FALLS, FALL SEQUELAE, AND HEALTHCARE USE IN THE COMMUNITY DWELLING ELDERLY WITH A HISTORY OF CANCER By Sandra Lee Spoelstra A DISSERTATION Submitted to Michigan State University in partial fulfilment of the requirments for the degree of DOCTORAL OF PHILOSOPHY Nursing 2010 ABSTRACT FALLS, FALL SEQUELAE, AND HEALTHCARE USE IN THE COMMUNITY DWELLING ELDERLY WITH A HISTORY OF CANCER By Sandra Lee Spoelstra Cancer survivors are living longer, but continue to encounter physical, psychosocial, and economic impacts of their cancer until the end of life. Of all types of injuries, falls pose the most serious threat to quality of life in the elderly. The hypothesis in this study was that communitydwelling elderly cancer survivors experience the influence of the disease or treatment of cancer and have increased falls and fall sequelae with increased use of health care subsequent to a fall. If it can be shown that elderly cancer survivors experience a higher rate of falls, a convincing case can be made for modification of nursing practice, taking cancer history into account when assessing and treating patients. There is a gap in the literature with few studies focusing on falls, fractures, or health care use in elderly survivors subsequent to cancer. This study was an analysis of data from the Michigan Home and Community Based Services program combined with information from the Cancer Registry; exploring, comparing, and contrasting in those with and without a cancer diagnosis, the clinical presentation of falls, fall sequelae, and use of healthcare, in a vulnerable, disparate community-dwelling elderly patient population. Specific aims were: 1) after adjusting for sociodemographic characteristics, medications, comorbidities, and frailty to determine the extent to which patients with a cancer diagnosis experience a greater number of falls, fractures, Emergency Room use, hospitalization, or nursing home placement, in the year following the cancer diagnosis compared with those patients with no diagnosis of cancer; and if there are differences in the number of falls among cancer patients according to site, stage, or cancer treatment; and 2) to examine if the effects of frailty variables on falls are different with respect to site, stage, or cancer treatment. An aging and nursing model of care were synthesized for use as a conceptual model to guide this study. This was a longitudinal, retrospective, cohort study comparing 865 with cancer to 8617 without cancer. Generalized Estimating Equations modeling was used. Findings include mean age was 77.1 years, 67.8% female, 74.0% Caucasian. Cancer diagnosis was stage 2 or later for 92.7%. Cancer survivors‘ fall rate was 32.7% compared to 29.4% in those without cancer. Adjusted odds ratios (ORs) of falls were: 1.16 (95% Confidence Interval [CI]=1.02, 1.33) for those with cancer versus those without cancer; 1.12 (95% CI=1.03, 1.22) for male versus female, 1.29 (95% CI=1.19, 1.40) for antidepressants versus none, 1.53 (95% CI=1.41, 1.65) short-term memory recall problems versus none, 1.45 (95% CI=1.32, 1.59) for evidence of pain daily versus none, 1.56 (95% CI=1.37, 1.77) for weight loss versus none, and 1.07 (95% CI=1.04, 1.12) for comorbidities versus none. ORs for increased fractures were 1.28 (95% CI= 1.17, 1.40) for daily pain versus none, and 1.61 (95% CI=1.11, 1.22) for comorbidities versus none. The odds were smaller with age OR= 0.95 (95% CI=0.87, 0.99), for males versus females 0.76 (95% CI=0.67, 0.69), African Americans versus Whites 0.36 (95% CI=0.26, 0.51), and short-term memory recall problems versus none 0.91 (95% CI=0.83, 1.00). Cancer survivors fell at a higher rate, and the risk of fall was higher closer to the date of cancer diagnosis; however, fractures did not occur more. As elderly cancer survivor‘s transition through life, clinicians need to be aware that these patients are prone to increased falls, assess risk, and implement fall prevention measures. Nurses can integrate fall risk assessment and behavioral and psychological interventions to prevent the onset of falls. Fall incidents should be used to prompt reassessment of the underlying cause, with subsequent interventions to prevent further falls. Copyright by SANDRA LEE SPOELSTRA 2010 DEDICATION This work is dedicated to all of the elderly individuals with cancer whom I have been privileged to care for; my aunt Char, aunt Barb, and cousin Debbie, each of whom were cancer survivors; and my Aunt Helen who has Alzheimer‘s—all of whom had the courage to get back up after a fall and keep going. v ACKNOWLEDGEMENTS I would like to thank the following individuals and organizations for their support of the study. Barbara Given, PhD, RN, FAAN, Dissertation Committee Chairperson and Charles (Bill) Given, PhD, MS Dissertation Committee Member for their mentorship, wisdom, requirement for excellence, and encouragement, which were essential to the completion of this study. I would also like to thank Alla Sikorskii, PhD, for her statistical expertise and responsiveness and support; and Debra Schutte, PhD, RN, for her research expertise and support; as Dissertation Committee Members. Denise Saint Arnault, PhD, RN, as Major Professor and Guidance Committee Chairperson. Furthermore, I would like to thank the many faculty members at Michigan State University, and at Iowa and Indiana universities where I was a traveling scholar, whom I had the privilege of interacting with. I would also like to thank the Institute for Health Care Studies at Michigan State University for their assistance with obtaining the data. A special thank you to Karen Burritt, MSN, RN, Lori Houghton-Rahrig, MSN, Chai Tai Hung, MSN, RN, RN, Melodee Vanden Bosch, MSN, RN, and Kari Wade, MSN, RN for sharing the doctoral experience with me and encouraging me along my journey. Most importantly, immeasurable gratitude is expressed to my husband, Craig, and children, Rachel, Philip, and Sarah, and my extended family, for their unending support and encouragement. Funding sources for this project included the National Institute of Nursing Research Grant Number 1F31NR011522 - 01A1, the State of Michigan Nurse Corp Fellowship, The Walther Cancer Institute, the Schuman family scholarship, Blue Cross Blue Shield of Michigan Foundation, the College of Nursing, and Graduate School. vi TABLE OF CONTENTS LIST OF TABLES.……………………………………………………………… xi LIST OF FIGURES..……………………………………………………………. xiii LIST OF ABBREVIATIONS………………………………………………. xiv CHAPTER 1……………………………………………………………………… 1 OVERVIEW.……………………………………………………………… Research Questions……………………………………………….. Research Question 1………………………………………. Research Question 2………………………………………. Exploratory Research Question 3…………………………. Significance to Nursing…………………………………………… Purpose of the Research ………………………………………….. 1 3 4 5 5 5 7 CHAPTER 2…………..…………………………………………………………. 9 THEORETICAL FRAMEWORK……………………..………………….. 9 Philosophical Foundation for This Study: Social Ecology……….............. 9 Theories for the Examination of Aging and Cancer Survivors.................... 10 The Health Related Quality of Life Model……….......................... 10 Model Elements………....................................................... 10 The Life Course Aging Model........................................................ 12 Model Elements………....................................................... 13 Rationale for Selecting Two Theories for Synthesis……………… 14 Conceptual Framework for the Research Study………………….. 15 Factors in the Study………………………………………………………. 16 Environmental Influences………………………….……........................... 16 Patient Characteristics…………………………….……................. 16 Biologic Factor and Key Variable of Interest…………………….. 17 Medications and Cancer Treatment……..………………………… 18 Frailty………………..……………………………………………. 19 Health Outcomes………………………………………………….. 19 Summary………………………………………………………………….. 20 CHAPTER 3…………..………………………………………………………. LITERATURE REVIEW……………………………………………… Background…………………………………………………..…............... Aging and Falls…………………………………………………… Cancer and Falls…………………………………………………... Aging and Frailty…………………………………………………. Frailty and Cancer………………………………………………… Significance……………………………………………………………….. vii 21 21 21 21 22 26 27 29 Review of the Literature………………………………………………….. Social and Environmental Factors………………………………… Patient Characteristics…………………………………………….. Biologic Factors…………………………………………………… Medications….……………………………………………………. Cancer Treatment….……………………………………………… Frailty…..…………………………………………………………. Limitations and Gaps in the Literature…………………………………… Interventions to Reduce Falls……………………………………………... Conclusion and Significance……………………………………………… 29 30 30 30 31 32 32 34 34 36 CHAPTER 4…………………………………………………………………… 39 METHODOLOGY………………………………………………………… 39 Research Questions……………………………………………………….. 39 Research Question 1………………………………………………. 39 Research Question 2………………………………………………. 39 Exploratory Research Question 3…………………………………. 40 Design…………………………………………………………….. 40 Sample…………………………………………………………….. 40 Description of Data Sources………………………………. 40 Setting…………………………………………………….. 42 Sample…………………………………………………….. 43 Measures………………………………………………………………….. 44 Social and Environmental Factors……..………………………….. 44 Characteristics: Age, Sex, and Race/Ethnicity……………………. 45 Biologic Factor of Interest: Cancer……………………………….. 45 Cancer Treatment Covariates……………………………………… 46 Medication Covariates…………………………………………….. 46 Frailty……………………………………………………………… 46 Comorbidity Covariates…………………………………………… 47 Health Outcome Variables………………………………………… 47 Analytic Sample…………………………………………………………… 47 Inclusion Criteria………………………………………………….. 47 Exclusion Criteria…………………………………………………. 48 Gender and Minority Inclusion……………………………………. 48 Data Collection……………………………………………………………. 49 Procedures………………………………………………………… 49 Quality Control and Data Management…………………………… 50 Data Analysis Plan………………………………………………………… 51 Analytic Techniques………………………………………………. 52 Analysis for Specific Aims……………………………………………….. 53 Aim 1……………………………………………………… 53 Aim 2……………………………………………………… 54 Aim 3………………………………………………………. 54 Alternative Methods………………………………………. 55 Power Analysis……………………………………………. 55 viii Effect Size………………………………………………… 56 Strengths and Limitations…………………………………………………. 56 Protection of Human Subjects…………………………………………….. 57 Internal Review Board……………………………………………. 57 Potential Risks and Protection Against Risk………………………. 58 Anticipated Benefits……………………………………………… 58 CHAPTER 5…………………………………………………………………... 59 RESULTS……………………………………………………………….. 59 Research Question 1………………………………………………………. 59 Research Question 2………………………………………………………. 59 Exploratory Research Question 3………………………………………….. 60 Preliminary Research……………………………………………………… 60 Data Management…………………………………………………………. 61 Sample Selection………………………………………………….. 61 Cancer Group……………………………………………… 61 Non-cancer Group………………………………………… 62 Final Sample Selection……………………………………………. 63 Variable Recategorization………………………………………… 65 Review of Missing Data…………………………………………... 65 Preliminary Work…………………………………………………………. 66 Activities of Daily Living…………………………………………. 66 Factor Loading Analysis………………………………….. 66 Exploratory Factor Analysis……………………… 67 Confirmatory Factor Analysis……………………. 68 Summed Activities of Daily Living Score………………… 69 Medications……………………………………………………….. 69 Comorbidities……………………………………………………… 70 Factor Loading Analysis………………………………………….. 70 Exploratory Factor Analysis……………………… 71 Confirmatory Factor Analysis……………………. 72 Summed Comorbidity Score……………………………… 73 Summary………………………………………………………….. 73 Descriptive Analysis of Sample…………………………………………… 74 Environmental and Social Status…………………………………. 74 Patient Characteristics……………………………………………. 76 Biologic Factors………………………………………………….. 79 Site of Cancer…………………………………………….. 79 Cancer Stage……………………………………………… 80 Medications………………………………………………………. 81 Cancer Treatment………………………………………………… 84 Frailty…………………………………………………………….. 84 Comorbidities…………………………………………….. 84 Symptoms…………………………………………………. 85 Cognition………………………………………………….. 86 Functional Status………………………………………….. 87 ix Health Outcomes………………………………………………….. 89 Summary…………………………………………………………… 91 Research Question Results………………………………………………… 91 Research Question 1………………………………………………. 91 Falls……………………………………………………….. 92 Fractures………………………………………………….. 95 ER Use……………………………………………………. 98 Hospitalization……………………………………………. 100 Nursing Home Placement………………………………... 102 Research Question 2……………………………………………… 106 Research Question 3 Exploration of Environment and Social….. 108 CHAPTER 6……………………………………………………………………… DISCUSSION…………………………………………………………… Main Findings In This Study.…………………………………………... Research Study Findings ………………………………………………. Research Question 1……………………………………………. Falls………….…………………………………………. Fractures …….…………………………………………. Health Care Use ……………………………………….. ER Use………….……………………………... Hospitalization….………………………………. Nursing Home Placement.……………………… Falls and Specific Analysis Cancer.……………………. Falls and Site of Cancer.……………………….. Falls and Stage of Cancer.……………………... Falls and Cancer Treatment.…………………… Summary………….……………………………………. Research Question 2…………………………………………… Site of Cancer, Frailty, and Falls………………………. Stage of Cancer, Frailty, and Falls…………………..... Cancer Treatments, Frailty, and Falls…………………. Summary………….…………………………………… Exploratory Research Question……………………………….. Study Limitations……………………………………………………. Implications for Nursing Practice……………………………………. Fall Prevention……………………………………………….. Implications for Research…………………………………………… Implications for Research…………………………………………….. Conclusion…………………………………………………………..... 110 110 111 111 112 112 121 124 125 125 125 127 127 128 129 129 132 132 136 138 138 139 141 142 143 149 153 153 APPENDICES…………………………………………………………………… 157 Appendix A………………………………………………………………. 157 Appendix B……………………………………………………………… 160 REFERENCES…………………………………………………………………… 170 x LIST OF TABLES Table 1. Theoretical Concepts, Operationalized Variables, and Tool Measures for this Study…………………………………….. 17 Descriptive Statistics of Social and Environmental Status, Marital Status, Housing, Living with Other Person, Time Alone During Day, Feelings of Loneliness, and Pets of Cancer and Non-Cancer Patients with Falls or No Falls …………………………………….. 75 Descriptive Statistics of Sociodemographic of Age, Gender, and Race/Ethnicity of Cancer and Non-Cancer Patients with Falls or No Falls ………………………………………………….. 77 Descriptive Statistics of Site of Cancer by Age Groups, Gender, Race/Ethnicity…………………………………………………….. 80 Descriptive Statistics of Stage of Cancer by Age Groups, Gender, Race/Ethnicity, and Type of Cancer……………………… 82 Descriptive Statistics of Number of Medications, Anti-psychotics, Anti-anxiety, Antidepressants, and Hypnotics of Cancer and Non-Cancer Patients with Falls or No Falls ……………………… 83 Descriptive Statistics of Comorbidities of Cancer and Non-Cancer Patients with Falls or No Falls …………………………………… 85 Descriptive Statistics of Evidence of Pain, Intensity of Pain, and Pain Disruption of Activity of Cancer and Non-Cancer Patients with Falls or No Falls……………………………………………… 86 Descriptive Statistics of Short and Long-Term Memory of Cancer and Non-Cancer Patients with Falls or No Falls ………………… 87 Descriptive Statistics of Functional Status Change and Activity of Daily Living Performance Items of Cancer and Non-Cancer Patients with Falls or No Falls…………………………………….. 88 Descriptive Statistics of Health Outcomes Fractures, ER Use, Hospitalization, and Nursing Home Placement of Cancer and Non-Cancer Patients with Falls or No Falls……………………… 90 Table 12. Factors Associated with Falls in the Final GEE Model………… .. 93 Table 13. Factors Associated with Falls in the Final GEE Model Adjusted for Days Since Cancer Diagnosis…………………………………. 95 Table 2. Table 3. Table 4. Table 5. Table 6. Table 7. Table 8. Table 9. Table 10. Table 11. xi Table 14. Factors Associated with Fractures in the Final GEE Model……… 97 Table 15. Factors Associated with Emergency Room Use in the Final GEE Model…………………………………………………………….. 98 Factors Associated with Hospitalization in the Final GEE Model……………………………………………………………. 101 Factors Associated with Nursing Home Placement in the Final GEE Model………………………………………………………. 103 Table 18. Summary of Findings for Research Question 1…………………. 130 Table 19. Summary of Findings for Research Question 2…………………. 138 Table 20. Independent Variables Source of Data, Data Format, and Data Codes for Study…………………………………………………………… 158 Dependent Variables Source of Data, Data Format, and Data Codes for Study……………………………………………………………… 159 Social Environmental Status Variables Source of Data, Data Format, and Data Codes for Study…………………………………………… 159 Multi-nomial Regression Models of Comorbidities, Medications, and Falls……………………………………………………………… 161 Generalized Linear Modeling Analysis of Least Mean and Standard Errors of Antidepressants and Anti-psychotics, Anti-Anxiety, and Hypnotics…………………………………………………………… 164 Generalized Linear Modeling Analysis of Least Mean and Standard Errors of Depression and Congestive Heart Failure, Coronary Artery Disease, Arthritis, and Cerebral Vascular Accident…………………… 165 Table 16. Table 17. Table 21. Table 22. Table 23. Table 24. Table 25. Table 26. Descriptive Statistics of Vision, Bladder, and Bowel Incontinence, Activity of Daily Living Performances Items of Cancer and Non-Cancer Patients with Falls or No Falls…………………………………………… 166 Table 27. Total Variance of Eigenvalues for each ADL Factor …..…………….. 167 Table 28. Rotated Factor Matrix Correlations of the 14 ADLS with Four Factors.. 168 Table 29. Rotated Factor Matrix Correlations of the 11 ADLS with Three Factors.. 168 Table 30. Rotated Factor Matrix Correlations of the 10 ADLS with Two Factors.. xii 169 LIST OF FIGURES Figure 1. The Conceptual Framework for the Study of Falls and Cancer in the Community Dwelling Elderly……………………… 16 Figure 2. Data Management Final Sample Selection for this Study………… 64 Figure 3. Study Schematic…………………………………………………… 65 Figure 4. Screen Plot of Factors and Eigenvalues ……………………………… 167 xiii LIST OFABBREVIATIONS ADL Activities of Daily Living β Beta CAD Coronary Artery Disease CCI Charlson Comorbidity Index CDC Centers for Disease Control CHF Congestive Heart Failure CFA Confirmatory Factor Analysis CVA Cerebral Vascular Accident EFA Exploratory Factor Analysis ER Emergency Room ES Effect Size GEE Generalized Estimating Equations GLM General Linear Model HCBS Home and Community Based Services HRQOL Health Related Quality of Life HMO Health Maintenance Organization IADL Instrumental Activities of Daily Living IRB Internal Review Board IOM Institute of Medicine LC Life-Course Conceptual Model of Aging MDS Minimum Data Set MSU Michigan State University NIA National Institute of Aging NCI National Cancer Institute NINR National Institute of Nursing Research ONS Oncology Nursing Society PE Parameter Estimates RN Registered Nurse SD Standard Deviation xiv CHAPTER 1: OVERVIEW Cancer is a commonly occurring life-threatening disease leading to higher health care utilization and costs. Cancer survivors are living longer but are continuing to suffer physical, psychosocial, and economic impacts of cancer and its treatment until the end of life (National Cancer Institute, 2008; Spoelstra, 2008). The likelihood that an elderly cancer survivor experiences falls and fall sequelae may be influenced by an association between the patient‘s vulnerability and frailty (Mohile et al., 2009) and cancer history (Alibhai, Gogov, & Allibhai, 2006; Bylow et al., 2008; Cheville, Troxel, Basford, & Kornblith, 2008; Deimling, Bowman, & Wagner, 2007; Deimling, Sterns, Bowman, & Kahana, 2007; Levy et al., 2008; Overcash, 2007; Pautex, Herrmann, & Zulian, 2008). The elderly are a very heterogeneous group with respect to underlying health status. The spectrum of impairment ranges from those who are independent, to those who are at moderate risk of health deterioration (i.e., vulnerability and pre-frailty), to those at high risk for functional decline and mortality (i.e., frail) (Balducci & Extermann, 2000a, 2000b; Wenger et al., 2003). Several clinical characteristics have synergistic influences on the frail vulnerable elderly, which may lead to a fall and may be different in those who have had cancer. Disparities, or an increased rate in the occurrence of falls among cancer survivors, are beginning to emerge in the literature (Fried, Ferrucci, Darer, Williamson, & Anderson, 2004; Hewitt, Rowland, & Yancik, 2003; Keating, Narredam, Landrum, Huskamp, & Meara, 2005; Koroukian, Murray, & Madigan, 2006; Luctkar-Flude, Groll, Tranmer, & Woodend, 2007; Snyder et al., 2008; Sweeney et al., 2006; Walston et al., 2006; Yabroff et al., 2007). However, studies examining whether the cancer diagnosis alters the rate of falls and fall sequelae in elderly survivors are rare. 1 The hypothesis in this study stated that community-dwelling elderly cancer survivors experience increased falls related to the influence of the disease or treatment of cancer and have increased fractures and health care use. If it is demonstrated that elderly cancer survivors experience a higher rate of falls and fall sequelae and health care use than those without a cancer diagnosis, a convincing case can be made for modification of gerontological nursing practices (Cope & Reb, 2006, p. 6; Hodgson, 2002; Kagan, 2004; Rowland & Yancik, 2006; Travis & Yahalom, 2008), taking cancer history into account when treating patients. If falls and fall sequelae and health care use in elderly cancer survivors are unique, as expected, then nurses could refine and standardize approaches to fall risk assessment (Chen, Kenefick, Tang, & McCorkle, 2004) and implement behavioral and/or psychological interventions to prevent the onset of falls (Cimprich et al., 2005), reducing adverse outcomes (Chen et al., 2004; Cope & Reb, 2006). Furthermore, the Institute of Medicine recommends that elderly patients receiving health care should be provided a comprehensive care summary and follow-up plan, which presents an opportunity to review survivors are unique, as expected, then nurses could refine and standardize approaches to fall risk assessment (Chen, Kenefick, Tang, & McCorkle, 2004) and implement behavioral and/or psychological interventions to prevent the onset of falls (Cimprich et al., 2005), reducing adverse outcomes (Chen et al., 2004; Cope & Reb, 2006). Furthermore, the Institute of Medicine (2005) recommends the risks and benefits of falls and fall prevention measures . Accordingly, if falls occur more often in cancer survivors, inclusion of falls and fall prevention measures in cancer survivors‘ plans of care may be required to meet the needs of elderly survivors (Travis & Yahalom, 2008). In this study, an examination of how the addition of a cancer diagnosis alters the rate of falls, fall sequelae, and use of health care among cancer survivors compared to those without a 2 cancer diagnosis is conducted. Furthermore, this study sheds light on whether the cancer stage, site, or treatment alters the rate of falls and fall sequelae in elderly survivors (Kagan, 2004; Travis & Yahalom, 2008). Research Questions Each year in the elderly population, about one-third of those over 65 fall (Center for Disease Control and Prevention, 2006; Davison & Marrinan, 2007), and fall-related fractures increase with age (Lim & Chutka, 2006). Falls are the leading cause of injury (Bennett, Winters, & Nail, 2007; Chen et al., 2004; Davison & Marrinan, 2007; de Rekeneire et al., 2003; French et al., 2007; Stel et al., 2003; Stel, Smit, Pluijm, & Lips, 2004; Tinetti, Allore, Araujo, & Seeman, 2005; van Helden et al., 2008), leading to functional decline, hospitalization, institutionalization, higher health care costs, and decreased quality of life (Chen, Chan, Kiely, Morris, & Mitchell, 2007; Chen et al., 2008; Davison & Marrinan, 2007; Tinetti et al., 2005; van Helden et al., 2008). Falls also rank as the sixth leading cause of death in older people (Center for Disease Control and Prevention, 2006). As the population ages, the numbers of vulnerable and frail elderly cancer survivors will increase dramatically (Balducci & Extermann, 2000b). The elderly are a special group in relation to cancer because of higher occurrence (Gillespie, Gillespie, Robertson, & al, 2004 ; Hewitt et al., 2003; Hodgson, 2002). Evidence is beginning to emerge revealing that individuals with cancer fall at a higher rate than individuals without a history of cancer (Overcash, 2007; Pautex et al., 2008; Pearse, Nicholson, & Bennett, 2004). To date, few studies have simultaneously evaluated the incidence of falls, fall sequelae (fractures), and health care use in cancer survivors. A need exists to better understand the 3 relationship between the disease or treatment of cancer and how it may contribute to falls in elderly survivors (Holley, 2002). In this study an analysis of data from the State of Michigan Home and Community Based Services (HCBS) program combined with information from the Michigan Cancer Registry and Medicaid claims files was examined. A compare and contrast of those with a cancer diagnosis to those without cancer was conducted to determine if the clinical presentation of falls, fall sequelae, and health care use in elderly cancer survivors differ. The purposes of this descriptive study are as follows. To determine if the addition of frailty (cognition, comorbidities, activities of daily living [ADL], pain, weight loss, and vision), certain medications, and a cancer diagnosis are associated with the rate of falls, fall sequelae (fractures), emergency room (ER) use, hospitalization, or nursing home placement, occurring among those 65 years of age and older in the HCBS program and to compare those with or without a diagnosis of cancer. Furthermore, to determine whether the rate of falls, fractures, ER, hospital, or nursing home use among cancer survivors varies before and after the diagnosis of cancer by site of cancer or cancer treatment. Finally, to examine social factors of the individuals in the HCBS program and report the findings. The following research questions were addressed in this study. Research question 1. After adjusting for sociodemographic characteristics (age, sex, race/ethnicity), medications, and frailty (ADLs, cognition, comorbidities, pain, weight loss, vision), to determine the extent to which patients with a cancer diagnosis experience a greater number of falls, sequelae of falls (fractures), ER use, hospitalization, or nursing home placement following the cancer diagnosis compared with those patients with no diagnosis of cancer; and to determine if there were differences in the number of falls among cancer patients in the year 4 following diagnosis according to site of cancer (breast, colon, prostate, lung, or other), stage of cancer (I-IV), or cancer treatment (chemotherapy and/or radiation). Research question 2. To examine if the effects of frailty variables (ADLs, cognition, comorbidities, pain, weight loss, and vision) on falls in the year after diagnosis are differential with respect to site of cancer (breast, colon, prostate, lung, or other), stage of cancer (I-IV), or cancer treatment (chemotherapy and/or radiation). Exploratory research question 3. The social environmental status will be described to understand the context of the population. These factors include: marital status, living environment, whom the patient lives with, time alone in a day, and feelings of loneliness. Significance to Nursing The Oncology Nursing Society (ONS), as well as individual researchers, called for studies of the physiologic and psychosocial consequence of cancer (Bender et al., 2008; Fried et al., 2004; Horning, 2008; Koroukian et al., 2006; Li, 2005; Travis & Yahalom, 2008; Walston et al., 2006). The consequence of cancer could be attributed to the cancer or the side effects of cancer treatment and may appear months or years after diagnosis or treatment has ended (Arroyave et al., 2008; Blaauwbroek et al., 2007; Fox & Lyon, 2007; Hawkins et al., 2008; Travis & Yahalom, 2008), which may include physical and psychological problems (Center for Disease Control, 2007; Daubman, 2008; Hewitt, Bamundo, Day, & Harvey, 2007; Horning, 2008; Stovall, 2008), ultimately leading to frailty and falls. The interface of aging, comorbidity, and cancer in survivors is now a focus of the National Cancer Institute (NCI) (Rowland & Yancik, 2006). In 2009 the ONS expanded its research agenda to focus on physical limitations of cancer survivors. Additionally in 2009, the Institute of Medicine (IOM) identified falls as a top research priority. The NIH, National Institute 5 of Nursing Research (NINR), along with the National Institute of Aging (NIA), NCI, and Centers for Disease Control (CDC) have called for better understanding of the influence of cancer or cancer treatment in elderly survivors (National Institute for Health and Clinical Excellence, 2004). The concern regarding physical functioning and cancer or the effects of cancer treatments are anticipated to increase as the prevalence of cancer increases, anti-cancer therapies become more complex, and the age threshold for active treatment continues to expand (with aged receiving treatment) (Hewitt et al., 2003; Snyder et al., 2008). Empirical research in the area of physical function in cancer survivors, specifically falls or fall sequelae (fractures), is underdeveloped. In summary, falls are a result of a interaction between an individual‘s intrinsic risk factors, the physical environment, and the riskiness of a person‘s behavior (Oliver, 2007). Often oncology nurses accept the occurrence of falls, in part because the goal of care may be focused on eradication of the cancer. This perspective neglects the persistent and deleterious effects of falls, such as fractures, hospitalization, nursing home placement, functional decline, and diminished quality of life. In addition to alterations in social and family roles as well as quality of life, the risks of increased falls or fall injuries require clinicians to assess fall risk factors before, during, and after cancer and cancer treatment and to intervene appropriately—an area where nursing can make a difference. Falls are a major concern for patients, their family, and caregivers and should become a safety priority for oncology nurses. Nursing assessments in the community-dwelling elderly should include identification of fall risk factors, with subsequent implementation of interventions. Fall incidents should therefore be used to prompt nurses to conduct a reassessment 6 of the underlying cause of the fall, with subsequent implementation of interventions to prevent further falls. This study makes an important contribution to nursing and oncology research agenda priorities, examining if the addition of a cancer diagnosis alters the rate of falls, fall sequelae (fractures), and health care use. If falls are found to be more probable in cancer survivors, then fall assessment and prevention interventions will become more important as the elderly population with cancer increases in size and health care costs continue to rise. Further, this research identifies if survivors of particular cancers or later stages of cancer are in greater need of these interventions than others. Descriptive research in cancer survivors on falls, fall sequelae, and health care use will lead to novel interventions ripe for development and testing, and subsequently translational research projects. Purpose of the Research The purpose of this research was to examine how the addition of a cancer diagnosis alters the rate of falls, fall sequelae, and health care use among cancer survivors, compared to those without a cancer diagnosis, as well as to shed light on whether the cancer diagnosis alters frailty that leads to falls in elderly survivors. This study examined variables within the context of a synthesized Life Course Aging Model (Elder, 1985) and the Health Related Quality of Life Model (Ferrans, Zerwic, Wilbur, & Larson, 2005) to accommodate the interface of aging and cancer diagnosis as well as the longitudinal design of the study within the selected database. In Chapter 2, a discussion of how the conceptual model was developed will be presented. In Chapter 3, a synthesized a review of the state-of-the science on nursing, aging, frailty, cancer, and fall and fracture literature is presented. Moreover, an explanation of the Home and Community Based Services (HCBS) Waiver program will be presented. Then dataset, sample, 7 measures, methods, and data analysis plan are presented in Chapter 4. The research findings are in Chapter 5. Finally, a discussion of the implications for clinical practice, study limitations, and future direction for nursing research based on this work is summarized in Chapter 6. This innovative research fills a gap in the scientific literature, providing knowledge on the impact of a cancer diagnosis on falls, fall sequelae, and health care use in cancer survivors. Ultimately, this research may lead to future studies intended to design and test effective interventions to reduce the occurrence of falls in community-dwelling elderly cancer survivors. The long-term goals of this research are to improve functional status and quality of life for elderly cancer survivors, ultimately reducing health care costs. 8 CHAPTER 2: THEORETICAL FRAMEWORK In this chapter, a description of the model used in this study is presented. The Health Related Quality of Life model (HRQOL) (Ferrans et al., 2005), as well as the inter- and intrarelatedness of the dynamic individual and environmental characteristics included in the LifeCourse Conceptual Model of Aging (LC) (Elder, 1985), are synthesized to formulate a new model for examination of the factors in this study. This synthesized model allows for the examination of the inter- and intra-relatedness of sociodemographic characteristics, cancer, cancer treatment, comorbidities, medications, cognition, pain, and ADLs, providing a framework for examination of falls, fall sequelae, and health care in community-dwelling elderly cancer survivors. In the next section, the philosophic underpinnings and the concepts of the HRQOL and LC models are explained. Next, the rationale for selecting the two theories are presented, followed by an explanation of the need to synthesize these models to examine the relationships among the variables within this study. Philosophical Foundation for This Study: Social Ecology The philosophy of social ecology was developed by Bookchin (1990) in the 1960s. The ―social‖ component of this philosophy posits that life‘s problems arise from social problems and that problems can be understood by examining the environment around the person (Bookchin, 1990). Social ecology is a grand theory that posits a continuous interaction between persons and their environment as dynamic components with layers in between as nested factors (Bronfenbrenner, 1977, 1979). This worldview lays the foundation for the inter- and intrarelatedness between people and their environment (Hawley, 1950) contained within the models chosen for synthesis in this study. The following is a description of the models. 9 Theories for the Examination of Aging and Cancer Survivors For this study, the HRQOL and LC theories were synthesized to develop a framework for the study of factors that may influence falls in the aged with cancer. A review of each of the theory‘s empiric origins follows, then model elements are described. Introduction to the HRQOL model. The first theory is the Wilson and Cleary HRQOL model published in 1995, which was later modified by Ferrans in 2005. The following is the empiric origins and model description. The HRQOL model is derived from the ecological model of McElroy and colleagues (Mc Elroy, Bibeau, Steckler, & Glanz, 1988) explicating layers of health outcomes at the individual and environmental level. Prior to the model being published, Ferrans (2005) had grouped 22 studies on quality of life into five broad categories to facilitate concept definition: normal life, happiness and satisfaction, achievement of personal goals, social utility, and natural capacity. Consequently, Wilson and Cleary published the HRQOL model in 1995 with the intent of guiding researchers in measuring the concept. Ferrans then modified that model in 2005, eliminating the non-medical aspects. Model elements. To begin, a brief look at the updated HRQOL model (Ferrans et al., 2005; Patrick, 1997; Wilson & Cleary, 1995) is presented. HRQOL is defined as the aspects of quality of life that relate specifically to a person‘s health (Wilson & Cleary, 1995). There are five main determinants of quality of life in this model, including biological function, symptoms, functional status, general health perceptions, and overall quality of life. Furthermore, characteristics of the individual and environment influence each of these determinants. Biological status includes the physiological processes that support life (Ferrans et al., 2005) as the fundamental determinant of health status (Wilson & Cleary, 1995). Biological 10 function focuses on the performance of cells and organ systems measured through lab tests, physical assessment, and medical diagnosis. Alterations in biological status impact subsequent determinants including symptoms, functional status, and general health perceptions. Symptoms include a patient‘s perception of an abnormal physical, emotional, or cognitive state (Wilson & Cleary, 1995). While symptoms are often related to biological function, they sometimes differ because biological changes do not produce symptoms, and symptoms may be perceived in the absence of a biological cause. This feature makes symptoms unique to the individual and important to measure individually. The next dimension in the model is functional status, which assesses the ability to perform certain tasks (Wilson & Cleary, 1995). Wilson and Cleary (1995) define functional status broadly as the ability to perform tasks in multiple domains such as physical, social, role, and psychological function. Physical functional status may be influenced by biological aging and by symptoms such as pain or fatigue. However, functional status must be measured separately because it may not be correlated with biological function or symptoms. The next dimension of the model is general health perceptions (Wilson & Cleary, 1995). General health perceptions are subjective in nature and allow for the individual to summarize the preceding concepts, placing value on the importance of each, to generate a summation of individual health. General health perception is commonly measured with a single global question of overall health. Each of these constructs ultimately impacts overall quality of life, which is a person‘s sense of well-being that stems from satisfaction or dissatisfaction with the important areas of his or her life (Ferrans et al., 2005). Due to the subjective nature of many of the antecedents such as symptoms or general health perception, overall quality of life is subjective and individualized. 