significantly associated with pain control (p=0.45) while opioid pain medications were
significantly associated with pain control. When compared to older adult HCBWP participants
with no prescribed pain medications, older adult HCBWP participants with prescribed opioid
pain medications were 0.22 times less likely to experience partial or full controlled by
medication. When the dichotomized measure of behaviors indicative of depression was added to
the opioid pain medication model, prescribed opioid pain medications continued to be a
significant negative predictor of partial or full control of pain by medication (p=0.03). Behaviors
indicative of depression was also a significant negative predictor of partial or full control of pain
by medication (p=0.01). When compared to older adult HCBWP participants with a DRS score
of <3, older adult HCBWP participants with a DRS score of > 3 were 0.37 times less likely to
experience partial or full control of pain by medication.

197

Conclusion of Results for Research Question 2 Analyses
Research question 2 examined the associations between pain, pain management strategies
and pain management outcomes in regards to diagnosis of cancer and how these associations
were influenced by age, sex, race/ethnicity, cognitive functioning, behaviors indicative of
depression and comorbid conditions. Diagnosis of cancer had a limited effect in Research
Question 2 analyses. Older adult HCBWP participants with daily pain and in the initial phase of
diagnosis of cancer were less likely to prescribed non-opioid and adjuvant pain medications
when compared to older adult HCBWP participants without cancer. Diagnosis of cancer was not
associated with the pain management outcomes of physical function or pain control. Pain and
factors within the domains of person and health and illness did have effects on pain management
strategies and pain management outcomes, while accounting for diagnosis of cancer in the
analytic models.
Pain was consistently associated with the pain management strategy of prescribed pain
medications. Older adult HCBWP participants with daily unusually intense pain were
significantly more likely than older adult HCBWP participants without daily unusually intense
pain to be prescribed non-opioid, opioid and adjuvant pain medications vs. no prescribed pain
medications as well as opioid pain medications vs. non-opioid pain medications. Pain was
significantly associated with physical function. Older adults with daily unusually intense pain
were more likely to experience dependencies in activities of daily living over time than older
adult HCBWP participants. Pain was not associated over time with pain control among older
adult HCBWP participants. Prescribed pain medications were not associated with the pain
management outcomes of physical function or pain control over time among older adult HCBWP
participants.

198

The influence of the personal and health and illness factors of age, sex, race/ethnicity,
cognitive functioning, behaviors indicative of depression and comorbid conditions on pain
management strategies and pain management outcomes were somewhat inconsistent. Regarding
age and sex, as age increased, the likelihood of being prescribed opioid pain medications
decreased and females were more likely than males to have opioid pain medications prescribed
vs. non-opioid pain medications. However, age and sex were not significantly associated with the
pain management outcomes of physical function and pain control.
Race/ethnicity had a consistent impact on pain management strategies and pain
management outcomes, with African American older adult HCBWP participants less likely to
have opioid and adjuvant pain medications prescribed as well as more likely to experience
dependencies in activities of daily living than white older adult HCBWP participants. Older adult
HCBWP participants who were cognitively impaired were less likely to be prescribed opioid
pain medications and more likely to experience more physical function dependencies than
cognitively intact older adult HCBWP participants, after adjusting for other covariates. Cognitive
function was not significantly associated with pain control. The measure of behaviors indicative
of depression was not associated with the prescription of pain medications but was associated
with the pain management outcomes of physical functioning and pain control, likely due to
depression’s disabling effect on physical function (L. Li & Conwell, 2009) and its worsening
effect on pain (Geerlings, et al., 2002; L. Li & Conwell, 2007; Reyes-Gibby, et al., 2002).
Finally, as the measure of comorbid conditions increased, the likelihood of an older adult
HCBWP participant being prescribed opioid pain medications vs. non-opioid pain medications
also increased

199

Research Question 3: How do the pain experience, pain management strategies and pain
management outcomes of older adult, HCBWP participants predict the admission and time to
admission of older adult HCBWP participants to a nursing home and how does this relationship
differ in regards to diagnosis of cancer while accounting for sex, age, race, comorbid conditions,
depression and cognitive functioning?
For research question 3, analyses were performed that included subjects who stayed in
the HCBWP and those who were admitted to a nursing home in order to investigate whether
variables defining the pain experience, pain management strategies and pain outcomes related to
admission to a nursing home and time to admission. Subjects who died or left the program for
reasons other than admission to a nursing home were excluded from this analysis.
Nursing Home Admission
Logistic regression was used for the first part of this analysis to determine significant
predictors of the “event” which was admission to a nursing home prior to 12/31/05. Predictors
included: diagnosis of cancer, age, sex, race, cognitive functioning, behaviors indicative of
depression, prescribed pain medications, hospice services, diagnosis of cancer, comorbid
conditions, physical function and pain control. Insignificant predictors were removed one by one
and the logistic regression model re-ran until a final, parsimonius model was achieved.
Significant predictors of admission to a nursing home included pain, age, race/ethnicity,
cognitive functioning and comorbid conditions (Table 35) . Diagnosis of cancer, while
insignificant, was left in the model as it was a variable of interest for the study. Regarding pain,
subjects experiencing daily, unusually intense pain were 0.75 times less likely than subjects with
no pain of being admitted to a nursing home. Age was a positive significant predictor of nursing
home admission: as subject age increased by one year, subjects were 1.07 times more likely to

200

be admitted to a nursing home, after adjusting for other covariates. African American subjects
were 0.62 times less likely to be admitted to a nursing home than white subjects, after adjusting
for other covariates. Cognitively impaired subjects were 1.23 times more likely to be admitted to
a nursing home than cognitively intact subjects, after adjusitng for other covariates. Comorbid
conditions was a significant predictor of nursing home admission: as the measure of comorbid
conditions increased by one unit, subjects were 1.35 times more likely of being admitted to a
nurisng home, after adjusting for other covariates.
Table 35
Coefficient, Standard Error, Odds Ratios and Confidence Interval from Logistic Regression
Model for Admission to Nursing Home

Variable

B

SE

Odds
Ratio

95% CI of
the Odds
Ratio

P-Value

No Cancer

Ref.

---------

-------------

----------

------------

Initial Phase

-0.23

0.24

0.79

0.49- 1.26

0.33

Continuing/Terminal
Phase

0.14

0.18

0.58

0.80-1.65

0.45

No Pain

Ref.

------------

-------------

-----------

------------

Less than daily pain

-0.13

0.10

0.88

0.72 – 1.07

0.19

Daily not unusually
intense

-0.18

0.10

0.84

0.69 – 1.02

0.08

Daily, unusually
intense

- 0.29

0.10

0.75

0.62 – 0.91

0.003

Diagnosis of
Cancer

Pain

201

Table 35 (Continued)

Variable

B

SE

Odds
Ratio

95% CI of
the Odds
Ratio

P-Value

Age

0.07

0.005

1.07

1.06-1.08

<0.001

White

Ref

-------

-------

-------

-------

African American

-0.48

0.09

0.62

0.52-0.73

<0.001

Other

0.14

0.19

1.16

0.79-1.69

0.45

Cognitively Intact

Ref.

------------

------------

-------------

----------

Cognitively
Impaired

0.21

0.07

1.23

1.07-1.42

0.004

Comorbid
Conditions

0.30

0.14

1.35

1.01-1.79

0.04

Race/Ethnicity

Cognitive
Functioning

Time to Nursing Home Admission
Next, Cox proportional hazards models were used to analyze the association between the
above significant predictors of nursing home admission, diagnosis of cancer and time to nursing
home admission. Proportional hazard assumptions were met as log minus log plots of time to
nursing home admission showed parallel lines for pain, race/ethnicity and cognitive functioning
(not shown in Figures). Kaplan-Meier models were used to determine the unadjusted mean
number of days to nursing home admission by categorical covariates.
To review, significant predictors of the logistic regression analysis examining admission
to a nursing home included pain, age, race/ethnicity, cognitive functioning and comorbid

202

conditions. Of the variables that were significant in predicting nursing home admission, only
pain did not contribute to time to nursing home admission as its overall significance increased to
0.091 from 0.034 in the logistic regression model (Table 36). The measure of pain was however,
left in the final Cox model as pain was a variable of interest for this study.
Table 36
Coefficient, Standard Error, Hazard Ratios and Confidence Intervals from Cox Proportional
Hazard Model for Time to Nursing Home Admission

Variable

B

SE

Hazards
Ratio

95% CI of
Hazard
Ratio

P-value

No Cancer

Ref.

---------

-------------

----------

------------

Initial Phase

-0.005

0.20

0.99

0.67- 1.47

0.98

Continuing/Terminal
Phase
Pain

0.51

0.14

1.05

0.80-1.38

0.71

No Pain

Ref.

------------

-------------

-----------

------------

Less than daily pain

-0.07

0.08

0.93

0.80 – 1.08

0.36

Daily not unusually
intense

-0.12

0.08

0.88

0.76– 1.03

0.11

Daily, unusually
intense

-0.19

0.08

0.83

0.71 – 0.96

0.09

Age

0.05

0.004

1.05

1.05-1.06

<0.001

Diagnosis of
Cancer

203

Table 36 (Continued)

Variable

B

SE

Hazards
Ratio

95% CI of
Hazard
Ratio

P-value

White

Ref

-------

-------

-------

-------

African American

-0.44

0.07

0.64

0.56-0.74

<0.001

Other

0.03

0.15

1.03

0.77-1.37

0.45

Cognitively Intact

Ref.

------------

------------

-------------

----------

Cognitively
Impaired

0.31

0.06

1.37

1.22-1.53

<0.001

0.30

0.11

1.35

1.08-1.67

0.008

Race/Ethnicity

Cognitive
Functioning

Comorbid
Conditions

Results showed that as subject age increased, the time to nursing home admission
decreased, or happened sooner. Comorbid conditions were positively associated with nursing
home admission, as the measure of comorbid conditions increased, time to nursing home
admission decreased. African American subjects experienced a decreased hazard of being
admitted to a nursing home when compared to white subjects. African American subjects on
average stayed in the HCBWP approximately 2.5 months longer prior to nursing home admission
than white subjects. Subjects who were cognitively impaired had an increased hazard of being
admitted when compared to cognitively intact subjects. Cognitively impaired subjects had on

204

average 2.3 less months in the HCBWP prior to admission to a nursing home than cognitively
intact subjects.
Conclusion of Research Question 3 Analyses
In conclusion, diagnosis of cancer was not significantly associated with the admission of
older adult HCBWP waiver program participants to a nursing home or time to admission to a
nursing home. Pain, age, race/ethnicity, cognitive functioning and comorbid conditions were
significantly associated with the admission of older adult HCBWP participants to a nursing
home. Older adult HCBWP participants with daily unusually intense pain were less likely than
those without pain to be admitted to a nursing home than those with no pain, although pain did
not have an impact on the time to nursing home admission. As the measure of comorbid
conditions increased among older adult HCBWP participants, hazard of being admitted to a
nursing home also increased. Older adult HCBWP participants who were African American
stayed longer in the HCBWP than white participants and cognitively impaired participants had
less time in the HCBWP before admission to a nursing home when compared to cognitively
intact older adult HCBWP participants.
In the following discussion section, the results of above analyses will be summarized and
discussed in relation to the findings of previous research and the study conceptual model.
Discussion of Study Results
This study was developed to examine the pain experience, pain management strategies,
pain management outcomes and nursing home admission of older adult HCBWP participants in
regards to diagnosis of cancer. Previous research had found that older adults with cancer are
more likely to experience pain when compared to older adults without cancer (Buchanan, et al.,
2005; Reyes-Gibby, Aday, et al., 2007; Rodin, 2008). Older adults who experience pain are at

205

risk for not having their pain assessed and managed appropriately for a variety of patient and
system-related reasons and are therefore likely to experience poor pain management and poor
pain management outcomes (American Geriatrics Society Panel on Persistent Pain in Older
Persons, 2002). Very little was known about the pain experience, pain management strategies
and pain management outcomes of older adult HCBWP participants with and without cancer or
how the pain experience, pain management strategies and pain management outcomes of older
adult HCBWP participants are associated with nursing home admission.
The conceptualization of the longitudinal associations between diagnosis of cancer and
the pain experience, pain management strategies, pain management outcomes and admission to a
nursing home as influenced by age, sex, race, cognitive functioning, behaviors indicative of
depression and comorbid conditions were depicted in the conceptual model developed for this
study (Figure 2). The development of the study conceptual model was guided by the Symptom
Management Theory (Dodd, et al., 2001; Humphreys & et al., 2008) which described interactive
relationships between the pain experience, pain management strategies, pain management
outcomes and the domains of person and health and illness. The following discussion of the
study findings will be presented in relation to the study conceptual model (Figure 2) and to
literature findings presented in Chapter 3.
The Pain Experience
The pain experience is the complex process of perceiving, evaluating, and responding to
a sensory input (Dodd, et al., 2001; Humphreys & et al., 2008; Kandle, et al., 2000; Turk &
Okifuji, 1999). The pain experience was conceptualized as being influenced by factors within the
domain of person (age, sex, race/ethnicity, cognitive functioning and behaviors indicative of
depression) and the domain of health and illness (diagnosis of cancer and comorbid conditions).

206

Research Question 1 focused on differences in the pain experience of older adult HCBWP
participants in regards to diagnosis of cancer and how this association was influenced by age,
sex, race, cognitive functioning, behaviors indicative of depression and comorbid conditions over
time.
In the present study, daily pain was experienced by over half of the subjects at each
assessment, with approximately 30% experiencing daily pain that was unusually intense. The
prevalence of pain has been reported to be between 28% and 72% in community-dwelling older
adults (Landi, et al., 2001; Reyes-Gibby, Aday, et al., 2007; Thomas, et al., 2004), Therefore,
subjects in the present study were in the middle range of pain prevalence reported in previous
research for community dwelling older adults. Time in the HCBWP was not significantly
associated with the pain experience, i.e. the pain experience did not change significantly over
time while adjusting for other covariates. However, the study was limited to a time span of about
15 months on average so changes in the pain experience may not have been detectable over this
amount of time.
Results from the present study found no significant difference in the pain experience of
older adult HCBWP participant in regards to diagnosis of cancer over time, after adjusting for
other covariates. This result was contrary to previous research that found that older adults with
cancer were more likely to experience pain when compared to older adults without cancer
(Buchanan, et al., 2005; Reyes-Gibby, Aday, et al., 2007; Rodin, 2008). The finding of no
significant difference in the pain experience in regards to diagnosis of cancer cannot simply be
attributed to the fact that the present study was longitudinal and previous research primarily
cross-sectional, i.e. that the pain experience is not different over time in regards to diagnosis of
cancer. Descriptive research at each time point also found no significant difference in the pain

207

experience in regards to diagnosis of cancer (Table 15). The ability to detect differences in the
pain experience in regards to diagnosis of cancer may have been compromised by the small
sample size of older adult HCBWP participant who had cancer, with 6% of subjects at each time
point having a diagnosis of cancer. This was less that the 11% of HCBWP participants found to
have a diagnosis of cancer by Fries, James and Aliaga (2004) and the 11% of nursing home
residents found to have a diagnosis of cancer at admission to a nursing home (Buchanan, et al.,
2005; Rodin, 2008).
Although there were no differences in the pain experience of older adult HCBWP
participants in regards to diagnosis of cancer, associations between the pain experience and age,
sex, race/ethnicity, cognitive functioning, behaviors indicative of depression and comorbid
conditions (factors within the domains of person and health and illness) that were found in
previous research were confirmed as well as clarified and expanded by the present study. Being
female increased the likelihood of experiencing daily pain over time. The presence of comorbid
conditions and behaviors indicative of depression were positively associated with the pain
experience: as each of the measures increased, the likelihood of the older adult HCBWP
participant experiencing daily pain increased over time. In comparison, age, being of African
American race and having impaired cognitive functioning were negatively associated with the
pain experience, such that the likelihood of experiencing daily pain decreased. Finally, not only
were age and behaviors indicative of depression significantly associated with whether pain was
daily or not daily, but also if pain was unusually intense vs. not unusually intense.
The results from the present study are consistent with previous findings as well as support
the association between the pain experience and domains of person and health and illness as
depicted in the study conceptual model (Figure 2). This study found that the associations