11 Characteristics in the HRQOL model are categorized as demographic, developmental, psychological, and biological factors that influence each of the HRQOL determinants (Ferrans et al., 2005). For example, the individual characteristic of age is hypothesized to influence cognitive status. Characteristics of the environment are either social or physical factors and influence each construct in the model (Ferrans et al., 2005). For example, the living environment, such as an assisted living complex or a nursing home, is hypothesized to influence quality of life. In summary, the HRQOL model depicts a unidirectional flow of factors toward overall quality of life. While these arrows represent the typical causal pathway (Ferrans et al., 2005), it is conceivable and probable that any arrow could point in the opposite direction, representing the complexity of the interactions among the various factors impacting quality of life. Additionally, each of the dimensions in the HRQOL model is broad, overlapping, and not operationalized, and each determinant has multiple parts. Therefore, the HRQOL does not provide adequate guidance for operationalization. Finally, the HRQOL model does not provide adequate support for research within the aging process. This includes such elements as medical care, medications prescribed, and the effect of aging on function. As a result, the HRQOL model was adapted to provide the guidance necessary to conduct the research in this study. Introduction to the life course aging model. During the late 1920s, longitudinal studies of children centered on human development across the lifespan were conducted at the University of California, Berkeley (Jones, Jones, Bayley, Macfarlane, & Honzik, 1971). Elders (1985), an emerging scientist, encountered these studies while using the data to examine careers of these children from the Great Depression. This work led to techniques of thinking about social change, life pathways, and individual development as models of behavioral continuity and change 12 representing distinctive pathways for exploration and the publication of the LC model, which outlines the factors that predict frailty in those who are aging. Model elements. The LC aging model (Elder, 1985) has an ecological foundation. The LC model operates under the premise that no single factor can explain why some people are at higher risk for frailty later in life while others are not (Van Voorhees, Walters, Prochaska, & Quinn, 2007). The framework places equal importance on interacting factors at different ecological levels, as well as the influence of factors within a single level (Mortimer & Shanahan, 2003; Van Voorhees et al., 2007). The approach helps explain how the interaction of factors promotes frailty as a person ages. The LC model addresses the influence of the dynamic patterns of ever-changing human development across the lifespan (Elder, 1985). An individual‘s health evolves through each life stage: early-, mid-, and late-stage as the person ages. The type of the individual‘s medical care, environmental exposures, health behaviors, and economic income influences these stages. The impact of these factors, combined with the normal aging process, may lead to frailty. These social and environmental factors contributing to frailty are dynamic, and the interaction of the factors influences the level of frailty as a person ages (Freedman, Martin, Schoeni, & Cornman, 2008; Van Voorhees et al., 2007). Relationships among factors (social, financial, and human capital; health behaviors and health status) contribute to health or frailty as a person ages (Freedman et al., 2008; Mortimer & Shanahan, 2003; Van Voorhees et al., 2007). Social capital is defined as connections within and between social networks and includes marital status, family, friends, and community. Financial capital represents a person‘s wealth and is the income level of the individual that influences health status. Human capital is defined as the stock of the individual‘s productive skills and 13 technical knowledge, including level of education, personality traits, and ability to cope. Finally, health behaviors are defined as actions taken by a person to maintain, attain, or regain good health and to prevent illness. All of these factors are hypothesized to influence each other in a dynamic and interactive way. Each is hypothesized to contribute directly to the health status of the individual. The LC model provides clearly defined concepts to guide operationalization of factors that enhance the overall view of HRQOL. This model also supports the development of a theoretical framework for the examination of aging factors. The next sections discuss the rationale for selecting two theories for synthesis and explain the need for a new model. Rationale for Selecting Two Theories for Synthesis The rationale for selecting the two theories for synthesis is as follows. The LC and HRQOL theories share the ecological philosophic worldview that there is an inter- and intrarelatedness between the individual and environment along with nested factors in between. The HRQOL model delineates the overall influence of the environment, biologic correlates on symptoms, which compound on functional status (Browall, Ahlberg, Persson, Karlsson, & Danielson, 2008; Coons, Chongpison, Wendel, Grant, & Krouse, 2007; Ferrucci et al., 2000; Goodwin, Black, Bordeleau, & Gantz, 2003). This compounding influence of the causal pathways provides guidance for developing hypotheses for this study, as well as biologic status, symptoms, and function influencing falls and fractures. Additionally, the HRQOL model provides guidance for how individual and environmental characteristics influence elements within the model, further supporting the hypothesis that age, sex, and race or ethnicity influence each determinant. 14 The LC model provides a dimension of theoretical guidance for how nested factors may interact within the context of the aging process. Including biologic factors and medical care in late-life health guides the delineation of frailty (Elder, 1985; Freedman et al., 2008; Mortimer & Shanahan, 2003; Van Voorhees et al., 2007) in elderly samples. The LC framework provides a mechanism for the study of inter- and intra-related variables to better understand falls in elderly cancer survivors (Freedman et al., 2008; Mortimer & Shanahan, 2003; Van Voorhees et al., 2007). The framework thus provides support for the examination of variables in the study within the context of the aging process. Finally, there was a need to synthesize the LC and HRQOL models to effectively examine the relationships among the variables in this study (age, sex, race/ethnicity, cancer, cancer treatment, medications, ADLs, comorbidities, cognition, pain and vision, falls, and fractures). Conceptual Framework for the Research Study The model for this study was synthesized from the LC (Elder, 1985; Freedman et al., 2008; Mortimer & Shanahan, 2003; Van Voorhees et al., 2007) and the HRQOL (Ferrans et al., 2005; Patrick, 1997; Patrick & Chiang, 2000; Wilson & Cleary, 1995) models. An overall model summary identifies pathways and inter-and intra-relationships within the models follows. In the framework, an integration of the environment, characteristics, biologic factors, medications and treatments, symptoms, and functional factors are proposed as a means to examine the health outcomes of falls, fall sequelae, and health care use. The new model incorporates those variables hypothesized to influence health outcomes based on literature review and clinical expertise. Figure 1 portrays the theoretical concepts, operationalized variables, and measures in this study. 15 Figure 1. The Conceptual Framework for This Study Marital status, living arrangement, caregiver, time alone during day Age, Sex, Race/Ethnicity Cancer— type, stage, & times since diagnosis Falls Fractures Comorbidity Cognition Medications & Cancer Treatments Weight Loss Pain Vision Hospitalization ER Use Nursing Home Placement ADLs: mobility, transferring, locomotion, gait, eating, dressing, toileting, bathing, walking, stair climbing In the following sections, each variable in this study as well as causal pathways in the model are described. Furthermore, a review of the literature supporting this model is presented. Social and environmental influences. An exploration of social and environmental influences in this study will occur to better understand this population. Social and environmental factors that are known to influence falls in the elderly include availability of caregivers, living arrangement, and time alone during the day (Gill, Williams, & Tinetti, 2000); thus the marital status, living arrangement, caregiver, time alone during day was examined in this study. Patient characteristics. The sociodemographic patient characteristics involved in this study are age, sex, and race or ethnicity. A multitude of evidence supports that age, sex, and race 16 or ethnicity influence falls in the elderly. A review of this literature follows in the next chapter (Gill, Williams, & Tinetti, 2000; Harrison, Booth, & Algase, 2001; Tinetti et al., 2005). Table 1 Theoretical Concepts, Operationalized Variables, and Categories for this Study Theoretical Concepts Environment/social Operationalized Variables Marital status Living arrangement Caregiver Time alone during day Categories Married, single, or divorced Home, apartment, or other Yes or no All day, part of day, never Date of birth Characteristics Age Sex Race or Ethnicity Male or female Caucasian, African American, other Presence or absence; stage; and site Biologic Cancer site Cancer stage Time since cancer diagnosis Treatments Medications Cancer treatment Presence of anti-psychotic, anxiety, depressant, hypnotics; chemotherapy, and/or radiation Frailty Cognition ADL performance Comorbidities Pain Weight loss Vision Cognition, ADL performance, comorbidity, pain, weight loss, and vision impairment Health Outcomes Falls Occurrence of a fall Fall sequelae ER use Hospitalization Nursing home Placement Fractures ER visits Hospitalization Nursing home placement 17 Biologic Factor and Key Variable of Interest in the Study. This study examined cancer, the key variable of interest. Cancer data included the date of diagnosis, site, and cancer stage. Emerging evidence is beginning to prove that those with a diagnosis of cancer may have reduced physical function and be prone to falls and fractures as a result of a fall (Overcash, 2007; Pautex et al., 2008; Pearse et al., 2004). Overall, based on the hypothesis in this study, the influence of cancer is expected to alter the health outcomes of falls and fractures compared to those without a cancer diagnosis directly or mediated by comorbidities, ADLs, cognition, weight loss, vision, and pain. Patient characteristics including age, sex, and race or ethnicity are expected to influence each construct, biologic factors, symptoms, and functional status factors individually and will eventually influence outcomes. Medications and cancer treatment. A covariate in this study is the influence of certain types of medication classifications and cancer treatments. This included the following medication groups: anti-anxiety, antipsychotics, antidepressants, and hypnotics. These four medication classifications were chosen as they are taken more often in the elderly in this sample. A plethora of evidence exists on the use of these medications and studies conducted on falls in the elderly (Agostini, Han, & Tinetti, 2004; Landi, Onder, Cesari, Barillaro et al., 2005; Leipzig, Cumming, & Tinetti, 1999b). The cancer treatments in this study include chemotherapy and radiation. Some evidence purports that cancer treatments influence functional limitations (Bylow et al., 2008; Cheville et al., 2008; Given, Given, Sikorskii, & Hadar, 2007; Goodwin, 2007), which may affect the outcomes in this study. Medications and cancer treatments will be influenced by biologic status, which in turn has been influenced by patient characteristics. Medications and cancer treatments can directly influence outcomes in this study or indirectly, mediated by comorbidities, ADLs, cognition, weight loss, vision, and pain. The treatments 18 examined in this study are expected to affect the constructs of symptoms and functional status, eventually influencing outcomes. Frailty. Frailty is a multidimensional construct commonly used to describe the condition of an older person who has chronic health problems, has lost functional ability, and is likely to deteriorate further (Inouye, et al., 2007). Frailty is commonly operationalized as the presence of any of the following characteristics: aged 85 years or older, a limitation in any ADL, any geriatric syndrome (dementia, memory loss, osteoporosis, weight loss, depression, falls, or incontinence), and one or more comorbidities (Balducci, 2007; Balducci & Extermann, 2000b; Ferrucci, Guralnik, & Cavazzini, 2003; Koroukian et al., 2006). In this study, frailty is considered to be a multi-dimensional construct (Fried et al., 2004) to include the covariates of cognition, ADLs, comorbidities, pain, weight loss, and vision. The compounding influence of frailty will be supported (Badger, Segrin, Dorros, Meek, & Lopez, 2007; Blaauwbroek et al., 2007; Walke, Gallo, Tinetti, & Fried, 2004). Furthermore, the influence of frailty on health status is expected to alter the health outcomes of falls and fractures (Fried et al., 2004; Fried, Tangen, & Walston, 2001; Roche et al., 2006). Frailty is expected to be influenced by patient characteristics, biologic status, and medications in this study. The hypothesis is that frailty would influence falls, fall sequelae, and health care use. Health outcomes. A fall is defined as an unexpected event in which a person comes to rest on the ground, floor, or lower level (Lamb, Jorstad-Stein, Hauer, & Becker, 2005) in the past 180 days (National Institute on Aging, 2008) (French et al., 2007; Hirdes et al., 2004; Vellas, Wayne, Romero, Baumgartner, & Garry, 1997). A fracture is defined as the separation of a bone into two or more pieces under the action of stress or a broken bone (Holley, 2002; Rubenstein, Kenny, Eccles, Martin, & Tinetti, 2002; Stel et al., 2004; Stevens & Sogolow, 2005) as identified 19 in claims files with the ICD-9 coding (Klabunde, Legler, Warren, Baldwin, & Schrag, 2007; Koroukian et al., 2006). Health care use is conceptually defined as when a patient visits the ER, has a hospital admission, or is placed in the nursing home. The health outcome variables of interest in this study are falls, fall sequelae (fractures), and health care use (ER use, hospitalization, and nursing home placement) (ASCO, 1995). In this framework, the patient characteristics influence each of the constructs, biologic status, frailty, and health outcomes. Additionally, biologic status influences medications and cancer treatments, which has a compounding influence on symptoms and functional status. All these determinants lead to an influence on falls, fall sequelae, and health care use. Summary Evidence is emerging that those with cancer may fall at a higher rate than those without a cancer diagnosis (Overcash, 2007; Pautex et al., 2008; Pearse et al., 2004). Further studies examining whether a cancer diagnosis alters the rate of falls, fracture, and health care use need to be conducted to determine if a cancer diagnosis alters the occurrence of falls, fall sequelae, and health care use. The conceptual framework provides theoretical guidance for the examination of the variables of interest in this study. The intent of this study is to test this new theoretical framework by examining the research questions. Another purpose of this study is to examine the extent to which a diagnosis of cancer alters the occurrence of falls, fall sequelae, and health care use in cancer survivors. Factors will also begin to be identified that may be modifiable by nurse clinicians to prevent or reduce the occurrence of falls (Given & Sherwood, 2005). In the next chapter, a review of the literature on aging, frailty, cancer, and falls is presented supporting each component of the theoretical framework. 20 CHAPTER 3: REVIEW OF THE LITERATURE The purpose of Chapter 3 is to review the state of the science concerning falls, fractures, and frailty in order to support the foundation of the methodology for examining falls, fall sequelae, and health care use in community-dwelling elderly cancer survivors, which is described in Chapter 4. A review of studies related to each component of the theoretical framework from left to right in Figure 1 is presented, as well as a discussion of how the literature is related to the research questions in this study. Additionally, this chapter will examine the significance of this research and identify gaps in existing literature that this study will begin to fill. Background Aging and falls. A major risk factor for falling is aging (Fauth, Zarit, Malmberg, & Johansson, 2007), and the risk of being seriously injured in a fall increases with age. Annually, one in three Americans, or 30% of older adults 65 and older, fall, and 40% of those 80 and older fall (Tinetti, Baker, & McAvay, 1994; Tinetti, Speeckley, & Ginter, 1988); many of these falls are recurrent (Campbell et al., 1990; Nevitt, Cummings, Kidd, & Black, 1989; Tinetti et al., 1988). The rates of fall injuries for elderly 85 and over were four to five times that of adults 65 to 75 (Stevens, Corso, Finkelstein, & Miller, 2006). Additionally, elderly individuals often have comorbidities and disabilities (Fried et al., 2004; Iezzoni & Freedman, 2008; Johnson & Wiener, 2006), which increase the occurrence of falls (Guilley et al., 2008; Iezzoni & Freedman, 2008; Johnson & Wiener, 2006; US Department of Health and Human Services, 2005). Twenty to 30% of the elderly who fall suffer moderate to severe injuries (Sterling, O'Connor, & Bonadies, 2001), decreasing the ability to perform ADLs such as walking, bathing, dressing, and eating (Bennett et al., 2007; Beswick et al., 2008). Especially among older adults, 21 falls increase disability, and those injured during a fall often do not return to their pre-fall level of physical functioning (Scaf-Klomp, Van Sonderen, Sanderman, Ormel, & Kempen, 2001), which diminishes quality of life (Bandeen-Roche et al., 2006; Guilley et al., 2008; Tinetti, McAvay et al., 2008). Consequently, a fall increases the likelihood of admission to a nursing home (Donald & Bulpitt, 1999). Physical injuries associated with falling include fractures, contusion, and lacerations (Davison & Marrinan, 2007). Most fractures among the elderly are caused by falls (Bell, TalbotStern, & Hennessy, 2000) with the most common fractures including the spine, hip, forearm, leg, ankle, pelvis, upper arm, and hand (van Helden et al., 2008). Falls are also the most common cause of traumatic brain injuries (Jager, Weiss, Coben, & Pepe, 2000), accounting for 46% of fatal falls among the elderly (Stevens et al., 2006). Nearly 85% of deaths from falls were among people 75 and older (Center for Disease Control, September 21, 2007). The rate of fall-related deaths among the elderly rose significantly over the past decade (Stevens et al., 2006). Of all types of injuries, falls pose the most serious threat to quality of life in the elderly (Davison & Marrinan, 2007; de Rekeneire et al., 2003; van Helden et al., 2008). Cancer and falls. Compounded with the general effects of aging, people are diagnosed with cancer at an older age (Bennett et al., 2007; Rowland & Yancik, 2006; Schmitz, Cappola, Stricker, Sweeney, & Norman, 2007; Yancik & Ries, 2000). Cancer survivors are living longer (Center for Disease Control, 2007) and usually have comorbidities (Wedding et al., 2007) and functional limitations (Sweeney et al., 2006). Some cancer treatments are linked to bone loss (Alibhai et al., 2006; Chen, Maricic, Pettinger et al., 2005; Delmas & Fontana, 1998; Friedlaender, Tross, & Doganis, 1984; Michaud & Goodin, 2006; Saarto et al., 2001) or vitamin D depletion (Overcash, 2008) long after treatment, which may lead to falls and fractures 22 (Coleman et al., 2007; Greenspan, Bhattacharya, Sereika, Brufsky, & Vogel, 2007; Waltman et al., 2006). Others have found a decrease in cognitive skills during cancer treatments, increasing the prevalence of falls (Pautex et al., 2008). A high incidence of vertebral fracture has been found in women with breast cancer (Kanis et al., 1999). There is a 15% higher rate of hip fractures from falls in breast cancer survivors than in women without a breast cancer diagnosis (Chen et al., 2008; Chen, Maricic, Bassford et al., 2005). Evidence has revealed that functional decline occurs among those with cancer. Multiple studies demonstrate cancer-related fatigue (Barsevick, Dudley, & Beck, 2006; Deimling, Bowman et al., 2007; Luctkar-Flude et al., 2007) and pain (Deimling, Bowman et al., 2007; Gulluoglu et al., 2006; Holen, Lydersen, Klepstad, Loge, & Kaasa, 2008), which influences functional status in elderly survivors (Deimling, Sterns et al., 2007) and increases the risk of falls. One study disentangling the effects of cancer-related pain and fatigue on the physical functioning of elderly long-term cancer survivors found that age-related factors accounted for 21% of the variance in physical functioning, whereas cancer-related factors contributed to 6% (Deimling, Bowman et al., 2007), thereby increasing fall risk for cancer survivors. In early-stage breast cancer (Mandelblatt et al., 2006), as well as those with metastatic breast cancer (Cheville et al., 2008; DeSanto-Madeya, Bauer-Wu, & Gross, 2007), there are evident patterns of physical function impairments as well as the possibility of falls and fall-related fractures. In prostate cancer, certain treatments have demonstrated evidence of long-term effects on physical function decline (Alibhai et al., 2006; Bylow et al., 2008), increasing the prevalence of falls (Bylow, Mohile, Stadler, & Dale, 2007; Clay et al., 2007; Levy et al., 2008). Although these studies seem 23 to indicate that physical function decline exists in cancer survivors, limited information specifically on falls is available. Evidence shows that those with cancer fall more often than others (Flood et al., 2006; Overcash, 2007; Pautex et al., 2008; Pearse et al., 2004) and that certain types of cancer (Pearce & Ryan, 2008; Waltman et al., 2006; Waltman et al., 2007) or metastatic disease (Pearce & Ryan, 2008) may be associated with particularly higher rates of falling and higher rates of fractures (Waltman et al., 2006). A prospective study of 165 elderly community-dwelling cancer patients with 45% receiving chemotherapy found 21% had a fall (Overcash, 2007). In this study, no comparison was made to non-cancer patients. Three studies on falls in cancer patients exist in palliative care. In the hospital setting, an exploratory study of the elderly (mean age 71) found the rate of falls in 198 cancer patients to be 18%, with delirium being a significant factor (Pautex et al., 2008). However, in another inpatient study of 119 hospice patients, only 10% had a fall; the low percent was attributed to a limited amount of time out of bed (Pearse et al., 2004). A prospective study of 102 hospice patients found 11.8% had a fall and 5.9% of those fell more than once (Flood et al., 2006). Significant risk factors in this study were cognitive impairment, low blood pressure, visual impairment, and age over 80 (Flood et al., 2006). In a retrospective study of 119 elderly (mean age 74.1), hospital patients who had an oncologic diagnosis, functional dependencies, and geriatric syndromes were examined, which revealed cognitive impairment, depression, weight loss, and use of high-risk medications, yet only 5% had falls (Flood et al., 2006) (the low percent may also be attributed to limited amount of time out of bed). 24 Evidence is also emerging on the relationship between specific types of cancers or cancer stage and falls. In a hospital study of 119 cancer patients, patients with lung cancer fell more often, followed by patients with head and neck cancer compared to other types of cancer; the presence of metastatic disease was evident in 64% of the fallers (Pearce & Ryan, 2008), demonstrating that this disease, particularly in lung cancer, increases the prevalence of falls. In a retrospective study of postmenopausal breast cancer, findings indicate cancer survivors were at higher risk for early bone loss and increased falls and fractures (Waltman et al., 2006). Longterm studies on childhood cancer survivors are underway (Blaauwbroek et al., 2007; Hawkins et al., 2008; Lee, Santacroce, & Sadler, 2008), yet limited information is available on the prevalence of falls since the time of cancer diagnosis. However, few if any studies compare falls and fractures of patients prior to and following the diagnosis of cancer. The disease or treatment of cancer may contribute to functional decline leading to disability, disablement, and falls in the elderly (Bylow et al., 2008; Holley, 2002; Overcash, 2007; Pautex et al., 2008), and evidence suggests that functional decline may exist. Both the rate of falls and risk of falling may be significantly higher for those living as cancer survivors (Overcash, 2007; Pautex et al., 2008). This is significant because cancer is a commonly occurring life-threatening disease leading to higher health care utilization and costs. Additionally, cancer survivors are living longer and experiencing physical, psychosocial, and economic impacts of cancer and its treatment until the end of life. Few studies on falls in elderly cancer survivors, comparing those with a cancer diagnosis to those without a cancer diagnosis, are present in the literature. A need exists for further research in a multivariate context (e.g., presence of the diagnosis; cancer types: breast, colon, prostate, lung; and stages: I-IV). Using a disparate and vulnerable population with poorer 25 functional status to examine frailty variables will allow us to distinguish differences between the two comparison groups—those with cancer compared to those without cancer. Furthermore, using this sample allows for the ability to distinguish what the magnitude of difference is between the two groups. Moreover, examining frailty in the elderly is an emerging science. There is a paucity of evidence examining frailty in elderly cancer survivors. What follows in the next sections is a discussion of frailty that will frame the examination of frailty variables in this study. Aging and frailty. Frailty is now known to be a potential state of vulnerability or a multisystem reduction in physiological capacity not related to a single disease process (BandeenRoche et al., 2006; Bartali et al., 2006; Fried et al., 2004; Guilley et al., 2008; Semba et al., 2006). As an emerging geriatric syndrome, the ―essence‖ of frailty has been defined as ―excess demand imposed upon reduced capacity‖ (Ahmed, Mandel, & Fain, 2007). Frailty is a bodywide set of linked deteriorations including but not confined to musculoskeletal, cardiovascular, metabolic, and immunologic systems with a pathway that leads to a decline in physical activity either as a result of habit or disease (Bortz, 2002) and may lead to falls. Frailty as a phenotype is supported by the Women‘s Health and Aging Study, which examined 1,002 women age 65 and older and found the distinct presence of a unique set of signs and symptoms indicted frailty (Roche et al., 2006). This included, weight loss, poor grip strength, muscle wasting, and poor nutrition (Roche et al., 2006). Balducci and Extermann propose a framework to describe frail elderly as the presence of any of the following characteristics: aged 85 years or older, a limitation in any ADL, any geriatric syndrome (dementia, memory loss, osteoporosis, weight loss, depression, falls, or incontinence), and one or more comorbidities (Balducci, 2007; 2000b; Ferrucci et al., 2003; 26 Koroukian et al., 2006). Fried and colleagues approach frailty as a distinct clinical entity that coexists with and is causally related to comorbidity and disability in the elderly (2004; Walston et al., 2006). The relationship between frailty, comorbidity, and disability are complicated, overlapping, and interrelated, with causal interconnectedness and co-occurrence (Chen et al., 2007; Crimmins, Saito, & Reynolds, 1997; Fried et al., 2004; Koroukian et al., 2006; Lee & Rantz, 2008; Li, 2005; Walston et al., 2006) that often predicts the use of health resources (Boult et al., 2008; Bouman et al., 2008; Huss, Stuck, Rubenstein, Egger, & Glough-Gorr, 2008; Koroukian et al., 2006; Melis, Adand et al., 2008; Melis, Van Eijken et al., 2008; Weiner et al., 2003). Frailty and cancer. Recent studies on cancer in the elderly have documented disparities in frailty by age and comorbid conditions (Yancik & Ries, 2000; Yancik, Wesley, & Ries, 2001). Three national-level studies (Hewitt & Rowland, 2002; Hewitt et al., 2003; Keating et al., 2005; Yabroff et al., 2007) seem to indicate that frailty components of functional (ADLs and IADLs) and cognitive status differ in those who have had cancer compared to those who have not had cancer. Studies of cancer-related outcomes in the elderly focus on the role of physical function and geriatric syndromes in addition to comorbidities. Although the definition of frailty is dynamic, one overarching theme within the various characterizations is an increasing risk for adverse health outcomes, such as functional or physical decline and/or mortality. There is emerging evidence on the topic of frailty in those who are aging. Examinations of frailty factors within a disease process such as cancer are rare. What‘s significant is that there is a scarcity of evidence concerning the extent to which frailty factors, cognition, ADLs, incontinence, depression, and vision impact falls in elderly cancer survivors compared to those 27 without cancer. Moreover, there is a paucity of evidence on if the frailty factors are different in those who have a cancer diagnosis compared to those without cancer. Furthermore, there is a paucity of evidence on if frailty is similar to cancer cachexia and if cancer cachexia exacerbates frailty in cancer survivors. This study begins to examine frailty factors within the context of the disease of cancer, beginning to add critical information to the literature. Cancer survivors are living longer but are continuing to suffer the physical, psychosocial, and economic impacts of cancer and its treatment until the end of life (National Cancer Institute, 2008; Spoelstra, 2008). The likelihood that an elderly cancer survivor experiences a fall may be influenced by an association between their characteristics, frailty (Mohile et al., 2009), cancer history, or treatment (Alibhai et al., 2006; Bylow et al., 2008; Cheville et al., 2008; Deimling, Bowman et al., 2007; Deimling, Sterns et al., 2007; Levy et al., 2008; Overcash, 2007; Pautex et al., 2008). The elderly are a very heterogeneous group with respect to underlying health status. The spectrum of impairment ranges from those who are independent, to those who are at moderate risk of health deterioration (i.e., vulnerability and pre-frailty), to those who are at high risk of functional decline and mortality (i.e., frail) (Balducci & Extermann, 2000a, 2000b; Wenger et al., 2003). Several clinical characteristics have synergistic influences on the frail vulnerable elderly, which may lead to a fall and fall sequelae and use of health care resources, and may be different in those who have had cancer. In this study, we examined the relationship between frailty factors and cancer and cancer treatment to identify if this synergistic effect is different in those with a cancer diagnosis compared to individuals without cancer. What follows in the next section is why it is important to study elderly cancer survivors. 28 Significance Studying the incidence and impact of falls is significant because 20% of the population in the United States are 65 years and older, and the most rapidly growing segment of our population is 85 and older (Johnson & Wiener, 2006; U.S. Census Bureau, 2008), who experience frailty and fall more often. In 2006, health care utilization for older adults resulting from fall injuries included more than 1.8 million persons treated in emergency departments and more than 421,000 persons hospitalized (Center for Disease Control and Prevention, 2006). According to the Centers for Disease Control, direct medical costs related to falls totaled $179 million for fatal falls and $19 billion for nonfatal falls (Center for Disease Control, September 21, 2007; Stevens et al., 2006). The known estimate of elderly cancer survivors in the United States is over 12 million (Ries et al., 2005). Fifty-six percent of all new cancer diagnoses are among people 65 or older (US Department of Health and Human Services, 2005). Innovations in medical technology have led to earlier diagnoses and improved treatment of cancer (Bender et al., 2008), and consequently, people are living longer and developing chronic conditions. As the elderly population expands, cancer survivorship and the impact of falls and fall injuries will affect families and pervade our economic, health care, and social systems. Review of the Literature The following review of the literature follows the components within the theoretical framework. A review of studies related to each component is in Figure 1 as shown in Chapter 2, starting with social and environmental factors and patient characteristics, then presented from left to right in the model to include biologic factors, medications and treatments, cognition, symptoms, and functional factors. 29 Social and environmental factors. The physical and social environments are both associated with falls in the elderly. In the physical environmental home setting for the elderly, an unsafe physical living environment (Fortinsky et al., 2008; Tinetti, Baker et al., 2008; Tinetti, McAvay et al., 2008) has repeatedly been associated with falls. Additionally, lack of caregiver support (Agostini et al., 2004; Harrison et al., 2001; Leipzig, Cumming, & Tinetti, 1999a; Rubenstein et al., 2002; Stel et al., 2003) are known to increase the risk of falling. Patient characteristics. There are many risk factors for falls. Women are 67% more likely than men to have a nonfatal fall injury (Center for Disease Control, September 21, 2007). Rates of fall-related fractures among the elderly are more than twice as high for women compared to men (Stevens et al., 2006), with 72% of hip fractures in women (Center for Disease Control, September 21, 2007). White women have significantly higher rates of fall-related hip fractures than African American women (Stevens & Sogolow, 2005). There is little difference in fatal fall rates between Caucasians and African Americans from ages 65 to 75. After age 75, Caucasian men have the highest fatality rates from falls, followed by Caucasian women, African American men, then African American women (Center for Disease Control, September 21, 2007). Additionally, among the elderly, non-Hispanics have higher fatal fall rates than Hispanics (Stevens & Sogolow, 2005). Patient characteristics of sex and race or ethnicities are known to influence falls and occasionally fractures in the elderly. However, what is not known is if these patient characteristics alter fall rates in cancer survivors at a higher rate than those without cancer due to the influence of the cancer diagnosis or cancer treatment. Biologic factors. Besides the known risk factors in aging, people with cancer often have multiple risk factors for falls (Holley, 2002; Keating et al., 2005). De-conditioning related to 30 cancer-related fatigue during cancer treatment is known to exacerbate fall risk factors (Braun, Greenberg, & Pirl, 2008; Holley, 2000; Janz et al., 2007; Mock et al., 2005; Phipps, Braitman, Stites, & Leighton, 2008). Furthermore, certain types of cancer, such as lung, are more likely to be related to functional decline than breast or colon cancer, which may lead to falls in the elderly (Bylow et al., 2008; Kanis et al., 1999; Waltman et al., 2006). Advancing stages of cancer are also known to lead to cachexia, to include with loss of appetite, weight loss, muscular wasting, and general mental and physical debilitation, all of which are likely to be lead to functional decline (Given, Given, Azzouz, & Stommel, 2001; Overcash, 2007). Moreover, the combined effect of type of cancer and later stages may influence functional decline leading to falls. In general, fall risk factors in those with cancer are emerging in the literature, and several are associated with fall risk factors in the elderly. However, studies on how a cancer diagnosis or cancer treatment influences the occurrence of a fall or the rate of falls in cancer survivors have not been conducted, and a study of this nature is integral in identifying if a difference exists between elderly with and without a cancer diagnosis. Medications. Certain types of medications (antidepressants, antipsychotics, and benzodiazepines) (Fortinsky et al., 2008; Tinetti, Baker et al., 2008; Tinetti, McAvay et al., 2008) and greater than four medications (Agostini et al., 2004; Harrison et al., 2001; Leipzig et al., 1999a; Rubenstein et al., 2002; Stel et al., 2003) are associated with falls in the elderly. Central nervous system depressants seem to affect balance, sleep patterns, and increase confusion. Polypharmacy has increased drug to drug interactions. Thus, taking certain types of medications or greater than four medications increases the risk of falls. 