208

between pain and age, sex, race/ethnicity, cognitive functioning and behaviors indicative of
depression were longitudinal in nature, whereas previous research was primarily cross-sectional.
Additionally, there was no significant time effect on pain experience: i.e. the pain experience
was not associated with time in the HCBWP, indicating that pain did not change significantly
over time. Whether this limited change is due to pain not actually changing over time or to an
insensitive measure of pain in the MDS-HC is unknown. However, pain was a significant
predictor of prescribed pain medication and pain control was highly sensitive to the Pain Scale,
as approximately 95% of those without pain per the measure of pain were also noted as having
no pain per the pain control measure over the four time points. Thus, the measure of pain appears
to be closely related to other indicators of pain in the study. The association between pain and
prescribed pain medications and pain control will be discussed further in the following section. It
is hoped that future versions of the MDS-HC include comprehensive and sensitive pain measures
that would allow for more detailed measures of pain frequency, pain severity and pain
interference.
Pain Management Strategies
The pain management strategies dimension included what is done for and by the older
adult HCBWP participant to manage pain (Dodd, et al., 2001; Humphreys & et al., 2008) and
included prescribed pain medications and hospice services. The pain management strategies
dimension was conceptualized as being influenced by factors within the domain of person (age,
sex, race/ethnicity, cognitive functioning and behaviors indicative of depression), the domain of
health and illness (diagnosis of cancer and comorbid conditions) and the pain experience. Pain
management strategies are based on the assessment of the older adult HCBWP participant’s pain
experience (Bruckenthal & D'Arcy, 2007; Dodd, et al., 2001). The first part of Research

209

Question 2 examined differences in the association between the pain experience and pain
management strategies in regards to diagnosis of cancer and how this association was influenced
by age, sex, race, cognitive functioning, behaviors indicative of depression and comorbid
conditions over time.
Hospice Services. Regarding hospice services, only 1% of subjects at each assessment
received hospice services which equated to 6-8 subjects per assessment. Only 1 subject who was
receiving hospice services had a diagnosis of cancer, the others had terminal conditions other
than cancer. There was no significant association between diagnosis of cancer and hospice
services among older adult HCBWP participants across time points. The very small sample of
older adult HCBWP participants receiving hospice services was a limitation to further examining
hospice services in regards to diagnosis of cancer and adjusting for covariates via GEE
modeling.
Researchers have found limited use of hospice services in long term care facilities by
those with terminal diagnoses (Buchanan, et al., 2005; Duncan, et al., 2009). Barriers to the use
of hospice services in long term care facilities include conflict between the hospice and nursing
home care models and shifting hospice demographics in the context of eligibility and payment
piclies (Buchanan, et al., 2005; Duncan, et al., 2008; Stevenson & Bramson, 2009). As the U.S.
population continues to age, end of life and hospice services will become an increasingly
important issue to address (Ersek & Wilson, 2003). If both Medicare and Medicaid paid claim
files could be used for future research then billing for hospice services could be assessed for and
may be preferable to the hospice services measure in the MDS-HC, leading to a more sensitive
measure of hospice services. The outcomes of hospice services, as a pain management strategy,

210

could therefore be assessed and assist in the development of policies regarding the inclusion of
hospice services in HCBWP.
Prescribed Pain Medications. This study found that older adult HCBWP participants
with two or more assessments of daily pain and in the initial phase of diagnosis of cancer were
less likely to be prescribed non-opioid and adjuvant pain medications vs. no prescribed pain
medications when compared to older adult HCBWP participants with two or more assessments
of daily pain and without a diagnosis of cancer. Why diagnosis of cancer had a negative
association with prescribed non-opioid and adjuvant pain medications is not understood. Results
of the initial descriptive research showed that diagnosis of cancer was not significantly
associated with other patient-related variables. Therefore, there were not any apparent
differences between those with and without cancer.
Based on previous research, it was hypothesized that older adults with cancer would have
more pain management strategies than older adults without a diagnosis of cancer, as older adults
with cancer are more likely to experience pain when compared to older adults without a history
of cancer in both community and nursing home settings (Buchanan, et al., 2005; Reyes-Gibby,
Aday, et al., 2007; Rodin, 2008).

However, the results of this present study have found not

only no significance differences in the pain experience of older adults in regards to diagnosis of
cancer but also a negative association between diagnosis of cancer and the prescription of nonopioid and adjuvant pain medications. Concerns about the ability of the measure of prescribed
pain medication to capture the prescription of non-opioid pain medications have been presented
earlier in this chapter. Because many non-opioid pain medications are available over the
counter, Medicaid would not have been billed for them and therefore not recorded. Future
research examining the pain medication use of older adult HCBWP participants would benefit

211

from the consistent use of a current medication log in the MDS-HC to document all medications,
including over the counter, the participant is on.
Pain was the most consistent predictor of the prescription of the different pain
medications over time among older adult HCBWP participants with daily pain. Subjects with
daily, unusually intense pain were significantly more likely to be prescribed non-opioid, opioid
and adjuvant pain medications over time than subjects without daily unusually intense pain.
These results are consistent with the conceptualization (Figure 2) that the pain management
strategy of prescribed pain medication is based on the patient report (verbal or behavioral) of his
or her pain experience. Therefore, the report of pain would warrant the management of pain, with
the increased severity of pain (unusually intense vs. not unusually intense) increasing the
likelihood of pain medication being prescribed. In addition to being influenced by the pain
experience, the measure of prescribed pain medication was associated over time with age, sex,
race/ethnicity, cognitive functioning, behaviors indicative of depression, comorbid conditions
and time in the HCBWP among older adult HCBWP participants with daily pain.
Age. Older adults are at increased risk of not receiving adequate prescribed pain
medications when compared to younger adults (Landi, et al., 2001; Shega, et al., 2006; Won, et
al., 2004; Zyczkowska, et al., 2007). The findings from the present study confirm findings of
previous research, as the age of older adult HCBWP participants increased, the likelihood of
being prescribed pain medications diminished over time.
Sex. Regarding sex, the present study found that female older adult HCBWP participants
with daily pain were more likely than males to have opioid pain medications prescribed. There
were no differences between the sexes in the prescription of non-opioid or adjuvant pain
medications. The results of previous research regarding the association of sex and the

212

prescription of pain medications have been inconsistent. While one study found lower opioid use
by older adult male nursing home residents (Won, et al., 2004), another found no difference in
pain medication use between sexes among community dwelling older adults (Soldato, et al.,
2007). The results of the present research describe a significant association between sex and
prescribed pain medications over time in older adult HCBWP participants.
Race/Ethnicity. Overall, in previous research white adults have been significantly more
likely to receive appropriate prescribed pain medications than adults from other racial groups and
Hispanics with similar pain levels (Cintron & Morrison, 2006; J. A. Cleeland, et al., 2005;
Green, et al., 2003; Rodin, 2008). Specific to older adults residing in nursing homes, Won et al.
(2004) found a lower use of opioid pain medications by black nursing home residents when
compared to white nursing home residents.
In the present study, race/ethnicity was associated with the prescription of opioid and
adjuvant pain medications. African American older adult HCBWP participants with daily pain
were less likely to be prescribed opioid and adjuvant pain medications over time when compared
to white older adult HCBWP participants in daily pain. One possible explanation for this finding
may have to do with the fact that adjuvant pain medications are medications that have a pain
relieving effect, but whose primary or initial indication was not for the treatment of pain
(American Pain Society, 2005). Therefore, primary care doctors may not be as familiar with the
use of adjuvant pain medications for pain management as pain specialists are. If there is
differential access to pain specialists among race then this may explain why black older adult
HCBWP participants were less likely to have adjuvant pain medications prescribed.
Cognitive Functioning. Cognitive function has been associated with the prescription of
pain medication for older adults in previous research. Reynolds, Hanson, Devellis Henderson

213

and Steinhauser (2008) found that among nursing home residents, as the degree of cognitive
impairment increased, the less likely the nursing home resident was to receive treatment for pain.
Eighty-percent of cognitively intact nursing home residents received pain medications, while
only 56% of those with cognitive impairment received pain medications (Reynolds, et al.). The
present study found that cognitively impaired older adult HCBWP were less likely to receive
opioid pain medications than cognitively intact subjects. There was no significant association
between cognitive functioning and the prescription of non-opioid pain medication or adjuvant
pain medications vs. no pain medications.
Pain management begins with the assessment of the patient’s pain experience (Dodd, et
al., 2001). Pain may be challenging for healthcare providers to accurately assess in older adults
with cognitive changes or communication difficulties if pain-related behaviors must be assessed
in place of verbal reports of pain (Delgado-Guay & Bruera, 2008; Goldstein & Morrison, 2005).
Cognitive impairment may result in underreported pain and a poor pain assessment. Pain
management strategies are conceptualized as being based on the patient’s report of their pain
experience. If the pain assessment is poor, the resulting pain management strategies may
therefore be poor as well. This effect was seen in the present study, as older adult HCBWP
participants with cognitive impairment were less likely to experience pain, as documented, and
less likely to receive opioid pain medications.
Comorbid Conditions. The presence of comorbid conditions influences the provision of
pain management among older adults (Duncan, et al., 2008; Goldstein & Morrison, 2005;
McNeill, et al., 2007). The presence of comorbid conditions in older adults may act as a barrier
to pain management, as clinicians may be more cautious in prescribing pain management due to

214

side effects of medications and interactions with medications prescribed for comorbid conditions
(Duncan, et al., 2008; Goldstein & Morrison, 2005; McNeill, et al., 2007).
However, results from the present study suggest that as the measure of comorbid
conditions increases, the likelihood of the prescription of opioid pain medications actually
increased. Pain occurs in older adults because of the presence of comorbid conditions that are
commonly associated with increased age and pain such as arthritis, diabetes and peripheral
vascular disease (Bruckenthal & D'Arcy, 2007; Davis & Srivastava, 2003; Freedman, 2002). As
the measure of comorbid conditions was a weighted index of conditions that were associated
with pain, these results suggest that the more pain-associated comorbid conditions a subject had,
the more likely he or she was to have opioid pain medications prescribed.
Time in the HCBWP. As time in the HCBWP increased the likelihood of being
prescribed opioid or vs. no pain medications and opioid pain medications vs. non-opioid pain
medications also increased. There was no significant time effect for the prescription of nonopioid pain and adjuvant pain medications. As stated earlier in this chapter, the measure of
prescribed pain medications may not have captured the use of non-opioid pain medications, as
many non-opioid pain medications are available over the counter and therefore, would not have
been billed to Medicaid. If not all non-opioid pain medications were accounted for then a timerelated effect may have been insignificant. Overall, the study results suggest that the likelihood
of pain medications stronger than non-opioid pain medications being prescribed may be due,
albeit in a small way, to the amount of time spent in the HCBWP. As the older adult HCBWP
participant spends time in the HCBWP he or she may have more access to health care services
than before they entered the HCBWP and therefore more likely to be prescribed opioid pain
medications.

215

The study results support findings from previous research that pain, age, sex,
race/ethnicity, cognitive functioning and comorbid conditions were all significantly associated
with the prescription of pain medications overtime among older adult HCBWP participants with
two or more assessments of daily pain. The results also are in agreement with the study
conceptual model, as factors within the pain experience and domains of person and health and
illness impact prescribed pain medications. However, diagnosis of cancer, as a health and illness
factor, was not significantly associated with prescribed pain medications over time among older
adult HCBWP participants. The small sample of hospice services was a limitation. Future
access to both Medicaid and Medicare paid claim files for this dual eligible population would
likely lead to a larger sample of HCBWP participants who have received hospice services.
Future research could then better examine the association of person and health and illness factors
and pain with hospice services.
These study results serve to clarify the associations between prescribed pain medications
and pain, age, sex, race/ethnicity, cognitive functioning, behaviors indicative of depression,
comorbid conditions and time in the HCBWP. The present research also adds to the science as
previous research was primarily cross-sectional and this study was longitudinal. Results from
this study note that the significant associations between prescribed pain medications, age, sex,
race/ethnicity, cognitive functioning, behaviors indicative of depression, pain and comorbid
conditions continue over-time in older adult HCBWP participants and are not just limited to a
single observation.
In the following section, discussion regarding the second part of research question 2 is
presented. The second part of Research Question 2 examined differences in the association
between pain management outcomes, the pain experience and pain management strategies in

216

regards to diagnosis of cancer and how this association was influenced by age, sex, race,
cognitive functioning, behaviors indicative of depression and comorbid conditions over time.
The following summarizes study findings in relation to the conceptual model and the literature
review in Chapter 3, beginning with the outcome of pain control.
Pain Management Outcomes
Outcomes are the end results of care (Patrick, 1997). The pain management outcomes
dimension, as depicted in Figure 2, includes the end results of pain management strategies,
which are influenced by pain management strategies and the pain experience (Dodd, et al., 2001;
Humphreys & et al., 2008). For this study, pain management outcomes included pain control and
physical function that were examined over time while the older adult participated in the
HCBWP. The pain management outcomes dimension was conceptualized as being influenced by
factors within the domain of person (age, sex, race/ethnicity, cognitive functioning and behaviors
indicative of depression), the domain of health and illness (diagnosis of cancer and comorbid
conditions) and the dimensions of the pain experience and pain management.
Pain Control. Pain control was conceptually defined as an outcome of pain management
by which pain is limited or decreased (Allard, et al., 2001; Christine Miaskowski, et al., 2002;
Oliver, et al., 2001; Shvartzman, et al., 2003). Initial descriptive analysis of the measure of pain
control indicated that pain control was highly sensitive to the measure of pain, as approximately
94% of those without pain per the measure of pain at each time point were also noted as having
no pain per the pain control measure. Of those who had medication offered with no control, on
average 45% had daily, unusually intense pain over the four time points (Table 21). Of those
subjects who received opioid pain medications at each time point, approximately 93% had partial
or full control of pain, indicating a good outcome from the prescription of opioid pain

217

medications (Table 22). Of subjects who reported no pain at each time point, on average only
13% had evidence of prescribed pain medications in the 30 days prior to each time point, leading
to the question if this sub-group’s lack of pain is due to not having pain instead of pain being
managed well by prescribed pain medication that was not captured in the 30 days before each
time period.
In the GEE model with pain control as the dependent variable, only behaviors indicative
of depression had a significant association with pain control, after controlling for other covariates
As behaviors indicative of depression increased, the likelihood of a subject experiencing partial
or full control of pain with medication decreased after adjusting for other covariates. While
evidence addressing the association between depression and pain control among older adults is
limited, pain-related research has found that pain is associated with depression among older
adults, with the presence of one increasing the likelihood of occurrence and worsening the
prognosis of the other (Geerlings, et al., 2002; L. Li & Conwell, 2007; Reyes-Gibby, et al.,
2002). Therefore, the presence of depression may cause pain to be refractory to treatment,
decreasing the likelihood of pain control. There was no significant association between
prescribed pain medications and pain control over time after adjusting for other covariates.
As noted earlier in Chapter 5, there were issues with the measure of pain control. Pain
control had to be dichotomized to use it as a dependent variable in the GEE models, but the
variable had to be re-dichotomized as the first dichotomization caused the hessian matrix to not
be positive definite. The revised measure of pain control may not have been sensitive to
differences across the covariates. Future research could evaluate the pattern of change in the
measure of pain overtime in response to the presence of prescribed pain medications. Pain
control is a very important factor in the treatment of pain. If pain management strategies do not

218

result in pain control, then pain management strategies must be altered and the pain experience
reevaluated to determine if pain control has occurred (NCCN, 2006).
Physical Function. The pain management outcome of physical function was represented
by a count of the following activities of daily living (ADL) that the HCBWP participant is
dependent in: dressing, personal hygiene, toilet use, bathing and eating. Physical function was
conceptualized as an outcome of pain management as previous research has shown that pain has
a significant, negative effect on physical function and activities of daily living (Onder, et al.,
2006; Reyes-Gibby, et al., 2002; Soldato, et al., 2007). Among older adults either residing in a
nursing home or in the community, pain has been positively associated with walking and
mobility problems, functional limitations and requiring assistance with activities of daily living.
(Jakobsson, et al., 2003). While previous research has noted that older adults who had had
cancer were more likely to have mobility and ADL deficiencies than older adults without cancer
(Keating, et al., 2005) the present study found that there was no significant association between
diagnosis of cancer and physical function over time. One possible explanation for this finding
could be that the older adult HCBWP population is a very frail population as they must meet
nursing home care criteria in order to enter the HCBWP. A diagnosis of cancer may therefore not
significantly add to the health-related burden they already experience. Pain, race, cognitive
functioning, behaviors indicative of depression and time in the HCBWP were found to be
significant predictors of physical function over time among older adult HCBWP participants
with daily pain.
Pain. Among older adults either residing in a nursing home or in the community, pain
has been positively associated with walking and mobility problems, functional limitations and
requiring assistance with activities of daily living. (Jakobsson, et al., 2003; Soldato, et al., 2007).

219

The present study found that the impact on physical function was limited to daily pain, when
compared to those with less than daily pain or no pain. Subjects with daily pain-unusually
intense were more likely than subjects with not daily unusually intense pain to experience an
increase in physical function dependencies over time, after adjusting for other covariates.
Race/Ethnicity. Cleeland, Palmer and Venzke (2005) reported that black adults
experience higher pain-related disability compared to white adults. Results from the present
study concur with Cleeland, Palmer and Venzke’s findings, as black HCBWP program
participants were more likely to experience an increase in physical function dependencies over
time when compared to white subjects, after adjusting for other covariates.
Differences in physical functioning between white and black older adult HCBWP are not
fully understood. In regards to the pain experience, black older adult HCBWP participants were
less likely to report daily pain than white older adult HCBWP participants. Regarding the
prescription of pain medications, black older adult HCBWP participants differed from white
older adult HCBWP participants in the prescription of opioid and adjuvant pain medications
being less likely. Therefore, there was less pain documented in African American older adult
HCBWP participants, but also less prescribing of pain medications among African American
older adult HCBWP participants when compared to white older adult HCBWP participants.
Cognitive Functioning. The significant impact of cognitive functioning on physical
functioning has been presented in previous research. Older adults who are cognitively impaired
are significantly more likely to be physically disabled when compared to cognitively intact older
adults (L. Li & Conwell, 2009; McGuire, Ford, & Ajani, 2006). In the present research, older
adult HCBWP participants with daily pain and cognitive impairment were more likely than
cognitively intact subjects to experience an increase in physical function dependencies over time,

220

after adjusting for covariates. Cognitive impairment has a negative impact on the ability and
independence of the older adult HCBWP participant to complete ADLs.
Behaviors indicative of depression. Among older adults in the community, depression
has a worsening effect on physical functioning (Callahan et al., 2005; L. Li & Conwell, 2009).
In the present study, the measure of behaviors indicative of depression had a positive, significant
association with physical function over time, such that as behaviors indicative of depression
increased, the likelihood of experiencing an increase in the number of physical function
dependencies also increased.
Time in the HCBWP. Finally, as time in the HCBWP increased by one month, the
likelihood of experiencing an increase in the number of physical function dependencies increased
over time. The significant time factor may signify that over Time 1 to Time 4, dependencies in
ADLs continue to worsen over time after adjusting for covariates and in spite of care services
received through the HCBWP
In summary, within the study conceptual model (Figure 2), the pain management
outcomes of pain control and physical function were conceptualized at being impacted by the
pain experience, pain management strategies and the domains of person and health and illness.
Pain control is the result of pain assessment and analgesic treatments (Allard, et al., 2001;
Shvartzman, et al., 2003). In the present study, pain control was not associated with either the
pain experience or pain management strategies over time. Pain control was associated with
behaviors indicative of depression over time, while controlling for other covariates. This finding
is easily understandable in light of the knowledge that pain is associated with depression among
older adults, with the presence of one increasing the likelihood of occurrence and worsening the

221

prognosis of the other (Geerlings, et al., 2002; L. Li & Conwell, 2007; Reyes-Gibby, et al.,
2002).
In regards to physical functioning, race, cognitive functioning, behaviors indicative of
depression, pain and time in the HCBWP were all found to be significant predictors of physical
function over time among older adult HCBWP participants. As with the measure of pain control,
the measure of prescribed pain medications was not associated with physical function over time
among older adult HCBWP participants. These study results serve to clarify the longitudinal
associations between pain management outcomes and factors within the dimensions of the pain
experience, pain management strategies and the domains of person and health and illness.
Admission to a Nursing Home & Time to Nursing Home Admission
Admission to a Nursing Home, as conceptualized in Figure 2, is conceptually defined as
the movement of older adult, HCBWP participants to a nursing home facility, without returning
to the HCBWP. Admission to a Nursing Home is part of the healthcare services continuum of
older adults, where older adults move from community to institutionalization (L. Li & Zullo,
2003; Williams, 2001). A goal of the HCBWP is to prevent or delay admission to a nursing
home (Fries, et al., 2002).
Time to Nursing Home Admission was conceptualized as the amount of time in months
that a HCBWP participant spends in the HCBWP prior to admission to a nursing home, without
returning to the HCBWP. Previous research has indicated that the strongest predictors of nursing
home placement among a general population of community-dwelling older adults were black
race or Hispanic ethnicity, ADL/physical deficiencies, behavioral issues, cognitive impairment,
prior nursing home use, urinary incontinence, depression and impaired peak flow (Gaugler, et al.,
2000; Gaugler, et al., 2009; McCallum, et al., 2005; Yaffe, et al., 2002). In the study conceptual

222

model (Figure 2), admission to a nursing home and time to admission to a nursing home were
conceptualized as being associated with the dimensions of pain experience, pain management
strategies and pain management outcomes as well as the domains of person and health and
illness. In the present study, diagnosis of cancer was not a significant predictor of the admission
of older adult HCBWP participants to a nursing home or time to nursing home admission. In the
sample, 6% of older adult HCBWP participants had a diagnosis of cancer. This compares to 11%
of nursing home residents who have a diagnosis of cancer at admission (Buchanan, et al., 2005;
Rodin, 2008).
Significant predictors of admission to a nursing home included pain, age, race/ethnicity,
cognitive functioning and comorbid conditions. Surprisingly, physical function was not a
predictor of nursing home admission among older adult HCBWP participants. As HCBWPs
supply assistance with ADLs, physical function needs may be met by HCBWPs so physical
function deficits are no longer a driving force for admission to a nursing home. Of the variables
that were significant in predicting nursing home admission, only pain did not contribute to time
to nursing home admission.
Pain. Regarding pain, subjects experiencing daily, unusually intense pain had a lower
likelihood than that of subjects with no pain of being admitted to a nursing home. Pain was not
significantly associated with time to nursing home admission
Age . Age was a positive significant predictor of nursing home admission. As age
increased by one year the likelihood of being admitted to a nursing home also increased. For
each year increase in subject age, the hazard of experiencing a nursing home admission increased
after adjusting for other covariates, meaning less time in the HCBWP before being admitted to a
nursing home. Luppa, Luck, Matschinger, Konig and Reidel-Heller (2010) also found increasing

223

age to decrease the amount of time to nursing home admission (or increasing the hazard of
admission) for community-dwelling older adults. These results are not surprising given that the
average age of nursing home admission for those 65 and older is 82.6 (Sahyoun, Pratt, Lentzner,
Dey, & Robinson, 2001).
Race/Ethnicity. Nursing home residents are 90% white (Agency for Healthcare
Research and Quality, 2000) while the sample of the present study for older adult HCBWP
participants was 74% white. African American older adult HCBWP participants were less likely
to be admitted to a nursing home than white subjects, after adjusting for other covariates.
African American subjects experienced a decreased hazard of being admitted to a nursing home
when compared to white subjects. African American subjects on average stayed in the HCBWP
approximately 2.5 months longer before admission to a nursing home than white subjects. The
present study findings are different from findings by Yaffe and colleagues (2002), who found
patient characteristics that were predictive of nursing home placement included black race or
Hispanic ethnicity, living alone, one or more ADL dependencies, high cognitive impairment and
one or more difficult behaviors.
Cognitive Functioning. In the literature the strongest predictors of nursing home
placement among a general population of community-dwelling older adults were ADL/physical
deficiencies, cognitive impairment, prior nursing home use, urinary incontinence, depression and
impaired peak flow (Gaugler, et al., 2007; McCallum, et al., 2005). The odds of being admitted
to a nursing home were higher for the cognitively impaired then cognitively intact subjects, after
adjusitng for other covariates. Subjects who were cognitively impaired had an increased hazard
of being admitted when compared to cognitively intact subjects. Cognitively impaired subjects
had on average 2.3 less months in the HCBWP prior to admission to a nursing home than

224

cognitively intact subjects. As cognitive functioning has an impact on physical functioning and
behaviors, it is logical that cognitive functioning would increase the likelihood of admission to a
nursing home as well as decrease the time the admission to a nursing home. HCBWP care
services including respite care for caregivers as well as education for caregivers in managing
patient behaviors may assist in delaying time to nursing home admission from a HCBWP.
Comorbid Conditions. The measure of comorbid conditions was a significant predictor
of nursing home admission. As the measure of comorbid conditions increased by one unit, the
odds of nursing home admission increased as well. Comorbid conditions were positively
associated with time to nursing home admission. As the measure of comorbid conditions
increased the time to nursing home admission was shortened. For the present study, the comorbid
conditions included comorbid conditions significantly associated with pain.
In summary, this study sought to determine if there were differences in the admission of
older adult HCBWP participants in regards to diagnosis of cancer and if the relationship is
influenced by sex, age, race, comorbid conditions, depression and cognitive functioning. The
study conceptual model presented the admission of older adult HCBWP participants as being
influenced by the pain experience, pain management strategies and pain management outcomes
over time, as influenced by personal an health and illness factors. Diagnosis of cancer was not
significantly associated with admission to a nursing home or time to admission to a nursing
home. As noted earlier in this chapter, HCBWP participants must meet nursing home-level needs
criteria and are therefore already compromised and frail. Cancer may not have added
significantly more burden to what the participant was already experiencing.
Analyses were centered on the role of pain, pain management strategies and pain
management outcomes on admission to a nursing home. Pain was significantly associated with

225

admission to a nursing home, but was not significantly associated with time to nursing home
admission. Prescribed pain medications and the pain management outcomes of physical function
and pain control were not significantly associated with admission of older adult HCBWP
participants to a nursing home. Age, race/ethnicity, cognitive functioning and comorbid
conditions were significantly asssociated with both admission to a nursing home and time to
admission to a nursing home. The present study results point to a diminished role for pain, pain
management strategies and pain management outcomes as predictors of nursing home admission
and an emphasis instead on the role of personal and health and illness factors on the admission of
older adult HCBWP participants to a nursing home.
Study Limitations
The study had limitations which are presented and discussed in the following. First, the
study was a secondary analysis of pre-existing data. Therefore, the analysis was limited to what
data was already collected and by what instruments the data was collected with. The pain items
in the MDS Version 1 data that were used for this study were quite limited, asking “Frequently
complains or shows evidence of pain in the past 7 days” with response as no pain, pain less than
daily or pain daily and “pain is unusually intense” with response as no or yes. Because of the
limited possible responses, the items may have not been sensitive to change over time or
specific. However, the MDS-HC pain items did assess multiple dimensions of pain and for the
present study frequency and intensity were examined. Measurement of the multiple dimensions
of pain is preferred to a measure of only pain intensity as pain is multidimensional and intensity,
the sensory aspect of pain, is only one dimension. Williamson and Hoggart (2005) state that “…
the reliance on pain intensity alone suggests that it is the only dimension of pain that is important
to assess and record, although this is not the case” (p. 799) supporting the need to utilize pain

226

measures that address more than pain intensity alone. More recent versions of the MDS-HC
have more detail in the possible responses to pain items, allowing for more sensitive
measurement of pain in future research.
The present study limited the analyses to MDS-HC assessments 2-5, for a total of four
assessments. These assessments were chosen as they would allow for approximately 12 -15
months of data. Analyses were limited to four assessments as the number of assessments among
the subjects varied greatly. Because these assessments were at the beginning of the older adults’
time in the HCBWP the study may have failed to capture changes in key study variables that
may take place after a longer period of time in the HCBWP. Future research should utilize more
time points, or perhaps examine change after a major health-related event such as diagnosis of
cancer to determine if changes in pain, pain management or pain management strategies occur
after such a health related change.
Pain management strategies were limited to what was in the Medicaid paid claim files.
While this limited information regarding the use of over the counter pain medications, there also
was a lack of information regarding the use of non-pharmacological pain management strategies
such as ice, heat, exercise, meditation or massage, for example. Non-pharmacological therapies
have an important role in pain management (JCAHO, 2000). The accounting for the use of nonpharmacological pain management strategies may assist researchers in explaining low or no pain
levels in those without prescribed pain medications among older adult HCBWP participants.
There was the small number of non-white older adult HCBWP participants represented in
the sample. As the study is a secondary analysis of existing data, the inclusion of minority groups
in the study were as they were already represented in the existing data set. In the present study,
74% of subjects were white and 22% were African American. The remaining minority groups

227

were represented by very small numbers and were consolidated into one group for analysis.
Therefore, generalizability of study results to older adult HCBWP participants may be limited.
The measure of physical functioning was limited to activities of daily living, including
Physical function is represented by a count of the following activities of daily living including
dressing, personal hygiene, toilet use, bathing and eating. The measure of physical functioning
did not measure Instrumental Activities of Daily Living (IADL) and therefore, changes in the
ability of the older adult HCBWP participant were not examined.
The researcher was not able to detect if a comorbid condition was the index disease or if
cancer was. Comorbid conditions were conceptualized as the “…the co-occurrence of health
conditions or diseases in reference to an index disease” (Yancik, et al., 2007, p. 276) and for the
purposes of this research the researcher conceptualized cancer as the index disease. The
researcher may therefore have underestimated the effect of comorbid conditions on the pain
experience, pain management strategies, pain management outcomes and admission of older
adult HCBWP participants to nursing homes.
Interactions between variables were not examined with the present research, as they were
outside the scope and time constraints of the present research. The research acknowledges the
importance of interactions between variables in explaining phenomena in nursing practice. Of
particular interest is to examine the interaction between age and cognitive functioning in relation
to the pain experience, pain management strategies, and pain management outcomes.
In Chapter 4, results from a pre-study power analysis were presented. For an actual power
of .90, the required total sample size was 6189 subjects: 928 in the group with a diagnosis of
cancer and 5261 in the group without a diagnosis of cancer. Effect size would be 0.11,
indicating a small effect (Cohen, 1988). For the present study, the final total sample was 4054