31 The additive influence of medications with other factors in the elderly may lead to the occurrence of a fall. However, what remains unknown is whether the addition of a cancer diagnosis or cancer treatment interacts with these medications to increase the rate of falls in cancer survivors, and further study is needed. Cancer treatment. As a result of cancer treatments, neurological and nutritional deficits increase the known risk factors for falls (Bennett et al., 2007). Factors influenced by cancer treatments increasing the risk for falls include: neurotoxicity (Holley, 2002; Visovsky, 2006), fatigue (Barsevick et al., 2006; Deimling, Bowman et al., 2007; Luctkar-Flude et al., 2007), depression (Overcash, 2007), postural hypotension (Bennett et al., 2007), hypoesthesia (Bennett et al., 2007), delirium (Pautex et al., 2008), cognitive function (Holley, 2002; Overcash, 2007), pain (Deimling, Bowman et al., 2007), gait and balance problems (Holley, 2002; O'Connell, Baker, Gaskin, & Hawkins, 2007), loss of bone density (Alibhai et al., 2006; Bennett et al., 2007), weight loss (Agostini et al., 2004; Limburg, 2007), reduced muscle strength (Alibhai et al., 2006; Bennett et al., 2007), vitamin D deficiency (Overcash, 2008), loss of physical endurance (Bennett et al., 2007; K. Bylow et al., 2007), and number of medications (Agostini et al., 2004; Holley, 2002). Emerging evidence reveals that reduced cognitive and physical performance occurs after cancer treatment. Further study is needed to directly examine falls in cancer survivors to determine if there is a difference compared to those without cancer; such studies will fill a critical gap in the literature. Frailty. Intrinsic factors such as gait and balance problems (Agostini et al., 2004; Chen et al., 2008; Clough-Gorr et al., 2008; de Rekeneire et al., 2003; Inouye, Studenski, Tinetti, & Kuchel, 2007; Moreland, Richardson, Goldsmith, & Classe, 2004; Sibbritt, Byles, & Regan, 32 2007; Stel et al., 2003; Szabo, Jannsen, Khan, Potter, & Lord, 2008; Tinetti, Williams, & Gill, 2000), cognitive function (Chen et al., 2008; Inouye et al., 2007; Sibbritt et al., 2007), and depression (Anstey, Burns, von Sanden, & Luszcz, 2008; Davison & Marrinan, 2007; Reid, Williams, Concato, Tinetti, & Gill, 2003; Scaf-Klomp et al., 2001) are associated with falls in the elderly. Additionally, poor vision (Szabo et al., 2008), incontinence (de Rekeneire et al., 2003), delirium (Davison & Marrinan, 2007; Inouye et al., 2007), weight loss (Agostini et al., 2004; Chen et al., 2008), peripheral neuropathy (Rubenstein et al., 2002), Parkinson‘s disease (Rubenstein et al., 2002), and stroke (Rubenstein et al., 2002) are associated with falls in the elderly. Certain comorbidities such as arthritis, heart disease, hypertension, and diabetes are known to increase the risk (Baldwin, Klabunde, Green, Barlow, & Wright, 2006; Hewitt et al., 2003; Keating et al., 2008; Klabunde, Harlen, & Warren, 2006; Klabunde et al., 2007; Koroukian et al., 2006; Yabroff et al., 2007; Yancik, 1997; Yancik, Havlik, & Wesley, 1996; Yancik & Ries, 2000). Furthermore, the combination of these frailty factors may lead to a decline in functional status and ability to perform ADLs. Frailty may lead to a decline in function in the elderly, which may lead to falls. However, what is not known is if in cancer survivors, a combination of frailty factors in addition to the cancer diagnosis or cancer treatments creates more functional decline compared directly to those without a cancer diagnosis. Multiple retrospective analyses support the assumption that most falls can be anticipated based on identifiable fall risk factors (Morse, Tylko, & Dixon, 1987). Fall risk factors are well known. However, studies directly examining falls in cancer survivors compared to those who do not have a cancer diagnosis do not exist, revealing a significant knowledge gap in the literature. Moreover, further study directly examining the difference in clinical presentation of frailly in cancer survivors is needed to determine if this should be a focus 33 for clinicians. Further discussion will occur regarding these limitations and gaps in the next section. Limitations and Gaps in the Literature Limitations of past cancer survivor research on falls is threefold. Primarily all studies occurred at the patient level examining a very small number of patients with similar types of cancer or cancer stage limiting generlizability of findings. Second, the study samples were limited to inpatient acute care hospital settings. Finally, no direct comparison of those with and without cancer occurred. Some of these gaps will be addressed by this study, while others must be addressed in future studies. If falls, fall sequelae, and health care use are found to be more probable in cancer survivors, then fall prevention interventions will become more important as the elderly population with cancer increases in size and health care costs continue to rise. This research will identify if survivors of particular cancers or if later stages of cancer are in greater need of these interventions than others, so that scarce resources could be tailored to patients most in need of care. Supplementary information will also be provided on the characteristics of the HCBS population. The following section will provide a discussion on what clinicians can do to reduce the risk of falls further supporting the importance of the research in this study. Interventions Interventions to reduce the risk of falls and fall-related fractures are emerging in the literature. International fall expert Tinetti (Tinetti, 2003; Tinetti & Williams, 1998) has focused on the development of fall assessment tools (Abruzzese, 1998), the relationship between number of medications and weight loss or impaired balance (Agostini et al., 2004), disability difficulty and interference leading to falls (Gill, Robison, & Tinetti, 1998; Koch, Gottschalk, Baker, 34 Palumbo, & Tinetti, 1994), modifiable disability impairment identification (Tinetti et al., 2005), the impact of chronic dizziness (Tinetti et al., 2000), the effect of medications on falls (Leipzig et al., 1999a), the effect of depressive symptoms on falls (Reid et al., 2003), and environmental hazards (Gill et al., 2000). Efforts have been multidisciplinary (Brown, Gottschalk, Van Ness, Fortinsky, & Tinetti, 2005) and focused on reducing injurious falls (Tinetti, Baker et al., 2008) and has led to the development of international evidence-based guidelines for fall prevention (Rubenstein et al., 2002). Efforts are concentrated on fall risk assessment and management in clinical practice (Fortinsky et al., 2004; Tinetti, 2003), multifactoral approaches to prevention strategies (Tinetti, 2008; Tinetti et al., 1994), and dissemination of evidence-based management strategies to reduce falls (Baker et al., 2005; Fortinsky et al., 2008). Once the incidence of falls in cancer survivors is better understood, we will recognize the importance of integrating evidence-based interventions to decrease falls into gerontological nursing practice (Hurria et al., 2007). Multifaceted fall interventions in the frail and elderly with multiple comorbidities are not the same as drug trials with single conditions and tight exclusion criteria (Cameron et al., 2005; Fogarty et al., 2002; Gillespie et al., 2003 ; Haines, Hill, Walsh, & Osborn, 2007; D. Oliver, 2008). Developing multi-modal fall interventions tailored to the most common fall risk factors has been effective at reducing fall rates (Fortinsky et al., 2004; Tinetti, 2003). Falls are usually a result of a synergistic interaction between an individual‘s intrinsic risk factors, the physical environment (home and equipment), and the riskiness of a person‘s behavior (D. Oliver, 2007). Systematic reviews (Myers, 2003; Oliver, Daly, & Martin, 2004; Scott, Votova, & Scanlan, 2007) have shown that the following factors are consistently identified in patients who fall: recent fall, postural instability, muscle weakness, behavioral 35 disturbance/agitation/confusions, urinary incontinence/frequency, and prescription of ―culprit‖ drugs, postural hypotension, or syncope. Additionally, comorbidities, poly-pharmacy, delirium, frailty, and functional impairment are highly prevalent and coexisting in the elderly (Fried et al., 2004; Leipzig et al., 1999b; Tinetti et al., 2005) and are known to increase falls. Finally, the occurrence of falls is usually a sign of underlying frailty or a medical illness, or a change in functional status, all of which merit further investigation (Clough-Gorr et al., 2008; Fried et al., 2004). Fall incidents should therefore be used to prompt reassessment of the underlying cause, with subsequent intervention implementation to prevent further falls. The balance and emphasis of fall prevention programs should be tailored to the case-mix of the population and secondarily to an individual‘s risk profile (Oliver, 2007). It may be more effective to focus on employing a multi-modal intervention for each common risk factor, bearing in mind that half of all falls are people who have already fallen, and modifying interventions after reassessment (Oliver, 2007). Interventions are effective at reducing the risk of falls, and nurses can influence this health outcome by taking action to prevent falls in cancer survivors. Conclusion and Significance This study will begin to fill some gaps in the cancer survivor literature to date and advance the science of cancer research in several ways. First, the study will begin to examine if a cancer diagnosis alters the rate of falls, fall sequelae, and health care use in frail communitydwelling elderly cancer survivors compared directly to those without cancer. By studying this directly, this research should be able to begin to address another limitation of past cancer survivor research, the paucity of studies investigating functional limitations in cancer survivors. Second, the study design and analytic technique allows for an accounting of the multiple variables, which has been a limitation in most studies of falls in cancer survivors to date. Third, 36 this study begins to examine frailty variables in elderly cancer survivors, directly comparing cancer survivors to those without cancer to evaluate if a difference exists. Finally, this study will fill a gap in the literature describing the characteristics of the HCBS population, as there is a paucity of information on patients in the HCBS community-dwelling setting. Although this study shares some limitations with other cancer survivor research on falls (i.e., a descriptive design), the previous sections illustrate ways in which the study will begin to fill gaps in the literature and advance the science of cancer survivor research. Both gaps in past cancer survivor research on falls, as well as the results from decades of studies on falls, were used to develop this research study. Of all types of accidents, falls pose the most serious consequence related to reducing physical functioning and quality of life in elderly cancer survivors (Bandeen-Roche et al., 2006; Dunn, Rudberg, Furner, & Cassel, 1992; Guilley et al., 2008; O'Connell, Wellman, Cockayne, & Baker, 2005; O'Connell et al., 2007; Tinetti, McAvay et al., 2008). The multi-factorial nature of falls requires coordination and a multifaceted approach in gerontological nursing that does not adhere to the traditional disease model driving most health care (Inouye et al., 2007). Elderly cancer survivors experience falls, fall sequelae, or health care use, with a unique set of consequences as expected in this study, and then a convincing case could be made for further nursing research to understand the influence of falls on cancer survivors. Furthermore, attention can be focused on fall risks and fall prevention strategies during the process of Cancer Survivor Care Planning (Institute of Medicine, 2005). Translational research could begin to identify a mechanism for gerontological nurses to refine and standardize approaches to fall-risk assessment (Chen et al., 2004; Goodwin, 2007; Hurria et al., 2007) and to integrate behavioral and/or psychological interventions to prevent the occurrence of falls into practice (Cimprich et 37 al., 2005; DeSanto-Madeya et al., 2007), reducing adverse outcomes (Chen et al., 2004; Cope & Reb, 2006; Knobf, Musanti, & Dorward, 2007). What follows in the next chapter is the methodology for this study and the impact if these research questions are answered. A detailed review of the sample and measures used in this study will be presented followed by a presentation of the methods and data analysis plan. 38 CHAPTER 4: METHODOLOGY In Chapter 4, the methodology for this study is presented and the design is described. The operational definitions as well as the measurement of key variables are presented. In the following sections, the target sample is explained, followed by the power of the study and study procedures. Key issues in data collection, including data management are discussed, and the methods employed to protect the rights of human subjects are explained. In the following section, the research questions are reiterated to set the stage for the methodological approach used in this study. Research Questions Research question 1. After adjusting for sociodemographic characteristics (age, sex, race/ethnicity), medications, frailty (ADLs, cognition, comorbidities, pain, weight loss, and vision), determine the extent to which: patients with a cancer diagnosis experience a greater number of falls, sequelae of falls (fractures), ER use, hospitalization, or nursing home placement following the cancer diagnosis compared with those patients with no diagnosis of cancer; and if there were differences in the number of falls among cancer patients in the year following diagnosis according to site of cancer (breast, colon, prostate, lung, or other), stage of cancer (IIV), or cancer treatment (chemotherapy and/or radiation). Research question 2. After adjusting for sociodemographic characteristics (age, sex, race/ethnicity), medications, frailty (ADLs, cognition, comorbidities, pain, weight loss, and vision), examine if the effects of frailty variables (ADLs, cognition, comorbidities, pain, weight loss, and vision) on falls in the year after diagnosis are differential with respect to site of cancer (breast, colon, prostate, lung, or other), stage of cancer (I-IV), or cancer treatment (chemotherapy and/or radiation). 39 Exploratory research question 3. To explore and describe the social environmental status of the sample to better understand the context of the population. These factors included marital status, living environment, whom the patient lives with, time alone per day, feelings of loneliness, and pets. Design This was a longitudinal retrospective cohort (Yabroff et al., 2007) study examining whether a cancer diagnosis alters the rate of falls, fall sequelae, and health care use in elderly community-dwelling cancer survivors; examining frailty variables in cancer survivors; and describing the social factors in this population. The study focused on falls, fall sequelae (fractures), and health care use (ER, hospitalization, and nursing home) over a period of a year; and in those diagnosed with cancer after January 1, 2000. The study also focused on persons with all types of cancer compared to those without cancer, after controlling for confounding variables known to influence falls (Cesari et al., 2002; Chen et al., 2008; Coussement et al., 2008; Davison & Marrinan, 2007). This study linked three sources of data previously collected, the Minimum Data SetHome Care (MDS), the Michigan Medicaid program submitted claims (bills), and the Michigan Cancer Registry. Connecting these datasets resulted in a unique population-based data resource that can be used by health service researchers to perform a wide variety of outcomes research. Each data source will be described individually in the following section. Sample Description of data sources. The MDS is a person-centered assessment with uniform standards for collecting minimum essential nursing data (Werley & Lang, 1988), enabling clinicians to assess multiple domains (Forsyth, 2000). The MDS in the HCBS program was 40 developed in 1993, a modified version of that used in the nursing home, to inform and guide comprehensive care and service planning for community-dwelling elderly (InterRai, 2008). The screening portion of the instrument covers key domains of service use (hospital, home care, hospice, therapy), function (ADL and IADL), health (physical, psychological, emotional), and social supports (family, friends, community) (Hirdes et al., 2004; Morris et al., 1997). The second portion of the MDS provides a standardized mechanism to identify persons who could benefit from further evaluation of a specific problem or risk for functional decline over time to support care planning (French et al., 2007; Harrison et al., 2001; Hirdes et al., 2004). The information on the MDS contains a combination of self-report by the patient and clinical validation by a registered nurse, which is collected in person in the patient‘s home upon entry into the HCBS program, and then every 180 days thereafter, or more often if the patient has a condition change (CMS, 2008). Due to the potential for cognitive impairment in this sample, patients‘ self-report is a limitation in this study. Cognition was evaluated for the group with cancer and the group without cancer. Cognition for each group was similar, or equally impaired, allowing for a comparison of like groups. The MDS validity and reliability is reported in an international trial, with independent dual assessment of 241 patients using 780 assessments, finding Kappa of .74 (Morris et al., 1997). The average inter-rater weighted Kappa reliability of the item falling on the MDS was a Kappa of .81, compared to the MDS used in the nursing home setting, which was a Kappa of .69 (Morris et al., 1997). The second data source for this study was Medicaid claims files. Medicaid claims data are accrued as services are utilized. The files are inclusive of paid claims submitted by providers of services to account for hospital, home care, HCBS, hospice, physicians, and outpatient 41 services (e.g., medications). HCBS patients in this investigation are dually eligible for Medicare and Medicaid, with Medicare as the first payer of record. The final data source for this study was the Cancer Registry. The Cancer Registry in the State of Michigan is collected and housed by the Department of Community Health and linked to the National Program of Cancer Registries. The Registry activates when an individual is diagnosed with cancer and includes data on histological determination and stage. The Cancer Registry began in 1985, and data is collected by Tumor Registry staff at hospitals across the state. The linked datasets provide a unique source of population-based data. Next, the study population will be described. Setting. Individuals in the State of Michigan Medicaid HCBS program are 65 years of age and older, low income below 300% of the Federal Poverty Level, and living in the community in a home-type setting. Additionally, each individual must meet State of Michigan nursing facility level of care requirements, which includes a combination of assistance required for ADL and IADL needs (CMS, 2008). The ADL needs include the things people normally do in daily living, including any daily activity performed for self-care, such as feeding themselves, bathing, dressing, grooming, or eating (CMS, 2008). The IADL needs comprise daily activities that allow an individual to remain living in the home or community setting, such as doing housework, preparing meals, taking medications, shopping, using the telephone, and managing (CMS, 2008). These combined factors determine if an individual is eligible for the nursing facility level of care requirements, thus eligible for services in the HCBS program. 42 Sample. The individuals in this study are community-dwelling elderly in the HCBS program in Michigan. The HCBS program offers an array of services that assist Medicaid beneficiaries in the community to avoid institutionalization (CMS, 2008). Services are provided in conjunction with informal supports (family, friends, and community) and include case management, personal care, homemaker, snow removal, lawn care, transportation, homedelivered meals, respite, equipment, counseling, and adult day care. The sampling frame or eligible subjects are all those participants enrolled in the HCBS program between January 1, 2002, and December 31, 2007. The Michigan Department of Community Health estimates that 20,000 people over 65 years of age participate in the HCBS program, with approximately 15% of those having cancer (Mc Nab, March 24, 2007). This study was a convenience sample, of data collected previously for a different use, and certain bias may occur. In a sample of dually eligible Medicare and Medicaid recipients who are low-income and vulnerable, generlizability may be a challenge. However, elderly who are dually eligible consume the greatest proportion of health care dollars and remain understudied, due to the challenge of recruiting low-income dually eligible recipients to participate in research. For this study, two distinct groups of patients were identified. The first group is those who have a cancer diagnosis. The second group is those without a cancer diagnosis. The cancer diagnosis was confirmed by using the data in the Cancer Registry, assuring that the inclusion criteria are met. The plan was to examine four MDS assessments were evaluated for each patient in this study. The collection of these MDS assessments will have occurred over a 2-year time period. For the group with cancer, we will evaluate two MDS assessments occurring prior to the diagnosis of cancer, as well as two assessments after the diagnosis of cancer, for a total of 4 assessments. For the group without cancer, four assessments will be evaluated, and then the two 43 groups will be compared. This will answer research question 1, examining if cancer survivors fall, have fractures, and have ER, hospital, or nursing home use at a higher rate than those without a cancer diagnosis. The answer will provide direction on whether clinicians need to focus on fall risk in cancer survivors. Then the frailty variables will be examined in the cancer group to identify if there are specific factors that may contribute to or influence the occurrence of a fall or fracture in this sample. This will answer research question 2, examining frailty factors in cancer survivors. The answer to this question will provide direction to practitioners on which fall risk factors must be included in assessment and, moreover, what types of patients have risk factors may lead to falls. Furthermore, descriptive statistics will be used to analyze social factors in this population. This will provide new information to fill a gap in the literature. Measures The operational definition and the level of measurement for variables of interest in this study are each described. In addition to the description, the variables, data source, format, and codes are listed in Appendix A (Table 20 independent variables, Table 21 dependent variables, and Table 22 the social factors). Social and environmental factors. The social and environmental factors in this study included marital status, type of living environment, caregivers, and the amount of time alone during the day. The marital status variable is coded categorically as 1 for married, 2 for divorced, and 3 for other. Type of living environment is coded 1 for a home or apartment and 2 for any other type of living environment. Caregivers are coded categorically as 1 for spouse, 2 for child, and 3 for any other type of caregiver. The amount of time alone during the day is coded 44 categorically as 0 for never or hardly ever alone, 1 for much or most of the day. Each variable is listed in Appendix A (Table 22). Characteristics: age, sex, and race/ethnicity. The individual characteristics included in this study are age, sex, race and ethnicity. Age is the date of birth and is a continuous measure. Sex is the gender, as male or female. The sex variable is coded categorically as 0 for male and 1 for female. Race or ethnicity as defined by the United States Census Bureau are the selfidentification of the race that a person most closely identifies with and an indication of whether or not they are of Hispanic or Latino origin (ethnicity). Race and ethnicity is coded categorically as 1 for Caucasian, 2 for African American, 3 for American Indian, 4 for Hispanic, and 5 for other. Biologic factor of interest in this study: cancer. There are two levels of measure for cancer. This includes the presence of the cancer as verified in the Cancer Registry by type or no presence of cancer. The diagnosis of cancer is defined as the presence of biopsy-confirmed cancer (with the exception of skin cancer), as provided by the Cancer Registry (Koroukian et al., 2006; National Cancer Institute, 2008) and categorized by site. The lack of the presence of cancer was coded categorically in this study. The number 0 means a cancer diagnosis is not present in the individual. Confirmed cancer was coded as 1. Cancer types were also identified and labeled. Stage of cancer is defined as a descriptor used to explain how much the cancer has spread, taking into account the size of the tumor, how deeply it has penetrated, whether it has invaded adjacent organs, how many lymph nodes it has metastasized to (if any), and whether it has spread to distant organs. Staging will be numerically coded 1 to 4 to match the staging codes of I through IV. 45 Cancer treatment covariates. Cancer treatments are chemotherapy or radiation. Chemotherapy is defined as treatment with drugs that kill cancer cells. Radiation treatment is defined as the use of high-energy radiation from X-rays, gamma rays, neutrons, protons, and other sources to kill cancer cells and shrink tumors. The lack of any treatment will be coded as 0; chemotherapy will be coded as 1; radiation will be coded as 2; and chemotherapy and radiation combined will be coded as 3. Medication covariates. In this study, the medications are prescriptions to include four classifications: antidepressants, antipsychotics, anti-anxiety, or hypnotics (ASCO, 1995). The lack of presence of any of the four listed prescriptions will be coded as a 0 and the presence of any one or more of the selected prescriptions will be coded as 0 for not taken and 1 for taken. Frailty. Each of the frailty variables will be defined. Cognition is defined as the mental capacity for processing information, applying knowledge, and changing preferences. Incontinence is defined as the involuntary leakage of urine. Depression is defined as a state of low mood or feelings of sadness. ADLs are defined as the daily activities performed for selfcare, such as bathing, feeding oneself, dressing, and grooming. Weight loss is defined as the loss of weight greater than 5 pounds in the past 6 months. Vision is defined as the ability to interpret information and surroundings from visible light that reaches the eye. Frailty is operationalized in this study as the variables on the MDS to include: cognition (decision-making, short- and long-term memory recall, change in mental function, or disorientation), incontinence, depression, ADLs (bathing, eating, toileting, transferring, locomotion, stair-climbing, and dressing), weight loss, and vision. Specific independent variables, type, source of data, format of data, and data codes are identified in Appendix A (Table 20). 46 Comorbidity covariates. The covariate comorbidities are discussed individually due to their complexity. Comorbidities are the concurrent presence of a minimum of one medically diagnosed disease of arthritis, coronary artery disease, diabetes, congestive heart failure, arthritis, or cerebral vascular accident in the same individual (Klabunde et al., 2007; Koroukian et al., 2006). The lack of presence of comorbidity from the select list will be coded as 0 and the presence of one or more of the comorbidities will be coded from 1 to 6 and counted in a manner as a cumulative effect of the number of comorbidities. Health outcome variables. Health outcome variables in this study include falls, fractures, ER use, hospitalization, and nursing home placement. The following is a description of each. Appendix A (Table 21) is a complete listing of the dependent variables in this study. Falls are measured at every 180-day interval when the registered nurse completes an MDS assessment. Falls will be coded categorically, beginning with 0 for no falls, followed by a simple count of falls from 1 to 9, with each number equaling the number of falls. If greater than nine falls occur, then the number 9 is coded. The lack of presence of a fracture will be coded categorically with the number 0 for no fracture, followed by a code of the number 1 for a fracture. Health care service utilization in this study includes ER, hospitalization, and nursing home placement, the most frequently used types of services. ER use and hospitalization are coded by use in the past 6 months. Nursing home placement is identified by location of Medicaid services in the level of care dataset by date of placement. Analytic Sample Inclusion criteria. Inclusion criteria comprised subjects who are 65 years of age and older and are enrolled in the HCBS program (meeting HCBS program eligibility requirements) between January 1, 2002, and December 31, 2007. Additionally, inclusion of subjects included 47 those who have had at least four MDS assessments administered between those dates over a 2year time period. Those diagnosed with cancer on or after January 1, 2000, were also included in this study. Exclusion criteria. Exclusion criteria included those not participating in the HCBS program, members of health maintenance organizations, individuals in the nursing home, and participants under 65. Additionally, exclusion comprises subjects with cancer who are diagnosed with or are deceased from cancer during the 12-month time period and have skin cancer. This will be discussed later as a limitation in this study. Gender and minority inclusion. A person‘s gender is included in both the conceptual framework and statistical models of the research. Males and females age 65 and older in the HCBS program during the study period were examined during this study. Inclusion of both males and females is necessary because both participate in the HCBS program. MDCH Longterm Care Division has reported that a greater number of females than males participate in the HCBS program because females live longer. Consequently, there are more women than men within the study. A person‘s race is included in both the conceptual framework and statistical models of the research. The State of Michigan collects data on race in the MDS utilizing standardized federal guidelines. Peoples of all races participate in the HCBS program during the study period and will be examined during this study. The Michigan Department of Community Health Longterm Care Division has reported that a greater number of African Americans than usual in the general population participate in the HCBS program due to the nature of the eligibility criteria. Because of this, there was a higher percentage of African Americans within the study, but this factor can be considered a strength of the study as the African American population is 48 understudied. Data collection. This is an analysis of previously collected data and relied on the inclusion and exclusion criteria implemented for each program. The MDS for this sample is collected by registered nurses (RNs) in HCBS provider agencies. The RN interviews the patient and his or her caregiver during a home visit while simultaneously conducting an assessment and completing the MDS form. The MDS is completed every 6 months or more often as the patient‘s condition warrants. The Medicaid Claims data is a compilation of paid claims for service providers for the time period of January 1, 2001, to December 31, 2007. The Medicare Part D pharmacy data is a compilation of paid claims for services for the time period of January 1, 2005, to December 31, 2007, to account for the new pharmacy program and obtain information for chemotherapy that may have been administered during that time period. The Cancer Registry Data are collected by the State of Michigan from hospitals and labs, physicians and clinics, other registries, and death certificates and has been available since 1985, stored in a data warehouse by the State of Michigan. Data are organized categorically with the date of birth. Procedures. A Data Use Agreement with the Department of Community Health at the State of Michigan was completed requesting the data for this study and included: the MDS; Medicaid claims files; and the Cancer Registry, as well as a Data Use Agreement with the Center for Medicare and Medicaid Services for Medicare Part D pharmacy files. Institutional Review Board (IRB) approval was obtained from both the State of Michigan and Michigan State University (MSU). Patients for this study were identified based on the inclusion and exclusion criteria. 49 The data project manager for the Institute for Health Care Studies at MSU obtained the data from the State of Michigan data warehouse in September 2008. The data sources were then linked using the State of Michigan unique patient identifier numbers, and then sensitive identifiers were removed. The data are stored in multiple sets of data files organized by section on the MDS form in a secure password-protected location at MSU College of Nursing. A code book was developed and organized by section on the MDS form corresponding to how the data are stored for easy retrieval, defining each item for this study. The codebook was developed in conjunction with the data project manager for the Institute for Health Care Studies at MSU. The accuracy of the code book was reviewed and verified by a statistician in the MSU College of Nursing. Quality control and data management. Quality control and data management integrity complied with the MSU Research Committee and MSU Data Management Safety Protocol (Michigan State University [MSU] IRB)). The data project manager for the Institute for Health Care Studies at MSU obtained all datasets and sensitive identifiers were removed prior to receipt of the data for this study. Additionally, the project manager for the Institute for Health Care Studies at MSU verified the accuracy of the de-identified data prior to release of the data to the investigator. This study did not involve any original data collection or contact with any study subjects. All information is available in existing databases that are publicly available and de-identified. Confidentiality of all Medicaid recipients was protected at all times. Data were maintained in a password-protected computer drive accessible to the primary investigator and statistician in this study. 50 Due to the large size of the dataset, the information was stored in the MSU College of Nursing password-protected warehouse and used as needed for analysis. This data warehouse is a secure password-protected server that is inaccessible via the Internet. Data Analysis Plan The statistical software SAS, 9.1 was used for data analysis and data management. The SAS software has the ability to create subsets of the data to analyze of specific variables for this study, simplifying data management of the large datasets. The major focus of this study was on the differences in the outcome variables of falls and fractures between the cancer and non-cancer populations, and among cancer patients by type of cancer, cancer stage, and cancer site, after controlling for all measured confounders. As an initial step, basic descriptive statistics were computed for variables of interest, including frequency distributions, measures of central tendency, skewness, and variability for the year before and year after the cancer diagnosis. Outcome measures included falls, fractures, ER use, hospitalizations, and nursing home placement. Covariates included demographic characteristics, medications, cancer treatment, comorbidities, and frailty. The next step involved the covariates (age, sex, race or ethnicity), medications, cancer treatment, comorbidities, and frailty (ADLs, cognition, depression, incontinence, weight loss, and vision), which were used to build regression models to better understand the influence of each variable on the outcome measures. Candidate variables that were considered for inclusion in each model include patient characteristics, cancer characteristics, and condition/treatment parameters. Patient characteristics included age, sex, and race or ethnicity. Cancer characteristics included stage and type of cancer. Condition or treatments included comorbidity conditions and 51 medications and cancer treatment such as chemotherapy or radiation. The race, sex, ADLs, cognition, depression, vision, weight loss, and incontinence were obtained from the MDS. The cancer diagnosis time period since date of diagnosis, stage, and type of cancer were obtained from the Cancer Registry. Comorbidity conditions were obtained from the MDS and Medicaid claims files. Medications, age, and fractures were obtained from the Medicaid claims files. Cancer treatments were obtained from the Medicare and Medicaid claims files. Appendix A (Table 20, 21, and 22) includes a list of each of the variables. Analytic techniques. Generalized Estimating Equation (GEE) for binary (whether the patient falls or has fractures), and Poisson outcome (number of falls or fractures during each measurement times), were used to carry out the analyses. GEE has become popular in all longitudinal studies in general and is especially widely used in analyzing categorical data (Ballinger, 2004). Longitudinal logistic regression analysis was implemented via GEE technique (Twisk, 2003). Generalized linear models with Poisson (Chelly et al., 2008; Fried et al., 2006; Walke et al., 2004) distributed errors were used to fit the statistical models in this study (McCullagh & Nelder, 1989; Wolfinger & O'Connell, 1993). A preliminary study found that a Poisson distribution provided a good fit for the dependent variable (falls) (Spoelstra, 2008). Since outcome variables are either binary or right-skewed count data with repeated measures, the ordinary linear regression models with normal assumption are not appropriate to be applied here. First the generalized linear model is specified as defined by Nelder and Wedderburn (1972), which allows the mean of a population to depend on a linear predictor through a nonlinear link function (logit for binary outcome; log for Poisson outcome). Second, 52 GEE technique allows to account for the correlation of multiple measurements with the same patients by the method Zeger and Liang (1986; 1988). For missing data, GEE can estimate the working correlation using all available pairs‘ method, in which all non-missing pairs of data are used in the estimators of the working correlation parameters. Kappa-like statistics were used to assess the model fit for the GEE categorical response models. The kappa-like statistic takes on a value of 0.0 for the interceptonly model and a value of 1.0 for the saturated model (Zeger & Liang, 1986; Zeger et al., 1988) GEE models are fit with main effects terms only, using Kappa-like statistic with .00-.20 poor fit, .21-.40 fair fit, .41-.60 good fit, and >.61 excellent fit. The focus was on estimating the average response over the population rather than the regression effect on a given individual. Analysis plan for specific aims. The following was the specific analysis plan for each aim. Aim 1. After adjusting for sociodemographic characteristics (age, sex, race/ethnicity), medications, frailty (ADLs, cognition, comorbidities, pain, weight loss, vision), determine the extent to which: patients with a cancer diagnosis experience a greater number of falls, sequelae of falls (fractures), ER use, hospitalization, or nursing home placement following the cancer diagnosis compared with those patients with no diagnosis of cancer; and if there were differences in the number of falls among cancer patients in the year following diagnosis according to site of cancer (breast, colon, prostate, lung, or other), stage of cancer (I-IV), or cancer treatment (chemotherapy and/or radiation). The primary aim of this analysis was to compare those with and without cancer in the year after diagnosis, to determine if a difference exists. Repeated measurements or observations of the outcome variable (Y [falls and fractures]) occurs on four occasions, thus a random vector 53 of responses for the ith individual (i=1, 2,…,N [subjects in the study]) at jth time (j=1 to 2). A correlation matrix was specified to account for the association of meaning for the same patient at different times. The following statistical model was implemented: the number of falls in the two 6-month time periods after diagnosis = (cancer or no cancer + [plus] covariates [age, race, sex, frailty, medications, and comorbidities]) described above. The essential parameter is associated with the coefficient for the group variable (cancer or no cancer). The test of the significance of this parameter was used to determine if the difference between groups existed. Aim 2. The second aim examined if the effects of frailty variables (ADLs, cognition, comorbidities, pain, weight loss, and vision) on falls in the year after diagnosis are differential with respect to site of cancer (breast, colon, prostate, lung, or other), stage of cancer (I-IV), or cancer treatment (chemotherapy and/or radiation). The primary aim of this analysis tested for the interaction between frailty and cancer on falls. The following statistical model was employed for those with a cancer diagnosis: falls in the 12 months after diagnosis = falls in the 12 months prior the cancer diagnosis + (plus) frailty + (plus) site of cancer + (plus) frailty time‘s site of cancer + (plus) other covariates. The essential parameters investigated for this aim are associated with the interaction term in the statistical model. When the interaction was not significant, it was removed and the additive effects of the site of cancer and frailty were explored. Aim 3. The social environmental status factors were examined using descriptive statistics. Chi-square or t-tests of differences are also conducted. This was used to better understand the population. 