228

individuals at each time point, with on average 243 individuals with cancer at each time point
included. The sample size and ratio of those with a diagnosis of cancer to those without cancer in
the present study were smaller than the sample size needed to detect differences predicted by the
pre-study power analysis. Therefore, the ability to detect differences in the pain experience in
regards to diagnosis of cancer may have been hampered by the smaller sample size. However,
the purpose of sampling is to be a representation of the population of interest so that the results
of the research (with the sample) can be then be applied to the population as a whole (Burns &
Grove, 2005; Israel, 2009). The present study examined data from all older adult HCBWP
participants from 1/1/2003 to 12/31/2005. Because the sample included all older adult HCBWP
participants the study sample can instead be considered a population, as no additional subjects
could be added to the sample. Therefore, the issue of sample size may not be appropriate for this
study.
Finally, the percentage of older adult HCBWP participant with cancer was approximately
6% over the four assessment time point. This was less than the 11% of older adult HCBWP
participants found to be diagnosed with cancer by Fries, James and Aliaga (2004). The measure
of diagnosis of cancer for the present study may have underestimated the number of older adult
HCBWP participants with cancer.
Conclusion of Chapter 5
In conclusion, study results suggest that diagnosis of cancer had a limited association
with the pain experience, pain management strategies, pain management outcomes and
admission to a nursing home among older adult HCBWP participants. The pain experience was
significantly associated with personal factors of age, sex, race/ethnicity, cognitive functioning,

229

behaviors indicative of depression and the health and illness factor of comorbid conditions. The
pain experience was not influenced by the diagnosis of cancer.
Regarding pain management strategies, due to the limited number of subjects receiving
hospice services at each time point, further analyses of hospice services with Generalized
Estimating Equations was not carried out. Prescribed pain medications were most consistently
associated with pain over time. Older adult HCBWP participants who were in the initial phase
of diagnosis of cancer were less likely to have non-opioid and adjuvant pain medications
prescribed than older adult HCBWP participants without a diagnosis of cancer. Significant
predictors of the prescription of opioid pain medications vs. no pain medications or non-opioid
pain medications included pain, age, sex, race/ethnicity, cognitive function, comorbid conditions
and time in the HCBWP.
Pain, race/ethnicity, cognitive functioning, behaviors indicative of depression and time in
the HCBWP were significantly associated with the physical functioning of older adult HCBWP
participants over time. The pain management outcome of pain control was significantly
associated with the measure of behaviors indicative of depression. Significant predictors of the
admission of older adult HCBWP participants to a nursing home included pain, age,
race/ethnicity, cognitive functioning and comorbid conditions. Time to nursing home admission
was associated with age, race/ethnicity, cognitive functioning and comorbid conditions.
Next, Chapter 6 will present the contributions to science made by this study as well as
clinical, research and policy implications.

230

CHAPTER 6
Contributions to Science
While many of these results were in agreement with previous research examining the
pain experience, pain management strategies and pain management outcomes of older adults, the
science was extended to include older adult HCBWP participants, a population at high risk for
cancer and poor pain management due to age and poverty.
The science of pain, pain management and pain management outcomes among older
adults was extended through the use of longitudinal analyses of the four assessment time points,
as previous research regarding pain, pain management and pain management outcomes has been
cross-sectional. The study results serve to clarify the associations between prescribed pain
medications and pain, age, sex, race/ethnicity, cognitive functioning, behaviors indicative of
depression, comorbid conditions and time in the HCBWP. The present research adds to the
science as previous research was primarily cross-sectional and this study was longitudinal.
Results from this study note that the significant associations between prescribed pain
medications, age, sex, race/ethnicity, cognitive functioning, behaviors indicative of depression,
pain and comorbid conditions continue over-time in older adult HCBWP participants and are not
just limited to a single observation.
This study utilized a novel combination of data from the MDS-HC, Medicaid paid claim
files, Michigan Cancer Registry and Death Certificate data to examine the pain experience, pain
management strategies, pain management outcomes of older adult HCBWP participants and
admission of older adult HCBWP participants to a nursing home.

231

Implications
Historically, nurses have been responsible for providing comfort and alleviating suffering
from pain. The nursing profession’s focus on the alleviation of pain continues to be at the
forefront of practice and research directives. Nurses are to address issues such as physical
comfort, discomfort and pain via nursing interventions (American Nurses Association, 2008,
2010). Research examining symptoms, such as pain, has been prioritized by the National
Institute of Nursing Research in order to improve patient quality of life (National Institute of
Nursing Research, 2006b). Results from the present study have important implications for
nursing clinical practice and research, which are presented in the following sections.
Implications for Clinical Practice
The results of the present study found that diagnosis of cancer had a limited association
with the pain experience, pain management strategies, pain management outcomes and
admission of older adult HCBWP participants to nursing homes. However, the assumption that
the diagnosis of cancer has no affect on the pain experience, pain management strategies, pain
management outcomes, and nursing home admission among older adult HCBWP participants
cannot be made based on the results of the present study alone as previous research has found
that older adults with cancer are more likely to experience pain when compared to older adults
without cancer (Buchanan, et al., 2005; Reyes-Gibby, Aday, et al., 2007; Rodin, 2008).
Personal factors and the presence of comorbid conditions were strongly associated with the
pain experience, pain management strategies, pain management outcomes of older adult
HCBWP participants over time. Pain is a highly personal experience, being “…whatever the
experiencing person says it is, existing whenever he/she says it does” (McCaffery, 1968).
Individual-related characteristics contribute to the multidimensional nature of pain (American

232

Pain Society, 2005; Armstrong, 2003; Dodd, et al., 2001; Humphreys & et al., 2008) and are
associated with the experience of pain among older adults. The following clinical implications
are focused on the nurse’s (including nurses in advanced practice, case management, clinical
settings, education, administration, education and policy) acknowledgement and promotion via
education and policy development of personal and health and illness-related characteristics that
influence the pain experience, pain management strategies, pain management outcomes of older
adults and admission of older adult HCBWP participants to nursing homes.


Nurses must foster an appreciation of how person and health and illness-related factors
impact the pain experience, pain management strategies and pain management outcomes
of older adult HCBWP participants.
o Before pain management strategies can be implemented the assessment of the
patient’s pain experience must be completed. While the patient’s verbal report is
the “gold standard” for pain assessment, poor cognitive functioning may impact
the ability of the patient to convey his or her pain experience. Results from the
present study found that those with cognitive impairment were less likely to have
daily pain when compared to cognitively intact older adult HCBWP participants
over time. Therefore, nurses must be aware of methods other than verbal report
when assessing for pain in the cognitively impaired including observation of
facial expressions (grimacing), changes in activity level, body movements
(guarding) that may indicate pain (Curtiss, 2010; Herr et al., 2006). Pain
assessment includes other methods beyond the numeric rating scale (pain on a 010 scale). Nurse faculty should include alternative methods of assessing pain,
such as facial expressions, body movements, in nursing education .

233

Nurses function as educators of both waiver staff and family members
assisting with the care of HCBWP participants. Nurses working with waiver
agencies can educate staff working with participants regarding differences in
assessment of the pain experience of cognitively impaired participants. Nurses
working for waiver agencies can educate family members about ways the
cognitively impaired HCBWP participant may express pain.
o Results from the present study noted that age was negatively associated with the
likelihood of experiencing daily pain and the likelihood of having adjuvant and
opioid pain medications prescribed for older adult HCBWP participants. Not only
must clinicians be aware of their own opinions regarding pain and pain
management among persons 65 and above but they must educate older patients
that pain is not “just to be expected” as part of aging and therefore not worthy of
proper management (Delgado-Guay & Bruera, 2008; Goldstein & Morrison,
2005). Careful assessment of the multiple dimensions of pain (severity,
frequency, interference) can ascertain the patient’s perception of the level of pain
as well as its impact on the lives of older adults. Nurses can educate older adult
HCBWP participants that pain is not to be expected as a result of aging and
therefore just “put up with”. Nurses can educate patients and family members
about pain management options for older adults.
o While the measure of behaviors indicative of depression was associated with an
increased likelihood of experiencing daily pain, it was not associated with the
likelihood of having prescribed pain medications. The measure of behaviors
indicative of depression was associated with pain control, with higher levels of

234

behaviors indicative of depression resulting in a decreased likelihood of having
partial or full control of pain via medications. Clinicians and nurses must be
aware of the effect of depression on the pain experience, as well as how the
presence of depression may make pain refractory to treatment with pain
management strategies. Strategies for clinicians and nurses would include
screening older adults that present with pain for depression, aggressively treating
depression and referring to pain management specialists if pain is not responsive
to pain management. If the HCBWP participant is exhibiting behaviors indicative
of depression, nurses working with waiver agencies could educate the participant
and family members to notify the health care provider for worsening symptoms or
encourage the participant to seek treatment for depression.


The goal of the HCBWP is to prevent or delay nursing home admission (Fries, et al.,
2002). The present study found that age, race/ethnicity, cognitive functioning and
comorbid conditions were significantly associated with the admission of older adult
HCBWP participants to a nursing home. Older age, race/ethnicity, cognitive functioning
and comorbid conditions were significantly associated with time to admission of older
adult HCBWP participants to a nursing home. Higher age, white race/ethnicity, impaired
cognitive functioning and comorbid conditions are risk factors for nursing home
admission among older adult HCBWP participants.
Clinicians working in the state programs overseeing the HCBWP as well as
working in the waiver agencies must be aware of factors that may lead to a nursing home
admission and develop care strategies to help mitigate the effect of these factors on the
admission of older adult HCBWP participants to nursing homes. Strategies that may

235

assist in delaying and/or preventing admission to a nursing home include aggressive
management of comorbid conditions. Nurses in the waiver agency can educate the patient
and family members to utilize medications/treatments and monitor comorbid conditions.
Older adult HCBWP participants who are at risk for nursing home admission should be
closely assessed for care needs in order to develop care plans to best support the
participant in his or her environment.
Summary of Implications for Clinical Practice. In summary, nursing implications from
this study arise from the impact of person and health and illness factors on the associations
between the pain experience, pain management strategies, pain management outcomes and
admission of HCBWP participants to a nursing home. Careful assessment of the pain
experience and development of pain management strategies and HCBWP care strategies
must be completed while taking into account personal and health and illness factors that
place the older adult HCBWP participant at risk for daily pain, poor pain management
outcomes and admission to a nursing home. Nurses assume a leadership role within the
waiver program as they assess the patient as well as family needs, utilizing the results of the
assessment to develop a plan of care including education of the participant and family
members to maximize the participant’s quality of life.
Implications for Research


Future research regarding the pain experience, pain management strategies, pain

management outcomes and admission of older adult HCBWP would involve testing
significant associations from the present study with alternative measures also found in the
MDS-HC. For example, preliminary work completed by the researcher has found that an
MDS-HC item asking if a participant is receiving treatment for depression may be a more

236

sensitive measure for depression than the DRS scale used to represent behaviors indicative of
depression.
In the present study, physical function was measured by the total number of dependent
ADL. This measure of physical function was utilized because it was a count variable and could
easily be used in GEE modeling with the specification of a Poisson distribution. Additionally,
concurrent research with the same data set was being carried out and also utilized the ADL count
variable as well. For consistency across studies, the researcher chose to use the same measure of
physical function.
An alternative measure of ADL is the ADL Scale, created by developers of the MDS.
The Activities of Daily Living Scale (ADLS) (Landi, et al., 2000; Morris, et al., 1999) is
comprised of the observed ability of the HCBWP participant in regards to mobility in bed,
transferring, locomotion, dressing, eating, toilet use and personal hygiene over the previous 7
days as measured by MDS-HC items (Section P, Question 2) comprising the ADLS (Landi, et
al., 2000; Morris, et al., 1999). Responses to each item address the amount of assistance the
MIChoice participant needs with mobility in bed, transferring, locomotion, dressing, eating,
toilet use and personal hygiene over the previous 7 days: 0=independent, 1=supervision,
2=limited assistance, 3=extensive assistance and 4=total dependence 5=Activity did not occur,
regardless of ability . Additional research testing associations found in the current study with
alternative measures will assist in determining the most sensitive measures within MDS-HC
data.
o Future Research Question


What is the association between the Activities of Daily Living Scale and
the pain experience, pain management strategies and pain management

237

outcomes and how is this association affected by age, sex, race/ethnicity,
cognitive functioning, behaviors indicative of depression and comorbid
conditions over time among older adult HCBWP participants.


The present study was limited to Medicaid paid claim files to examine pain management

strategies as Medicare paid claim files were not available to the researcher. In a dual eligible
population such as older adult HCBWP participants, pain management strategies such as pain
management procedures (i.e. nerve injections or other pain clinic-related procedures and physical
therapy) would be billed primarily to Medicare. Access to Medicare paid claim files would allow
the researcher to determine if pain management procedures had taken place in addition to any
prescribed pain medications billed to Medicaid paid claim files and Medicare Part D (after
1/1/2006).
o Future Research Question


What pain management procedures did older adult HCBWP participants
receive, as evidenced in Medicare claim files? Do older adult HCBWP
participants who received pain management procedures also have
prescribed pain medications? What differences exist in prescribed pain
medications between older adult HCBWP participants with and without
pain management procedures?



What is the pain experience among older adult HCBWP participants who
received pain management procedures?



What differences exist in the pain experience of older adult HCBWP
participants who receive pain management procedures when compared to

238

older adult HCBWP participants who received only prescribed pain
medications?


The original plan for this dissertation was to compare pain, pain management and pain
management outcomes between older adult HCBWP participants and older adult nursing
home residents. Prior research had noted poor pain management and pain management
outcomes in nursing homes, despite high rates of pain among older adult nursing home
residents (American Geriatrics Society Panel on Persistent Pain in Older Persons, 2002;
Fisher, et al., 2002; Reynolds, et al., 2008; Sawyer, et al., 2006; Teno, et al., 2001; Won,
et al., 2004). HCBWPs provide nursing home-level services for persons who would
otherwise be admitted to nursing homes for care with the goal of the HCBWP being to
prevent or delay nursing home admission (Fries, et al., 2002).
Nursing homes and HCBWP are supposed to both draw from the same
population, therefore person and health-related characteristics between the two groups
should be similar (Fries, James, Hammer, et al., 2004; L. Li & Zullo, 2003). Research is
needed to examine how the pain experience, pain management strategies and pain
management outcomes among older adult HCBWP differs from the pain experience, pain
management strategies and pain management outcomes of older adults in nursing homes.
If HCBWPs are to be a preferred method of caring for frail older adults over nursing
home care, knowledge is needed to determine if pain management services provided in
the HCBWP are preferable to nursing home pain management services, as well as to
determine what pain management services are needed. However, at the time of
dissertation development nursing home MDS data was not accessible. The dissertation
was revised to work with the data that was available at that time. If in the future there is

239

access to nursing home MDS data, this comparative study should be completed. Not only
would this research examine differences in the pain experience, pain management
strategies and pain management outcomes between HCBWP participants and nursing
home residents, but research results may assist governmental agencies in better targeting
HCBWP services and delaying nursing home admission.
o Future Research Questions


How does the pain experience differ between older adult HCBWP
participants and nursing home residents at admission? Are differences in
the pain experience between the two groups associated with sex, age, race,
comorbid conditions, depression and cognitive functioning over time?



What differences exist in pain management strategies among older adult
HCBWP participants and nursing home residents over time? Are
differences in pain management strategies between the two groups
associated with sex, age, race, comorbid conditions, depression and
cognitive functioning over time?



What differences exist in the associations between the pain experience,
pain management strategies and pain management outcomes among older
adult HCBWP participants and nursing home residents over time? Are
differences in the associations between the pain experience, pain
management strategies and pain management outcomes of the two groups
associated with sex, age, race, comorbid conditions, depression and
cognitive functioning over time?