54 The overarching goal was to identify if the falls and fractures are higher with the addition of a cancer diagnosis and among those with cancer; and if variation existed by stage or site (Bender et al., 2008). The alternative methods identified prior to analysis are as follows. Alternative methods. For analysis of longitudinal repeated-measures data, the three statistical methods included the random effects model, fixed effects model, and GEE (Gardiner, Luo, & Roman, 2009). An alternative to GEE modeling is generalized linear mixed effect modeling where associations between repeated measures within subjects are modeling by inclusion of the random effect (of the subjects). Under full likelihood specification for linear models, both subject-specific and population-average estimates of the β coefficients are possible. Additionally, various correlation structures (e.g., auto-regression of order 1, exchangeable) were considered. Power analysis. A power analysis (the probability of rejecting the null hypothesis when the hypotheses should be rejected [avoiding Type II error]) was conducted to assure that when a statistical difference was found between the cancer and non-cancer group that the difference actually existed. The power analysis used was based on a-priori analysis for regression by using predictor variables, as discussed by Cohen (Cohen, 1988). Power analysis was performed using findings from a preliminary study (Spoelstra, 2008). In total there were 7,448 in the sample; 967 were diagnosed as cancer patients, and 6,481 as non-cancer patients. The proportion of the noncancer group compared to the cancer group was 6.7 in the sample. For those with a cancer diagnosis, 363 (3.8%) reported they had a fall compared to 1,762 (27%) in the non-cancer group. With G Power International Version statistical power analysis program (Erfelder, Faul, & Buchner, 1996) with level of significance set at 0.05, power set at 0.80, finding the required sample size for the cancer group was 165 and for the non-cancer group was 1,106. Thus, at a 55 minimum, a total sample size of 1,271 was needed for this study to detect if there was a statistical difference between those with a cancer diagnosis and those without cancer. However, a statistical difference is more than likely to occur since such a large sample size was used in this study. Thus, the magnitude of the difference must be evaluated by examining the effect size, to determine if a meaningful difference between those with a cancer diagnosis and those without cancer is present. Effect size. To know if the observed difference is not only statistically significant but also important or meaningful, the effect size was calculated. The effect size is a measure of the strength of the relationship between the variables. In this study, a small effect size was expected with near certainty due to the large sample size. According to Cohen, the 0.16 value found regarding falls in this study is within the moderate range of ES for multiple regression analysis (<0.1 trivial ES; .1 - .3 small ES; .3 - .5 moderate ES; .5 > large ES) (Cohen, 1988). A small to moderate ES finding would be consistent with other findings in psychosocial studies in adult cancer patients (Rehse & Pukrop, 2003). A small to moderate ES of .1 to .5 for falls or fractures in this study would be a clinically important difference in the community dwelling elderly who already fall at a high rate. Even a small ES would be considered to be clinically significant, and should demand attention from clinicians caring for cancer survivors. Strengths and Limitations A major strength in this study was verifying the diagnosis of cancer and using self-report for the outcome measurement. This was conducted by cross-checking claims-based indications of cancer and the cancer registry. It was hypothesized that the most likely missed cancers were those not under active treatment and therefore another limitation is the inability to include cancer treatment in the modeling. Furthermore, the longitudinal nature of the data was strength. 56 Additionally, cancer diagnosis was known from the Cancer Registry, but it may be reoccurrence of a primary or secondary cancer. A major limitation in this study was the ability of an elderly individual to recall a fall event. As stated in the literature, the elderly are able to recall the occurrence of a fall for approximately 3 months; however, the assessment occurs every 6 months, which may lead to underreporting falls. Cognitive impairment and recall ability were expected to be equally present in the group with cancer and the group without cancer. The analytic technique included examination of cognitive impairment in both groups, those with or without cancer, and all analysis included cognition as a time-varying covariate to adjust for possible recall bias. Postdiagnosis cognitive impairment may be greater than at the time of diagnosis in the cancer group, but was controlled for by inclusion of the appropriate variable as a time-varying covariance. An examination of the distribution of cognition variables was conducted in the cancer group compared to those in the non-cancer group and a determination was made that is was not necessary to control for differences in this sample. Another limitation was the unknown severity of comorbidities. The analytic technique included examination of comorbidities in both groups, those with or without cancer, and all analysis included comorbidities as a time-varying covariate to adjust for possible differences. Furthermore, those who are deceased within 12 months are excluded, which may lead to an underrepresentation of advanced-stage cancers, most likely lung cancer. One final limitation was the potential influence of chemotherapy on cognition. There was no opportunity to control for this limitation in this study, and it is simply acknowledged. Protection of Human Subjects Institutional Review Board. Approval of the research study was received from the IIRB 57 prior to obtaining the dataset (IRB# X08-768M/ APP# i031013). The study aims were modified and resubmitted to the IRB in January 2010 and approval was again obtained (IRB# X x10067M/ APP# i034918). The investigator adhered to and maintained the human subject‘s certification provided by the University Committee on Research Involving Human Subjects. Potential risks and protection against risk. The human subjects in this research are those individuals receiving services in the HCBS program. Since the study was a secondary data analysis and did not involve subjects in an investigational intervention, patients are placed at minimal risk. All datasets obtained from the data project manager for the Institute for Health Care Studies MSU were de-identified prior to our receipt of the data. This study did not involve any original data collection or contact with any study subjects. All information is available in existing databases that are publicly available. Confidentiality of all Medicaid recipients was protected at all times and data was de-identified prior to receipt. Patient and other anticipated benefits. Participants are unlikely to directly benefit from participation in this study. A secondary data analysis has the potential to lead to indirect benefits that individuals in this study may experience in the future. The potential gain of the anticipated benefits from the research study significantly outweighs any possible risk. What follows in the next chapter are the results from this study. Any modifications to the analysis will be explained in the results section. The descriptive data are reported, followed by preliminary work, and each research aim is then answered. 58 CHAPTER 5: RESULTS This descriptive study secondary analysis was designed to determine if communitydwelling elderly cancer survivors experience the influence of the disease or treatment of cancer, and have increased falls and fall sequelae with subsequent increased use of health care. Furthermore, this study sheds light on whether the cancer stage or site alters the rate of falls and fractures in the elderly. This study answered the following questions. Research Question 1 After adjusting for sociodemographic characteristics (age, sex, race/ethnicity), medications, frailty (ADLs, cognition, comorbidities, pain, weight loss, and vision), to determine the extent to which patients with a cancer diagnosis experience a greater number of falls, sequelae of falls (fractures), ER use, hospitalization, or nursing home placement following the cancer diagnosis compared with those patients with no diagnosis of cancer. Furthermore, if there were differences in the occurrence of falls among cancer patients according to site of cancer (breast, colon, prostate, lung, or other), stage of cancer (I-IV), or cancer treatment (chemotherapy and/or radiation). Research Question 2 To examine if the effects of frailty variables (ADLs, cognition, comorbidities, pain, weight loss, and vision) on falls in the year after diagnosis are differential with respect to site of cancer (breast, colon, prostate, lung, or other), stage of cancer (I-IV), or cancer treatment (chemotherapy and/or radiation). 59 Exploratory Research Question 3 The social environmental status is described to better understand the context of the population. This includes marital status, living environment, with whom the patient lives with; time alone per day, and feelings of loneliness. This study used data from the State of Michigan HCBS program from the MDS assessments during the time period of 2002 to 2007, combined with information from a cancer registry and claims files. First, preliminary research, data management, and preliminary work conducted prior to performing the analysis are presented. This will be followed by descriptive analysis. Finally, a discussion of the study results will be presented by research question. Preliminary Research Prior to completing this study, three retrospective, cross-sectional research studies examining the community-dwelling elderly 65 and older in the HCBS program using the MDS assessment during the time period of 2002 to 2007 were conducted. The first analysis examined 7,448 and found no difference in falls between cancer and non-cancer groups (p > .05) (Spoelstra, Given, von Eye, & Given, 2009, 2010b). The second analysis of 6,912 found that fall risk had a strong positive correlation with age and cognition, moderate correlation with pain, depression, IADL, ADLs, and strong negative correlation with pain and cognitive status, age, and depression; however, no association was found between cancer and falls (Spoelstra, Given, von Eye, & Given, 2010a). The third analysis examined interactions among falls and medications in 13,751 finding the number of medications or the use of antidepressants, hypnotics, or psychotics was associated with increased falls (p = .00) (Spoelstra, Given, & C. Given, 2010a; Spoelstra, Given, & C. Given, 2010; Spoelstra, Given, & C. Given, 2010b). These findings informed this study and led to obtaining the 60 Michigan Cancer Registry. Data Management Several data management functions were performed prior to initiating the analysis. This included selecting the sample for analysis, combining variable categories for analysis, and reviewing missing data. Each is described in detail. Sample selection. The purpose of this study was to examine if the addition of a cancer diagnosis altered falls, fractures, and use of health care after controlling for the number of falls in the year prior to the cancer diagnosis. The intent was to control for falls prior to the cancer diagnosis by selecting two MDS assessments prior to the cancer diagnosis and two MDS assessments after the cancer diagnosis. The sample of patients who were age 65 or older in the HCBS program in 2002 to 2007 (excluding those in HMOs) was 18,234. The cancer diagnosis was confirmed using the Michigan Cancer Registry and the sample was divided into two parts, those with cancer (n=2,239) and those without cancer (n=16,035). Figure 2 depicts the sample selection. Cancer group. The types of cancer in the cancer group (n=2,239, 15.4%) were explored. After consideration and review of the literature, a decision was made to exclude skin cancer (n=46) because most skin cancers are not treated, nor are they terminal (American Cancer Society, 2008), which led to a sample including all other cancer types (n=2,193). Then, the two MDS assessments prior to cancer diagnosis and two after cancer diagnosis (n=322) were identified and analyses were conducted. A determination was made that this sample was not adequate size for analysis. Next, the cancer group sample of one MDS assessment prior to the cancer diagnosis and two MDS assessments after cancer diagnosis (n=351) was identified and analyses was 61 conducted; finding the sample was still too small for analysis. Consequently, the analysis of an MDS assessment prior to cancer diagnosis was abandoned and a decision was made to evaluate the MDS assessments in the cancer group based on times since cancer diagnosis. Advances in cancer treatment have led to different types of treatment compared to a decade ago (American Cancer Society, 2008). The cancer group (n=2,193) was separated into patients diagnosed with cancer on or after January 1, 2000 (n=1,463), and those diagnosed prior to that date (n=730) and used the first group for the next step. The first two MDS assessments after the cancer diagnosis (n=977 [diagnosed on/after January 1, 2000]) were identified and analyses were conducted. This sample (n=977) size was adequate for analysis. All patients who had confirmed date of death within 2 months (n=79) of cancer diagnosis were removed from the study because these terminal patients could have created bias in the study findings (American Cancer Society, 2008). As shown in Figure 2, the final cancer group consisted of 864 patients. Non-cancer group. Then appraisal of the non-cancer group was conducted. The intent was to identify patients with two MDS assessments and then to match the non-cancer group to the cancer group based on age, sex, and race. All non-cancer patients (n=16,035) in the HCBS program during 2002 to 2007 having two MDS assessments were identified (n=10,600) and the first two assessments available were selected. The final step in the selection of the sample for this study was to match the non-cancer group (n=10,600) to the cancer group (n=864). Matching is a commonly used statistical technique that leads to data reduction and removal of some bias in observational studies (Anderson, Kish, & Cornell, 1978; Kupper, Karon, Kleinbaum, Morgenstern, & Lewis, 1981; Rubin, 1973). 62 An analysis of the frequency of age, sex, and race was conducted. Those in the noncancer group who were over age 105 were removed from the sample because they did not match any patients in the cancer group, and frequencies were reexamined. A large percentage of those age 85 and older, female, and Caucasian were present in the non-cancer group; as a result, Caucasian women over age 92 were removed in the non-cancer group to match the non-cancer group to the cancer group. A large percentage of those age 85 and older, female, and Caucasian race continued to be prevalent, so Caucasian women over age 87 were removed in the noncancer group, again to match the non-cancer group to the cancer group. A satisfactory age, race/ethnicity, and somewhat similar sex match was attained. The final non-cancer group consisted of 8,617 (Figure 2). Final sample selection. The final sample for analysis in this study consists of 9,483 individuals who were 65 years of age or older in the HCBS program during 2002 to 2007 with two MDS assessments. Those in HMOs were excluded. The cancer group consisted of 864 patients with all cancer types, except for skin cancer, and those who were deceased within 2 months of the MDS assessment. The non-cancer group comprised 8,617. The final sample was a 10:1 match of non-cancer to cancer group. A match of this magnitude is considered ample for research when comparing and contrasting a cancer group to a non-cancer group (Hewitt et al., 2003; Yabroff et al., 2007). 63 Figure 2 Data Management Final Sample Selection for this Study Patients >65 in HCBS Program who are not in an HMO during 2002 to 2007 N=18,274 No Cancer N=16,035 With 2 MDS N=10,600 Match age, race & sex N=8,617 Cancer N=2,239 Without 2 MDS N=5,435 No Match N=1,983 Comparison All except skin cancer N=2,193 Cancer except skin diagnosed 2000 or later N=1,463 With 2 MDS N=977 Not deceased within 2 months N=864 Skin Cancer N=46 Cancer except skin diagnosed before 2000 N=730 With 1 MDS N=519 Deceased within 2 months N=79 The study schematic illustrated in Figure 3 reflects the sample selected for analysis in this study. In the cancer group patients diagnosed with cancer were selected and two MDS assessments after the cancer diagnosis were analyzed over a one-year time period. In the noncancer group two MDS assessments were selected and analyzed over a one-year time period. The cancer group was then compared to the non-cancer group. 64 Figure 3 Study Schematic Cancer group O Diagnosis of Cancer X 1-year time period X st nd 1 MDS 2 MDS Comparison: cancer to noncancer Claims and pharmacy files for both groups Non-cancer group X X st nd 1 MDS 2 1-year time period MDS Variable recategorization. The MDS variables are categorical with three exceptions: age, assessment dates, and time since cancer diagnosis. Several of the variables were recategorized prior to analysis. Date of birth or age at time of MDS assessment was grouped into categories of 5-year time periods, age 65 to 69, 70 to 74, 75 to 79, 80 to 84, and age 85 and older. Race was categorized into Caucasian, African American, and other. ADLs were categorized as 0 for none to some performance, and 1 for assistance to totally dependent—those with greater performance needs have functional deficits (Yabroff et al., 2007). Comorbidities were categorized as 0 for not present and 1 as present. Falls, fractures, ER use, and hospitalization were categorized as 0 for none and 1 for any occurrence. Review of missing data. The rate of missing data found in this dataset was 2.1%. The small number of missing values did not require imputation due to the small percentage, as it did not influence the outcome of the analysis(Stommel & Wills, 2004) (Zeger et al., 1988). Furthermore, for missing data, GEE can estimate the working correlation using all available 65 pairs‘ method, in which all non-missing pairs of data are used in the estimators of the working correlation parameters (Zeger et al., 1988). Data management steps prior to beginning analysis in this study were completed. The following section will describe the sample. Preliminary Work Preliminary analyses prior to examining the study aims were conducted. This was done to determine if data reduction could occur and to better understand which variables should be entered into models. These analyses were conducted using SAS ® Version 9.1 (SAS software, 2003), SPSS 14.0 (SPSS, 2005), and Mplus. Activities of daily living. Factor analyses were conducted on ADL variables to determine which variables to use in the study; and to evaluate how to enter the variables into the models. As an initial step, Cronbach‘s Alpha and inter-item correlations among the ADL variables were conducted. Cronbach‘s Alpha measures internal consistency reliability, or the inter-item-correlation and estimates of the proportion of the variance that accounts for the common factor (Cronbach, Gleser, Nanda, & Rajaratnam, 1972). Cronbach‘s Alpha was .883 (CI .88 to .86; p = .00). A score above .70 suggests the ADL items are measuring the same concept. This confirmed that factor analysis was needed to determine if all the ADL variables could be grouped for entry into the models. A two step factor analysis was conducted (Cronbach et al., 1972). Factor analysis. A factor loading of the following variables: functional status change, weight loss, vision, bladder and bowel incontinence, mobility, transfer, locomotion, dressing, eating, toileting, bathing, personal care, unsteady gait, walking, and stair-climbing was conducted. Construct validity testing of the unidimensionality of the ADL variables using structural equation modeling (SEM) techniques of exploratory factor analysis (EFA) and 66 confirmatory factor analysis (CFA) with principal axis factoring varimax (orthogonal) rotation was assessed. SEM is an analytical tool that provides an alternative to experimentation for examining the plausibility of hypothesized models (Kline, 2005). The data were 97.9% complete, probably due to missing at random. Imputation was not conducted because the remaining cases were more than sufficient power for generalizability (Kline, 2005). Cases with missing data were removed prior to analysis. Random assignment occurred and the dataset used in this study was split in half. One half of the dataset was used for EFA and the other half for the CFA. Crossvalidation of a dataset strengthens predictive validity (Vandenberg, 2006). Exploratory analysis. First, an EFA with the 16 ADLs examined in this study was conducted. Sampling adequacy tests were conducted using the Kaiser-Meyer-Olkin (KMO) and Bartlett‘s test. KMO is a measure of sampling adequacy comparing the magnitudes of the observed correlations coefficients to the correlation coefficients, which should be greater than .50 for a satisfactory factor analysis to proceed (Kline, 2005). Bartlett‘s is an indicator of the strength of the relationship among variables testing if the correlation matrix is uncorrelated and whether the correlation matrix is an identity matrix; a significance level must be small enough to conclude that the association of the relationship among the variables is strong (Kline, 2005). The KMO was .890 and Bartlett‘s test was significant (p = .00), demonstrating that the correlation matrix was not an identity matrix. that structure existed and the strength of the relationship among the variables was large enough for factor analysis (Kline, 2005). The total variance explained was examined using the initial Eigenvalues and percent of variance as depicted in Appendix B (Table 27). There were four factors in the total variance that had a value greater than one, demonstrating that a three- or four-factor solution would be most plausible. Also, none of the percent variances were less than one, demonstrating all 14 of the 67 factors could be used in a one-dimensional factor solution. Appendix B (Figure 4 identifies Eigenvalues equal to one at the fourth point so the process of evaluation of a factor matrix started with four factors. The four-factor model (Appendix B, Table 28) demonstrated five, three, and two items grouped above .40, a few cross-loaded, and a few not loaded, eliminating this model. The factors that did not load (functional status change, weight loss, and vision) were removed; then a threefactor model (Appendix B, Table 28). A KMO of .89 and significant Bartlett‘s (p = .00) demonstrating a three-factor model could be used. However, ensuring that the groupings made sense conceptually as well as clinically was crucial. A two-factor model (Appendix B, Table 29 and 30) finding a KMO of .896 and significant Bartlett‘s test (p = .00) was also developed; demonstrating that a two-factor was a poor fit. A one-factor model demonstrated that 13 ADLs sufficiently load on one factor. CFA was then conducted to confirm factor loadings. Confirmatory analysis. Second, a CFA which included 13 ADL variables (functional status change, vision, and weight loss were removed during EFA) was conducted to confirm the EFA findings. A two-step approach was taken, testing the measurement model for fit before testing the full structural model (Joreskog & Sorbom, 2004). Two indictors were used when examining CFA: comparative fit index (CFI), a relative fit index with values ≥ .95 indicating good fit; and the root mean square error of approximation (RMSEA), an indicator of the discrepancy in fit per degree of freedom adjusted for sample size, with values smaller than 0.08 providing a reasonable approximation of the factor loading (Kline, 2005). The model converged, and the fitting measures indicated a good fit (RMSEA = .06; CFI = .945), confirming that a relationship between the 13 ADLs existed. Summary of factor loading analyses. The EFA model demonstrated that 13 ADL 68 factors sufficiently loaded as a one-dimensional factor or that three subscales could be used and the CFA modeling confirmed that the theoretical constructs were related. However, the belief was that multicollinearity could occur among the ADLs. Multicollinearity is the extent to which predictor variables in a regression analysis are correlated with one another and cause problems, making it difficult to study separate effects of independent variables (Stommel & Wills, 2004). Consequently, a decision was made to sum the other 13 ADLs into one variable for entry into the models. The new summed ADL variable ranged from 0-52, incorporating the categorical score of each ADL into the sum. Finally, a determination was made to enter weight loss and vision separately in the GEE model. Functional status change was not used in this study, because it was thought to be a global indicator of overall function rather than an individual ADL performance measure. Summed activities of daily living score. The summed ADL score was calculated for each patient. This included the following ADL variables: bladder and bowel incontinence, mobility, transfer, locomotion, dressing, eating, toileting, bathing, personal care, unsteady gait, walking, and stair-climbing. The ADLs scored from 0 to 5 were categorized prior to summation, so that the 5 values became 4 values. The mean of the summed ADL score was 16.90 (SD 11.19; range of 0.0 to 52.0). The weight loss and vision variables were entered separately, as is in the model. Medications. The medication variables were then analyzed to determine if data reduction could occur and to better understand which variables should be entered into models. As an initial step, the association among medications (antipsychotics, anti-anxiety, antidepressants, or hypnotics) in relation to falls was examined. Pearson product-moment correlation coefficients testing the strength of the association were conducted. Antidepressants were strongly associated 69 with falls (r = .09), while hypnotics (r = .03), antipsychotics (r = .20), and anti-anxiety (r = .22) have a moderate association. Next, generalized linear modeling (GLM) among the medications association were investigated. Appendix B (Table 25) depicts minimal variation after examining the Least Square mean difference between antidepressants and antipsychotics (p < .00), antidepressants and anti-anxiety medications (p < .00), and antidepressants and hypnotics (p < .00). A decision was made to enter each of the medications in the model to better understand which drug classification may influence falls in this population. Although an excessive number of medications were taken (9+) by patients in this sample, a decision was made not to include this variable in the model because there was no ability to distinguish which medication may have increase falls. Comorbidities. The comorbid condition variables were then analyzed to determine if data reduction could occur and to better understand which variables should be entered into models. There are four reasons for adjusting for comorbid conditions in this study (de Groot, Beckerman, Lankhorst, & Bouter, 2003): first, to correct for confounding and improve the internal validity of the study; second, to identify effect modifications; third, to take co-occurring comorbidities into consideration; and finally, for statistical efficiency. Next, a factor loading occurred. Factor analysis. A factor analysis was conducted to determine is all of the comorbid conditions should be considered in the models. This included the following comorbidities: diabetes, hyperthyroidism, hypothyroidism, arteriosclerosis, cardiac dysrhythmias, CHF, CAD, deep vein thrombosis, hypertension, hypotension, peripheral vascular disease, cardiovascular disease, musculoskeletal arthritis, osteoporosis, Alzheimer‘s, aphasia, cerebral palsy, CVA, dementia, hemiplegic, multiple sclerosis, paraplegia, Parkinson‘s Disease, quadriplegia, seizure 70 disorder, transient ischemic attack, traumatic brain injury, anxiety disorder, depression, manic depressive, schizophrenia, other psychiatric diagnosis, asthma, emphysema, cataracts, diabetic retinopathy, glaucoma, macular degeneration, allergies, anemia, and HIV. Construct validity testing of the unidimensionality of the comorbid conditions using EFA and CFA with principal axis factoring varimax (orthogonal) rotation was assessed. Random assignment occurred and the dataset was split in half for cross-validation, one for EFA and one for CFA. Exploratory analysis. EFA was conducted to identify is the concepts grouped into a construct. This included the 41 comorbidities from the MDS. The KMO was .736 and Bartlett‘s test was significant (p = .00), demonstrating that the correlation matrix was not an identify matrix and that structure existed and the strength of the relationship among the variables was large enough for factor analysis (Kline, 2005). The total variance explained was examined using the initial Eigenvalues and percent of variance. There were 14 factors in the total variance that had a value greater than one, demonstrating a 13- or 14-factor solution would be most effective. Also, as expected, 27 of the items had percent variances that were less than one as conditions had varying biologic or clinical presentation, demonstrating that some of the factors were not helpful in loading on a onedimensional factor. Consequently, the 27 items from the factor analysis were removed and a new solution was attempted. In the second solution, eight items had percent variances less than one, demonstrating that some of the factors were not helpful in loading on a one-dimensional factor, so they were removed. The remaining six items included diabetes, CHF, CAD, depression, arthritis, and CVA. A six-factor solution found a KMO of .70 and significant Bartlett‘s (p = .00), demonstrating that a one-dimensional scale could be used. This grouping of comorbid conditions 71 made sense conceptually as well as clinically. CFA was then conducted to confirm factor loadings. Confirmatory analysis. CFA was conducted to better understand how to enter the variables into the model. This included the six comorbid conditions identified in the EFA. A two-step approach was taken, testing the measurement model tested for fit before testing the full structural model (Joreskog & Sorbom, 2004). The model converged and the fitting measures indicated a good fit (RMSEA = .08; CFI = .911). Summary of factor analyses. The EFA modeling demonstrated that six comorbidities sufficiently loaded as a one-dimensional factor. Consequently, a decision was made to sum the comorbidities into one variable, minimizing multicollinearity among the comorbidities and attaining the goal of data reduction. To affirm this decision, GLM models were developed. Appendix B (Table 24) depicts Least Square means using depression as the reference point, finding a difference with diabetes (p = .92), CHF (p = .04), depression and CAD (p < .00), arthritis (p < .00), and a CVA (p = .00). Pearson product-moment correlation coefficients were then conducted to examine if an individual comorbid condition influenced the occurrence of falls, finding depression strongly associated with falls (r = .09), while diabetes (r = .04) and CAD (r = 0.37) had a moderate association, and CVA (r = .02), arthritis (r = .02), and CHF (r = .01) had weak association. Finally, a summed count of the comorbid conditions was compared to a comorbidity index to assure the most effective approach for entering comorbidities into the model. A systematic review identified four comorbidity indices with reliability and validity (de Groot et al., 2003). The Charlson Comorbidity Index (CCI) has demonstrated statistical prognostic for clinical research in oncology (Extermann, 2000a, 2000c), with inter-rater reliability of 0.74 72 among a cohort of older oncology patients and 0.95 in a group of elderly breast cancer patients (Extermann, 2000b). The comorbid conditions were summed using the CCI domains then compared to a summed count. GLM models were used to examine the sensitivity of the summed scores‘ predictive ability for ADLs (Charlson & Sax, 1987; Deyo & Inui, 1984; Elixhauser, Steiner, Harris, & Coffey, 1998). The summed comorbidities were more predictive of ADL dependence (r = .17 compared to r = .12) than the CCI. This confirmed using a summed count of comorbidities as the best approach for this study. Summed comorbidity score. The summed comorbidity score was calculated to include the following variables: diabetes, CAD, CVA, CHF, arthritis, and depression. The comorbidity scored from 0 to 2, categorized prior to summation, so that the 2 values became a value of 1. Summary. This preliminary work provided guidance on how to enter the MDS variables into the models. Prior to model-building, variables associated with falls in the elderly that were cited in the literature and linked in the conceptual framework were evaluated in a univariate model to assess their unadjusted effect on the outcome. Based on this preliminary work, determinations were made on data reduction and how to enter variables into the model. For ADLs, 13 variables were summed and will be entered in the model as a one variable totaling the 13 items. Weight loss and vision are entered separately. For medications, the four classifications will be entered separately and the count of medications will not be used in modeling. For comorbidities, a simple count of the six primary comorbid conditions will be entered in the model as a single item. The following variables were included in the model. The sociodemographic variables included age, gender, and race/ethnicity. Cancer variables included cancer type, cancer stage, and time since cancer diagnosis. Medications variables included antidepressants, anti-anxiety, 73 antipsychotics, and hypnotics. The summed comorbid and ADL scores, along with individual variables of weight loss, vision, short-term memory recall, and evidence of pain, were included. Descriptive Analysis of Sample A total of 9,482 community-dwelling elderly patients age 65 and older in this study were examined. A 10:1 matched group of non-cancer patients (86.5%, n=8,617) and cancer patients diagnosed on or after 2000 (13.5%, n=864) was compared. The conceptual framework as shown in Figure 1 in Chapter 2 guides how the findings will be described in this chapter (environment, patient characteristics, biologic variables, medications and cancer treatment, frailty, and health outcomes) to answer the research questions. Research Question 1 examined if those with a cancer diagnosis fall or have fractures at a higher rate, or have more frequent ER visits, hospitalization, or nursing home placement than those without cancer after taking sociodemographic characteristics (age, sex, and race or ethnicity), certain medications, comorbidities, ADLs, and frailty into consideration; and if the rate of falls differed by cancer site or stage. Research Question 2 examined if frailty (cognition, ADLs, comorbidity, pain, weight loss, and vision) influenced the rate of falls with respect to cancer type and stage. Finally, the exploratory aim described the social and environmental status of the population. Environmental and Social status. Items examined in this study are depicted in Table 2. The majority of patients without cancer (90.1%) lived alone, while most with cancer (60.5%) lived with another person. However, there is a higher rate of falls in those with cancer (59.9% compared to 11.5%) when living with another person. Minimal differences were found in marital status, with most patients being widowed, or in the type of living environment, with most patients living in a house or apartment. 74 Table 2 Descriptive Statistics of Social and Environmental Status, Marital Status, Housing, Living with Other Person, Time Alone During Day, and Feelings of Loneliness of Cancer and Non-Cancer Patients with Falls or No Falls Fallers Cancer n (%)* No Falls Non-cancer n (%)* Cancer n (%)* Non-cancer n (%)* Lives with other person No 122 (40.1) Yes 182 (59.9) 2358 (88.5) 307 (11.5) 210 (38.8) 332 (61.2) 5346 (92.3) 445 (7.7) Marital Status Married 55 (24.2) Widowed 124 (54.6) Other 48 (21.2) 566 (27.1) 1127 (53.9) 399 (19.0) 95 (23.2) 225 (54.9) 90 (21.9) 1022 (22.5) 2613 (57.6) 905 (19.9) Housing House Apartment Other 1625 (60.4) 871 (32.4) 194 (7.2) 316 (59.0) 165 (30.8) 55 (10.2) 3459 (54.6) 1973 (33.8) 387 (11.6) 998 (37.2) 322 (12.0) 823 (30.6) 543 (20.2) 192 (35.7) 60 (11.2) 175 (32.5) 111 (20.6) 2119 (36.3) 675 (11.6) 1913 (32.8) 1127 (19.3) Feels Lonely No 228 (77.3) Yes 67 (22.7) 2015 (76.2) 629 (23.8) 425 (79.4) 107 (20.1) 4692 (81.6) 1059 (18.4) Pets No Yes 1141 (60.1) 757 (39.9) 216 (64.5) 119 (35.5) 1993 (63.2) 1162 (36.8) 171 (57.0) 104 (34.7) 25 (8.3) Time Alone During the Day Never 134 (44.8) 1 hour 24 (8.0) Long periods 95 (31.8) All the time 46 (15.4) 146 (61.9) 90 (38.1) *Some of the total values do not equal as some of the data are missing 75 The amount of time patients spent alone during the day was examined. The number of cancer patients who were never left alone and fell was higher (44.8% compared to 37.2%) than those without cancer who had a fall. In contrast, a lower rate of falls occurred in those with cancer who were alone all of the time (15.4% to 20.2%). No difference between the cancer and non-cancer groups was found when examining whether the patient had feelings of loneliness. However, a higher frequency of loneliness was found in those who fall (23.7%, n = 686) compared to those who did not fall (18.6%, n = 1,166). No difference among the groups was found on whether the patient had a pet; however, overall a higher frequency of pets in the home was found in those who fall (39.7%, n = 847) compared to those who did not fall (36.7%, n = 1,281). The amount of time patients spent alone during the day was examined. The number of cancer patients who were never left alone and fell was higher (44.8% compared to 37.2%) than those without cancer who had a fall. In contrast, a lower rate of falls occurred in those with cancer who were alone all of the time (15.4% to 20.2%). No difference between the cancer and non-cancer groups was found when examining whether the patient had feelings of loneliness. However, a higher frequency of loneliness was found in those who fall (23.7%, n = 686) compared to those who did not fall (18.6%, n = 1,166). No difference among the groups was found on whether the patient had a pet; however, overall a higher frequency of pets in the home was found in those who fall (39.7%, n = 847) compared to those who did not fall (36.7%, n = 1,281). Patient characteristics. The sociodemographic characteristics of the participants are presented in Table 3. The mean age of the non-cancer group was 77.04 years (SD = 6.57) and the mean age in the cancer group was 77.06 years (SD = 7.53). Three of the age groups, 65-69, 75- 76 79, and >85, are similar. However, in the 70 to 74 age group more patients fall in the cancer group (23.7% to 17.5%), and in the 80 to 84 age group, fewer patients fall (17.1% to 26.1%). However, overall age was associated with falls (p = .02), but age was not associated with the presence of a cancer diagnosis (p = .67) in this sample. Furthermore, aging was a predictor of falls in both the non-cancer (p = .02) and cancer groups (p = .00) when examined separately. Table 3 Descriptive Statistics of Sociodemographic of Age, Gender, and Race/Ethnicity of Cancer and Non-Cancer Patients with Falls or No Falls Fallers No Falls Cancer n (%)* Non-cancer n (%)* Cancer n (%)* Non-cancer n (%)* Age Groups 65-69 70-74 75-79 80-84 >85 Total 54 (17.8) 72 (23.7) 72 (23.7) 52 (17.1) 54 (17.7) 304 (100)* 472 (17.4) 475 (17.5) 660 (24.3) 707 (26.1) 400 (14.7) 2714 (100)* 97 (17.9) 118 (21.8) 125 (23.1) 107 (19.8) 95 (17.4) 542 (100)* 1005 (17.1) 1018 (17.4) 1424 (24.3) 1533 (26.2) 882 (15.0) 5862 (100)* Sex Female Male Total 177 (62.4) 114 (37.6) 291 (100)* 1827 (69.9) 815 (30.1) 2642 (100)* 318 (60.1) 211 (39.9) 529 (100)* 4080 (71.1) 1636 (28.9) 5716 (100)* 2114 (77.9) 391 (72.1) 4226 (72.1) 495 (18.2) 105 (3.9) 2714 (100)* 131 (24.2) 20 (3.8) 542 (100)* 1425 (24.3) 211 (3.6) 5862 (100)* Race or Ethnicity Caucasian 231 (76.0) African American 64 (21.0) Other 9 (3.0) Total 304 (100)* *Some of the total values do not equal as some of the data are missing. 77 The sample is predominantly female (61.8%); however, there were more males with cancer than in the non-cancer group (38.6% to 29.2%). Gender was not a significant predictor of falls (p = .08). To further examine this, a regression model to include cancer and gender found gender was not associated with falls in cancer patients (p =.80), however, it was a predictor of falls in the non-cancer group (p = .01), with females falling more often. Overall, the predominant race was Caucasian (non-cancer 73.9%) followed by African American as the second most predominant (23.6%). The ―other‖ category included Asian Pacific Islander, American Indian, Hispanic, and multiracial (3.3%). Race alone was a significant predictor of falls (p = .00), and Caucasians associated with a higher fall occurrence compared to African Americans or the other groups. To further examine this, a regression model that included cancer diagnosis, race, and their interaction found that race was associated with falls in the non-cancer group (p = .00), but not in the cancer group (p = .05). Race was not a significant predictor of cancer (p = .80). One other item was notable; African Americans with cancer fall at a somewhat higher rate (21.0%) than African Americans without cancer (18.2%). When fractures was examined, age (p = .00) and race/ethnicity (p = .00), were predictors of fractures, while gender (p = .64) was not. Age was a predictor of fractures in both the noncancer (p = .00) and cancer groups (p = .00), while race/ethnicity was a predictor of fractures in the non-cancer group (p = .00), but not a predictor in the cancer group (p = .76). When ER use was examined, age (p = .00) and gender (p = .03) were predictors of use, while race/ethnicity was not (p = .10). Those who were aged and White had higher ER use. When comparing groups, aging (p = .00) and gender (p = .00) were predictors of ER use in the cancer group, but race was not (p = .98), and in the non-cancer group aging (p = .00), gender (p = 78 .00), and race/ethnicity (p = .00), were predictors of ER use with aged, White, females having more ER use. When hospitalization was examined, age (p = .00) was a predictor of use, while gender (p = .58) and race/ethnicity were not (p = .41). Finally, age (p = .00) and race/ethnicity (p = .00), were predictors of nursing home placement while gender was not (p = .21), with aged Whites having higher placement. Biologic Factors Site of cancer. As depicted in Table 4, 64.4% (n = 865) of the cancer patients had a solid tumor. The most common type was breast (n = 179, 20.8%), followed by colon (n=144, 16.7%), prostate (n = 119, 13.8%), and lung cancer (n = 113, 13.1%). The remaining patients were grouped and analyzed as other cancers. This included kidney (n = 80, 9.3%), leukemia (n = 48, 5.6%), uterine (n = 45, 5.2%), lymphoma (n = 23, 2.7%), and cervical cancer (n = 19, 2.2%), followed by stomach (n = 15, 1.7%), pancreatic (n = 12, 1.4%), mouth or tongue (n = 8, 1.0%), larynx (n = 7, 1.0%), esophagus (n = 4, 0.4%), thyroid (n = 4, 0.4%), liver (n = 4, 0.4%), and other (n = 41, 4.7%). Somewhat older (> 70 years) females with breast cancer (n = 151, 17.5%) and males with prostate cancer (n = 119, 13.7%) represented the largest percentage of the sample. To examine the time since cancer diagnosis, the mean number of days between the date of cancer being diagnosed and the first MDS assessment was analyzed. In this study, the mean number of days from the date of the cancer diagnosis was 465.3 days or 15.3 months (SD 546.8 days or 18.0 months). Frailty (ADLs, cognition, comorbidities, pain, vision, and weight loss) by cancer type was examined and a difference (p < .05) was evident among unsteady gait, weight loss, eating, or short-term memory recall. No difference (p > .05) was evident for vision, bladder and bowel 79 incontinence, transferring, dressing, locomotion, toileting, personal hygiene performance, walking, stair-climbing, evidence of pain, or long-term memory recall. Falls did not differ by cancer type (p = .17). However, the number of falls (one to > nine) was significantly higher for those with breast or another type of cancer (p = .00) than for colon, lung, or prostate cancer. Table 4 Descriptive Statistics of Site of Cancer by Age Groups, Gender, Race/Ethnicity Type: Breast n (%)* Colon n (%)* Lung n (%)* Prostate n (%)* Other n (%)* Total n (%)* Cancer Type by Age Groups 65-69 26 (13.2) 20 (15.2) 26 (23.1) 70-74 32 (18.9) 26 (18.1) 31 (27.4) 75-79 41 (22.9) 27 (19.8) 31 (27.4) 80-84 44 (24.6) 32 (23.2) 17 (15.0) >85 36 (20.4) 39 (22.1) 8 (7.1) Total 179 (100)* 144 (100)* 113 (100)* 21 (17.6) 62 (20.1) 34 (28.6) 70 (27.7) 33 (27.7) 70 (27.7) 14 (11.8) 55 (17.8) 17 (14.3) 52 (16.7) 119 (100)* 309 (100)* 155 (17.9) 193 (22.4) 202 (23.4) 162 (18.7) 152 (17.6) 864 (100)* Cancer Type by Gender Female 151 (84.4) 86 (61.1) 65 (59.0) Male 28 (15.6) 53 (38.9) 45 (41.0) Total 179 (100)* 139 (100)* 110 (100)* 0 (0) 119 (100) 119 (100)* 206 (68.0) 97 (32.0) 303 (100)* 508 (56.0) 342 (44.0) 850 (100)* 76 (63.8) 220 (71.2) 637 (73.6) 17 (15.0) 35 (29.4) 82 (26.5) 6 (5.4) 8 (6.8) 7 (2.3) 113 (100)* 119 (100)* 309 (100)* 197 (22.9) 30 (3.5) 864 (100)* Cancer Type by Race/ethnicity Caucasian 139 (77.6) 112 (77.8) African American 35 (19.6) 28 (19.4) Other 5 (2.8) 4 (2.8) Total 179 (100)* 144 (100)* 90 (79.6) *Some of the total values do not equal as some of the data are missing data Cancer stage. Cancer stage in this study is depicted as I to IV. Refer to Table 5 for cancer stage by age, gender, and race or ethnicity. Only 5.3% of the patients were diagnosed at 80 stage I. The largest group of patients was 44.9% who were diagnosed at stage II, followed by 20.3% diagnosed at stage III. Finally, 28.7% of the patients were diagnosed at stage IV. Overall, the majority of the patients (n = 423, 49.0%) experienced later stage disease (III and IV). The majority of the patients (59.7%) were diagnosed at age 75 or older, and female Caucasians were diagnosed at a later stage. Fall occurrence varied by cancer stage (p = .00), and the number of falls (one to > nine) was somewhat higher for those with stage II through IV cancer (p = .00). Medications. In this study, medications most often known to increase the occurrence of falls (Agostini et al., 2004; Landi, Onder, Cesari, Barillaro et al., 2005; Landi, Onder, Cesari, Russo et al., 2005; Leipzig et al., 1999b) were evaluated. This included antipsychotics, antianxiety, antidepressants, and hypnotics. Overall, 90.4% of the patients in this study take five or more medications and 69.5% take nine or more medications. However, those with cancer as a whole took fewer medications. Although the rate of falls for cancer patients was lower for those who took one to four medications (6.4% to 6.9%) and five to eight medications (22.2% to 23.5%), those with cancer taking nine or more medications fall at a somewhat higher rate (71.4%) than those without cancer (69.2%). The most commonly reported medication taken was antidepressants (n = 3,469, 37.2%), followed by anti-anxiety medications (n = 1,995, 21.4%), hypnotics (n = 797, 8.7%), and antipsychotics (n = 722, 7.4%). Many of the patients took a combination of these medications: 37.2% (n = 3,469) took one of the four; 14.2% (n = 1,311) took two of the four; 3.1% (n = 284) took three of the four; and 0.4% (n = 16) took all four medications. 81 Table 5 Descriptive Statistics of Stage of Cancer by Age Groups, Gender, Race/Ethnicity, and Type of Cancer Stage-I n (%)* Stage-II n (%)* Stage-III n (%)* Stage-IV n (%)* Cancer Stage by Age Groups 65-69 5 (10.9) 75 (19.3) 70-74 8 (17.4) 78 (20.1) 75-79 9 (19.6) 92 (23.5) 80-84 14 (30.4) 73 (18.8) >85 10 (21.7) 70 (18.3) Total 46 (100)* 388 (100)* 32 (18.3) 44 (25.1) 41 (23.4) 33 (18.9) 25 (14.3) 175 (100)* 46 (18.4) 61 (24.4) 58 (23.2) 38 (15.2) 47 (18.8) 250 (100)* 155 (18.0)* 193 (22.3)* 203 (23.3)* 162 (18.6)* 152 (15.1)* 864 (100)* Cancer Stage by Gender Female 29 (64.4) Male 16 (35.6) Total 45 (100)* 105 (62.5) 63 (37.5) 168 (100)* 157 (72.6) 89 (27.4) 246 (100)* 507 (60.3)* 332 (39.7)* 839 (100)* 139 186 (75.5) 638 (74.0)* 28 (16.0) 8 (4.6) 175 (100)* 56 (23.6) 5 (0.9) 248 (100)* 196 (22.5)* 30 (3.5)* 865 (100)* Cancer Stage by Type of Cancer Breast 20 (43.5) 100 (25.7) 37 (21.1) Colon 7 (15.2) 59 (15.2) 56 (32.0) Lung 0 (0.0) 45 (11.6) 30 (17.1) Prostate 0 (0.0) 86 (22.2) 3 (1.7) Other 19 (41.3) 98 (25.3) 48 (27.4) Total 46 (100)* 388 (100)* 174 (100)* 22 (8.9) 22 (8.9) 38 (15.3) 28 (11.3) 138 (55.6) 248 (100)* 179 (23.4)* 44 (14.8)* 113 (11.6)* 117 (8.2)* 303 (42.0)* 856 (100)* 211 (56.7) 161 (43.3) 372 (100)* Cancer Stage by Race or Ethnicity Caucasian 29 (63.0) 278 (71.6) African American 16 (34.8) 94 (24.2) Other 1 (2.2) 16 (4.2) Total 46 (100)* 388 (100)* (79.4) Total n (%)* * Some of the total values do not equal as some of the data are missing data An increase in falls was found for all four drug classifications: antipsychotics (p = .02), anti-anxiety (p = .02), antidepressants (p < .00), and hypnotics (p = .00). Table 6 shows these data. As shown in Appendix B (Table 24), generalized linear models were used to examine 82 relationships between the four medication classifications and falls; and multi-nomial regression models were used to examine these medications and comorbidities (Appendix B, Table 23) with no significant findings (p > .05). Table 6 Descriptive Statistics of Medications of Cancer and Non-Cancer Patients with Falls or No Falls Fallers Cancer n (%)* No Falls Non-cancer n (%)* Cancer n (%)* Non-cancer n (%)* Number of Medications Taken None 0 (0.0) 12 (0.4) 1-4 19 (6.4) 204 (6.9) 5-8 66 (22.2) 572 (23.5) >9 212 (71.4) 1856 (69.2) Total 297 (100)* 2644 (100)* 4 (0.7) 53 (9.8) 138 (25.7) 344 (63.8) 539 (100)* 40 (0.7) 465 (8.0) 1365 (23.4) 3949 (67.9) 5819 (100)* Antipsychotics None 269 (91.8) >1 24 (8.2) Total 293 (100)* 2401 (90.7) 245 (9.3) 2646 (100)* 513 (96.3) 20 (3.7) 533 (100)* 5377 (92.8) 680 (7.2) 6057 (100)* Anti-anxiety None 224 (77.0) >1 67 (23.0) Total 291 (100)* 2059 (77.6) 595 (22.4) 2654 (100)* 433 (80.8) 103 (19.2) 536 (100)* 4601 (79.4) 1197 (20.6) 5798 (100)* Antidepressants None 170 (57.8) >1 124 (42.2) Total 294 (100)* 1487 (55.7) 1181 (44.3) 2668 (100)* 380 (70.8) 157 (29.2) 537 (100)* 3819 (65.8) 1985 (34.2) 5804 (100)* Hypnotics None >1 Total 2375 (90.1) 61 (9.9) 2636 (100)* 492 (92.3) 41 (7.7) 533 (100)* 5306 (91.8) 473 (8.2) 5779 (100)* 263 (90.4) 28 (9.6) 291 (100)* * Some of the total values do not equal as some of the data are missing data 83 Overall, cancer patients took fewer antipsychotics (7.7% to 7.3%), anti-anxiety (2.7% to 21.4%), antidepressants (33.7% to 37.6%), and hypnotics (7.2% to 8.7%) than non-cancer patients. Additionally, the rate of falls was somewhat lower in the cancer group for antipsychotics (8.2% to 9.3%), antidepressants (42.2% to 44.3%), and hypnotics (9.6% to 9.9%) medications and somewhat higher for cancer patients taking anti-anxiety medications (23.0% to 22.4%). Cancer treatment. Cancer treatment was defined as chemotherapy, radiation, or the combination of both in this study. In this sample, only 49 (5.7%) of the patients (Given, Sikorskii, Spoelstra, & You, 2010 ) as being treated upon examining the claims files were identified. Due to the small number of patients under treatment in this sample, examining the influence of treatment on the outcome variables was not feasible. Frailty Comorbidities. Multiple comorbidities for this study were examined as depicted in Table 7. In this sample, 10.7% (n = 978) of the patients do not have comorbidities and 24.3% (n = 2,221) of the patients experienced only one of these comorbidities. However, 28.6% (n = 2,636) had two, 20.8% (n = 1,899) had three, 11.1% (n = 1,011) had four, 3.8% (n = 351) had five, and 0.5% (n = 43) had six comorbid conditions. The most commonly reported comorbid condition was arthritis (79.6%), followed by diabetes (42.1%), depression (41.6%), CHF (38.0%), CAD (34.4%), and CVA (30.8%). The mean number of comorbid conditions was 2.11 (SD 1.32, range 0 to 6). A significant difference between having a comorbid condition and falling for patients with diabetes (p = .00), CHF (p = .05), CAD (p = .03), CVA (p = .01), depression (p < .00), and arthritis (p = .04) was found. In comparing the cancer group to the non-cancer group, the only 84 comorbidity that occurred more frequently in the cancer group was CAD (33.3% to 31.2%). A higher rate of falls was found in cancer patients with the following comorbidities: diabetes (46.5% to 43.6%), CAD (37.9% to 36.2%), and depression (48.8% to 48.1%). Table 7 Descriptive Statistics of Comorbidities of Cancer and Non-Cancer Patients with Falls or No Falls Fallers Cancer n (%)* Non-cancer n (%)* No Falls Cancer Non-cancer n (%)* n (%)* Diabetes Mellitus No 161 (53.5) Yes 140 (46.5) 1520 (56.4) 1175 (43.6) 303 (61.0) 211 (39.0) 3420 (58.8) 2395 (41.2) Congestive Heart Failure No 191 (63.5) Yes 110 (36.5) 1649 (61.4) 1035 (38.6) 351 (65.5) 185 (34.5) 3615 (62.4) 2181 (37.6) Coronary Artery Disease No 187 (62.1) Yes 114 (37.9) 1706 (63.8) 968 (36.2) 365 (68.7) 166 (31.3) 3836 (66.5) 1936 (33.5) Cerebral Vascular Accident No 228 (76.0) 1794 (67.1) Yes 72 (24.0) 881 (32.9) 401 (74.5) 137 (25.5) 4027 (69.7) 1754 (30.3) Depression No Yes 154 (51.2) 147 (48.8) 1384 (51.9) 1282 (48.1) 352 (65.4) 186 (34.6) 3572 (61.6) 2227 (38.4) Arthritis No Yes 228 (76.5) 70 (23.5) 743 (27.7) 1942 (72.3) 413 (76.3) 128 (23.7) 1512 (26.1) 4286 (73.9) * Some of the total values do not equal as some of the data are missing data 85 However, a lower rate of falls was found in cancer patients with CHF (36.5% to 38.6%), CVA (24.0% to 32.9%) and arthritis (23.5% to 72.3%). Overall, arthritis occurred at a much higher rate in the non-cancer group (73.4%) compared to the cancer group (23.6%). As shown in Appendix B (Table 25), generalized linear models examining the relationships among the six comorbid conditions were conducted. Symptoms. Evidence of pain, intensity of pain, and the incidence of pain disrupting an activity were examined as depicted in Table 8. A higher rate of daily pain (59.8% to 51.5%; p < .00), intensity of pain (46.1% to 42.4%; p < .00), and disruption of activity due to pain (50.7% to 47.0%; p = .00) in those who fall was discovered. Table 8 Descriptive Statistics of Evidence of Pain, Intensity of Pain, and Pain Disruption of Activity of Cancer and Non-Cancer Patients with Falls or No Falls Fallers Cancer n (%)* No Falls Non-cancer n (%)* Cancer n (%)* Non-cancer n (%)* Evidence of Pain No pain 53 (17.7) Less than daily 58 (19.3) Pain daily 189 (63.0) 572 (21.2) 546 (20.2) 1582 (58.6) 131 (24.3) 128 (23.7) 280 (52.0) 1570 (26.8) 1263 (21.6) 3016 (51.6) Intensity of Pain No 115 (48.3) Yes 123 (51.7) 1138 (55.2) 925 (44.8) 210 (55.4) 169 (44.6) 2380 (57.9) 1731 (42.1) Pain disrupts activity No 109 (46.2) Yes 127 (53.8) 1042 (50.7) 1015 (49.3) 197 (51.8) 183 (48.2) 2184 (53.1) 1927 (46.9) * Some of the total values do not equal as some of the data are missing data 86 When examining pain, daily pain (55.9% to 57.5%) and pain less than daily (21.1% to 21.3%) occurred at a lower rate in the cancer group, while intensity of pain (47.7% to 43.1%) and disruption of activity due to pain (50.5% to 47.7%) occurred at a higher rate in the cancer group. However, in general, those with cancer who have pain daily (63.0% to 58.6%), pain intensity problems (51.7% compared to 44.8%), or disrupted activity due to pain (53.8% compared to 49.3%) fall at a higher rate. Cognition. Short- and long-term memory recall was evaluated as depicted in Table 9. Overall, a higher rate of short-term memory problems was found in patients who fall (n = 1,642, 59.2% to n = 3,300, 50.0%; p < .00), but no difference existed in the rate of falls for those with long-term memory problems (p = .47). A lower rate of falls was found in cancer patients for both short-term memory recall (n = 164, 54.6% and n = 1,565, 58.1%) and long-term memory recall (n = 643, 21.8% and n = 619, 23.0%) compared to non-cancer patients. Table 9 Descriptive Statistics Memory of Cancer and Non-Cancer Patients with Falls or No Falls Fallers Cancer n (%)* No Falls Non-cancer n (%)* Cancer n (%)* Non-cancer n (%)* Short-term Memory: Recall after 5-minutes OK 135 (45.2) 1130 (41.9) Problem 164 (54.6) 1565 (58.1) Total 299 (100)* 2695 (100)* 311 (57.4) 231 (42.6) 542 (100)* 2933 (50.2) 2906 (49.8) 5839 (100)* Long-term Memory: Appears to Recall Long Past OK 230 (78.2) 2067 (77.0) Problem 64 (21.8) 619 (23.0) Total 294 (100)* 2686 (100)* 447 (83.6) 88 (16.4) 535 (100)* 4525 (77.8) 1293 (22.2) 5818 (100)* * Some of the total values do not equal as some of the data are missing data 87 Functional status. The 14 ADL performance items for this study were examined. Each ADL item is described individually in this section. As depicted in Table 10,, significant differences (p < .05) were found comparing those who fall and those who do not fall for each ADL, except for vision. Additionally, a higher rate of functional status change both in those who fall (n = 1,224, 44.6% to n = 2,167, 30.4%) and in those with cancer (n = 827, 48.9% to n = 3,500, 38.9%) was found. A higher level of ADL dependence was found in those with cancer. This included bathing (80.1% compared to 77.9%); dressing (51.9% compared to 47.1%); toilet use (38.7% compared to 35.8%); transferring (42.9% compared to 39.2%); locomotion (57.8% compared to 53.1%); and walking (91.9% compared to 91.0%). Weight loss had the largest difference (21.9% compared to 9.4%). However, less need for assistance in those who fall with cancer to include vision (11.5% to 13.6%); bladder incontinence (41.8% to 45.0%); bowel incontinence (12.5% to 24.8%); mobility in bed (21.9% to 25.4%); eating (20.5% to 22.1%); and stair-climbing (93.6% compared to 95.2%) was found (Appendix B, Table 26). Overall, cancer patients who fell had greater disability in functional status change and weight loss, and moderate differences in bathing, dressing, toileting, transfers, and locomotion compared to those who did not have cancer and fall. Further analysis of ADLs will be reported in the preliminary work section, where development of the ADL index used in this study will be reported. Further analysis of the ADLs will also be reported when answering the research aims. 88 Table 10 Descriptive Statistics of Functional Status Change and Activity of Daily Living Performance Items of Cancer and Non-Cancer Patients with Falls or No Falls Fallers Cancer n (%)* No Falls Non-cancer n (%)* Cancer n (%)* Independent to some supervision in bathing performance 47 (19.9) 596 (22.1) 152 (28.2) Limited assistance to total dependence in bathing performance 251 (80.1) 2104 (77.9) 388 (71.8) Independent to some supervision in dressing performance 143 (48.1) 1410 (52.9) 318 (59.3) Limited assistance to total dependence in dressing performance 154 (51.9) 1260 (47.1) 218 (40.7) Independent to some supervision in toilet use performance 178 (61.3) 1702 (64.2) 391 (71.3) Limited assistance to total dependence in toilet use performance 112 (38.7) 946 (35.8) 144 (28.7) Non-cancer n (%)* 1424 (24.4) 4424 (75.6) 3235 (55.9) 2559 (44.1) 3796 (66.1) 1947 (33.9) Independent to some supervision in transferring performance 165 (57.1) 1625 (60.8) 374 (70.0) 3791 (65.6) Limited assistance to total dependence in transferring performance 124 (42.9) 1048 (39.2) 160 (30.) 1986 (34.4) Weight Loss No 232 (78.1) Yes 65 (21.9) 2434 (90.6) 253 (9.4) 461 (85.8) 76 (14.2) Independent to some supervision in walking performance 24 (8.1) 243 (9.0) 88 (16.3) Limited assistance to total dependence in walking performance 271 (91.9) 2455 (91.0) 453 (83.7) Independent to some supervision in locomotion performance 125 (42.2) 1252 (46.9) 293 (54.8) Limited assistance to total dependence in locomotion performance 171 (57.8) 1416 (53.1) 241 (45.2) 5500 (94.5) 323 (5.5) 908 (15.5) 4939 (84.5) 2981 (51.6) 2796 (48.4) * Some of the total values do not equal as some of the data are missing data 89 Health outcomes. Five health outcomes were evaluated for this study to include falls, fractures, ER use, hospitalization, and nursing home placement. Table 11 depicts initial descriptive analyses of these health outcomes. Further interactions were tested in the full GEE models which will be reported by specific research aim. Table 11 Descriptive Statistics of Health Outcomes Fractures, ER Use, Hospitalization, and Nursing Home Placement of Cancer and Non-Cancer Patients with Falls or No Falls Fallers No Falls Cancer n (%)* Non-cancer n (%)* Cancer n (%)* Non-cancer n (%)* Falls No Yes Total 591 (67.3) 273 (32.7) 864 (100)* 5827 (70.4) 2790 (29.6) 8617 (100)* Fractures No Yes Total 239 (81.3) 55 (18.7) 294 (100)* 2119 (80.1) 525 (19.9) 2644 (100)* 464 (86.7) 71 (16.3) 535 (100)* 4919 (85.9) 807 (14.1) 5726 (100)* ER Use No Yes Total 233 (80.3) 57 (19.7) 290 (100)* 2210 (82.4) 473 (17.6) 2683 (100)* 468 (87.8) 65 (12.2) 533 (100)* 5353 (92.1) 460 (7.9) 5813 (100)* Hospitalization No 151 (51.2) Yes 144 (48.8) Total 295 (100)* 1833 (68.0) 861 (32.0) 2694 (100)* 320 (59.8) 215 (40.2) 535 (100)* 4659 (73.2) 1167 (26.8) 5826 (100)* Nursing Home Placement No 112 (66.7) 938 (50.2) 198 (61.7) 1972 (53.6) Yes 56 (33.3) 929 (49.8) 123 (38.3) 1708 (46.4) Total 168 (100)* 1867 (100)* 542 (100)* 3680 (100)* * Some of the total values do not equal as some of the data are missing data 90 The overall fall rate in this sample was 29.6% (n = 2,790). A significant difference (p = .01) in the occurrence of falls was found in those with cancer compared to those who did not have cancer. Furthermore, a higher rate of falls was found in those with cancer (32.7%, n = 273) compared to those without cancer (29.4%, n = 2,517). The fracture rate was significantly higher (p < .00) in those who had a fall. However, the overall fracture rate was lower in the cancer group at 18.7% (n = 55) compared to the non-cancer group at 19.9% (n = 525). A higher rate of ER use for those who had a fall (p < .00) and for those who had cancer (p < .00) was found. A higher rate of ER use in the cancer group who had a fall (19.7%) compared to the non-cancer group who had a fall (17.6%) was also found. Similarly, hospitalization rates occurred at a higher frequency for those who had a fall (n = 670, 24.2% compared to n=1,339, 20.3%; p < .00). Hospitalization rates were also higher in those with cancer (48.8% compared to 32.0%; p < .00). The overall nursing home placement occurred at a lower rate in those who had cancer (n=179, 36.2% compared to n = 2,637 47.5%) and in those with cancer who had a fall (33.3% compared to 49.8%). Summary. The descriptive findings in this study have been presented and described according to the conceptual framework in Figure 1. This includes patient characteristics (age, sex, and race/ethnicity), biologic (cancer type and stage), medications (antidepressants, antipsychotics, anti-anxiety, and hypnotics), cancer treatment, and frailty variables (symptoms [pain], function [ADLs], comorbidities, and cognition). The following section will discuss preliminary research findings that support this study, followed by preliminary work conducted prior to analysis of the specific aims. 91 Research Question Results GEE model-building was conducted in a standard manner recommended in the literature (Ballinger, 2004; Liu, Dixon, Qiu, Tian, & McCorkle, 2009; Zeger et al., 1988; Zheng, 2000). The GEE model for dichotomous outcome data looks similar to a logistic regression model, including a correlation between repeated outcomes from the same patient, assuming a covariance structure for estimation of model parameters using quasi-likelihood estimations (Zeger et al., 1988; Zheng, 2000). Odds Ratios (ORs) and their confidence intervals are reported making use of empirical robust estimators in the model (Ballinger, 2004). All candidate variables were entered into the GEE model. The model was evaluated for fit, and the variable that was least significant was removed in a step-wise manner before running the next model. In all instances changes in β coefficients were examined after removal of a factor, to identify if items interacted, or if clinical significance of certain factors precluded their removal from the model. Modelfitting continued until a final parsimonious model was identified, with all variables statistically significant (p < .05); and in some instances a decision was made to keep factor in the model that was not significant due to its clinical importance. Research Question 1. The first research question was to determine the extent to which patients with a cancer diagnosis experience a greater occurrence of falls, fractures, ER use, hospitalization, or nursing home placement, compared with those patients with no diagnosis of cancer; and if there are differences in the number of falls among cancer patients according to site of cancer or stage of cancer; and if there is a difference in falls by cancer site or stage. The primary aim of this analysis was to compare those with and without cancer, to determine if a difference existed. Repeated measurements or observations of the outcome variable (Y [falls, fractures, ER use, nursing home placement, hospitalizations]) occur on two 92 occasions, thus a random vector of responses for the ith individual (i=1, 2,…,N [subjects in the study]) at jth time (j=1 to 2). A correlation matrix was specified to account for the association for the same patient at different times. The following statistical models were implemented: health outcome (falls, fractures, ER use, hospitalization, or nursing home placement [yes or no]) each treated separately = (equals) cancer (yes or no) + (plus) covariates. Covariates included age, race, sex, medications (antipsychotics, antidepressants, anti-anxiety, and hypnotics [yes or no]), and frailty (ADL summed score, comorbidity summed score, evidence of pain, short-term memory loss, weight loss, and vision). The essential parameter is associated with the coefficient for the groups (cancer [yes or no]). The test of the significance of this parameter was used to determine if a difference between groups (cancer [yes or no]) existed. A model was built for each of the outcomes in this study. A new model was developed to evaluate if time since cancer diagnosis was a factor for both falls and fractures. Falls. Table 12 summarizes the final GEE model identifying factors that are associated with falls. While conducting backward elimination, changes in ES were examined after each factor was removed to determine if the factor that was removed greatly influenced the ES of the remaining factors or was clinically significant and should be retained in the model. Often no ES change size occurred and in a few instances small ES changes (0.01 to 0.02) occurred. Consequently, only significant factors were retained in the final model. Findings indicate that cancer patients were more likely to fall compared to those without cancer (OR = 1.16; and 95% CI: 1.02, 1.33; p = .03). Males are more likely to fall compared to females; Caucasians were more likely to fall compared to other races; taking antidepressants compared to not taking antidepressants; having short-term memory recall problems compared to 93 little or no memory recall problems; evidence of pain daily compared to no pain daily; those who experienced weight loss compared to no weight loss; or with additional comorbidities were more likely to fall (ORs = 1.03 to 1.56; 95% CI: 0.83 to 1.41, 1.12 to 1.77). Table 12 Factors Associated with Falls in the Final GEE Model Explanatory variable (reference level) Odds Ratio Wald 95% Confidence Interval p-value Cancer (vs. no cancer) 1.16 1.01 1.33 0.03* Male (vs. female) 1.12 1.03 1.22 0.01* African American (vs. other) 0.76 0.61 0.96 0.02* Antidepressant (vs. none) 1.29 1.19 1.40 <.00* Memory recall (vs. no problem) 1.53 1.41 1.65 <.00* Pain 85 with cancer who have a fracture 4.4% (n = 38) compared to non-cancer with fractures 3.9% (n = 323);  Females with cancer who have a fracture 66.7% (n = 86) compared to Females without cancer who have a fracture 70.0% (n = 864);  African Americans with cancer who have a fracture 19.2% (n = 8) compared to African Americans without cancer who have a fracture 22.6% (n = 194);  Short term memory recall problems with cancer who have a fracture 49.4% (n = 19) compared to those with short term memory recall problems without cancer who have a fracture 52.9% (n = 443);  Evidence of daily pain with cancer who have a fracture 77.6% (n = 30) compared to those who have evidence of daily pain without cancer who have a fracture 77.7% (n = 643); 97  ADL dependence sum mean with cancer who have a fracture 16.52 compared to without cancer who have a fracture 14.36; and  Comorbidities mean with cancer and a fracture 1.91 compared to without cancer who have a fracture 2.14. Table 14 Factors Associated with Fractures in the Final GEE Model Explanatory variable (reference level) Odds Ratio Wald 95% Confidence Interval p-value Cancer (vs. no cancer) 0.01 0.00 0.01 0.97 Age 65-69 (vs. >85) Age 70-74 (vs. >85) Age 75-79 (vs. >85) Age 80-84 (vs. >85) 0.78 0.44 0.68 0.98 0.74 1.11 0.66 0.95 1.16 1.13 0.98 1.43 0.75 0.30 0.45 0.13 Male (vs. female) 0.76 0.67 0.69 <.00* African American (vs. other) 0.36 0.26 0.51 <.00* Memory recall (vs. no problem) 0.91 0.83 1.00 <.04* Pain daily (vs. no pain) 1.28 1.17 1.40 <.00* ADL dependent (vs. independent) 1.01 1.01 1.02 <.00* Comorbidity (vs. none) 1.61 1.11 1.22 <.00* *p <0.05 . ER use. Table 15 summarizes the final multivariate GEE model identifying factors associated with ER use. Changes in ES were examined after each factor as it was removed from the model. 98 Table 15 Factors Associated with Emergency Room Use in the Final GEE Model Explanatory variable (reference level) Odds Ratios Wald 95% Confidence Interval p-value Cancer (vs. no cancer) 1.24 1.04 1.48 0.02* Age 65-69 (vs. >85) Age 70-74 (vs. >85) Age 75-79 (vs. >85) Age 80-84 (vs. >85) 0.73 0.84 0.52 0.71 0.60 0.71 0.40 0.51 0.88 1.00 0.77 0.99 0.00 0.05 0.00 0.05 Memory recall (vs. no problem) 1.51 1.03 1.29 0.01* Pain daily (vs. no pain) 1.39 1.21 1.60 <.00* Weight loss (vs. none) 1.54 1.29 1.84 <.00* Comorbidity (vs. none) 1.12 1.07 1.16 <.00* *p <0.05 The ES size had no or small change after each factor until age were removed, then moderate ES changes (.20 to .40) were observed in the remaining factors. Consequently, age was retained in the final model and clinically as age increased a person was more likely to use the ER. Findings indicate that cancer patients were more likely to use the ER than those without cancer (OR = 1.24; and 95% CI: 1.04, 1.48). Furthermore, those with short-term memory recall problems compared to those with little or no recall problems, with daily pain compared to those without daily pain; with weight loss compared to no weight loss; and with more comorbid conditions had higher ER use (OR = 1.12; 95% CI: 1.03 to 1.29, 1.16 to 1.80). A similar model taking time since cancer diagnosis into account was developed; however, the number of days since the cancer diagnosis was not part of the final model and ES did not change in the remaining factors. 99 The frequencies of each factor in the final model in relation to ER use are as follows:  ER use in those with cancer14.1% (n = 122) compared to ER use in those without cancer 10.8% (n = 933);  Age: o Age 65-69 with cancer who use the ER 0.9% (n = 8) compared to non-cancer ER use 1.0% (n = 79), o Age 70-74 with cancer who use the ER 2.2% (n = 24) compared to non-cancer ER use 2.2% (n = 201), o Age 75-79 with cancer who use the ER 2.9% (n = 23) compared to non-cancer ER use 2.7% (n = 220), o Age 80-84 with cancer who use the ER 3.7% (n = 29) compared to non-cancer ER use 2.6% (n = 289), and o Age >85 with cancer who use the ER 4.4% (n = 38) compared to non-cancer ER use 2.3% (n = 144);  Short term memory recall problems with cancer who use the ER 49.3% (n = 54) compared to short term memory recall problems without cancer 46.8% who use the ER (n = 348);  Evidence of daily pain with cancer who use the ER 85.0% (n = 94) compared to evidence of daily pain without cancer who use the ER 76.7% (n = 570);  Weight loss with cancer who use the ER 22.4% (n = 25) compared to weight loss without cancer who use the ER 13.2% (n = 101); and  Comorbidities mean with cancer who use the ER 1.91 compared to without cancer who use the ER 2.14. 