240



This study found that subjects experiencing daily, unusually intense pain had a lower
likelihood than that of subjects with no pain of being admitted to a nursing home. Pain
was not significantly associated with time to nursing home admission. Because those
with cognitive impairment were less likely to experience pain and older adult HCBWP
participants with daily unusually intense pain were less likely to be admitted to a nursing
home , future research should examine the possible mediating or moderating effect of
cognitive function on the association between the pain experience and the admission of
older adult HCBWP participants to a nursing home.
o Future Research Question


Does cognitive functioning have a mediating or moderating effect on the
association between pain and admission to a nursing home among older
adult HCBWP participants?



In the present study, race/ethnicity was associated with the prescription of opioid and
adjuvant pain medications. African American older adult HCBWP participants with daily
pain were less likely to be prescribed opioid and adjuvant pain medications over time
when compared to white older adult HCBWP participants in daily pain. One possible
explanation for this finding may have to do with the fact that adjuvant pain medications
are medications that have a pain relieving effect, but whose primary or initial indication
was not for the treatment of pain (American Pain Society, 2005). Therefore, primary care
doctors may not be as familiar with the use of adjuvant pain medications for pain
management as pain specialists are. If there is differential access to pain specialists
among race then this may explain why black older adult HCBWP participants were less
likely to have adjuvant pain medications prescribed. Future research examining the

241

billing of pain specialist services to Medicaid and Medicare by race would be beneficial
in understanding the prescribing of pain medications in regards to race.
o Future Research Questions:


What differences are there among older adult HCBWP participants in the
prescribing of adjuvant pain medications by race and/or ethnicity?

Implications for Policy
The present study found that over half of older adult HCBWP participants reported daily
pain and 40% of those with daily unusually intense pain had no prescribed pain medications in
the 30 days prior to each time point (assessment). The financial costs of pain management to the
health care system have been reported to exceed $4000.00 per year for persons with chronic pain
(Turk, 2002). If cost containment within HCBWP is of concern to policy makers pain, pain
management and pain management outcomes among HCBWP participants would benefit from
changes in the MDS-HC pain measures and more thorough documentation of both over the
counter and prescribed pain medication.
For the present study, data from the MDS-HC Version 1 was used. Michigan’s HCBWP
was supposed to update to Version 3 and has not done so yet. While the MDS-HC Version 3
includes a more detailed pain measures than Version 1, HCBWP-related research would benefit
more from a pain measure, such as the numeric pain scale, which has been found to be reliable
and valid among various populations (Jensen, 2003). The numeric rating scale would be a more
sensitive measure of pain and changes in pain over time than the pain items used in the MDS-HC
Version 1.
Research examining pain management strategies would benefit from improved
documentation of all medications (including over the counter pain medications) that a participant

242

is taking. The present study was limited to examining prescribed pain medications only. Changes
in policies regarding the documentation of all medications would enable researchers to examine
all medications the participant is taking at each assessment, allowing research addressing
associations between the pain experience and pain medications. In summary, in order to further
examine issues of pain, pain management and pain management outcomes among older adult
HCBWP participants policy must change the measures used to document the assessment of pain
and use of over the counter pain medications and prescribed pain medications.
Conclusion of Dissertation
The primary purpose of this study was to examine longitudinal differences in the pain
experience, pain management strategies and pain management outcomes among older HCBWP
participants with respect to diagnosis of cancer while participating in the HCBWP. The
secondary purpose of this study was to determine what differences exist in how the pain
experience, pain management strategies and pain management outcomes among older adult,
HCBWP participants associates with the admission of older adult HCBWP participants to a
nursing home, with respect to diagnosis of cancer, over the course of time while participating in
the HCBWP.
Diagnosis of cancer, as measured in the present study, had limited association with the
pain experience, pain management strategies and pain management outcomes of older adult
HCBWP participants. Diagnosis of cancer was negatively associated with the prescription of
non-opioid and adjuvant pain medications among older adult HCBWP participants. These results
are contrary to previous research which found that persons with cancer were more likely to
experience pain than persons with no diagnosis of cancer (Buchanan, et al., 2005; Reyes-Gibby,
Aday, et al., 2007; Rodin, 2008). Significant findings of associations between the pain

243

experience, pain management and pain management outcomes and person and health and illness
factors were consistent with findings from primarily cross-sectional pain-related research among
older adult populations.
While findings from this study were consistent with previous findings, the present study
added to the science by focusing on older adult HCBWP participants, a population of which very
little is known regarding pain, pain management or pain management outcomes. Additionally,
the present study was longitudinal, where other pain-related research among older adults has
been primarily cross-sectional. Longitudinal associations between pain, pain management and
pain management outcomes and person-related factors (age, sex, race/ethnicity, cognitive
functioning and behaviors indicative of depression) and health and illness factors (comorbid
conditions) were noted by findings in the present study. This study utilized a novel data set
comprised of data from the Minimum Data Set Home Care, Michigan Medicaid Paid Claim
Files, Michigan Cancer Registry and Michigan Death Certificate data.

244

APPENDICES

245

Appendix A
Demographic Items from MDS-HC Face Sheet
Age: Section A Item 2-“Date of Birth” =empty space to write in date of birth
Sex: Section A Item 3- “Gender” select circle corresponding to “Male” or “Female”
Race/Ethnicity: Section A Item 7a- “Race fill only one White, Asian/Pacific Islander, Black,
American Indian/Eskimo/Aleut”.

246

Appendix B
Table 37
Items of the Adapted Cognitive Performance Scale Utilizing MDS-HC Items
MDS-HC Item
Memory
Short term memory OKseems/ appears to recall
after 5 minutes?

Response
0=Memory OK
1=Memory problem

Intent of Item
To determine client’s
ability to recall what
was learned or known
after 5 minute.

Cognitive skills for daily
decision making
How well client made
decisions about organizing
the day (when to get up or
have meals, which clothes to
wear or activities to do)?

0=Independentdecisions consistently
reasonable
1=Modified
independence-some
difficulty in new
situations
2=moderately
impaired-decisions
poor, cues,
supervision required
3=Severely impairednever/rarely made
decisions

To determine client’s
ability to make
everyday decisions
about the task’s or
activities of daily
living.

Communication: Making
self understood
Expressing information
content-however able

0=Understood-client
expressed ideas
clearly
1=Usually
understood-difficulty
finding words or
finishing thoughts
2=Sometimes
understood-ability
limited to making
concrete requests
3=Rarely/never
understood

To determine
resident’s ability to
express or
communicate requests,
needs, opinions,
urgent problems and
social conversation
(whether in speech,
writing, sign language
or a combination of
these.

247

Table 37 (Continued)

MDS-HC Item

Response

Intent of Item

Self-Performance in eating
Eating-including taking in
food by any method,
including tube feedings

Requires:
0=Independent-no help
or oversight or
help/oversight provided
only 1 or 2 times in the
last 7 days
1=SupervisionOversight,
encouragement or cueing
provided 3 or more times
the last 7 days or
supervision (3 or more
times) plus physical
assistance provided
2=Limited assistanceClient highly involved in
activity; received
physical help in guided
maneuvering of limbs or
other non-weight bearing
assistance 3 or more
times
3=Extensive assistanceWhile client performed
part of activity, over past
7-day period, help of the
following type(s) were
provided 3 or more
times: weight bearing
support OR full
performance by another
during part, but not all of
the past 7 days.
4=Total dependence-Full
performance of activity
by another during entire
7 days
5= Activity did not occur
during the entire 7 days

How client eats and drinks
(regardless of skill).

248

Appendix C
Answered “3”
moderately
impaired on item
E2

Answered
0-2 on item
E2
(not
severely
impaired)

0

Decision Making
Yes
4

Impairment Count?
(Number of the
following)
 Decision making:
Not independent=1-2
 Making self
understood: Not
independent=1-3
 Short term memory:
Not OK=1

0
1

Level 1
Borderline
Intact

No
0-3

2 or 3

Severe Impairment
Count?
Number of the
following
 Decision making:
mod. Impaired=2
 Making self
understood:
Sometimes/Never
=2-3

2

1
Level 3
Moderate
Impairment

Level
0
intact

Level 2
Mild
Impairment

Total
Dependent
Eating?

Level 5
Severe
Impairment

Level 6
Very Severe
Impairment

Level 4
Mod. Severe
Impairment

Figure 8. Diagram Depicting Scoring Rubric for Cognitive Performance Scale
Note. From “MDS Cognitive Performance Scale” by J.N. Morris, Fries, B.E., Mehr, D.R.,
Hawes, C., Phillips, C., Mor, V. and Lipsitz, L.A., 1994, Journal of Gerontology, 49, p. M 178.
Reproduced with permission from Oxford University Press and authors.

249

Appendix D
The variable “Frequently complains or
shows evidence of pain in the past 7 days”,
coded: 0=no pain, 1=pain less than daily,
and 2=pain daily

No Pain in Past 7
Days
Coded 0
“No Pain”

Less than daily
pain in past 7 Days
Coded 1
“Mild”

Daily Pain that is
Not Unusually
Intense
Coded 2
“Moderate”

Figure 9. Diagram Depicting the Scoring of Pain Scale

250

Daily Pain in Past
7 Days

Daily pain that is
Unusually Intense
Coded 3
“Intense”

Appendix E
MDS-HC Items for Pain Management Outcomes of Physical Function and Pain Control
Physical Function


MDS-HC Section P titled “Physical Functioning”
o Item 2 titled “ADL Self Performance”
 “The following address the client’s physical functioning in routine
physical activities of daily life, for example dressing, eating, etc. in
the last 7 days considering all episodes of these activities. For
clients who performed an activity independently be sure to
determine and record whether others encouraged the activity or
were present to supervise or oversee the activity
0 =Independent-no help or oversight or help/oversight provided
only 1 or 2 times in the last 7 days.
1= Supervision-Oversight, encouragement or cueing provided 3 or
more times during the last 7 days or supervision (3 or more times)
plus physical assistance provided only 1 or 2 times during the last
7 days.
2=Limited Assistance-Client highly involved in activity, received
physical help in guided maneuvering of limbs or other non-weight
bearing assistance 3 or more times.
3=Extensive Assistance-While client performed part of activity
over last 7 day period, help of the following types were provided 3
or more times: Weight bearing support or Full performance by
another during part but not all of the last 7 days
4=Total Dependence- Full performance of activity by another
during the entire 7 days
5=Activity did not occur- during the entire 7 days, regardless of
ability”
2d: Dressing: Including laying out of clothes, retrieving clothes
from closet, putting clothes on and taking clothes off. Choose
performance 1-5 (as noted above).
2e) Eating: Including taking food by any method, including tube
feedings Choose performance 1-5 (as noted above).

251

Appendix E (Continued)
2f) Toileting: Including using the toilet room of commode,
bedpan, urinal, transferring on/off toilet, cleaning self after toilet
use, changing pad, managing special required (ostomy,catheter)
and adjusting clothes. Choose performance 1-5 (as noted above).
2g) Personal Hygiene: Including combing hair, brushing teeth,
shaving, applying makeup, washing/drying face and hands and
perineum (excludes baths and shower). Choose performance 1-5
(as noted above).
o Item 3 Titled “Bathing”
Bathing-in the last 7 days (include shower, full tub or sponge bath;
exclude washing back or hair. Choose performance 1-5 (as noted above).
Pain Control


MDS-HC Section J titled “Health Conditions and Preventative Health Measures”
o Item 8e: “Pain controlled by medication”
 0=no pain
 1=medication offered, no control
 2=pain is partially or fully controlled with medication

252

REFERENCES

253

References
Administration on Aging. (2007). Statistics on the aging population Retrieved 7/15/09, from
http://www.aoa.gov/AoARoot/Aging_Statistics/index.aspx
Agency for Healthcare Research and Quality. (2000). Has the Gender or Racial Composition of
Nursing Homes Changed? Chart on Nursing Home Trends Retrieved 09/7/10, from
http://www.ahrq.gov/news/press/pr2000/nurshmpr.htm
Albert, S. M., Simone, B., Brassard, A., Stern, Y., & Mayeux, R. (2005). Medicaid home care
services and survival in New York City. The Gerontologist, 45(5), 609-616.
Allard, P., Maunsell, E., Labbe, J., & Dorval, M. (2001). Educational interventions to improve
cancer pain control: A systematic review. Journal of Palliative Medicine, 4(2), 191-203.
American Geriatrics Society Panel on Persistent Pain in Older Persons. (2002). Clinical
guideline: The management of persistent pain in older persons. Journal of the American
Geriatric Society, 50, S205-S224.
American Nurses Association. (2008). Code of Ethics for Nurses with Interpretive Statements.
Washington , D.C.: American Nurses Publishing.
American Nurses Association. (2010). Nursing's Social Policy Statement: The Essence of the
Profession (3rd ed.). Washington, D.C.: American Nurses Association.
American Pain Society. (2005). Guideline for the management of cancer pain in adults and
children. Glenview, IL: American Pain Society.
Anderson, R. L., Buckwalter, K. C., Buchanan, R. J., Maas, M. L., & Imhof, S. L. (2003).
Validity and reliability of the Minimum Data Set Depression Rating Scale for older
adults in nursing homes. Age and Aging, 32(4), 435-438.
APS. (2005). Guideline for the management of cancer pain in adults and children. Glenview, IL:
American Pain Society.
Armstrong, T. S. (2003). Symptoms Experience: A Concept Analysis. Oncology Nursing
Forum, 30(4), 601-606.
Backonja, M.-M. (2003). Defining Neuropathic pain. Anesthesia and Analgesia, 97, 785-790.
Baker, T. A., & Green, C. R. (2005). Intrarace differences among black and white Americans
presenting for chronic pain management: The influence of age physical health and
psychosocial factors. Pain Medicine, 6(1), 29-38.
Baldwin, L. M., Klabunde, C. N., Green, P., Barlow, W., & Wright, G. (2006). In search of the
perfect comorbidity measure for use with administrative claims data. Medical Care,
44(8), 745-753.