100 Hospitalization. Table 16 summarizes the final multivariate GEE model identifying factors associated with hospitalizations. While conducting backward elimination, changes in ES were examined after each factor was removed to determine if the factor that was removed greatly influenced the ES of the remaining factors or was clinically significant and should be retained in the model. Often no ES change size occurred and in a few instances small ES changes (0.01 to 0.02) occurred. Consequently, only significant factors were retained in the final model. Table 16 Factors Associated with Hospitalization in the Final GEE Model Explanatory variable (reference level) Odds Ratios Wald 95% Confidence Interval p-value Cancer (vs. no cancer) 1.89 1.66 2.16 <.00* Age 65-69 (>85) Age 70-74 (>85) Age 75-79 (>85) Age 80-84 (>85) 1.30 1.40 1.69 1.96 1.12 1.21 1.25 1.46 1.51 1.62 2.29 2.63 0.00* <.00* 0.00* <.00* Anti-anxiety (vs. none) 1.14 1.03 1.26 0.01* Antidepressants (vs. none) 0.83 0.77 0.92 0.00* Pain daily (vs. no pain) 1.31 1.18 1.45 <.00* Weight loss (vs. none) 2.29 2.00 2.63 <.00* ADL dependent (vs. independent) 1.02 1.02 1.03 <.00* Comorbidity (vs. none) 1.23 1.19 1.27 <.00* *p <0.05 Those with cancer (OR = 1.89; and 95% CI: 1.66, 2.16), aging, taking anti-anxiety medication compared to taking none, with daily pain compared to less than daily pain, weight 101 loss compared to no weight loss, ADL dependence compared to independence or minimal dependence, and those with increasing comorbidities were hospitalized more often (OR = 1.02 to 2.29; 95% CI: 1.02 to 2.0, 1.26 to 2.63).Those taking antidepressants were hospitalized less often (OR = 0.83; 95% CI: 0.77, 0.92). A similar model taking days since cancer diagnosis into account was developed; however, the number of days since cancer diagnosis did not influence the rate of hospitalizations in this sample. The frequencies of each factor in the final model in relation to hospitalization are as follows:  Those with cancer with hospitalizations 41.6% (n = 359) compared to without cancer with hospitalizations 23.5% (n = 2028);  Age: o Age 65-69 with cancer with hospitalizations 6.7% (n = 58) compared to noncancer with hospitalizations 6.3% (n = 149), o Age 70-74 with cancer with hospitalizations 9.0% (n = 78) compared to noncancer with hospitalizations 4.7% (n = 401), o Age 75-79 with cancer with hospitalizations 11.7% (n = 101) compared to noncancer with hospitalizations 4.6% (n = 398), o Age 80-84 with cancer with hospitalizations 14.1% (n = 122) compared to noncancer with hospitalizations 5.7% (n = 493), and o Age >85 with cancer with hospitalizations 16.2% (n =140) compared to noncancer with hospitalizations 6.8% (n = 587); 102  Those taking antianxiety medications with cancer with hospitalizations 23.9% (n = 72) compared to those taking antianxiety medications without cancer with hospitalizations 19.3% (n = 107);  Evidence of daily pain with cancer with hospitalizations 80.7% (n = 244) compared to evidence of daily pain without cancer with hospitalizations 76.2% (n = 423);  Short term memory recall problems with cancer with hospitalizations 46.0% (n = 139) compared to short term memory recall problems without cancer with hospitalizations 47.4% (n = 264);  ADL dependence sum mean with cancer with hospitalizations 16.52 compared to without cancer with hospitalizations 14.36; and  Comorbidities mean with cancer with hospitalizations 1.91 compared to without cancer with hospitalizations 2.14; and  Those taking antidepressants medications with cancer with hospitalizations 35.9% (n = 107) compared to those taking antidepressants medications without cancer with hospitalizations 32.6% (n = 181). Nursing home placement. Table 17 summarizes the final multivariate GEE model identifying factors associated with nursing home placement. ES were examined after each factor was removed. Often no ES change size occurred and in a few instances small ES changes (0.01 to 0.02) occurred. Consequently, only significant factors were retained in the final model. Findings indicate that cancer patients were less likely to be placed in a nursing home (OR = 0.61; and 95% CI: 0.50, 0.74). Aging increased nursing home placement, and African Americans and Caucasians had a lower rate of nursing home placement than other races (OR = 0.54 to 0.91; and 95% CI: 0.40 to 0.68, 0.72 to 1.24). Patients taking hypnotics or anti-anxiety 103 medication compared to not taking any; who had short-term memory recall problems compared to minimal or no problems; , evidence of daily pain compared no daily pain; or poor vision compared to good vision were more likely to be placed in a nursing home (OR = 1.00; and 95% CI: 1.00, 1.00). A similar model taking time or the count of days since cancer diagnosis into account was developed, as well as nursing home placement. The final model did not retain any factors associating the time since cancer diagnosis and nursing home placement. Table 17 Factors Associated with Nursing Home Placement in the Final GEE Model Explanatory variable (reference level) Odds Ratios Wald 95% Confidence Interval p-value Cancer (yes) 0.61 0.50 0.74 <.00* Age 65-69 (>85) Age 70-74 (>85) Age 75-79 (>85) Age 80-84 (>85) 0.99 0.99 0.97 0.98 0.99 0.99 0.97 0.97 0.99 0.99 0.98 0.99 <.00* <.00* <.00* <.00* African American (vs. other) 0.54 0.40 0.72 <.00* Anti-anxiety (no) 1.00 1.00 1.00 0.00* Hypnotics (no) 1.00 1.00 1.00 0.00* Evidence of pain present (severe) Evidence of pain severe (severe) 1.00 1.00 1.00 1.00 1.00 1.00 0.03* <.00* Memory recall (vs. no problem) 1.00 1.00 1.00 <.00* Vision (vs. no problem) 1.00 1.00 1.00 0.00* *p <0.05 104 The frequencies of each factor in the final model in relation to nursing home placement are as follows:  Those with cancer with nursing home placement 20.7% (n = 179) compared to those without cancer with nursing home placement 30.6% (n = 2637);  Age: o Age 65-69 with cancer with nursing home placement 2.7% (n = 23) compared to non-cancer with nursing home placement 3.3% (n = 286), o 7 Age 70-74 with cancer with nursing home placement 2.8% (n = 24) compared to non-cancer with nursing home placement 5.8% (n = 497), o Age 75-79 with cancer with nursing home placement 3.2% (n = 28) compared to non-cancer with nursing home placement 6.7% (n = 578), o Age 80-84 with cancer with nursing home placement 5.3% (n = 46) compared to non-cancer with nursing home placement 7.3% (n = 632), and o Age >85 with cancer with nursing home placement 6.7% (n =58) compared to non-cancer with nursing home placement 7.5% (n = 644);  African American with cancer with nursing home placement 44.8% (n = 64) compared to African American without cancer with nursing home placement 40.4% (n = 170);  Hypnotic medications taken with cancer with nursing home placement 9.8% (n = 13) compared to hypnotic medications taken without cancer with nursing home placement 6.7% (n = 53);  Antianxiety medications taken with cancer with nursing home placement 21.5% (n = 29) compared to antianxiety medications taken without cancer with nursing home placement 20.6% (n = 82); 105  Short term memory recall problems with cancer with nursing home placement 52.5% (n = 73) compared to short term memory recall problems without cancer with nursing home placement 49.9% (n = 203);  Evidence of daily pain with cancer with nursing home placement 73.8% (n = 100) compared to evidence of daily pain without cancer with nursing home placement 79.1% (n = 316);  Vision problems with cancer with nursing home placement 32.6% (n = 44) compared to vision problems without cancer with nursing home placement 41.7% (n = 166);  ADL dependence sum mean with cancer with nursing home placement 16.52 compared to ADL dependence without cancer with nursing home placement 14.36; and  Comorbidities mean with cancer with nursing home placement 1.91 compared to comorbidities mean without cancer with nursing home placement 2.14; and  Antidepressants medications taken with cancer with nursing home placement 43.7% (n = 59) compared to antidepressant medications taken without cancer with nursing home placement 32.8% (n = 127). Cancer type, stage, and falls. Next, models to answer the second part of Research Question 1 were developed asking if cancer type or stage influenced falls. Models for each will be described. To examine if the type of cancer influenced falls, the following statistical model was implemented: falls (yes) = cancer type (breast, prostate, colon, lung, or other) + (plus) covariates (age, race, sex, medications, and frailty [ADL summed score, comorbidity summed score, evidence of pain, short-term memory loss, weight loss, and vision]). Type of cancer was not retained in the model (p > .05) and did not influence falls in this sample. 106 To examine if the type of cancer (breast, colon, prostate, lung, or other) or the cancer stage influenced the health outcome falls in this study additional GEE models were developed. The following statistical model was implemented: falls (yes) = cancer stage (I, II, III, or IV) + (plus) covariates (age, race, sex, medications, and frailty [ADL summed score, comorbidity summed score, evidence of pain, short-term memory loss, weight loss, and vision]). Cancer stage was not retained in the model (p > .05) and did not influence falls in this sample. Research Question 2. Research Question 2 is as follows: to examine if the effects of frailty variables on falls was different with respect to type or stage of cancer. The following is a summary of the factors examined for this question.  Cancer stages: o Stage I 5.3% (n = 46), o Stage II 44.9% (n = 388), o Stage III 20.3% (n = 175), and o Stage IV 28.7% (n = 250);  ADL summed score with cancer 16.52 and 14.36 without cancer;  Comorbidity summed score with cancer 1.91 and without cancer 2.14;  Evidence of pain with cancer 21.5% (n = 186) and without cancer 21.0% (n = 1809);  Short-term memory loss with cancer 45.5% (n = 393)and without cancer 51.9% (n = 4471);  Weight loss with cancer 16.3% (n = 141)and without cancer 6.7% (n = 576); and  Vision with cancer 10.9% (n = 94) and without cancer 13.0% (n = 1121). The following statistical models were employed to examine if the influence of frailty differs by type of cancer. For those with a cancer diagnosis: falls = (equal) frailty (ADL summed score, 107 comorbidity summed score, evidence of pain, short-term memory loss, weight loss, and vision) plus site of cancer + (plus) frailty (ADL summed score, comorbidity summed score, evidence of pain, short-term memory loss, weight loss, and vision) x (times) site of cancer + (plus) other covariates (age, race, sex, and medications). The interaction between frailty x (times) the type of cancer (frailty [ADL summed score, comorbidity summed score, evidence of pain, short-term memory loss, weight loss, and vision] x [times] site of cancer) was not significant (p > .05) in the first model, so it was removed. Then the additive effects of the site of cancer and frailty were explored. After six iterations of the model, the site of cancer was the least significant variable and was removed from the model. The effect of the frailty variables on falls was no different (p > .05) with respect to site of cancer in this sample. The following statistical models were employed to examine the influence of the cancer stage. For those with a cancer diagnosis: falls = (equal) frailty (ADL summed score, comorbidity summed score, evidence of pain, short-term memory loss, weight loss, and vision) + (plus) site of cancer + (plus) frailty (ADL summed score, comorbidity summed score, evidence of pain, short-term memory loss, weight loss, and vision) x (times) stage of cancer + (plus) other covariates (age, race, sex, and medications). The essential parameters investigated for this aim are associated with the interaction term in the statistical model. The interaction between frailty x (times) the stage of cancer (frailty [ADL summed score, comorbidity summed score, evidence of pain, short-term memory loss, weight loss, and vision] x [times] stage of cancer) was not significant (p > .05), so it was removed from the model. Then the additive effects of the stage of cancer and frailty were explored. After seven iterations of 108 model-building, the stage of cancer was the least significant variable. The effect of the frailty variables on falls was not different (p > .05) with respect to stage of cancer in this sample. Research Question 3 Exploration of environment and social status. An exploratory analysis was conducted examining the social variables described to better understand the population. These variables include living arrangement, relationship with caregiver, length of time alone per day, desire for another living environment, and pets. The factors were examined using descriptive statistics and Chi-square. As shown in Table 11, the majority of patients without cancer (90.1%) were living alone, while most with cancer (60.5%) live with another person, a significant difference (p = .00). However, there is a higher rate of falls in those with cancer (59.9% to 11.5%) than in those without cancer when living with another person (p = .00). No differences were found in marital status between those with cancer compared to those without cancer, with most patients being widowed (55.8% to 55.5%) or married (26.6% to 23.5%) in both groups. Furthermore, minimal differences between groups were found in type of living environment, with most patients living in a house (59.6% to 60.2%) or apartment (33.6% to 30.9%) respectively (p = .67). The amount of time a patient spent alone during the day was then examined. Cancer patients who were never left alone fall at a higher rate (44.8% to 37.2%; p = .00). Conversely, a lower frequency of falls occurred in patients with cancer who are alone all of the time compared to those without cancer (15.4% to 20.2%). No difference among any of the groups was found when examining whether the patient had feelings of loneliness (p = .83). Overall, a higher frequency of loneliness was found in those who fall (23.7%, n = 686) compared to those who did not fall (18.6%, n = 1166). Finally, no difference (p = .22) among the groups was found on whether the patient had a pet. 109 What follows in the next chapter is a detailed discussion of these findings. This will be followed by discussion of implications for nursing practice and research. Finally, conclusions will be drawn from this research. 110 CHAPTER 6: DISCUSSION The elderly are at high risk for functional decline and frailty, which may be influenced by their cancer history (Alibhai et al., 2006; Barsevick et al., 2006; Bylow et al., 2006; Diemling et al., 2007; Levy et al., 2008; Overcash, 2007; Pautex et al., 2008) and may lead to falls or fractures with subsequent health care use. Evidence is emerging in the literature examining disparities and an increased rate of falls among cancer survivors (Hewitt et al., 2003; Keating et al., 2005; Koroukian et al., 2006; Luktar-Flude, 2007; Walston et al., 2006; Yabroff et al., 2007). The majority of this research has focused on global differences in function and was either conducted in the hospital setting, after certain types of treatment, or with terminal patients, and did not directly examine if falls or fractures differ in cancer survivors compared to those without cancer. The purpose of this research was to conduct an examination of how the addition of a cancer diagnosis alters the rate of falls, fractures, and use of health care among cancer survivors, compared to those without a cancer diagnosis. Furthermore, this study sheds light on whether the cancer stage, site, or treatment alters the rate of falls and fractures in elderly survivors. In this chapter a summary of the research study findings will be presented. Each research question will be discussed and literature will be presented supporting the findings in this study, and when literature is absent a rationale to support the findings will be provided. This includes falls, fractures, health care use (ER, hospital, and nursing home); falls and site, stage, or treatment; site and stage of cancer, frailty, and falls; and the exploratory question. Each section will end with a summary of findings and conclusions. Study limitations will then be presented, and nursing and research implications identified. Finally, conclusions drawn from this study will be presented. 111 Main Findings in this Study Cancer survivors were more likely to fall than those without a cancer diagnosis, and a more recent cancer diagnosis increased the rate of falls supporting the belief that some cancer survivors experience a higher rate of falls and that the fall rate increased closer to the date of cancer diagnosis. However, the rate of falls did not differ with respect to site of cancer or cancer stage, or frailty level. Although the rate of falls was somewhat higher, cancer survivors had a lower rate of fractures than those without a cancer diagnosis. Cancer survivors used the ER and hospital more often; while a lower rate of nursing home placement was found in cancer survivors compared to those without cancer. Research Study Findings Findings from this study of cancer survivors compared to those without cancer will be presented after adjusting for sociodemographic characteristics (age, sex, race or ethnicity), medications, and frailty (ADLs, cognition, comorbidities, pain, weight loss, and vision). This is followed by a presentation of cancer-specific findings examining if there were differences in the number of falls among cancer patients in the year following diagnosis according to site of cancer (breast, colon, prostate, lung, or other), stage of cancer (I-IV), or cancer treatment (chemotherapy and/or radiation). Cancer-specific findings will be presented after adjusting for sociodemographic characteristics (age, sex, race or ethnicity), days since cancer diagnosis, medications, and frailty (ADLs, cognition, comorbidities, pain, weight loss, and vision). During analysis an overriding constraint in the distribution of the data in this study was identified, which is discussed in this chapter as it affects analysis of this data. In this sample, many of the patients experienced similar comorbidities and frailty leading to functional disabilities, as ADL and IADL needs are required for entry into the HCBS program. In some 112 instances statistical differences in ES were found between cancer survivors and those without a cancer diagnosis in this study. However, these findings were derived from data skewed around a small range of potential variance limiting the ability to detect statistical differences in ES between cancer survivors and those without a cancer diagnosis. When this constraint influenced the ability to examine a factor, it will be reported. If this study were to be conducted again, use of a more diverse group of patients with a broader range of functional abilities or a tool with more precise measurement capabilities may be needed to identify the extent to which cancer survivors differ from those without a cancer diagnosis. Summary of findings for Research Question 1. The first Research Question 1 asked: Did patients with a cancer diagnosis experience a greater number of falls, fractures, ER use, hospitalization, or nursing home placement, compared with those patients with no diagnosis of cancer? The second part of Research Question 1 asked: Are differences in the occurrence of falls among cancer patients according to type of cancer (breast, colon, prostate, lung, or other), stage of cancer (I-IV), or cancer treatment (chemotherapy and/or radiation)? Findings for each question are presented in separate sections to include falls, fractures, healthcare use, cancer site, stage, and treatment, and exploratory findings. At the end of each section a summary and conclusions are provided. Falls. In this study, the overall fall rate was 29.6%. This fall rate is similar to the fall rate found in other studies of the elderly (Davison & Marrinan, 2007). When conducting GEE modeling, several factors were associated with a higher rate of falls (Table 12). This included the presence of a cancer diagnosis compared to those with no cancer; being male compared to being female; being White compared to African American. In addition, taking antidepressants compared to not taking antidepressants, short-term memory recall problems compared to little or 113 no memory recall problems, evidence of pain daily compared to no pain or less than daily pain, weight loss compared to no weight loss, and increasing occurrence of comorbidities. However, the covariates of age, ADLs, vision, antipsychotics, anti-anxiety, and hypnotics did not remain in the final model (p > .05). Each of these findings will be discussed individually. Cancer and falls. Cancer survivors‘ fall rate was 32.7% compared to 29.4% in those without a cancer diagnosis, a significant difference (p = .01) (Table 11). In the GEE model, a small ES difference (OR = 1.16; 95% CI: 1.01, 1.33) was found in those having a cancer diagnosis compared to those without cancer. Consequently, these findings imply that the prevalence of falls tended to increase reflecting variation in the fall rate of cancer survivors when compared directly to those without a cancer diagnosis. An increased rate in the occurrence of falls among cancer survivors is beginning to emerge in the literature (Hewitt et al., 2003; Keating et al., 2003; Keating et al., 2005; Koroukian et al., 2006). Results are in agreement with a study of community-dwelling elderly cancer patients, where 21% had a fall (Overcash, 2007); and with the three studies that exist in palliative care, where 18% (Pautex, et al., 2008), 11.8% (Flood, et al., 2006), and 10% (Pearse, et al., 2004) had a fall. However, in each of these previous studies, no comparison was made to those who did not have cancer to distinguish if the fall rate in the cancer patients was the norm in that population. Furthermore, the low fall rate in the palliative care studies may have been attributed to a limited amount of time the patients were out of bed. Results in this study extend what is known providing new information on the fall rate of cancer survivors in the community-dwelling setting and directly comparing cancer survivors to those without a cancer diagnosis. These findings provide useful information for clinicians who care for elderly cancer survivors regarding the increased prevalence of falls. 114 Sex and falls. Sex contributed to the prevalence of falls, with males more likely to fall than females. Overall, there were more females (68.7%) in this study; however, there were more males present in the cancer group (38.4%) compared to the non-cancer group (29.9%) (Table 2). In general being male was associated with an increased rate of falls (p = .00), and being male with cancer was associated with an increased rate of falls (p = .00). These results are consistent with the literature. For example, although elderly women are more likely to have a fall injury, men are more likely to have a fall (Stevens, et al., 2006). Furthermore, males undergoing treatment for prostate cancer (Alibhai et al., 2006; Chen et al., 2005; Michaud & Goodin, 2006) have increased functional limitations that may lead to falls. The potential additive effect of being male with a cancer diagnosis or cancer treatment may interact to limit functional abilities and increase falls in cancer survivors. Race or ethnicities and falls. Race or ethnicities also contributed to the prevalence of falls, with Whites more likely to fall than African Americans or those of other races (p = .00). Results are in agreement with the literature, with Whites more likely to fall than African Americans or other races or ethnicities (Stevens & Sogolow, 2005; Centers for Disease Control, September 21, 2007). The potential additive effect of race or ethnicities with a cancer diagnosis or cancer treatment may interact to increase falls in cancer survivors. Antidepressants and falls. Antidepressant medication use contributed to the increased occurrence of falls in general and in this study. The fall rate in cancer survivors who take antidepressants was 42.2% compared to the fall rate of 29.2% in those who do not take antidepressants (Table 5). Antidepressant use is strongly associated with falls in the elderly (Agostini et al., 2004; Harrison et al., 2001; Liepzing et al., 1999a; Rubenstein et al., 2003) and results in this study are in agreement with these previous findings. The potential additive effect 115 of antidepressants with a cancer diagnosis or cancer treatment may interact to increase falls in cancer survivors. Evidence of daily pain and falls. Evidence of daily pain contributed to the prevalence of falls in this study. Overall, a higher rate of falls was found in those overall who had daily pain (p < .00) and in cancer survivors with daily pain (p < .00) (Table 7). In the final GEE model, the presence of daily pain increased the odds of falls (Table 12). The occurrence of pain in combination with fatigue is known to influence health deterioration (Barsevick et al., 2006; Deimling et al., 2006; Luctkar-Flude et al., 2007; Moreh, Jacobs, & Stessman, 2010; Shega, Weiner, Paice, Bilir, Rockwood, Herr, K., et al., 2010), functional decline (Deimling et al, 2007), and disability (Balducci & Extermann, 2000a, 200b; Wenger et al., 2003), which may lead to falls. Deimling and colleagues (2007) found a 21% variation in function related to age and a 6% variation in function related to cancer, supporting the findings in this study that the additive effect in elderly cancer survivors leads to decline in function and increased falls. Short-term memory recall and falls. Short-term memory recall problems contributed to the prevalence of falls (Table 12). Overall, more short-term memory recall problems were found in patients who fall (p < .00), and an increased rate of falls was associated with cancer survivors with short-term memory recall problems (p < .00) (Table 8). Cognitive status decline is associated with falls in the elderly (Chen et al., 2008; Inouye et al., 2007; Sibritt et al., 2007). Cognitive impairment in conjunction with pain has also been associated with decrease in function (Shega et al., 2010). Furthermore, one study found cognitive impairment to be associated with an increased rate of falls and also reoccurring falls in oncology patients (Flood et al., 2006). 116 Results in this study, short term memory was associated with falls, are in agreement with these previous studies (Chen et al., 2008; Inouye et al., 2007; Sibritt et al., 2007). Although evidence exists in the literature on the effect of chemotherapy treatment on cognitive impairment and functional decline, limited information is available on long-term cancer survivor cognitive impairment. This study extends the literature providing new information. However, further study is needed on cognitive impairment in cancer survivors. Weight loss and falls. Those who experienced weight loss problems (5% or more in the past 30 days or 10% or more in the past 180 days) were more likely to fall (21.9% in cancer survivors compared to 9.4% in those without a cancer diagnosis) (Table 9). Overall, weight loss was strongly associated (p < .00) with falls among cancer survivors. Weight loss is a key component of frailty in the elderly (Agostini, 2004; Chen et al., 2008), a common ailment during cancer treatment or at the end stage cancer (Pautex et al., 2008; Pearce & Ryan, 2008), and often labeled as the syndrome of cachexia. Weight loss is strongly associated with an increase in the occurrence of falls, and results in this study are in agreement with previous studies in the elderly and in those with cancer. However, further study is needed on weight loss in cancer survivors to determine if this occurs during treatment, if it is a long-term effect of cancer treatment, or if it only occurs during the terminal phase of care. Comorbidities and falls. Finally, the number of comorbidities (diabetes, CHF, CAD, CVA, arthritis, and depression) was found to contribute to the prevalence of falls. In this study, 89.3% of the patients had a comorbid condition. Cancer survivors had similar rates of diabetes, CHF, CAD, and depression as those without cancer; however, CVAs and arthritis occurred more often in those without cancer (Table 6). Overall, an increased rate of falls was associated with 117 arthritis (p = .04), diabetes (p = .00), depression (p < .00), CHF (p = .03), CAD (p = .03), and a CVA (p = .01). Results in this study are in agreement with evidence from studies where heart disease, arthritis, diabetes, strokes, and depression are known to increase the risk of falls (Anstey et al., 2008; Baldwin et al., 2006; Hewitt et al., 2003; Keating et al., 2008; Klabunde et al., 2006; Klabunde et al., 2007; Koroukian et al., 2006; Yabroff et al., 2007; Yancik, 1997; Yancik et al., 1996; Yancik & Ries, 2000). What is not known is if the additive effect of comorbidities and a cancer diagnosis increases the rate of falls in cancer survivors and further study is needed. This study provides new information and adds to the literature on comorbidities in cancer survivors and the interactive effect of comorbidities on falls in this elderly sample. Age, activities of daily living, vision, and antipsychotics, anti-anxiety, and hypnotic medications. Conversely, the covariates of age, ADLs, vision, antipsychotics, anti-anxiety, and hypnotics did not remain in the final model examining falls. Evidence in the literature presents a different view from that which was found in this study, and each will be discussed individually (Agostini et al., 2004; Chen et al., 2008; Davison, 2007; Fauth et al., 2007; Glough-Gorr, et al., 2008; Inouye et al., 2007). Age and falls. Overall, age (p = .02) was associated with falls, and age was associated with falls in the cancer group (p = .01). A major risk factor for falling is aging (Davison, 2007; Fauth et al., 2007), and evidence exists that those aged 85 years or older have increased frailty (Fried et al., 2004; Iezzoni & Freedman, 2008; Stevens et al., 2006), which may lead to falls. Evidence also exists to the contrary, with some studies of frailty proposing that the level of functional disability, not age, is associated with decline in functional status (Ahmed et al., 2007; Yancik & Ries, 2000; Yancik et al., 2001), and results in this study support this evidence. 118 Activities of daily living and falls. The level of ADL dependence did not contribute to an increased occurrence of falls in this study. When examining the ADL items individually, a moderate difference in bathing, dressing, toileting, transfer, and locomotion was found in cancer survivors compared to those who did not have cancer (Table 9). However, those without a cancer diagnosis had greater ADL dependence needs for bladder and bowel incontinence, mobility in bed, eating, and stair climbing compared to those who had cancer (Appendix B, Table 26). When comparing the summed ADL score (range 0-52) little variation and large standard deviations were found. The cancer survivors had a mean ADL summed score of 13.23 (SD 9.08) and those without a cancer diagnosis had a mean ADL summed score of 14.59 (SD 10.75). Findings in multiple studies suggest that functional decline in the elderly (Agostini et al., 2004; Chen et al., 2008; Glough-Gorr, et al., 2008; Inouye et al., 2007) and among those with cancer (Barsevick et al., 2006; Diemling et al., 2007; Holley, 2002; Keating et al., 2005) may lead to a fall. As stated previously, a constraint in the data for this sample occurred, with limited variance. Use of a more diverse group of patients with a broader range of functional abilities is needed to identify if cancer survivors more closely follow the evidence in the literature. Vision and falls. Vision did not contribute to falls in this study. Overall, a low rate (12.6%) of vision problems was evident in this sample (Appendix B, Table 23). Vision problems are strongly associated with falls in the elderly (Badger et al., 2007; Szabo et al., 2008; Tinetti et al., 2003). Furthermore, preliminary work conducted prior to this study found vision contributed to an increased rate of falls (Spoelstra et al., 2010). The findings in this study were contrary to the literature on the influence of vision problems on falls in the elderly (Badger et al., 2007; Szabo et al., 2008; Tinetti et al., 2003). It is proposed that collinearity with other variables in this study may have been present, thus vision did not remain in the final GEE model. 119 Antipsychotics, anti-anxiety, and hypnotics medications and falls. Finally, antipsychotics, anti-anxiety, and hypnotics did not contribute to falls in this study. Overall, an increased rate of falls associated with each of these medications (antipsychotics [p = .02], anti-anxiety medications [p = .02], and hypnotics [p = .00]) (Table 5) was found. Antipsychotics, anti-anxiety medications, and benzodiazepines are each associated with an increased rate of falls in the elderly (Agostini et al., 2004; Harrison et al., 2001; Liepzing et al., 1999a; Rubenstein et al., 2003); study findings were contrary to the literature. It is proposed that collinearity may have occurred with comorbidities or with antidepressants. GEE modeling of the number of days since cancer diagnosis. The number of days since the cancer diagnosis were included in a second model to determine if cancer survivors with more recent diagnoses had a higher rate of falls (Table 13). The number of days since the cancer diagnosis, short-term memory recall versus no or minimal recall problems, evidence of pain daily versus no or less than daily pain, weight loss versus no weight loss, and comorbidities versus no or fewer comorbidities contributed to falls, while sex, race or ethnicity, and antidepressants did not. In this GEE model examining the number of days since cancer diagnosis, the odds ratios for the covariates, short-term memory recall, pain daily, weight loss, and comorbidities increased dramatically after adjusting for the number of days since the cancer diagnosis. Deimling and colleagues (2007) found that cancer-related factors contributed to 6% of the variation in individual functioning, which may increase fall risk in cancer survivors. Furthermore, in those with early-stage breast cancer, there is evidence of physical impairments and the possibility of increased falls (Cheville et al., 2008; Desanto-Madeya et al., 2007). In a non-cancer study, Shega and colleagues (2010) examined the interactive effect of pain and 120 cognition on function in an elderly population, finding decreased function. Although these studies found functional decline, falls were not reported on in the literature. Results in the current analysis provide useful information for clinicians that the closer a person is to the date of their cancer diagnosis, the higher the risk of falls. Furthermore, the factors found to influence falls in this sample, short-term memory recall problems, daily pain, weight loss, and comorbidities are somewhat similar to cachexia (weight loss, muscle atrophy, fatigue, and weakness), a common cancer syndrome often found in those undergoing cancer treatment or in end stage cancer. Consequently, findings in this study provide direction for future research examining the influence of time since cancer diagnosis on both function and falls. Summary of falls and conclusions. Results provide useful input to inform clinicians that an increased prevalence of falls in cancer survivors may exist and that the amount of time since the cancer diagnosis increases the risk for falls. The additive effect of being male and White, experiencing short-term memory recall problems, daily pain and comorbidities, and taking antidepressants contributed to the increased falls, thus these risk factors should be taken into account by clinicians who care for elderly cancer survivors. However, anti-anxiety, antipsychotics, or hypnotics medications did not follow previous evidence in the literature where both antidepressants, antipsychotics, and benzodiazepines) (Fortinsky et al., 2008; Tinetti, Baker et al., 2008; Tinetti, McAvay et al., 2008) and greater than four medications (Agostini et al., 2004; Harrison et al., 2001; Leipzig et al., 1999a; Rubenstein et al., 2002; Stel et al., 2003) are associated with falls in the elderly, which may have occurred due to collinearity. Further research is needed to examine specific medications and combinations of medications in this population to effectively determine which medications may interact in cancer survivors and increase the rate of falls. In this study ADLs and vision problems did not support 121 the current evidence and a more diverse group of patients with variation or refined measurement in ADLs and vision problems is needed to examine if this finding is true. Fractures. In this section, the fall sequelae (fractures) will be discussed. The fracture rate was higher in those who fall (p < .00); however, an 18.7% fracture rate was found in cancer survivors compared to a 19.9% rate in those without a cancer diagnosis (Table 11). When conducting the GEE modeling, cancer (p > .05) did not remain in the final model. Factors that were associated with a higher rate of fractures included evidence of daily pain compared to no pain or less than daily pain, ADL dependence compared to independence or limited ADL dependence, and comorbidities compared to no or fewer comorbidities (Table 14). Those who were aged were less likely to have a fracture; however, as age increased, the likelihood of a fracture increased. Males, African Americans, and those with short-term memory recall problems were also less likely to have a fracture. The covariates of vision and medications (p > .05) did not remain in the model. Cancer and fractures. The findings in this study demonstrate that the prevalence of fractures did not occur at a higher rate in cancer survivors compared to those who do not have a cancer diagnosis. Some cancer treatments are linked to bone loss (Alibhai et al., 2006; Chen et al., 2005; Delmas & Fontana, 1998; Michaud & Goodin, 2006; Saarto et al., 2001) or vitamin D depletion (Overcash 2008) long after treatment, which may lead to fractures. Additionally, a higher incidence of vertebral fracture (Kanis et al., 1999) and hip fractures (Chen et al., 2008; Chen et al., 2005) in breast cancer survivors has been established. As reported earlier, this study did not include an examination of current cancer treatment due to the small number of patients (n = 49) undergoing treatment, so there was no opportunity to take this factor into consideration. Study findings were contrary to the literature, as some 122 cancer treatments or vitamin D depletion long after treatment is associated with higher incidence of fracture (Alibhai et al., 2006; Chen et al., 2005; Delmas & Fontana, 1998; Michaud & Goodin, 2006; Overcash 2008; Saarto et al., 2001). It is possible patients were admitted to the HCBS program for care for a fracture; consequently, a higher rate of fractures was evident in the noncancer group. Age, sex, and race or ethnicity and fractures. Overall, aging (p = .00), sex (p = 0.00), and race or ethnicity (p = 0.00) were associated with an increased rate of fractures. However, in the GEE model, sex and race or ethnicity contributed to the prevalence of fractures in this study, while aging had a negative effect. Study findings are consistent with the literature on sex and race or ethnicity with females (CDC, 2007) and Caucasians (Stevens et al., 2007) being more likely to have a fracture. However, the literature also supports those who are aged, especially those over age 85 to have more fractures (Sterling et al., 2001). Findings in this analysis are contrary to other studies where those who were older having increased fractures (Sterling et al., 2001). It is proposed that due to the nature of the HCBS program, some older more resilient patients remain in the home setting and do not have fractures. Further study is needed examining the older old in the HCBS program. Short-term memory recall and fractures. Short-term memory recall problems also contributed to the prevalence of fractures, but again not as expected. Overall, more short-term memory recall problems were found in patients who have fractures (p < .00); however, a 54.6% memory recall problem rate was found in those with cancer who had a fall compared to 58.1% in those without a cancer diagnosis (Table 8). When comparing the groups, the short-term memory recall covariate had little variation and large standard deviations. The cancer survivors had a mean of 0.47 (SD 0.50), and those without a cancer diagnosis had a mean of 0.53 (SD 0.50). 123 Contrary to the finding on falls, those with short-term-memory recall problems had fewer fractures. Cognition is known to influence functional status, which is associated with a decline in functional status (Chen et al., 2008; Inouye et al., 2007; Sibritt et al., 2007) possibly leading to fractures in the elderly. Results did not support the evidence in the literature. As stated previously, this may have been due to a constraint in the data with a small range of potential variance in those with cognitive problems so it was difficult to detect differences when comparing the patients in this sample. Therefore future work should look at Cognition in cancer survivors with a more refined measure of the construct. Evidence of daily pain, activities of daily living, and comorbidities and fractures. Those factors contributing to the prevalence of fractures were pain, ADLs, and comorbidities. Evidence in the literature suggests that pain (Balducci & Extermann, 2000a, 200b; Wenger et al., 2003), functional decline (Agostini et al., 2004; Chen et al., 2008; Glough-Gorr, et al., 2008; Inouye et al., 2007), and comorbidities (Anstey et al., 2008; Baldwin et al., 2006; Hewitt et al., 2003; Keating et al., 2008; Klabunde et al., 2006; Klabunde et al., 2007; Koroukian et al., 2006; ScafKlomp et al., 2001; Yabroff et al., 2007; Yancik, 1997; Yancik et al., 1996; Yancik & Ries, 2000) increase falls, which may lead to fractures in the elderly. Results in this analysis support the evidence that pain, functional decline, and comorbidities (Agostini et al., 2004; Balducci & Extermann, 2000a, 200b; Chen et al., 2008; Glough-Gorr, et al., 2008; Inouye et al., 2007; Wenger et al., 2003) increase fractures. Vision, medications, and fractures. Finally, vision and medications did not contribute to fractures in this study. As stated previously, vision problems are a factor associated with falls in the elderly (Badger et al., 2007; Szabo et al., 2008; Tinetti et al., 2003). Antipsychotics, antianxiety medications, and benzodiazepines are also associated with falls in the elderly (Agostini 124 et al., 2004; Harrison et al., 2001; Liepzing et al., 1999a; Rubenstein et al., 2003). Consequently, in the elderly both vision problems (Badger et al., 2007; Szabo et al., 2008; Tinetti et al., 2003) and medication use (Agostini et al., 2004; Harrison et al., 2001; Liepzing et al., 1999a; Rubenstein et al., 2003) would be expected to influence the occurrence of fractures. Results in this study do not support this evidence. It is proposed that due to collinearity with comorbidities and ADLs these factors were not retained in the final GEE Summary of fractures and conclusions. The covariates for fractures were similar to those found in the GEE model for falls, to include sex, race, short-term memory recall problems, daily pain, weight loss, and comorbidities. However, two additional covariates were found in those with fractures, age, and ADLs. These results do not add any new information to the existing literature in regard to the occurrence of fractures in cancer patients and affirm what is in the literature in regard to fractures in the community-dwelling elderly (Alibhai et al., 2006; Chen et al., 2005; Delmas & Fontana, 1998; Michaud & Goodin, 2006; Overcash 2008; Saarto et al., 2001). Activity levels in the elderly are known to influence falls and consequently fractures (Peeters et al., 2010). Future studies need to incorporate measures of the type and intensity of activity level prior to a fracture compared to those without a fracture to better understand what types of interventions may be effective at reducing the occurrence of fractures. Health care use. In this section a discussion of health care use to include ER, hospitalization, and nursing home placement will be presented. This will be followed with how cancer survivors used health care. Finally, a comparison of health care use in cancer survivors who fall to those without a cancer diagnosis will be presented. A summary and conclusions will then be provided. 125 Emergency room use. Overall, 18.2% of those who fall use the ER compared to 8.8% who do not fall (Table 11). Of those, 19.7% with cancer use the ER compared to 17.6% without cancer who fall (Table 11). Several factors were associated with ER use in the final GEE model (Table 15). This included cancer compared to no cancer, short-term memory recall problems compared to no or minimal recall problems, evidence of daily pain compared to no pain or less than daily pain, weight loss compared to no weight loss, and increasing comorbidities. The covariates of age, sex, race or ethnicity, ADLs, vision, and medications (p > .05) did not remain in the model. Hospitalization. Overall, 39.4% of those who fall were hospitalized compared to 29.8% who do not fall (Table 11). Of those, 48.8% with cancer were hospitalized compared to 32.0% without cancer who fall (Table 11). Factors contributing to hospitalization in the final GEE model (Table 16) included cancer versus no cancer, aging, anti-anxiety medication versus no medication, evidence of daily pain versus no or less than daily pain, weight loss versus no weight loss, ADL dependence versus independence or minimal dependence, and comorbidities versus no or less comorbidities. The patients taking anti-depressants versus no antidepressants were hospitalized less often. The covariates of sex, race or ethnicity, and vision did not remain in the model. Nursing home placement. Overall, nursing home placement occurred at a high rate in both groups, 48.5% for those who fall compared to 47.8% who do not fall (Table 11). Of those, 33.3% with cancer who fall were placed in a nursing home compared to 49.8% without cancer who fall (Table 11). Factors contributing to nursing home placement in the final GEE model (Table 17) included anti-anxiety medications compared to no anti-anxiety medications, hypnotics compared to no hypnotics, evidence of daily pain compared to no pain or less than daily pain, 126 short-term memory recall problems compared to no or minimal memory recall problems, and vision problems compared to no vision problems. Those with cancer compared to no cancer, aging, and male compared to female had a lower rate of nursing home placement. The covariates of sex, ADLs, antidepressants, antipsychotics, and comorbidities did not remain in the model. Summary of health care use and conclusions. Falls often lead to functional decline, hospitalization, institutionalization, and higher health care costs (Chen et al., 2007; Chen et al., 2008; Davison & Marrianan, 2007; Tinetti et al., 2005; van Helden et al., 2008). Evidence demonstrates that frailty, disability, and aging are predictive of the use of health resources (Boult et al., 2008; Huss et al., 2008; Koroukian et al., 2006; Melis et al., 2008; Mudge, O'Rourke, & Denaro, 2010; Weiner et al., 2003). Functional decline has been found to be associated with increased use of the ER and subsequent hospitalizations in cancer survivors in a recent study my Mudge and colleagues (2010). Finally, limited evidence exists (French et al., 2007; Kerber, Dyck, Culp, & Buckwalter, 2005; Paquay, De Lepeleire, Schoenmakers, Ylieff, Fontaine, & Buntinx, 2007; Sawyer, Lillis, Bodner, & Allman, 2007) examining HCBS recipients‘ nursing home placement. The intent of the HCBS program is to assist recipients to remain in their home setting in lieu of nursing home placement. This study fills a gap in the literature providing new information on nursing home placement rates of HCBS recipients. Results in this study affirm the evidence in the literature, with cancer survivors who fall having a higher rate of ER use and increased hospitalizations providing new information regarding use of health care services by cancer survivors (Chen et al., 2007; Chen et al., 2008; Davison & Marrianan, 2007; Tinetti et al., 2005; van Helden et al., 2008). However, results on nursing home placement are somewhat different from the literature. Thus studies need to be 127 conducted in the future examining nursing home placement in the elderly cancer survivors using a more varied age group to see what is correct… It is proposed that due to the finding that the cancer survivors have a caregiver with whom they live, that the supports are available for these individuals to remain in the home setting longer than those without cancer—a unique finding for those in the HCBS program. These results also provide new information regarding the use of health care services. In future studies, claims data ICD-9 diagnosis codes and procedure codes should be utilized to determine if a relationship exists between a fall and the use of health care resources in this population. Falls and cancer-specific analysis. In this section cancer-specific analysis will be presented. This includes an examination of falls, site of cancer, and cancer stage. Cancer treatments analysis will also be reported. A summary and conclusion is presented for each. Falls and site of cancer. In this study, 64.4% of the cancer patients had a solid tumor (Table 3). No difference in falls (p = .09) was found based on type of cancer (breast, colon, lung, prostate, or other). When conducting GEE modeling to determine if the site of cancer influenced falls, no association existed between any site of cancer, breast, colon, prostate, lung, or others, and the rate of falls in this sample (p > .05). Furthermore, when conducting GEE modeling including the number of days since the cancer diagnosis, again no association was present between any site of cancer, breast, colon, prostate, lung, or others, and the rate of falls in this sample (p > .05). The findings in this study were somewhat contrary to the literature with no difference found in the rate of falls by site of cancer. Evidence exists that those with breast (Mandelblatt et al., 2006; Cheville et al., 2008; DeSanto-Madeya et al., 2008), lung (Tearce & Ryan, 2008), and prostate cancer (Alibhai et al., 2006; Bylow et al., 2008) have had patterns of physical 128 impairment, which may lead to disability, frailty, and falls. Studies directly examining falls in cancer patients have found that certain types of cancer are associated with higher rates of falls (Pearce & Ryan, 2008; Waltman et al., 2006; Waltman et al., 2007). Future studies should focus on physical function in various sites of cancer across the cancer trajectory. Although the site and stage of the cancer were known in this study, the findings were inconsistent with the literature. This study did not have available information on the treatment of the cancer and if the cancer was a reoccurrence or secondary site. It is proposed that more refined measures may be needed to detect if differences exist in cancer survivors by site and stage of cancer. Future research on falls in cancer survivors needs to incorporate specific details to determine whether the site of the cancer is a secondary site or reoccurrence, how the cancer has progressed or responded to treatment, and whether treatment occurred to effectively measure if falls vary by cancer site. Falls and stage of cancer. Only 50.7% were diagnosed with stage I-II cancer, while 49.3% were diagnosed with stage III-IV (Table 4). When conducting generalized linear models, a difference in falls (p = .03) based on stage of cancer was found, with the rate of falls increasing as the stage of the disease progressed. When conducting GEE modeling controlling for covariates, the cancer stage did not remain in the final model in this sample (p > .05). When conducting GEE modeling including the number of days since the cancer diagnosis, again no association was found between any stage of cancer and the rate of falls in this sample (p > .05). Evidence exists that those with early-stage breast cancer (Mandelblatt et al., 2006) and metastatic breast cancer (Cheville et al., 2008; DeSanto-Madeya et al., 2008) have had patterns of physical impairment that may lead to an increase in falls. One study of metastatic disease found a higher rate of falls (Waltman et al., 2006). Study findings were somewhat contrary to the 129 literature, finding no difference in falls based on stage of cancer. Although the stage of the cancer was known, it is proposed that due to the lack of information on the treatment of the cancer and if it was a reoccurrence or secondary site, this study was unable to detect if the cancer stage influenced the fall rate in this sample. It is also proposed the collinearity may have occurred due to the similarity of the level of frailty across all patients in this sample. Falls and cancer treatment. In this sample, only 49 (1.7%) of the patients with cancer (Given et al., 2010) were identified as being treated (chemotherapy, radiation, or both chemotherapy and radiation) when examining the claims files. Due to the small number of patients under treatment in this sample, examining the influence of treatment on the outcome variables was not feasible. Some cancer treatments are linked to bone loss (Alibhai et al., 2006) or vitamin D depletion after treatment, which may lead to falls and fractures (Coleman et al., 2007; Greenspan et al., 2007; Waltman et al., 2006). The literature (Alibhai et al., 2006; Coleman et al., 2007; Greenspan et al., 2007; Waltman et al., 2006) provides support for future research examining the additive effect of cancer treatments on falls or fractures in cancer survivors. Use of claims and pharmacy data to incorporate detailed information regarding the type and amount of treatment is needed to explore how cancer treatments might influence falls and fractures to identify if cancer survivors who had recently been treated, or have specific types of treatment follow the evidence in the literature and have increased falls. Furthermore future research needs to examine how type of, duration of, and intensity of cancer treatments influence falls during and after treatement (both short-term [1year] and long-term [>1 year]). Summary of Research Question 1. Research Question 1 asked: Did patients with a cancer diagnosis experience a greater number of falls, fractures, ER use, hospitalization, or 130 nursing home placement, compared with those patients with no diagnosis of cancer? Findings for Research Question 1 are summarized in Table 18. This study agreed with and extends the science finding that an increased prevalence of falls in cancer survivors may exist. These findings provide useful information for clinicians who care for elderly cancer survivors regarding risk factors to be aware of. This includes being male, Caucasian, using antidepressants, having short-term memory recall problems, daily pain, weight loss, or having comorbidities. Furthermore, the closer a person is to the date of their cancer diagnosis, the higher the risk of falls. Table 18 Summary of Findings for Research Question 1 Outcomes Key Variable Significant Covariates Falls Cancer Sex, race or ethnicities, antidepressants, short-term memory recall, daily pain, weight loss, & comorbidities Cancer & Days Since Cancer Diagnosis Short-term memory recall, daily pain, & comorbidities Fractures - Age, sex, race or ethnicities, short-term memory recall, daily pain, ADLs, & comorbidities, ER Use Cancer Short-term memory recall, daily pain, weight loss & comorbidities Hospitalization Cancer Age, anti-anxiety medications, antidepressants, daily pain, weight loss, ADLs, & comorbidities Nursing Home Placement - Age, race or ethnicities, anti-anxiety medications, hypnotics, daily pain, short-term memory recall, & vision 131 This study did not find an increased occurrence of fractures in cancer survivors when compared directly to those without a cancer diagnosis in this sample. Finally, results in this study provided new information on utilization of the ER, hospital, and nursing home for cancer survivors. Cancer survivors were found to use the ER and hospital services more often than those without cancer. Limited evidence in the literature existed regarding HCBS recipients‘ use of health care, so these results provided new information. As shown in Table 18, this study revealed that short-term memory recall problems, evidence of daily pain, and weight loss in conjunction with comorbidities influenced falls, fractures, and health care use. This may be early evidence that a constellation of indicators have a synergistic effect on elderly cancer survivors, limiting their ability to function leading to falls, and in some instances fractures and increasing health care use. This constellation of factors (memory recall problems, evidence of daily pain, weight loss, and comorbidities) warrants future research. The second part of the question asked: Are there differences in the occurrence of falls among cancer patients according to type of cancer (breast, colon, prostate, lung, or other) or stage of cancer (I-IV)? Although evidence exists that certain types of cancer and later stages of cancer have increased functional capacity (Cheville et al., 2008; DeSanto-Madeya et al., 2008; Mandelblatt et al., 2006; Pearce & Ryan, 2008; Waltman et al., 2006), no evidence was found in this study that the cancer type and cancer stage influenced falls in this sample. Using more refined measures and more detailed information on the status of the cancer (i.e., primary site or secondary site) could lead to different findings. Finally, some evidence exists that cancer treatments may lead to falls (Coleman et al., 2007; Greenspan et al., 2007; Waltman et al., 2006) and further study is needed using more refined data (i.e., types and intensity of treatments). 132 Summary of findings for Research Question 2. Research Question 2 asked: Are the effects of frailty (ADLs, cognition, comorbidities, pain, weight loss, and vision) on falls different with respect to site of cancer (breast, colon, prostate, lung, or other), stage of cancer (I-IV), or cancer treatment (chemotherapy and/or radiation)? This examination tested the interaction between frailty and cancer site on falls and the interaction between frailty and stage of cancer. This Research Question examined cancer survivors and did not compare the sample to those without a cancer diagnosis. Frailty included comorbidities (summed comorbid score [arthritis, CVA, CAD, depression, CHF, and diabetes]), cognition (short-term memory recall), symptoms (evidence of pain), functional status (summed ADL score [bladder incontinence, bowel incontinence, mobility, transfer, locomotion, dressing, eating, toileting, bathing, personal care, unsteady gait, walking, and stair-climbing]), weight loss, and vision problems. Site of cancer, frailty, and falls. Site of cancer and the influence of frailty on falls were examined. In this study 64.4% had a solid tumor. The most common type of cancer was breast (n = 179, 20.8%), followed by colon (n = 144, 16.7%), prostate n=119, 13.8%), and lung (n = 113, 13.1%) (Table 3). The remaining cancers were categorized as other. In this section, each type of cancer is examined individually then compared to those without cancer to better understand if a component of frailty is evident by cancer site. Short-term memory recall, daily pain, and site of cancer. Those with breast cancer (21.2% compared to 21.5%) and lung (10.8% to 14.7%) had fewer problems with short-term memory recall, while patients with colon (16.9% to 15.9%) and prostate (19.6% to 8.3%) had more problems with short-term memory recall. A higher level of daily pain was found in breast (21.9% to 18.1%) and lung (13.4% to 11.1%) cancer patients, while higher evidence of daily pain was not found in those with colon (15.8% to 21%) and prostate (13.0% to 15.8%) cancer. In 133 general, evidence exists in the literature that cognitive impairment and pain influence function (Shega et al., 2010), which may increase falls. In cancer patients, cognitive impairment is evident during cancer treatment (Coleman et al., 2007; Greenspan et al., 2007; Waltman et al., 2006). Moreover, pain is associated with many types of cancer (i.e., bone cancer) and cancer treatments (Waltman et al., 2006). However, most studies relate the type and intensity of the treatment to the impairment and not the type of cancer. The findings in this study follow the literature in general. In future studies more refined measurement of pain and cognitive impairment would provide more detailed information to determine if functional capacity is different by site of cancer. Combined with refined measures, more data on the cancer site (i.e., primary or secondary) would allow for better understanding of the influence of cancer site on falls in cancer survivors. Activities of daily living and site of cancer. In regard to ADLs, several findings were interesting. Persons with lung (21% to 11.5%) and colon (17.1% to 16.4%) had weight loss while persons with breast (11.9% to 22.4%) and prostate (7.5% to 14.5%) did not (p = .00). Bladder incontinence was only worse in those with breast cancer (22.5% to 17.3%) (p = .00) while only those with prostate cancer had worse bowel incontinence (13.9% to 12.6%) (p = .00). Mobility was worse for persons with colon (16.9% to 16.2%) and prostate (13.9% to 12.5%) patients (p = .00), while transferring (14.2% to 12.3%) (p = .01) and locomotion (14.1% to 12.8%) (p = .01) was slightly worse in persons with prostate. Persons with breast (20.7% to 20.5%) and prostate (14.5% to 11.9%) cancer had slightly more difficulty dressing (p = .00), while difficulty eating (14.3% to 12.5%) (p = .00) and toileting (14.5% to 10.5%) (p = .01) was only found in persons with prostate. Persons with breast (21.5% to 19.5%) and prostate (14.2% to 10.9%) cancer had difficulty with bathing (p = .00). Similarities were found in difficulty with walking and stair 134 climbing across all cancer sites. Vision problems were slightly more evident in those with colon cancer (16.4% to 16%) (p = .01). Although these findings are interesting, they provided limited information on how ADL dependencies are influenced by the site of cancer in elderly cancer survivors. It is proposed that these findings were influenced by the constraint in the data. Evidence exists in general that the disease or treatment of cancer may contribute to functional decline affecting ADLs (Bylow et al., 2008; Holley, 2002; Overcash, 2007; Pautex et al., 2008). However, few studies exist on falls in cancer survivors, and fewer still examining falls by site of cancer. Future studies could extend the science by examining site of cancer by specific ADL dependencies so that clinicians could better understand what types of assistance cancer survivors may need to remain in the home setting. Comorbidities and site of cancer. In regard to comorbidities several findings were also interesting. Persons with breast (21.9% to 20.2%) and colon (19.4% to 14.2%) cancer had higher prevalence of diabetes while those with lung (12.1% to 13.3%) and prostate (11.2% to 15.4%) cancer had lower prevalence of diabetes (p = .00). Persons with lung (13.8% to 12.3%), breast (22.1% to 20.5%), and colon (19.6% to 14.6%) cancer had higher prevalence of CHF while those with prostate cancer (10.2% to 15.3%) had fewer problems with CHF (p = .00). Those with breast (21.3% to 21.1%), lung (13.0% to 12.5%), colon (18.1% to 15.6%), and prostate (13.9% to 13.4%) cancer had higher prevalence of CAD (p = .00). Persons with breast (21.5% to 18.9) had more problems with arthritis while lung (12.4% to 14.5%), colon (15.9% to 18.0%), and prostate (12.4% to 17.1%) cancer had fewer problems with arthritis (p = .00). Those with breast (19.3% to 21.6%), lung (10.9% to 13.4%), and colon (15.5% to 16.6%) cancer had lower prevalence of CVAs while prostate (17.7% to 12.0%) had higher prevalence of 135 CVAs (p = .02). Persons with breast (20.9% to 19.4%) and lung (12.6% to 12.2%) cancer had higher prevalence of depression while colon (15.9% to 17.4%) cancer lower prevalence of depression, and no difference was evident in those with prostate cancer (13.5%) (p = .00). In general, the evidence supports comorbidities influencing functional limitations in the elderly (Wedding et al., 2007). Limited evidence in the literature exists examining the prevalence of comorbidities in cancer survivors. these findings may be influenced by the constraint in the data. Future studies of cancer survivors need to examine if comorbidities are more prevalent and if the severity of the comorbidity is exacerbated in those who have had cancer or cancer treatment. GEE modeling of cancer site, frailty, and falls. GEE modeling was conducted first testing the interaction model between frailty and the site of cancer and falls, and no associations among any of the factors were found (p > .05). The interaction was then removed from the model, and an exploration of the additive effects of the site of cancer and frailty on falls occurred in a new GEE model. No association between the level of frailty and site of cancer to include breast, colon, prostate, lung, or other cancer and falls (p > .05) was found. Cancer survivors are living longer and usually have more comorbidities (Wedding et al., 2007), loss of cognitive skills after treatment (Pautex et al., 2008), cancer-related pain (Diemling et al., 2007), and functional limitations (Alibhai et al., 2006; Bylow et al., 2008; Cheville et al., 2008; DeSanto-Madeya et al., 2007; Sweeney et al., 2006). Recent studies on cancer in the elderly have documented functional problems in cancer patients (Yancik & Ries, 2000; Yancik et al., 2001) and three national-level studies seem to indicate that frailty differs in those who have had cancer compared to those who do not have cancer (Hewitt & Rowland, 2002; Hewitt et al., 2003; Keating et al., 2005; Yabroff et al., 2007). Furthermore, certain types of cancer have been 136 related to functional decline (Bylow et al., 2008; Kanis et al., 1999; Waltman et al., 2006). Those patients with lung or head and neck cancer have had an increased rate of falls (Pearce & Ryan, 2008). Postmenopausal breast cancer patients have also had a higher rate of falls (Waltman et al., 2006). It is proposed that, due to the constraint in the data, differences were not detected in levels of frailty. What is not known is which component of the frailty construct (weight loss, ADLs, cognitive impairment) is influencing the functional problems in cancer patients. Future studies need to use measures that adequately capture the components of frailty to better identify which factors are contributing to the functional decline in cancer survivors. Summary of site of cancer, frailty, and falls and conclusion. Results (Table 19) in this study do not provide any new evidence. Due to constraints in the data it is proposed that the level of frailty in this sample of patients made it difficult to detect differences among sites of cancer. Use of a more diverse patient sample with a broader range of frailty or more refined measures of frailty factors is needed to further examine if the level of frailty in combination with the site of cancer influence falls. Stage of cancer, frailty, and falls. Stage of cancer and frailties‘ influence on falls in this sample was examined. In this study 5.3% of the patients were diagnosed at stage I, 44.9% at stage II, 20.3% at stage III, and 28.7% at stage IV (Table 4). In this study more problems were found with short-term memory recall in those who fall compared to those who did not fall for stage I (53.6% to 46.4%), stage II (54.5% to 45.5%), and stage IV (51.2% to 48.8%); and stage III had no difference. However no significant difference was found (p = .20) among the stages. Evidence of daily pain was higher for each stage of the diseases, however no difference was found among the stages (p = .26). In regard to ADLs some differences were found. Weight loss was higher for each stage of the disease (p = .00). Bladder 137 incontinence was slightly higher for stage I (.5% to .4%) (p = .00) and bowel incontinence was slightly higher for stage III (.8% to 1.7%) (p = .02). Mobility was slightly worse at stage IV (1.5% to 1.4%) but not significant (p = .68); transferring worse at stage II, III, and IV (p = .00); locomotion at stage II (p = .00); dressing at stage III (p = .13); eating was the same across stages (p = .64); toileting was slightly worse in stage II, III, and IV (p = .40); bathing was worse in stage I, III, and IV (p = .22); walking was worse in stage I and IV (p = .00); and stair climbing was worse in stage I and IV (p = .21). Minimal differences were found in vision across stages (p = .12). In regard to comorbidities some differences were also found. Stage I (28.6% to 71.4%), stage II (49.5% to 50.4%), and stage III (41.7% to 53.3%) had a lower rate of diabetes, while stage IV (55.6% to 44.4%) had a higher rate of diabetes (p = .03). The cancer patients had a lower rate of CHF (p = .04), CAD (p = .04), CVA (p = .00), depression (p = .30); and conversely a higher rate of arthritis (p = .01). GEE modeling of cancer, frailty, and falls. The interaction model between frailty and cancer stage, and falls was tested, and no associations among any of the factors were found (p > .05). The interaction was then removed from the model. Then the additive effects of cancer stage and frailty on falls was explored, again controlling for covariates, and no association was found between frailty and any cancer stage I, II, III, or IV and falls (p > .05). Table 19 summarizes these findings. Evidence exists in the literature that the advancing stage of cancer is likely to be associated with functional decline (Given et al., 2001; Overcash, 2007), which may be caused by frailty. Furthermore, those with metastatic cancer have established patterns of physical impairment (Cheville et al., 2008; DeSanto-Madeya et al., 2007), which may also be due to 138 frailty. Results do not support this evidence. Again, as stated previously, it is proposed a more diverse patient sample with a broader range of frailty is needed to further examine if the level of frailty in combination with the stage of cancer influence falls and more closely follow the evidence in the literature. Cancer treatment(s), frailty, and falls. This study was not able to examine if cancer treatment(s) and frailty influenced falls in this sample (Table 19). Due to the small number of patients under current treatment in this sample, examining the influence of treatment on the outcome variables was not feasible. As a result of cancer treatment, functional, cognitive, pain, and endurance decline have occurred (Barsevick et al., 2006; Bennett et al., 2007; Diemling et al., 2007; Holley, 2002; O‘Connell et al., 2007; Overcash, 2007; Pautex et al., 2008). The literature provides support for future research examining the additive effect of cancer treatments on frailty and falls in cancer survivors. Summary of Research Question 2. The question was examining if the effects of frailty on falls differs with respect to site of cancer or stage of cancer. Although the GEE models did not retain cancer type or site of cancer and cancer treatment could not be examined, some interesting observations were made that contribute to future research. Table 19 Summary of Findings for Research Question 2 Outcomes Key Variable Significant Covariates Falls & frailty Site of cancer None Falls & frailty Stage of cancer None Fails & frailty Cancer treatment Unable to evaluate 139 Frailty is a complex multi-dimensional construct that is known to be a multisystem reduction in physiological capacity not related to a single component (Bandeen-Roche et al., 2006; Bartali et al., 2006; Fried et al., 2004; Guilley et al., 2008; Semba et al., 2006). Moreover, frailty is a body-wide construct to include the musculoskeletal, cardiovascular, metabolic, and immunologic systems (Bortz, 2002) with overlapping, interrelated, interconnected co-occurring signs and symptoms (Chen et al, 2007; Crimmins et al., 2007; Fried et al., 2004; Koroukian et al., 2006; Lee & Rant, 2008; Li, 2005; Walston et al., 2006). Recent studies on cancer in the elderly have found disparities in frailty by age and comorbid condition (Yancik & Ries, 2000; Yancit et al., 2001). Furthermore, the three nationallevel studies indicate that ADLs and cognition, both components of frailty, differ in those who have had cancer (Hewitt & Rowland, 2002; Hewitt et al., 2003; Keating et al., 2005; Yabroff et al., 2007). Studies directly comparing cancer survivors to those who do not have a cancer diagnosis are needed to identify if frailty factors are different. Additionally, since a constellation of factors was found when examining Research Question 1, short-term memory recall problems, evidence of daily pain, weight loss, combined with comorbidities, an examination of cachexia (weight loss, muscle atrophy, fatigue, and weakness) and frailty (weight loss, ADL dependency, memory loss, osteoporosis, depression, and incontinence) needs to be examined and a comparison make between cancer survivors and similar patients without a cancer. This type of research would provide new information for clinicians on what physical function aspects to focus on while caring for cancer survivors. Summary of findings for the exploratory research question. The exploratory research question asked: What is the social environmental status of the population? The intent of this exploration was to better understand the environment of this population. 140 The majority of patients had similar living environments, with more than 90% in a house or apartment (Table 10). Although most of the patients in this study were widowed, 60.8% of the cancer survivors lived with another person, notably different than the 10.8% without a cancer diagnosis who lived with another person. Although cancer survivors lived with another person, a fall rate of 59.9% was notably higher than the fall rate of 11.5% in those without a cancer diagnosis who lived with another person. When examining time alone during the day, the cancer survivors fall at a notably higher rate even though they were never left alone (44.8%), when compared to those without cancer who had a fall and were never left alone (37.2%). There is limited information on the social and environmental status of the communitydwelling elderly cancer survivors and even less for those in the HCBS program. It is proposed that in this sample of patients with a high fall rate, cancer survivors with a caregiver were less likely be left alone and the fall occurred due to the constellation of factors found in this study: weight loss, evidence of daily pain, and depression (use of antidepressants). What is not known is if this was due to their cancer, comorbidities, aging, or a combination of all three, and further study is needed. Summary of exploratory research question and conclusion. Results in this study provide new information on the environment of dually eligible, low-income, community-dwelling elderly cancer survivors. Further study is needed to more fully understand the social environment of elderly cancer survivors, their caregivers in this population, and what level of care the caregivers provide; and why the fall rate is so much higher for cancer survivors with caregivers than for those who do not have a cancer diagnosis. 