254

Barkin, R. L., Barkin, S. J., & Barkin, D. S. (2005). Perception, assessment, treatment and
management of pain in the elderly. Clinics in Geriatric Medicine, 21, 465-490.
Basbaum, A., Bushnell, C., & Devor, M. (2005). Pain: Basic mechanisms. In D. M. Justins (Ed.),
Pain 2005: An Updated Review. Seattle: IASP Press.
Bennett, J. A., Winters-Stone, K., & Nail, L. (2006). Conceptualizing and measuring physical
functioning in cancer survivorship studies. Oncology Nursing Forum, 33(1), 41-49.
Beydoun, A., & Backjona, M. M. (2003). Mechanistic stratification of antineuralgic agents.
Journal of Pain and Symptom Management, 25(58), S18-S30.
Blyth, F. M., Rochat, S., Cumming, R. G., Creasey, H., Handelsman, D. J., Le Couteur, D. G., et
al. (2008). Pain, frailty and comorbidity on older men: The CHAMP study. Pain, 140,
224-230.
Bourbonniere, M., & Van Cleave, J. H. (2006). Cancer care in nursing homes. Seminars in
Oncology Nursing, 22(1), 51-57.
Bradley, C. J., Given, C. W., Luo, Z., Roberts, C., Copeland, G., & Virnig, B. A. (2007).
Medicaid, Medicare, and the Michigan Tumor Registry: A linkage strategy. Medical
Decision Making, 27(4), 352-363.
Brown, M. L., Riley, G. F., Potosky, A. L., & Etzioni, R. D. (1999). Obtaining long-term disease
specific costs of care: Application to Medicare enrollees. Medical Care, 37(12), 12491259.
Bruckenthal, P. (2008). Assessment of pain in the elderly adult. Clinics in Geriatric Medicine,
24, 213-236.
Bruckenthal, P., & D'Arcy, Y. M. (2007). Assessment and management of pain in older adults: A
review of the basics. Topics in Advanced Nursing Practice, 7(1), 1-8.
Bryant, L. L., Grigsby, J., Swenson, C., Scarbro, S., & Baxter, J. (2007). Chronic pain increases
the risk of decreasing physical performance in older adults: The San Luis Valley Health
and Aging Study. The Journals of Gerontology, 62A(9), 989-996.
Buchanan, R. J., Barkley, J., Wang, S., & Kim, M. S. (2005). Analysis of nursing home residents
with cancer at admission. Cancer Nursing, 28(5), 406-414.
Bureau of Medicaid Policy and Actuarial Services of Michigan Department of Community
Health. (2004). Michigan Medicaid long-term care client characteristics.
Bureau of Medicaid Policy and Actuarial Services of Michigan Department of Community
Health. (2006). Michigan's Long Term Care Client Populations.
Burns, S., & Grove, S. K. (2005). The Practice of Nursing Research: Conduct, Critique and
Evaluation (5th ed.). St. Louis, MO: Elsevier Saunders.
255

Burrows, A. B., Morris, J. N., Simon, S. E., Hirdes, J. P., & Phillips, C. (2000). Development of
a Minimum Data Set-based depression rating scale for use in nursing homes. Age and
Aging, 29(2), 165-172.
Callahan, C. M., Kroenke, K., Counsell, S. R., Hendrie, H. C., Perkins, A. J., Katon, W., et al.
(2005). Treatment of depression improves physical functioning. Journal of the American
Geriatric Society, 53, 367-373.
Carpenter, G. I., Hastie, C. L., Morris, J. N., Fries, B. E., & Ankri, J. (2006). Measuring changes
in activities of daily living in nursing home residents with moderate to severe cognitive
impairment. BMC Geriatrics, 6(7). Retrieved from
http://www.pubmedcentral.nih.gov/picrender.fcgi?artid=1522014&blobtype=pdf
Centers for Medicare & Medicaid Services. (2005). Medicaid at-a-glance 2005: A Medicaid
information source. Department of Health and Human Services Retrieved from
http://www.cms.hhs.gov/MedicaidEligibility/Downloads/MedicaidataGlance05.pdf.
Centers for Medicare & Medicaid Services. (2008). Medicare eligibility tool Retrieved 7/14/09,
from
http://www.medicare.gov/MedicareEligibility/Home.asp?dest=NAV|Home|GeneralEnroll
ment#TabTop
Centers for Medicare & Medicaid Services. (2009a). MDS 2.0 Public Quality Indicator and
Resident Reports Retrieved 7/23/2009, from
http://www.cms.hhs.gov/MDSPubQIandResRep/
Centers for Medicare & Medicaid Services. (2009b, 7/7/2009). Medicaid program: General
information Retrieved 7/12/2009, from
http://www.cms.hhs.gov/MedicaidGenInfo/03_TechnicalSummary.asp#TopOfPage
Chang, V. T., Hwang, S. S., & Kasimis, B. (2002). Longitudinal documentation of cancer pain
management outcomes: a pilot study at a VA medical center. Journal of Pain & Symptom
Management, 24(5), 494-505.
Chang, V. T., Janjan, N., Jain, S., & Chau, C. (2006). Update in Cancer Pain Syndromes. Journal
of Palliative Medicine, 9(6), 1414-1434.
Charlson, M. E., Pompei, P., Ales, K. L., & MacKenzie, C. R. (1987). A new method of
classifying prognostic comorbidity in longitudinal studies: Development and validation.
Journal of Chronic Disease, 40(5), 373-383.
Chou, K. L., & Chi, I. (2007). Reciprocal relationship betwen pain and depression in older
adults: Evidence from the English longitudinal study of ageing. Journal of Affective
Disorders, 102(1-3), 115-123.
Chu, L., Schnelle, J. F., & Osterwell, D. (2004). Prescription analgesic and antidepressant
utilization and cost among elderly Medicaid beneficiaries beofre and after nursing home
admission. Journal of the American Medical Directors Association, 75-81.
256

Cintron, A., & Morrison, R. S. (2006). Pain and ethnicity in the United States: A systematic
review. Journal of Palliative Medicine, 9(6), 1454-1473.
Cleeland, C. S., & Ryan, K. M. (1994). Pain assessment: Global use of the Brief Pain Inventory.
Annals of the Academy of Medicine, Singapore, 23, 129-138.
Cleeland, J. A., Palmer, J. A., & Venzke, J. W. (2005). Ethnic differences in pain perception.
Physical Therapy Reviews, 10, 113-122.
Cleland, J. A., Palmer, J. A., & Venzke, J. W. (2005). Ethnic differences in pain perception.
Physical Therapy Reviews, 10(113-122).
Cohen-Mansfield, J. (2004). The adequacy of the Minimum Data Set assessment of pain in
cognitively impaired nursing home residents. Journal of Pain and Symptom Management,
27(4), 343-351.
Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ:
Lawrence Erlbaum Associates.
Curtiss, C. P. (2010). Challenges in pain assessment in cognitively intact and cognitively
impaired older adults with cancer. Oncology Nursing Forum, 37(5), S7-S15.
D'Souza, J. C., James, M. L., Szafara, K. L., & Fries, B. E. (2009). Hard times: The effects of
financial strain on home care services use and participant outcomes in Michigan. The
Gerontologist, 49(2), 154-165.
Davis, M. P., & Srivastava, M. (2003). Demographics, assessment and management of pain in
the elderly. Drugs Aging, 20(1), 23-57.
Deane, G., & Smith, H. S. (2008). Overview of pain management in older persons. Clinics in
Geriatric Medicine, 24, 185-201.
Deimling, G. T., Bowman, K. F., & Wagner, L. J. (2007). The effects of cancer-related pain and
fatigue on functioning of older adult long-term cancer survivors. Cancer Nursing, 30(6),
421-433.
Delgado-Guay, M. O., & Bruera, E. (2008). Management of pain in the older person with cancer.
Oncology, 22(1), 56-61.
Delgado, P. L. (2004). Common pathways of depression and pain. journal of Clinical Psychiatry,
65(Suppl 12), 16-19.
Deyo, R. A., Cherkin, D. C., & Ciol, M. A. (1992). Adapting a clinical comorbidity index for use
with ICD-9-CM administrative databases. Journal of Clinical Epidemiology, 45(6), 613619.

257

Diatchenko, L., Nackley, A. G., Slade, G., Bhalang, K., Belfer, I., Max, M. B., et al. (2006).
Catechol-O-methyltransferase gene polymorphisms are associated with multiple painevoking stimuli. Pain, 125, 216-224.
Diatchenko, L., Slade, G., Nackley, A. G., Bhalang, K., Sigurdsson, A., Belfer, I., et al. (2005).
Genetic basis for individual variations in pain perception and the development of a
chronic pain conditions. Human Molecular Genetics, 14(1), 135-143.
Dodd, M. J., Janson, S., Facione, N., Faucett, J., Froelicher, E. S., Humphreys, J., et al. (2001).
Advancing the science of symptom management. Journal of Advanced Nursing, 33(5),
668-676.
Duckett, M. J., & Guy, M. R. (2000). Home and community-based waivers. Health Care
Financing Review, 22(1), 123-125.
Duncan, J. G., Bott, M. J., Thompson, S. A., & Gajewski, B. J. (2009). Symptom occurence and
associated clinical factors in nursing home residents with cancer. Research in Nursing
and Health, 32, 453-464.
Duncan, J. G., Forbes-Thompson, S., & Bott, M. J. (2008). Unmet symptom management needs
of nursing home residents with cancer. Cancer Nursing, 31(4), 265-273.
Dworkin, R., Backonja, M.-M., Rowbotham, M. C., Allen, R. R., Argoff, C. R., Bennett, G. J., et
al. (2003). Advances in neuropathic pain. Archives of Neurology, 60, 1524-1534.
Dworkin, R. H., Turk, D. C., Farrar, J. T., Haythornthwaite, J. A., Jensen, M. P., & Katz, N. P.
(2005). Core outcome measures for chronic pain clinical trials: IMMPACT
recommendations. Pain, 113, 9-19.
Edwards, C. L., Fillingim, R. B., & Keefe, F. (2001). Race, ethnicity and pain. Pain, 94, 133137.
Edwards, R. R., Fillingim, R. B., & Ness, T. J. (2003). Age-related differences in endogenous
pain modulation: A comparison of diffuse noxious inhibitory controls in healthy older
and younger adults. Pain, 101, 155-165.
Elliott, A. M., & Horgas, A. L. (2009). Effects of an analgesic trial in reducing pain behaviors in
community-dwelling older adults with dementia. Nursing Research, 58(2), 140-145.
Erikson, C., Salsberg, E., Forte, G., Bruinooge, S., & Goldstein, M. (2007). Future supply and
demand for oncologists: Challenges to assuring access to oncology services. Journal of
Oncology Practice, 3, 79-86.
Ersek, M., & Wilson, S. A. (2003). The challenges and opportunities in providing end-of-life
care in nursing homes. Journal of Palliative Medicine, 6(1), 45-57.
Extermann, M. (2007). Interaction between comorbidity and cancer. Cancer Control, 14(1), 1322.
258

Ezenwa, M. O., Ameringer, S., Ward, S. E., & Serlin, R. C. (2006). Racial and ethnic disparities
in pain management. Journal of Nursing Scholarship, 225-233.
Farrar, J. T., Portenoy, R. K., Berlin, J. A., Kinman, J. L., & Strom, B. L. (2000). Defining the
clinically important difference in pain outcome measures. Pain 88, 287-294.
Faul, F., Erdfelder, E., Lang, A.-G., & Buchner, A. (2007). G*Power 3: A flexible statistical
power analysis program for the social, behavioral, and biomedical sciences. Behavior
Research Methods, 39, 175-191.
Ferrans, C. E., Zerwic, J. J., Wilbur, J. E., & Larson, J. L. (2005). Conceptual model of healthrelated quality of life. Journal of Nursing Scholarship, 37(4), 336-342.
Fisher, S. E., Burgio, L. D., Thorn, B. E., Allen-Burge, R., Gerstle, J., Roth, D. L., et al. (2002).
Pain assessment and management in cognitively impaired nursing home residents:
Association of certified nursing assistant report, Minimum Data Set Report and analgesic
medication use. Journal of the American Geriatric Society, 50, 152-156.
Fitzmaurice, G. M., Laird, N. M., & Ware, J. H. (2004). Applied Longitudinal Analysis.
Hoboken, NJ: John Wiley & Sons, Inc.
Fortner, B. V., Demarco, G., Irving, G., Ashley, J., Keppler, G., Chavez, J., et al. (2003).
Description and predictors of direct and indirect costs of pain reported by cancer patients.
Journal of Pain and Symptom Management, 25(1), 9-18.
Freedman, G. M. (2002). Chronic pain. Clinical management of common causes of geriatric
pain. Geriatrics, 57(5), 36-41.
Fries, B. E., James, M., & Aliaga, P. (2004). A report to the Medicaid long-term care task force:
Persons using Medicaid-funded nursing home and home and community based waiver
services: Institute of Gerontology, University of Michigan.
Fries, B. E., James, M., Hammer, S., Shugarman, L. R., & Morris, J. N. (2004). Is telephone
screening feasible? Accuracy and cost-effectiveness of idetifying people medically
eligible for home-and community-based services. The Gerontologist, 44(5), 680-688.
Fries, B. E., Shugarman, L. R., Morris, J. N., Simon, S. E., & James, M. (2002). A screening
system for Michigan's home- and community-based long-term care programs. The
Gerontologist, 42(4), 462-474.
Fries, B. E., Simon, S. E., Morris, J. N., Flodstrom, C., & Bookstein, F. L. (2001). Pain in U.S.
nursing homes: Validating a pain scale for the Minimum Data Set. The Gerontologist,
41(2), 173-179.
Fuentes, M., Hart-Johnsom, T., & Green, C. R. (2007). The association among neighborhood
socioeconomic status, race and chronic pain in black and white older adults. Journal of
the National Medical Association, 99(10), 1160-1169.

259

Gagliese, L., & Farrell, M. (2005). The neurobiology of aging, nociception and pain: An
integration of animal and human experimental evidence. In S. J. Gibson & M. Weiner
(Eds.), Pain in Older Persons. Seattle: IASP Press.
Gaugler, J. E., Duval, S., Anderson, K. A., & Kane, R. L. (2007). Predicting nursing home
admission in the U.S.: A meta-analysis. BMC Geriatrics, 7(13). Retrieved from
http://www.pubmedcentral.nih.gov/tocrender.fcgi?iid=141469
Gaugler, J. E., Edwards, A. B., Femia, E. E., Zarit, S. H., Stephens, M. P., Townsend, A., et al.
(2000). Predictors of institutionalization of cognitively impaired elders: Family help and
the timing of placement. Journal of Gerontology: Psychological Sciences, 55B, P247P255.
Gaugler, J. E., Yu, F., Krichbaum, K., & Wyman, J. F. (2009). Predictors of nursing home
admission for persons with dementia. Medical Care, 47(2), 191-198.
Geerlings, S. W., Twisk, J. W., Beekman, A. T., Deeg, D. J., & van Tilburg, W. (2002).
Longitudinal relationship between pain and depression in older adults: Sex, age and
physical disability. Social Psychiatry and Psychiatric Epidemiology, 37(1), 23-30.
Gibson, S. J., & Farrell, M. (2004). A review of age differences in the neurophysiology of
nociception and the perceptual experience of pain. Clinical Journal of Pain, 20(4), 227239.
Gibson, S. J., & Helme, R. D. (2001). Age-related differences in pain perception and report.
Clinics in Geriatric Medicine, 17(3).
Given, B., Given, C., Azzouz, F., & Stommel, M. (2001). Physical functioning of elderly cancer
patients prior to diagnosis and following intitial treatment. Nursing Research, 50(4), 222232.
Given, C., Given, B., Azzouz, F., Kozachik, S., & Stommel, M. (2001). Predictors of pain and
fatigue in the year following diagnosis among elderly cancer patients. Journal of Pain
and Symptom Management, 21(6), 456-466.
Given, C., Spoelstra, S., You, M., Haque, R., & Given, B. (2010). Explaining transfers from
Home and Community Based Waiver Program to Nursing homes: Can patients be
retained at home? Manuscript submiited for publication.
Goldstein, N. E., & Morrison, R. S. (2005). Treatment of pain in older patients. Critical Reviews
in Oncology and Hematology, 54(2), 157-164.
Green, C. R., Anderson, K. O., Baker, T. A., Campbell, L. C., Decker, S., Fillingim, R. B., et al.
(2003). The unequal burden of pain: Confronting racial and ethnic disparities in pain.
Pain Medicine, 4, 227-294.
Greener, M. (2009). Chronic pain: nociceptive versus neuropathic. Nurse Prescribing, 7(12),
540-546.
260

Greenspan, J. D., Craft, R. M., LeResche, L., Arendt-Nielsen, L., Berkley, K. J., Fillingim, R. B.,
et al. (2007). Studying sex and gender differences in pain and analgesia: A consensus
report. Pain, 132, S26-S45.
Gruber-Baldini, A. L., Zimmerman, S. I., Mortimore, E., & Magaziner, J. (2000). The validity of
the Minimum Data Set in measuring the cognitive impairment of persons admitted to
nursing homes. Journal of the American Geriatrics Society, 48(12), 1601-1606.
Hanlon, J. T., Guay, D. R., & Ives, T. J. (2005). Oral analgesics: Efficacy, mechanism of action,
pharmacokinetics, adverse effects, drug interactions, and practical recommendations for
use in older adults. In S. J. Gibson & M. Weiner (Eds.), Pain in older adults. Seattle:
IASP Press.
Harrel, F. (1996). Regression coefficients and scoring rules. Journal of Clinical Epidemiology,
49(7), 819.
Harris, R. E., Williams, D. A., McLean, S. A., Sen, A., Hufford, M., Gendreau, R. M., et al.
(2005). Characterization and consequences of pain variability in individuals with
fibromyalgia. Arthritis and Rheumatisim, 52(11), 3670-3674.
Heinricher, M. M. (2005). Central processing of pain: Modulation of nociception. In D. M.
Justins (Ed.), Pain 2005: An Updated Review. Seattle: IASP.
Helme, R. D., & Gibson, S. J. (2001). The epidemiology of pain in elderly people. Clinics in
Geriatric Medicine, 17(3), 417.
Henly, S. J., Kallas, K. D., Klatt, C. M., & Swenson, K. K. (2003). The notion of time in
symptom experiences. Nursing Research, 52(6), 410-417.
Herr, K., Coyne, P. J., Key, T., Manworren, R., McCaffery, M., Merkel, S., et al. (2006). Pain
assessment in the nonverbal patient: Position statement with clinical practice
recommendations. Pain Management Nursing, 7(2), 44-52.
Hollenack, K. A., Cranmer, K. W., Zarowitz, B. J., & O'Shea, T. (2006). The application of
evidence-based principles of care in older persons: Pain management. Journal of
American Medical Directors, 7(8), 514-522.
Horgas, A. L., Yoon, S. L., Nichols, A. L., & Mansiske, M. (2008). The relationship between
pain and functional disability in black and white older adults. Research in Nursing and
Health, 31, 341-354.
Hospice Association of America. (2006). Hospice facts and Statistics. Retrieved from
http://www.nahc.org/hospicefs06.pdf
Huff, R. M., & Kline, M. V. (1999). Promoting health in multicultural populations. London:
Sage.

261

Humphreys, J., & et al. (2008). Theory of Symptom Management. In M. J. Smith & P. R. Liehr
(Eds.), Middle Range Theory for Nursing. New York: Springer Publishing Company.
Im, E. O., Chee, W., Guevara, E., Liu, Y., Lim, H. J., Tsai, H. M., et al. (2007). Gender and
ethnic differences in cancer pain experience. Nursing Research, 56(5), 296-306.
Institute of Medicine. (2001). Crossing the quality chasm. Washington D.C.: National Academy
Press.
International Association for the Study of Pain. (2009). IASP pain terminology Retrieved
05/21/09, from http://www.iasppain.org/AM/Template.cfm?Section=Pain_Definitions&Template=/CM/HTMLDisplay.c
fm&ContentID=1728
Israel, G. (2009). Determining sample size. Gainesville, FL: University of Florida.
Jakobsson, U., Klevsgard, R., Westergren, A., & I.R., H. (2003). Old people in pain: A
comparative study. Journal of Pain and Symptom Management, 26, 625-636.
JCAHO. (2000). Pain assessment and management: An organizational approach. Oakbrook
Terrace, IL: Joint Commission.
Jensen, M. P. (2003). The validity and reliability of pain measures in adults with cancer. The
Journal of Pain, 4(1), 2-21.
Jost, T. S. (2000). Medicare and Medicaid financing of pain management. The Journal of Pain,
1(3), 183-194.
Kalso, E. (2005). Pharmacological management of pain: anticonvulsants, antidepressants, and
adjuvant analgesics. In D. M. Justins (Ed.), Pain 2005: An Updated Review. Seattle:
IASP Press.
Kandle, E., Schwartz, J., & Jessell, T. (2000). The perception of pain (4th ed.). New York:
McGraw-Hill.
Keating, N. L., Norredam, M., Landrum, M. B., Huskamp, H. A., & Meara, E. (2005). Physical
and mental health status of older long-term cancer survivors. Journal of the American
Geriatric Society, 53, 2145-2153.
Kitchener, M., Carrillo, H., & Harrington, C. (2003). Medicaid community-based programs: A
longitudinal analysis of state variation in expenditures and utilization. Inquiry, 40(4),
375-389.
Kitchener, M., & Harrington, C. (2003). Long-term care in the U.S.: A dialectic analysis of
institutional dynamics. San Francisco: University of California.

262

Kitchener, M., Hernandez, M., Ng, T., & Harrington, C. (2006). Residential care provision in
Medicaid home- and community-based waivers: a national study of program trends.
Gerontologist, 46(2), 165-172.
Kitchener, M., Ng, T., & Harrington, C. (2004). Medicaid 1915(c) home and community-based
services waivers: A national survey of eligibility criteria, caps and waiting lists. Home
Health Care Serv Quarterly, 23(2), 55-69.
Kitchener, M., Ng, T., & Harrington, C. (2007). Medicaid home and community-based services
for the elderly: Trends in programs and policies. Journal of Applied Gerontology, 26,
305-324.
Klabunde, C. N., Harlan, L. C., & Warren, J. L. (2006). Data sources for measuring comorbidity:
A comparison of hospital records and Medicare claims for cancer patients. Medical Care,
44(10), 921-928.
Klabunde, C. N., Potosky, A. L., Legier, J. M., & Warren, J. L. (2000). Development of a
comorbidity index using physician data. Journal of Clinical Epidemiology, 53, 12581267.
Kwan, C. W., Chi, I., Lam, T. P., & Chou, K. L. (2000). Validation of Minimum Data Set for
Home Care Instrument for Hong Kong Chinese elders. Clinical Gerontologist, 21(4), 3547.
Laird, B., Colvin, L., & Fallon, M. (2008). Management of cancer pain: Basic principles and
neuropathic pain. European Journal of Cancer 44, 1078-1082.
Landi, F., Onder, G., Cesari, M., Gambassi, G., Steel, K., Russo, A., et al. (2001). Pain
management in frail, community-living elderly patients. Archives of Internal Medicine,
161, 2721-2724.
Landi, F., Onder, G., Cesari, M., Russo, A., Barillaro, C., & Bernabei, R. (2005). Pain and its
relation to depressive symptoms in frail older people living in the community: An
observational study. Journal of Pain and Symptom Management, 29(3), 255-262.
Landi, F., Tua, E., Onder, G., Carrara, B., Sgadari, A., Rinaldi, C., et al. (2000). Minimum data
set for home care: A valid instrument to assess frail older people living in the community.
Medical Care, 38(12), 1184-1190.
Larson, P. J., Carrieri-Kohlman, V., Dodd, M. J., Douglas, M., Faucett, J., Froelicher, E., et al.
(1994). A model for symptom management. Image: Journal of Nursing Scholarship, 26,
272-276.
Lee, K., Kang, S., Liu, X., & Seo, D. (2010). Likelihood-based approach for analysis of
longitudinal nominal data using marginalized random effects models. Journal of Applied
Statistics. Retrieved from
http://pdfserve.informaworld.com/821820_918013288_927505137.pdf

263

Lenz, E. R., Pugh, L. C., Milligan, R. A., Gift, A. G., & Suppe, F. (1997). The middle-range
theory of unpleasant symptoms: An update. Advances in Nursing Science, 19(3), 14-27.
Levin, C. A., Wei, W., Akincigil, A., Lucas, J. A., Bilder, S., & Crystal, S. (2007). Prevalence
and treatment of diagnosed depression among elderly nursing home residents in Ohio.
Journal of the American Medical Directors Association, 8, 585-594.
Li, L. (2005). Predictors of ADL disability tranjectories among low-income frail elders in the
community. Research on Aging, 27, 615-641.
Li, L., & Conwell, Y. (2007). Mental health status of home care elders in Michigan. The
Gerontologist, 47(4), 528-534.
Li, L., & Conwell, Y. (2009). Effects of changes in depressive symptoms and cognitive
functioning on physical disability in home care elders. The Journals of Gerontology,
64A(2), 230-236.
Li, L., & Zullo, R. (2003). A functioning profile of Medicaid waiver participants in Michigan:
Does the program admit the high functioning? Care Management Journals, 4(1), 31-36.
Liang, K. Y., & Zeger, S. L. (1986). Longitudinal data analysis using generalized linear models.
Biometrika, 73(13-22).
Lin, J. T., & Mathew, P. (2005). Cancer pain management in prisons: a survey of primary care
practitioners and inmates. Journal of Pain & Symptom Management, 29(5), 466-473.
Lipman, A. G., & Gautier, G. M. (1997). Pharmacology of opioid drugs: Basic principles. In R.
K. Portenoy & E. Bruera (Eds.), Topics in Palliative Care (Vol. 1, pp. 137-161). New
York: Oxford University Press.
Lipsitz, S. H., & Kim, K. Z., L. (1994). Analysis of repeated categorical data using generalized
estimating equations. Statistics in Medicine, 13, 1149-1163.
Liu, S., Dixon, J., Qiu, G., Tian, Y., & McCorkle, R. (2009). Using generalized estimating
equations to analyze longitudinal data in nursing research. Western Journal of Nursing
Research, 31, 948-964.
Locher, J. L., Kilgore, M. L., Morrisey, M. A., & Ritchie, C. S. (2006). Patterns and predictors of
home health and hospice use by older adults with cancer. Journal of the American
Geriatric Society, 54, 1206-1211.
Luppa, M., Luck, T., Matschinger, H., Konig, H., & Reidel-Hedel, S. G. (2010). Precitors of
nursing home admission of individuals without a dementia diagnosis before admission:
Results from the Leipzig Longitudinal Study of the Aged. BMC Health Services
Research, 10(186), 1-8.

264

Mao, J. J., Armstrong, K., Bowman, M. A., Xie, S., Kadakia, R., & Farrar, J. T. (2007).
Symptom burden among cancer survivors: Impact of age and comorbidity. Journal of the
American Board of Family Medicine, 20(5), 434-443.
Marek, K. D., Popejoy, L., Petroski, G., Mehr, D., Rantz, M., & Lin, W. C. (2005). Clinical
outcomes of aging in place. Nursing Research, 54(3), 202-211.
Martin, L., Poss, J. W., Hirdes, J. P., Jones, R. N., Stones, M. J., Fries, B. E., et al. (2008).
Predictors of a new depression diagnosis among older adults admitted to complex
continuing care: Implications for the depression rating scale. Age and Aging, 37, 51-56.
McCaffery, M. (1968). Nursing practice theories related to cognition, bodily pain, and manenvironment interactions. Los Angeles: University of California at Los Angeles Students
Store.
McCaffery, M. (1972). Nursing management of the patient with pain. Philadelphia: J.B.
Lippincott.
McCaffery, M., & Beebe, A. (1989). Pain: Clinical Manual for Nursing Practice. St. Louis,
MO: C.V. Mosby.
McCallum, J., Simons, L. A., Simons, J., & Friedlander, Y. (2005). Patterns and predictors of
nursing home placement over 14 years: Dubbo study of elderly Australians. Australasian
Journal of Aging, 24(3), 169-173.
McGuire, L. C., Ford, E. S., & Ajani, U. A. (2006). The impaact of cognitive functioning on
mortality and the development of functional disability in older adults with diabetes: The
second longitudinal study on aging. BMC Geriatrics, 6(8), 1-7.
McKenzie, D. A., Semradek, J., McFarland, B. H., Mullooly, J. P., & McCamant, L. E. (2000).
The validity of Medicaid pharmacy claims for estimating drug use among elderly nursing
home residents: The Oregon experience. Journal of Clinical Epidemiology, 53, 12481257.
McNeill, J. A., Reynolds, J., & Ney, M. L. (2007). Unequal quality of cancer pain management:
Disparity in perceived control and proposed solutions. Oncology Nursing Forum, 34(6),
1121-1128.
Miaskowki, C. (2004). Recent advances in understanding pain mechanisms provide future
directions for pain management. Oncology Nursing Forum, 31(4), 25-35.
Miaskowski, C. (2009). Understanding the genetic determinants of pain and pain management.
Seminars in Oncology Nursing, 25(2), S1-S7.
Miaskowski, C., Mack, K. A., Dodd, M. J., West, C., Paul, S. M., Tripathy, D., et al. (2002).
Oncology patients with pain from bone metastasis require more than round the clock
dosing of analgesics to achieve adequate pain control. Journal of Pain, 3(1), 12-20.

265

Michigan Department of Community Health. (2009). Health statistics and reports Retrieved
7/9/2009, from http://www.mdch.state.mi.us/pha/osr/index.asp
Miller, M. E., Davis, C. S., & Landis, J. R. (1993). The analysis of longitudinal polytomous data:
Generalized estimating equations and connections with weighted least squares.
Biometrics, 49, 1033-1044.
Miller, N. A., Elder, K. T., Kitchener, M., Kang, Y., & Harrington, C. (2008). Medicaid 1915(c)
waiver use and expenditures for persons living with HIV/AIDS. Medical Care Research
and Review, 65(3), 338-355.
Mishra, S., Bhatnagar, S., Chaudhary, P., Pratap, S., & Rana, S. (2009). Breakthrough cancer
pain: Review of prevalence, characteristics and management. Indian Journal of Palliative
Care, 15(1), 14-18.
Molenberghs, G., & Verbeke, G. (2005). Models for discrete longitudinal data. New York:
Springer.
Morris, J. N., Fries, B. E., Bernabei, R., Steel, K., Ikegami, N., Carpenter, I., et al. (1997). RAIHome Care Assessment Manual
Morris, J. N., Fries, B. E., Mehr, D., Hawes, C., Phillips, C., Mor, V., et al. (1994). MDS
cognitive performance scale. Journal of Gerontology, 49(4), M174-M182.
Morris, J. N., Fries, B. E., & Morris, S. A. (1999). Scaling ADLs within the MDS. Journal of
Gerontology, 54A(11), M546-M553.
Morris, J. N., Fries, B. E., Steel, K., Ikegami, N., Bernabei, R., Carpenter, I., et al. (1997).
Comprehensive clinical assessment in community setting: Applicability of the MDS-HC.
Journal of American Geriatrics, 45, 1017-1024.
Morris, J. N., Hawes, C., & Fries, B. E. (1990). Designing the national resident assessment
instrument for nursing homes. Gerontologist, 30(3), 293-307.
Mossey, J. M., & Gallagher, R. M. (2004). A 2-year, prospective study of the effects of
depression and pain on physical functioning and healthcare utilization in elderly persons
in a retirement community. Pain Medicine, 5(4), 335-348.
NANDA. (2005). Nursing Diagnoses: Definitions and Classification. Philadelphia: North
American Nursing Diagnosis Association.
National Comprehensive Cancer Network. (2006). Practice guidelines in oncology: Adult cancer
pain Retrieved from http://www.nccn.org/professionals/physician_gls/PDF/pain.pdf
National Institute of Nursing Research. (2006a, June 16, 2008). Executive Summary, from
http://www.ninr.nih.gov/NR/rdonlyres/5AC1595B-08BF-4A6C-A08B3B2B2FF82829/4750/ExecutiveSummary.pdf

266

National Institute of Nursing Research. (2006b). The NINR strategic plan: 2006-2010, from
http://www.ninr.nih.gov/AboutNINR/NINRMissionandStrategicPlan/
NCCN. (2006). Practice guidelines in oncology: Adult cancer pain Retrieved from
http://www.nccn.org/professionals/physician_gls/PDF/pain.pdf
NCCN. (2010). Practice guidelines in oncology: Adult cancer pain Retrieved from
http://www.nccn.org/professionals/physician_gls/PDF/pain.pdf
NCI. (2006). Pain (PDQ) Health professional version. Retrieved from
http://www.cancer.gov/cancertopics/pdq/supportivecare/pain/healthprofessional/allpages/
NCI. (2010a). Dictionary of Cancer Terms Retrieved 7/7/10, from
http://www.cancer.gov/dictionary/?CdrID=44324
NCI. (2010b). Pain (PDQ) Health professional version. Retrieved from
http://www.cancer.gov/cancertopics/pdq/supportivecare/pain/healthprofessional/allpages
#Section_655
Oliver, J. W., Kravitz, R. L., Kaplan, S. H., & Meyers, F. J. (2001). Individualized patient
education and coaching to improve pain control among cancer outpatients. Journal of
Clinical Oncology, 19(7), 2206-2212.
Onder, G., Cesari, M., Russo, A., V., Z., Bernabei, R., & Landi, F. (2006). Association between
daily pain and physical function among old-old adults living in the community: Results
from the ilSirente study. Pain, 121, 53-59.
Paquay, L., De Lepeleire, J., Schoenmakers, B., Ylieff, M., Fontaine, O., & Buntinx, F. (2007).
Comparison of the diagnostic accuracy of the Cognitive Performance Scale (Minimum
Data Set) and the Mini-Mental State Exam for the detection of cognitive impairment in
nursing home residents. International Journal of Geriatric Psychiatry, 22, 286-293.
Pasero, C. (2004). Pathophysiology of Neuropathic Pain. Pain Management Nursing, 5(Suppl.
1), 3-8.
Patrick, D. L. (1997). Finding health-related quality of life outcomes sensitive to health-care
organization and delivery. Medical Care, 35(11), NS49-ns57.
Procter, W. R., & Hirdes, J. P. (2001). Pain and cognitive status among nursing home residents
in Canada. Pain Research and Management, 6(3), 119-125.
Reyes-Gibby, C. C., Aday, L. A., & Cleeland, C. S. (2002). Impact of pain on self-rated health in
the community-dwelling older adults. Pain, 95, 75-82.
Reyes-Gibby, C. C., Aday, L. A., Todd, K. H., Cleeland, C. S., & Anderson, K. O. (2007). Pain
in aging community-dwelling adults in the United States: Non-Hispanic whites, nonHispanic blacks and Hispanics. The Journal of Pain, 8(1), 75-84.

267

Reyes-Gibby, C. C., Shete, S., Rakvag, T., Bhat, S. V., Skorpen, F., Bruera, E., et al. (2007).
Exploring joint effect of genes and clinical efficacy of morphone for cancer pain:
OPRM1 and COMT gene. Pain, 130, 25-30.
Reynolds, K. S., Hanson, L. C., DeVellis, R. F., Henderson, M., & Steinhauser, K. E. (2008).
Disparities in pain management between cognitively intact and cognitively impaired
nursing home residents. Journal of Pain & Symptom Management, 35(4), 388-396.
Riley, G. F., Potosky, A. L., Lubitz, J. D., & Kessler, L. G. (1995). Medicare payments from
diagnosis of death for elderly cancer patients by stage at diagnosis. Medical Care, 33(8),
828-841.
Riley, J. L., Wade, J. B., Myers, C. D., Sheffield, D., Papas, R. K., & Price, D. D. (2002).
Racial/ethnic differences in the experience of chronic pain. Pain, 100, 291-298.
Rodin, M. (2008). Cancer patients admitted to nursing homes: What do we know? Journal of the
American Medical Directors Association, 9(3), 149-156.
Rollman, G. B., Abdel-Shaheed, J., Gillespie, J. M., & Jones, K. S. (2004). Does past pain
influence current pain: Biological and psychosocial models of sex differences. European
Journal of Pain, 8, 427-433.
Romano, P. S., Roos, L. L., & Jollis, J. G. (1993). Adapting a clinical comorbidity index for use
with ICD-9 administrative data: Differing perspectives. Journal of Clinical
Epidemiology, 46(10), 1075-1079.
Rosso, A. L., Gallagher, R. M., Lubrosky, M., & Mossey, J. M. (2008). Depression and selfrated health are proximal predictors of episodes of sustained change in pain in
independently living, community-dwelling elders. Pain Medicine, 9(8), 1036-1049.
Sahyoun, N. R., Pratt, L. A., Lentzner, H., Dey, A., & Robinson, K. N. (2001). The changing
profile of nursing home residents 1985-1997. Retrieved from
http://www.cdc.gov/nchs/data/ahcd/agingtrends/04nursin.pdf.
Samad, T. A. (2004). New understandings of the link between acute pain and chronic pain: Can
we prevent long-term sequelae? . In D. B. Carr (Ed.), The Spectrum of Pain (pp. 16-27).
New York: McMahon.
Sands, L. P., Xu, H., Weiner, M., Rosenman, M. B., Craig, B. A., & Thomas, J. (2008).
Comparison of resource utilization for Medicaid dementai patients using nursing homes
versus home and community based waivers for long term care. Medical Care, 46(4), 449453.
SAS System for Windows (Version 9.3.1) (Version 9.1.1.). (2002). Cary, NC: SAS Institute.
Sawyer, P., Lillis, J. P., Bodner, E. V., & Allman, R. M. (2006). Substantial daily pain among
nursing home residents. Journal of the American Medical Directors Association, 8(158165).
268

Schug, S. A. (2005). Clinical pharmacology of non-opioid and opioid analgesics. In D. M.
Justins (Ed.), Pain 2005: An Updated Review. Seattle: IASP Press.
Sengupta, M., Bercovitz, A., & Harris-Kojetin, L. D. (2010). Prevelance and management of
pain, by race and dementia among nursing home residents.
Shega, J. W., Hougham, G. W., Stocking, C. B., Cox-Hayley, D., & Sachs, G. A. (2006).
Management of noncancer pain in community-dwelling persons with dementia. Journal
of the American Geriatric Society, 54, 1892-1897.
Shugarman, L. R., Fries, B. E., & James, M. (1999). A comparison of home care clients and
nursing home residents: Can community based care keep the elderly and disabled at
home? Home Health Care Services Quarterly, 18(1), 25-45.
Shvartzman, P., Friger, M., Shani, A., Barak, F., Yoram, C., & Singer, Y. (2003). Pain control in
ambulatory cancer patients -- can we do better? Journal of Pain & Symptom
Management, 26(2), 716-722.
Singer, J. D., & Willett, J. B. (2003). Applied Longitudinal Analysis. Oxford: Oxford Univeristy
Press.
Smith, B. D., Smith, G. L., Hurria, A., Hortobagyi, G. N., & Buchholz, T. A. (2009). Future of
cancer incidence in the United States: Burdens upon an aging, changing nation. Journal
of Clinical Oncology, 27(17).
Social Security. (2009). Find an answer to your question Retrieved 7/15/09, from http://ssacusthelp.ssa.gov/cgibin/ssa.cfg/php/enduser/std_adp.php?p_faqid=14&p_created=955050304&p_sid=zyF4UCj&p_accessibility=0&p_redirect=&p_lva=&p_sp=cF9zcmNoPTEmcF9zb3J0X2
J5PSZwX2dyaWRzb3J0PSZwX3Jvd19jbnQ9MjMsMjMmcF9wcm9kcz0mcF9jYXRzPT
MsMSZwX3B2PSZwX2N2PTIuMSZwX3NlYXJjaF90eXBlPWFuc3dlcnMuc2VhcmNo
X25sJnBfcGFnZT0x&p_li=&p_topview=1
Soldato, M., Liperoti, R., Landi, F., Finne-Soveri, H., Carpenter, I., Fialova, D., et al. (2007).
Non-malignant daily pain and risk of disability among older adults in home care in
Europe. Pain, 129, 304-310.
Spoelstra, S., Given, B., von Eye, A., & Given, C. (2010). Falls in the community-dwelling
elderly with a history of cancer. Cancer Nursing, 33(2), 149-155.
Staud, R., Price, D. D., Robinson, M. E., Mauderli, A. P., & Vierck, C. J. (2004). Maintenance of
windup of second pain requires less frequent stimulation in fibromyalgia patients
compared to normal controls. Pain, 110, 689-696.
Stevenson, D. G., & Bramson, J. S. (2009). Hospice care in the nursing home setting: A review
of the literature. Journal of Pain and Symptom Management, 38(3), 440-451.

269

Sun, V. C., Borneman, T., Ferrell, B., Piper, B., Koczywas, M., & Choi, K. (2007). Overcoming
barriers to cancer pain management: an institutional change model. Journal of Pain &
Symptom Management, 34(4), 359-369.
Sweeney, C., & Bruera, E. (2003). Opioids. In R. Melzack & P. D. Wall (Eds.), Handbook of
Pain Managment: A Clinical Companion to Wall and Melzack's Textbook of Pain. New
York: Churchill Livingstone.
Sykes, N., Fallon, M. T., & Patt, R. B. (2003). Clinical Pain Management: Cancer Pain.
London: Arnold.
Teno, J. M., Kabumoto, G., Wetle, T., Roy, J., & Mor, V. (2004). Daily pain that was
excruciating at some time in the previous week: Prevalence, characteristics, and
outcomes in nursing home residents. Journal of the American Geriatric Society, 52, 762767.
Teno, J. M., Weitzen, S., Wetle, T., & Mor, V. (2001). Persistent pain in nursing home residents.
JAMA, 285(16), 2081.
Thomas, E., Peat, G., Harris, L., Wilkie, R., & Croft, P. R. (2004). The prevalence of pain and
pain interference in general population of older adults: Cross-sectional findings from the
North Staffordshire Osteoarthritis Project. Pain, 110, 361-368.
Tilly, J., & Kasten, J. (2001). Home and community based services for older people and young
adults with physical disabilities in Michigan. Washington D.C.: The Urban Institute.
Tonner, M. C., & Harrington, C. (2003). Nursing facility and community based service need
criteria in the United States. Home Health Care Service Quarterly, 22(4), 65-83.
Tsang, A., VonKorff, M., Lee, S., Alonso, J., Karam, E., & Angermeyer, M. C. (2008). Common
chronic pain conditions in developed and developing countries: Gender and age
differences and comorbidity with depression-anxiety disorders. Journal of Pain, 9(10),
883-891.
Turk, D. C. (2002). Clinical-effectiveness and cost-effectiveness of treatments for patients with
chronic pain. Clinical Journal of Pain, 18, 355-365.
Turk, D. C., Dworkin, R. H., Allen, R. R., Bellamy, N., Brandenburg, N., Carr, D., et al. (2003).
Core outcome domains for chronic pain clinical trials: IMMPACT recommendations.
Pain, 106, 337-345.
Turk, D. C., & Okifuji, A. (1999). Assessment of patients' reporting of pain: an integrated
perspective. The Lancet, 353, 1784-1788.
Turk, D. C., & Okifuji, A. (2002). Psychological factors in chronic pain: Evolution and
revolution. Journal of Counseling and Clinical Psychology, 70(3), 678-690.

270

Valeberg, B. T., Rustoen, T., Bjordal, K., Hanestad, B. R., Paul, S., & Miaskowski, C. (2008).
Self-reported prevalence, etiology, and characteristics of pain in oncology outpatients.
Europeon Journal of Cancer, 12, 582-590.
Walke, L. M., Byers, A. L., McCorkle, R., & T.R., F. (2006). Symptom assessment in
community-dwelling older adults with advanced chronic disease. Journal of Pain &
Symptom Management, 31(1), 31-37.
Ward, E., Jemal, A., Cokkinides, V., Singh, G., Cardinex, C., Ghafoor, C., et al. (2004a). Cancer
disparities by race/ethnicity and socioeconomic status. CA: A Cancer Journal for
Clinicians, 54, 78-94.
Ward, E., Jemal, A., Cokkinides, V., Singh, G., Cardinex, C., Ghafoor, C., et al. (2004b). Cancer
disparities by race/ethnicity and socioeconomic status. Cancer: A Cancer Journal for
Clinicians, 54, 78-93.
Washington, L. J., Gibson, S. J., & Helme, R. D. (2000). Age-related differences in endogenous
analgesia to repeated cold water immersion in human volunteers. Pain, 89, 89-96.
Webster, L. R. (2008). Pharmacogenetics in pain management: The clinical need. Clinics in
Laboratory Medicine, 28(4), 569-579.
Weiner, M., & Hanlon, J. T. (2001). Pain in nursing home reisdents: Management strategies.
Drugs and Aging, 18(1), 13-29.
Weissman, D. E., Griffie, J., Muchka, S., & Matson, S. (2001). Improving pain management in
long-term care facilities. Journal of Palliative Medicine, 4, 567-573.
WHO. (1996). Cancer pain relief: With a guide to opioid availability (2nd ed.). Geneva: World
Health Organization.
Wiesenfeld-Hallin, Z. (2005). Sex differences in pain perception. Gender Medicine, 2(3), 137145.
Williams, K. O. (2001). Predicting the site of long-term care services in an elderly population
using demographic and functional characteristics. PhD, George Mason University,
Fairfax, VA.
Williamson, A., & Hoggart, B. (2005). Pain: A review of three commonly used pain rating
scales. Journal of Clinical Nursing, 14, 798-804.
Won, A. B., Lapane, K. L., Vallow, S., Schein, J., Morris, J. N., & Lipsitz, L. A. (2004).
Persistent nonmalignant pain and analgesic prescribing patterns in elderly nursing home
residents. Journal of the American Geriatric Society, 52, 867-874.
World Health Organization. (2002). Lessons for Long-term Care Policy Retrieved from
http://www.who.int/chp/knowledge/publications/ltc_policy_lessons.pdf

271

Yabroff, K. R., Warren, J. L., Knopf, K., Davis, W. W., & Brown, M. L. (2005). Estimating
patient time costs associated with colorectal cancer care. Medical Care, 43(7), 640-648.
Yabroff, K. R., Warren, J. L., Schrag, D., Mariotto, A., Meekins, A., & Brown, M. L. (2009).
Comparison of approaches for estimating incidence costs of care for colorectal patients.
Medical Care, 47(7), S56-S63.
Yaffe, K., Fox, P., Newcomer, R., Sands, L., Lindquist, K., Dane, K., et al. (2002). Patient and
caregiver characteristics and nursing home placement in patients with dementia. JAMA,
287(16), 2090-2097.
Yancik, R., Ershler, W., Satariano, W., Hazzard, W., Cohen, H. J., & Ferrucci, L. (2007). Report
of the National Institute on Aging Task Force on Comorbidity. Journal of Gerontology,
62A(3), 275-280.
Yancik, R., Ganz, P. A., Varricchio, C. G., & Conley, B. (2001). Perspectives on comorbidity
and cancer in older patients: Approaches to expand the knowledge base. Journal of
Clinical Oncology, 19(4), 1147-1151.
Yancik, R., & Ries, L. A. (2000). Aging and cancer in America: Demographic and
epidemiologic perspectives. Hematology/Oncology Clinics of North America, 14, 17-23.
Yong, H. H., Gibson, S. J., de L. Horne, D. J., & Helme, R. D. (2001). Development of pain
attitudes questionnaire to assess stoicism and cautiousness for possible age differences.
Journal of Gerontology: Psychological Sciences, 56B(5), P279-P284.
Zheng, Z., Gibson, S. J., & Helme, R. D. (2000). Age-related differences in the development of
capsaicin-induced hyperalgesia. Pain, 85, 51-58.
Zyczkowska, J., Szczerbinska, K., Jantzi, M. R., & Hirdes, J. P. (2007). Pain among the oldest
old in community and institutional settings. Pain, 129, 167-176.

272