141 Study Limitations Overall, limitations associated with use of existing data occurred in this study. A major limitation in this study was the ability of an elderly individual to recall a fall event. As stated in the literature, the elderly are able to recall the occurrence of a fall for approximately 3 months (Fauth et al., 2007). However, the assessment occurs every 6 months, which may lead to underreporting falls. Cognitive impairment and recall ability were equally present in cancer survivors and in the group without a cancer diagnosis. The analytic technique examined cognitive impairment in both groups, those with or without cancer, and all analysis included cognition as a covariate to adjust for possible recall bias. One limitation was the distribution of the frailty covariates in this study. This included ADLs, cognition, and comorbidities. As stated previously, the intent of the HCBS program is to maintain elderly vulnerable disparate individuals in their home setting in lieu of nursing home placement. This created frailty data that was skewed around a small range of potential variance making it difficult to detect if differences existed when comparing the cancer survivors to those without cancer, and when examining cancer site and stage of cancer. Use of a more diverse population of patients is needed for future research. Another limitation is the unknown severity of comorbidities. Elderly individuals often have comorbidities and disabilities (L. P. Fried et al., 2004; Iezzoni & Freedman, 2008; Johnson & Wiener, 2006), which increase the occurrence of falls (Guilley et al., 2008; Iezzoni & Freedman, 2008; Johnson & Wiener, 2006; US Department of Health and Human Services, 2005). Cancer survivors are living longer (Center for Disease Control, 2007) and usually have comorbidities (Wedding et al., 2007) and functional limitations (Sweeney et al., 2006). The 142 analytic technique included examining comorbidities in both groups, those with or without cancer, and all analysis included comorbidities as a covariate to adjust for possible differences. Those who were deceased within 2 months are excluded. Those diagnosed at death may lead to underrepresenting advanced-stage cancers, most likely lung cancer. Additionally, although the cancer diagnosis was known from the Cancer Registry, it may be reoccurrence of a primary or secondary cancer. There was no opportunity to control for this limitation in the study or to get current status of the patient. One final limitation is the potential influence of chemotherapy on cognition. Some have found a decrease in cognitive skills during cancer treatments, increasing the prevalence of falls (Pautex et al., 2008). In this study, there was no opportunity to control for this limitation. Implications for Nursing Practice Data from this study provide several implications for clinical practice in regard to lowincome elderly cancer survivors. It was demonstrated that the presence of a cancer diagnosis may have altered the rate of falls in elderly cancer survivors and consequently increased ER use and hospitalizations. It was also demonstrated that the closer a patient was to their time of cancer diagnosis, the more likely they are to fall. Therefore, a convincing case can be made for modification of gerontological nursing practice, taking cancer history into account when treating patients. This finding supports other scientists who recommend taking cancer history into account when treating patients (Cope & Reb, 2006, p. 6; Hodgson, 2002; Kagan, 2004; Rowland & Yancik, 2006; Travis & Yahalom, 2008). Falls are a major concern for patients, their families, and caregivers. Based on the findings in this study, falls need to become a major safety priority for nurses caring for cancer survivors. The key message for these nurses is that a cancer diagnosis alters the rate of falls and 143 that the closer a patient is to his or her cancer diagnosis, the more likely he or she will fall. Additionally, even if the cancer survivor had a caregiver and is not left alone for any length of time, they are still likely to fall. Two groups of nurses need to be aware of the findings in this study through publications, presentations, and training. The first group is oncology nurses who care for patients before, during, and after treatment for cancer in hospitals, physicians‘ offices, home health care, and outpatient clinics. The second group is gerontological nurses who care for cancer survivors in various health care settings. Once the increased incidence of falls in cancer survivors is better understood, nurses will need to integrate evidence-based falls risk interventions into practice (Hurria et al., 2007). Fall Prevention. There are international evidence-based guidelines for fall prevention (Rubenstein et al., 2002) that nurses caring for cancer survivors can adopt. These guidelines include a three-step approach. The first step is to conduct a fall risk assessment to identify factors specific to the individual that may increase the risk of falls (Chen et al., 2004). The fall risk assessment should occur on a regular basis dependent on the condition of the patient and the environment in which they reside. The second step is implementation of multi-modal behavioral and psychological fall prevention interventions that are matched to the individual‘s fall risk factors from the assessment and the environment in which the person resides (Cimprich et al., 2005). The final step is post-fall follow-up. When a fall occurs, often there is an underlying cause, and subsequent modification of the interventions is necessary to prevent further falls. The following is a review of evidence based fall prevention measures that are known to be effective at reducing falls in the community-dwelling elderly. Fall risk assessment. Falls are usually a result of a synergistic interaction between an 144 individual‘s intrinsic risk factors, the physical environment (home and equipment), and the riskiness of a person‘s behavior (Oliver, 2007). A fall risk assessment is a specially designed nursing assessment that is used to evaluate a patient‘s intrinsic risk factors, the environment, and behavior. The purpose of a fall risk assessment is to identify the fall risk factors that can then be used to guide the nurse in selecting fall prevention interventions (Oliver, 2007; Oliver et al., 2008). Reductions in falls rates have been achieved when using assessments to identify fall risk with subsequent implementation of interventions (Healey et al., 2004). Systematic reviews (Myers, 2003; Oliver et al., 2004; Scott et al., 2007) have shown that the following factors are consistently identified in patients who fall: a recent fall, postural instability, muscle weakness, behavioral disturbance, agitation, confusion, urinary incontinence or frequency, and prescription of ―culprit‖ drugs, postural hypotension, or syncope. Additionally, comorbidities, poly-pharmacy, delirium, frailty, and functional impairment are highly prevalent and coexisting in the elderly (Fried et al., 2004; Leipzig et al., 1999b; Tinetti et al., 2005) and are known to increase falls. In order to provide a comprehensive fall risk assessment in cancer survivors, it is essential to be aware of the additional risk factors that influence falls in this population. This study found that sex, race or ethnicity, short-term memory, evidence of pain, antidepressants, weight loss, and comorbid conditions influenced the rate of falls in cancer survivors. There are established fall risk assessments, such as the Morse Fall Risk Assessment (1989) that could be modified to incorporate factors known to increase falls in cancer survivors, creating a tool that would fit the needs of this population. Fall prevention interventions. Once the fall risk assessment is complete, interventions to prevent falls can be tailored to meet the need of the individual. The CDC (2008) recommended a 145 multi-modal fall prevention program incorporating evidence-based interventions into everyday life. This includes education, exercise programs, medication review and management, vision checking and improvement, and home hazard assessment and modification to create a safer living environment. The following are the CDC recommendations to prevent falls in the elderly. When conducting fall prevention, it is important to include an educational component. While education alone has not proved effectively to reduce falls among the elderly, it is typically combined with one of the other multi-modal interventions to enhance their effectiveness (Von Renteln-Kruse & Krause, 2007; Dacenko-Grawe & Holm, 2008). Individual education sessions may work better for people who are hearing- or vision-impaired or have special needs, and sessions should be tailored to the attention span and cognitive ability of the elderly. Visual aids such as brochures, fact sheets, and checklists will help facilitate the education of written information and have demonstrated effectiveness at reducing falls (Dacenko-Grawe & Holm, 2008). Education needs to be provided on the increased rate of falls in the elderly, and in cancer survivors. Once this is understood, patients, family, and caregivers must be educated on factors known to increase falls. Then a review of the agreed upon physical activity, medication management, vision impairment, and home hazards fall prevention interventions must be understood and an implementation plan developed. In this study, there was no opportunity to evaluate if these patients, family, or caregivers were being educated on fall prevention measures. Increasing physical activity through exercises designed to improve mobility, strength, and balance such as individualized exercise session, a home exercise program, group exercise classes, or Tai Chi are interventions known to reduce falls (CDC, 2008). To be effective at reducing falls, exercises must be performed at least twice weekly on an ongoing basis and progress in difficulty throughout the program (CDC, 2008). Patients can be taught the exercises 146 by a trained exercise instructor or physical or occupational therapist, either in one-on-one sessions or in group session, before performing them independently at home (CDC, 2008). Older adults should be assessed for strength, balance, and fitness at the beginning and end of each new exercise program. In this study, there was no opportunity to evaluate if an exercise program was being conducted. Medication review to identify side effects or drug interactions that may contribute to falls needs to be conducted on an ongoing basis, so that adjustments to or changes in the medications can be provided. The purpose of medication review and management is to identify and eliminate medication side effects and interactions, such as dizziness or drowsiness that can increase the risk of falls. Many elderly are unaware that their daily medications may increase their fall risk. Aging affects the absorption, distribution, metabolism, and elimination of medications. Age can also increase sensitivity to potential side effects (CDC, 2008). Older adults may also get prescriptions from multiple doctors who do not communicate with each other (CDC, 2008). Fall risk increases with the total number of prescription and over-the-counter medications. Psychoactive medications also increase fall risk (CDC, 2008). These include antidepressants, tranquilizers, antipsychotics, anti-anxiety drugs, and sleep medications, similar to findings in this study. Other medications that may cause problems include those prescribed for seizure disorders, hypertension, cholesterol-lowering medications, heart medications, and painkillers (CDC, 2008). Drug side effects that can contribute to falling include blurred vision, hypotension leading to dizziness and lightheadedness, sedation, decreased alertness, confusion and impaired judgment, delirium, compromised neuromuscular function, and anxiety. Medication reviews are recommended for elderly taking four or more medications and those taking any psychoactive medications. Medication review and management may also include assessing the need for 147 vitamin D and calcium supplements as well as osteoporosis treatment (CDC, 2008). In this study, medication played a significant role in increasing the fall rate. However, there was no opportunity to evaluate if a medication management program was a part of the care for these elderly patients. Vision changes and vision loss associated with aging are common fall risk factors, and can contribute to falls by disturbing balance, and by obscuring tripping and slipping hazards (CDC, 2008). Many vision conditions, such as cataracts, glaucoma, and macular degeneration, are gradual and painless. However, if these conditions are diagnosed early, they can be managed and minimize vision loss. After the age of 60, vision exams are recommended at least every 2 years, and more frequently if an eye condition has been diagnosed (CDC, 2008). Although in this study vision did not influence the fall rate, other studies in elderly have repeatedly demonstrated that vision problems increased the occurrence of falls, which warrants inclusion in preventive interventions. Home safety assessments to identify and modify home hazards that can increase risk of falling are effective for identifying needed home modifications. Environmental factors play a part in approximately half of all falls that occur at home (CDC, 2008). Falls can be caused by slipping and tripping hazards, poor lighting, or the lack of needed home modifications such as bathroom grab bars, handicapped showers, stair railings, and ramps (CDC, 2008). Patients with a history of falls or with mobility or balance difficulties have the greatest need for such an assessment. In addition to home modifications, some elderly may need to use personal assistive safety and mobility devices and be trained to use these devices properly (CDC, 2008). In this study there was no opportunity to evaluate the home safety environment and its potential influence of falls. 148 Nurses should focus on employing multi-modal fall prevention interventions for each identified fall risk factor to reduce rates of falls in community-dwelling elderly cancer survivors ( Oliver, 2007). Interventions are effective at reducing falls, and nurses can influence this health outcome by taking action to prevent falls in cancer survivors. Post-fall follow-up and modification of interventions. Half of all falls involve people who have already fallen—and modification of the interventions may be needed each time a fall occurs (Oliver, 2007). Consequently, fall incidents should be used to prompt reassessment of the underlying cause, with subsequent intervention modification and implementation to prevent further falls. Cancer survivors plan of care. In 2005 the IOM recommended that elderly patients receiving health care after the diagnosis of cancer should be provided a comprehensive care summary and follow-up plan. This presents an opportunity for nurses caring for cancer survivors to include fall prevention interventions in the ―Cancer Survivors‘ Plan of Care‖ (Ganz, Casillas, & Hahn, 2008). The intent of the care plan is to improve communication among the multiple providers of health care assuring that all aspects of a cancer survivor‘s health needs are addressed. Currently the recommended ―Cancer Survivors Plan of Care‖ addresses nutrition and exercise interventions and the additional fall preventive measures would fit naturally (Ganz et al., 2008). However, contrary to the nutrition and exercise interventions that are similar for all cancer survivors, the fall prevention interventions should be based on a fall risk assessment specifically designed to assess elderly cancer survivors. Summary. Often nurses know who is at highest risk and accept the occurrence of falls, in part because the goal of care may be focused on eradicating the cancer. This perspective neglects the persistent and deleterious effects of falls, such as fractures, hospitalization, nursing home 149 placement, functional decline, and diminished quality of life. In addition to alterations in social and family roles as well as quality of life, the risks of increased falls or fall injuries require nurses to assess fall risk factors before, during, and after cancer and cancer treatment and to intervene appropriately—an area where nursing can make a difference when patients are at increased risk. Nurses play an important role in the implementation and delivery of survivorship care plans. Incorporation of cancer prevention and health promotion such as fall prevention interventions is a natural activity for nurses as the entire health care system focuses on improving care for cancer survivors. Implications for Research Data from this study provide several implications for future research. This includes additional descriptive studies, novel interventions ripe for development and testing, and translational research projects. Empirical research in the area of physical function in cancer survivors, specifically falls or fractures, is underdeveloped. The interface of aging, comorbidities, and cancer in survivors is now a focus of the NCI (Rowland & Yancik, 2006). The ONS, as well as individual researchers, have called for studies of the physiological and psychosocial consequences of cancer (Bender et al., 2008; L. P. Fried et al., 2004; Horning, 2008; Koroukian et al., 2006; Li, 2005; Travis & Yahalom, 2008; Walston et al., 2006). In 2009, the IOM identified falls as a top research priority, and the ONS expanded its research agenda to focus on physical limitations of cancer survivors. The NIH to include NIA, NCI, NINR, along with the CDC have called for better understanding of the influence of cancer or cancer treatment on elderly survivors (National Institute for Health and Clinical Excellence, 2004). 150 The literature offers support for advancing age to have a negative effect on falls (Fauth et al., 2007; Davison, 2007) and fractures (Sterling et al., 2001) and that White female cancer patients (Stevens et al., 2007) have a higher fracture rate. This current study supported some of this evidence. Patient characteristics should be included in future research when examining cancer survivors. Likewise, researchers have found that cancer type or stage may influence functional status (Barsevick et al., 2006; Diemling et al., 2007; Holley, 2002; Keating et al., 2005) and increase the risk of falls. Although this current study did not support the literature findings to date, future research examining site of cancer and cancer stage and falls, and the influence of frailty by site of cancer and cancer stage on falls is recommended. This research should include if the cancer is a primary or secondary site, if the cancer is reoccurring, and all types of cancer treatments to include chemotherapy, radiation, and surgical interventions. Use of claims and pharmacy data could enhance and refine research by using drug, diagnosis, and procedure codes to examine what is occurring. Also, using a more diverse patient sample with a broader range of age, treatment, functional abilities, and level of frailty would be recommended to identify if differences are evident. What other diseases could be used to address these questions? Due to the small number of patients under treatment in this current study, examining the influence of treatment was not feasible. Some cancer treatments are linked to bone loss (Alibhai et al., 2006) or vitamin D depletion after treatment, which may lead to falls and fractures (Coleman et al., 2007; Greenspan et al., 2007; Waltman et al., 2006), supporting the need for future research examining the additive effects of cancer treatments on falls or fractures in cancer survivors. Use of claims and pharmacy data could be used for further exploration of cancer treatments‘ influence on falls and fractures to identify if cancer survivors are different. This 151 could include examination of all types of chemotherapy, procedures, treatments, and surgical interventions. The additive effects of certain medications such as antidepressants, anti-anxiety, or hypnotics and cancer treatments on the rate of falls could be further explored by identifying specific drug classifications in the pharmacy data. Time since cancer diagnosis played a key role as a predictor of increased falls in this current study. The symptoms created from side effects of chemotherapeutic medications or treatments may also increase falls. This unique finding that time since cancer diagnosis may play a role in the level of frailty or disability requires future research. Additionally, the symptoms from side effects of treatments need to be included in future studies to better understand how they may increase falls. This study limited inclusion of cancer patients to those diagnosed after January 1, 2000. Future research could focus on examination of falls over longer periods of time, to evaluate if age, certain types of cancer treatments, or both influences falls or fractures in cancer survivors. Similar to findings in this study, several researchers have identified certain medications that are associated with falls in the elderly (Agostini et al., 2004; Harrison et al., 2001; Liepzing et al., 1999a; Rubenstein et al., 2003). My analysis of the influence of medications in this current study supports literature findings to date, with these medications influencing falls. It is proposed that because the non-cancer group had higher usage of these types of medications that lead to increased falls and fractures, that the full extent of the medications‘ influence on falls was not detected in cancer survivors. These findings for the effect of medications on falls in this current study, as well as findings in the literature, provide support for future research examining the additive effect of medications with a cancer diagnosis or if cancer treatments interact with these medications to 152 increase the rate of falls or fractures in cancer survivors. Use of claims or pharmacy data would provide the detailed medication information necessary for further exploration to identify the extent to which these types of medications influence falls or fractures and if cancer survivors follow the evidence in the literature. Frailty factors were associated with increased falls, fractures, and health care use, which is consistent with the literature (Agostini et al., 2004; Anstey et al., 2008; Baldwin et al., 2006; Chen et al., 2008; Glough-Gorr, et al., 2008; Hewitt et al., 2003; Inouye et al., 2007; Klabunde et al., 2006; Moreland et al., 2004; Szabo et al., 2000; Yabroff et al., 2007; Yancik, 1997). This included short-term memory recall problems, evidence of daily pain, depression (use of antidepressants), and comorbidities. Each of the four items needs to be examined individually, and as a constellation of factors known to increase falls in cancer survivors. This current study did not find the effect of frailty on falls, supporting future research. Specifically, a study of comorbid conditions using claims data ICD-9 codes and medical record files is necessary for further exploration to identify if specific comorbid conditions in cancer survivors may increase the occurrence of falls and fractures. In addition, descriptive studies could be conducted collecting data on frailty factors thought to increase falls. Additionally, descriptive studies could compare cancer to other chronic diseases such as CAD or CHF. This would provide deeper understanding of the influence of cancer on falls, fractures, and use of health care services. Likewise, in the current study, the effects of pain provided support for future research using pain measures that are more sensitive to identify at what level of pain in cancer survivors that an increased rate of falls occurs. Similarly, the effect of short-term memory recall problems in the current study provided support for research using measures of cognition that are more 153 sensitive and for research incorporating the treatment of cancer patients into the modeling to identify if cancer survivors follow the evidence in the literature. Finally, this sample of patients may have experienced similar functional limitation, and using a more diverse patient sample with a broader range of functional abilities and level of frailty is in need of further exploration to identify if cancer survivors more closely follow the evidence in the literature. Ultimately, the hope is that this research may lead to future studies intended to design and test effective translation of fall prevention interventions to reduce the occurrence of falls in elderly cancer survivors. Interventions that would translate fall risk prevention measures and evidence to reduce frailty could support and enhance the ability of elderly community-dwelling elderly to remain in their home setting. If successful, this will eventually lead to improved functional status, reduced nursing home placement and hospitalizations, improved quality of life, and reduced health care costs overall in cancer survivors. Implications for Policy After additional research has validated the finding in this study on falls, a policy should be developed to include fall prevention in the ―Cancer Survivors‘ Plan of Care‖. This new rule would address the 2005 IOM recommendation for elderly patients with a diagnosis of cancer being provided a comprehensive follow-up plan. The intent of the guideline would be to improve communication among the multiple providers of health care to assure that all aspects of a cancer survivor‘s health needs are addressed Conclusion This study supports emerging evidence that the likelihood of an elderly cancer survivor experiencing falls may be influenced by their cancer history (Alibhai et al., 2006; Bylow et al., 2008; Cheville et al., 2008; Deimling, Bowman et al., 2007; Deimling, Sterns et al., 2007; Levy 154 et al., 2008; Overcash, 2007; Pautex et al., 2008). The consequence of cancer could be attributed to the cancer or the side effects of cancer treatment and may appear months or years after diagnosis or treatment has ended (Arroyave et al., 2008; Blaauwbroek et al., 2007; Fox & Lyon, 2007; Hawkins et al., 2008; Travis & Yahalom, 2008). The consequence of cancer may include physical and psychological problems (Center for Disease Control, 2007; Daubman, 2008; Hewitt et al., 2007; Horning, 2008; Stovall, 2008), ultimately leading to frailty and falls. The purpose of this research was to examine how the additive effect of a cancer diagnosis altered the rate of falls, fractures, and health care use among cancer survivors, compared to those without a cancer diagnosis, as well as to shed light on whether the cancer diagnosis altered frailty leading to falls in elderly survivors. This study did support that cancer survivors fall at a higher rate than those without a cancer diagnosis. However, no difference was found in the level of frailty among cancer survivors compared to those without a cancer diagnosis. Finally, no difference in fall rates was found by type of cancer or stage. The intention of the current study was to test a synthesized LC (Elder, 1985) and the HRQOL (Ferrans et al., 2005) model by examining the research questions. This synthesized model allowed for the examination of the inter- and intra-relatedness of sociodemographic characteristics, cancer, cancer treatment, medications, and frailty, to identify factors that may be modifiable by nurse clinicians to prevent or reduce the occurrence of falls (Given & Sherwood, 2005). This study makes an important contribution to nursing and oncology research agenda priorities, examining if the addition of a cancer diagnosis alters the rate of falls, fractures, and health care use. Further, this research demonstrated that the closer a person is to their diagnosis of cancer, the greater the need for fall prevention interventions. As falls were found to be more 155 probable in cancer survivors, fall assessment and prevention interventions will become more important as the elderly population with cancer increases in size and health care costs continue to rise. An immediate next step for this research project is to disseminate findings from the current study to inform nurses that elderly cancer survivors have an increased rate of falls. Future research with this dataset and other datasets with broader variation in frailty levels will continue to evaluate the influence of time since cancer diagnosis on falls to advance this emerging field of study of function in cancer survivors. Future research comparing cancer survivors to other chronic conditions will also advance this emerging field of study. Additionally, research needs to be conducted on how the time since an individual‘s cancer diagnosis may influence functional ability in cancer survivors. Finally, examining the constellation of factors that seemed to consistently affect the outcomes in this study, short-term memory recall problems, weight loss, evidence of daily pain, combined with comorbidities may lead to better understanding of cancer survivors‘ ability to remain in the community-dwelling setting. The concern regarding physical functioning and cancer or the effects of cancer treatments are anticipated to increase as the prevalence of cancer increases, anti-cancer therapies become more complex, and the age threshold for active treatment continues to expand (Hewitt et al., 2003; Snyder et al., 2008). Studying falls, fracture, and health care use may lead to innovative discoveries that help cancer survivors remain in the community-dwelling setting. 156 APPENDIX A 157 Table 20 Independent Variables Source of Data, Data Format, and Data Codes for Study Variable Source of Data Format of Data Data Codes Age in years MDS Face sheet genderraceagebd Continuous Date 00/00/00 Sex Gender_code MDS Face Sheet Genderraceagebd Categorical 0=male; 1=female Race Race_code MDS Face sheet Genderraceagebd Categorical 1=Caucasian, 2=African American 3=American Indian, 4=Hispanic; 5=other Cancer Site & Stage Cancer Registry ICD-9 0=Non-cancer; 1=Cancer Each code & Each stage Pain MDS Assessment ‗G‘ Categorical 0=none to some; 1=evident to daily Comorbidity MDS Assessment ‗I‘ Categorical 0=Not present; 1=Present Medications MDS Assessment ‗R‘ Categorical rmedications 0=Not taking 1=Taking Cancer Treatments MDS Assessment ‗Q‘ Categorical Radiation & Chemotherapy treatments 0=No 1=Yes Cognition: MDS Assessment ‗E‘ Categorical 0=OK; 1=Problem Frailty: Weight loss Vision MDS Assessment ‗K‘ Categorical MDS Assessment ‗M‘ Categorical 0=No; 1=Yes 0=Adequate, 1=Impaired 2=Moderately Impaired, 3=Highly & 4=Severely Impaired Frailty: MDS Assessment ‗P‘ Categorical Bathing, Eating - Dressing, Toileting Transferring Performance Stair Climbing, Walking, Gait Bladder Continence MDS Assessment ‗O‘ Categorical Bowel Continence continence 158 0=Independent 1=Supervision 2=Limited Assistance 3=Extensive Assistance 4=Total Dependence 0=Continent, usually continent 1=Occasionally & Incontinent Table 21 Dependent Variables Source of Data, Data Format, and Data Codes for Study Variable Source of Data Format of Data Data Codes Falls MDS Assessment ‗J‘ Categorical 0=None; 1—>9=1—>9 falls Fractures MDS Assessment ‗J‘ Categorical 0=None; 1=fracture ER Use MDS Assessment ‗Q‘ Categorical 0=None; 1—>9=1—>9 ER visits Hospitalization MDS Assessment ‗Q‘ Admits Categorical 0=None; 1—>9=1—>9 Hospital Nursing Home LOC 22 Placement Categorical 0=None; 1=Placement Table 22 Social Environmental Status Variables Source of Data, Data Format, and Data Codes for Study Variable Source of Data Format of Data Data Codes Marital Status MDS Telephone Screen Categorical 0=Single; 1=Married; 2=Widowed; 3=Separated 4=Divorced; 5=Other Housing MDS Assessment ‗D‘ Categorical 0=House: 1=Apartment 2=Group Home; 3=Other Time Alone During day MDS Assessment ‗B‘ Categorical 0=Never; 1=1-hour 2=Long periods; 3=All of the time Feeling of Sadness MDS Assessment ‗G‘ Categorical 0=Not in last 3- days 1=Up to 5 days a week 2=Daily or almost daily Pets MDS Assessment ‗D‘ Categorical 0=No; 1=Yes 159 APPENDIX B 160 Table 23 Multi-nomial Regression Models of Comorbidities, Medications, and Falls DF Β Standard Error Significance Exp (B) Confidence Interval Depression, Antidepressants and Falls Comorbid 1 -0.14 .04 Med 1 -0.26 .05 0.00** 0.00** .87 .77 .81 .70 .93 .85 Depression, Antipsychotics and Falls Comorbid 1 -0.26 .02 Med 1 -0.14 .06 0.00* 0.01* .77 .87 .74 .78 .80 .97 Depression, Hypnotics and Falls Med 1 -0.13 .06 Comorbid 1 -0.26 .02 0.00* 0.00* .77 .74 .74 .79 .80 .98 Depression, Anti-anxiety and Falls Med 1 -0.02 .04 Comorbid 1 -0.27 .02 0.64 0.00* .98 .76 .91 .73 1.06 .80 Diabetes, Antidepressants and Falls Med 1 -0.41 .03 Comorbid 1 -0.11 .02 0.00* 0.00* .66 .90 .62 .86 .71 .94 Diabetes, Antipsychotics and Falls Med 1 -0.21 .06 Comorbid 1 -0.11 .02 0.00* 0.00* .81 .90 .73 .86 .91 .93 Diabetes, Hypnotics and Falls Med 1 -0.22 .06 Comorbid 1 -0.11 .02 0.00* 0.00* .80 .90 .72 .86 .89 .94 Diabetes, Anti-anxiety and Falls Med 1 -0.13 .04 Comorbid 1 -0.11 .02 0.00* 0.00* .88 .90 .82 .87 .95 .94 161 Table 23 Continued Multi-nomial Regression Models of Comorbidities, Medications, and Falls DF Β Standard Error Significance Exp (B) Confidence Interval CAD, Antidepressants and Falls Med 1 -0.40 .03 Comorbid 1 -0.10 .02 0.00* 0.00* .67 .91 .63 .87 .71 .95 CAD, Antipsychotics and Falls Med 1 -0.21 .06 Comorbid 1 -0.12 .02 0.00* 0.00* .81 .89 .73 .85 .93 .93 CAD, Hypnotics and Falls Med 1 -0.20 .06 Comorbid 1 -0.11 .02 0.00* 0.00* .82 .90 .74 .86 .92 .94 CAD, Anti-anxiety and Falls Med 1 -0.10 .04 Comorbid 1 -0.11 .02 0.00* 0.00* .90 .90 .84 .86 .98 .94 CVA, Antidepressants and Falls Med 1 -0.41 .03 Comorbid 1 -0.08 .02 0.00* 0.00* .66 .93 .62 .88 .71 .97 CVA, Antidepressants and Falls Med 1 -0.41 .03 Comorbid 1 -0.08 .02 0.00* 0.00* .66 .93 .62 .88 .71 .97 CVA, Antipsychotics and Falls Med 1 -0.20 .06 Comorbid 1 -0.09 .02 0.00* 0.00* .82 .92 .73 .87 .92 .96 CVA, Hypnotics and Falls Med 1 -0.20 .06 Comorbid 1 -0.09 .02 0.00* 0.00* .82 .92 .73 .87 .91 .96 CVA, Anti-anxiety and Falls Med 1 -0.13 .04 Comorbid 1 -0.09 .02 0.00* 0.00* .88 .92 .82 .87 .95 .96 162 Table 23 Continued Multi-nomial Regression Models of Comorbidities, Medications, and Falls DF Β Standard Error Significance Exp (B) Confidence Interval Arthritis, Antidepressants and Falls Med 1 -0.41 .03 Comorbid 1 -0.06 .02 0.00* 0.01* .66 .94 .62 .90 .71 .99 Arthritis, Antipsychotics and Falls Med 1 -0.21 .06 Comorbid 1 -0.06 .02 0.00* 0.00* .81 .93 .73 .90 .91 .98 Arthritis, Hypnotics and Falls Med 1 -0.20 .06 Comorbid 1 -0.06 .02 0.02* 0.01* .82 .94 .73 .90 .91 .98 Arthritis, Anti-anxiety and Falls Med 1 0.11 .04 Comorbid 1 -0.06 .02 0.00* 0.01* .90 .94 .83 .90 .97 .99 CHF, Antidepressants and Falls Med 1 -0.40 .03 Comorbid 1 -0.04 .02 0.00* 0.06 .67 .96 .63 .92 .71 1.00 CHF, Antipsychotics and Falls Med 1 -0.21 .06 Comorbid 1 -0.05 .02 0.00* 0.02* .81 .95 .73 .91 .91 .99 CHF, Hypnotics and Falls Med 1 -0.20 .06 Comorbid 1 -0.05 .02 0.00** 0.03** .82 .96 .74 .92 .91 .99 CHF, Anti-anxiety and Falls Med 1 -0.11 .04 Comorbid 1 -0.04 .02 0.00** 0.05 .90 .96 .83 .92 .97 1.00 163 Table 24 Generalized Linear Modeling Analysis of Least Mean and Standard Errors of Antidepressants and Antipsychotics, Anti-Anxiety, and Hypnotics Anti- psychotic Anti- anxiety Hypnotic F Value DF* Mean Square 1 1 1 12.95 39.17 21.75 58.31 39.17 97.98 Least Square Mean P-Value Anti- psychotic No (vs. yes) 0.47 <.00* Anti- anxiety No (vs. yes) 0.46 <.00* Hypnotic No (vs. yes) 0.46 <.00* 164 P-Value <.00* <.00* <.00* Table 25 Generalized Linear Modeling Analysis of Least Mean and Standard Errors of Depression and Congestive Heart Failure, Coronary Artery Disease, Arthritis, and Cerebral Vascular Accident Diabetes CHF CAD Arthritis CVA F Value P>F DF* Mean Square 1 1 1 1 1 0.00 2.37 12.16 11.12 2.73 0.01 9.95 50.44 46.56 11.42 0.92 0.02* <.00* <.00* 0.00* Least Square Mean P-value Diabetes No (vs. yes) 0.46 0.92 CHF No (vs. yes) 0.44 0.02* CAD No (vs. yes) 0.42 <.00* Arthritis No (vs. yes) 0.42 <.00* CVA No (vs. yes) 0.44 0.00* 165 Table 26 Descriptive Statistics of Vision, Bladder, and Bowel Incontinence, Activity of Daily Living Performances Items of Cancer and Non-Cancer Patients with Falls or No Falls Fallers Cancer n (%)* No Falls Non-cancer n (%)* Cancer n (%) * Non-cancer n (%) * Vision None to slight impairment 263 (88.5) 2314 (86.4) Moderate to severe impairment 34 (11.5) 362 (13.6) 476 (88.8) 5058 (86.9) 60 (11.2) 759 (13.1) Bladder Incontinence Continent to usually continent 172 (58.2) 1481 (55.0) Occasionally to usually incontinent 124 (41.8) 1208 (45.0) 354 (65.4) 3574 (61.4) 187 (34.6) 2244 (38.6) Bowel Incontinence Continent to usually continent 227 (77.5) 2013 (75.2) Occasionally to usually incontinent 66 (12.5) 958 (24.8) 423 (78.3) 4434 (76.3) 117 (21.7) 1378 (23.7) Mobility in Bed Performance No help to some help 225 (78.1) 1978 (74.6) Moderate to full help 63 (21.9) 673 (25.4) 417 (78.4) 4362 (75.8) 115 (21.6) 1125 (24.2) Eating Performance Independent to some supervision 236 (79.5) 2093 (77.9) Limited assist to total dependence 61 (20.5) 595 (22.1) 441 (82.0) 4537 (78.1) 97 (18.0) 1274 (21.9) Stair Climbing Performance Independent to some supervision 19 (6.4) 182 (6.8) Limited assist to total dependence 277 (93.6) 2512 (95.2) 56 (10.7) 518 (8.9) 484 (89.3) 5334 (91.1) 166 Table 27 Total Variance of Eigenvalues for each ADL Factor Factor Total Percent Variance Cumulative Percent 1 2 3 4 4.814 1.290 1.123 1.043 34.4 9.2 8.0 7.5 34.4 43.6 51.6 59.1 5 6 7 8 9 10 11 12 13 14 .949 .905 .779 .668 .564 .519 .446 .389 .311 .201 6.8 6.5 5.6 4.8 4.0 3.7 3.2 2.8 2.2 1.4 65.9 72.3 77.9 82.6 86.7 90.4 93.6 96.3 98.6 100 Figure 4 Screen Plot of Factors and Eigenvalues Eigenvalues 4 2 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 Factor Number 167 Table 28 Rotated Factor Matrix Correlations of the 14 ADLS with Four Factors One Eating Performance Toileting Bathing Functional Status Change Unsteady Gait Weight Loss Vision Bladder Incontinence Bowel Continence Walking Stair Climbing Locomotion Transferring Mobility in Bed Factors Two Three .562 .737 .362 .024 -.148 .017 .088 .159 .310 .432 .162 .246 .761 .630 .139 .325 .319 .015 .125 -.014 .038 .132 .098 .628 .532 .859 .387 .303 .382 .350 .244 -.012 .002 .020 .113 .511 .719 .152 .080 .133 .175 .239 Table 29 Rotated Factor Matrix Correlations of the 11 ADLS with Three Factors One Eating Performance Toileting Bathing Unsteady Gait Bladder Incontinence Bowel Continence Walking Stair Climbing Locomotion Transferring Mobility in Bed Factors Two Three .506 .746 .468 -.004 .135 .288 .730 .467 .708 .804 .665 .428 .443 .291 -.008 .627 .642 .175 .076 .121 .284 .303 -.240 -.142 .098 .230 .130 -.097 .242 .349 .427 -.098 -.153 168 Four -.017 .102 .218 .132 .452 .088 .126 .119 -.068 .067 .103 -.024 .151 -.011 Table 30 Rotated Factor Matrix Correlations of the 10 ADLS with Two Factors Factors One Eating Performance Toileting Bathing Bladder Incontinence Bowel Continence Walking Stair Climbing Locomotion Transferring Mobility in Bed Two .667 .742 .404 .452 .648 .345 .117 .221 .618 .591 .221 .468 .385 .111 .113 .719 .560 .808 .556 .422 169 REFERENCES 170 REFERENCES Abruzzese, L. D. (1998). The Tinetti performance oriented mobility assessment tool. American Journal of Nursing, 98(12), 16J-L. Agostini, J. V., Han, L., & Tinetti, M. E. (2004). The relationship between number of medications and weight loss or impaired balance in older adults. Journal of the American Geriatrics Society, 52(10), 1719-1723. Ahmed, N., Mandel, R., & Fain, M. (2007). Frailty: An emerging geriatric syndrome. American Journal of Medicine, 120, 748-753. Alibhai, S. M. H., Gogov, S., & Allibhai, Z. (2006). Long-term side effects of androgen deprivation therapy in men with non-metastatic prostate cancer: A systematic literature review. Critical Reviews in Oncology/Hematology, 60(3), 201-215. American Cancer Society. (2008). Cancer facts and figures. American Cancer Society, 57th ed. Anderson, D., Kish, L., & Cornell, R. G. (1978). On stratification, grouping, and matching. Scandinavian Journal of Statistics, 7, 61-66. Anstey, K. J., Burns, R., von Sanden, C., & Luszcz, M. (2008). Psychological well-being is an independent predictor of falling in an 8-year follow-up of older adults. Journal of Gerontology, 63B(4), 249-257. Arroyave, W., Clipp, E., Miller, P., Jones, L., Ward, D., Bonner, M., et al. (2008). Childhood cancer survivor's perceived barriers to improving exercise and dietary behaviors. Oncology Nursing Forum, 35(1), 121-130. ASCO. (1995). American Society of Clinical Oncology: Outcomes of cancer treatment for technology assessment and cancer treatment guidelines. Journal of Clinical Oncology, 14(2), 671-679. Badger, T., Segrin, C., Dorros, S., Meek, P., & Lopez, A. (2007). Depression and anxiety in women with breast cancer and their partners. Nursing Research, 56(1), 44-53. Baker, D. I., King, M. B., Fortinsky, R. H., Graff, L. ., Gottschalk, M., Acampora, D., et al. (2005). Dissemination of an evidence-based multicomponent fall risk-assessment and management strategy throughout a geographic area. Journal of the American Geriatrics Society, 53(4), 675-680. Balducci, L. (2007). Aging, frailty, and chemotherapy. Cancer Control, 14(1), 7-12. Balducci, L., & Extermann, M. (2000a). Cancer and aging: An evolving panorama. Hematology/Oncology Clinics of North America, 14, 1-16. 171 Balducci, L., & Extermann, M. (2000b). Management of cancer in the older person: A practical approach. Oncologist, 5, 224-237. 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: Does it exist? Medical Care, 44(8), 745-753. Ballinger, G. A. (2004). Using generalized estimating equations for longitudinal data analysis. Organizational Research Methods, 7, 127-150. Bandeen-Roche, K., Xue, Q., Ferrucci, L., Walston, J., Guralnik, J. M., Chaves, P., et al. (2006). Phenotype of frailty: Characterization in the Women's Health and Aging Studies. Journals of Gerontology Series A: Biological Sciences & Medical Sciences, 61A(3), 262266. Barsevick, A. M., Dudley, W. N., & Beck, S. L. (2006). Cancer-related fatigue, depressive symptoms, and functional status: A mediation model. Nursing Research, 55(5), 366-372. Bartali, B., Frongillo, E. A., Bandinelli, S., Lauretani, F., Semba, R. D., Fried, L. P., et al. (2006). Low nutrient intake is an essential component of frailty in older persons. Journals of Gerontology Series A: Biological Sciences & Medical Sciences, 61A(6), 589-593. Bell, A. J., Talbot-Stern, J. K., & Hennessy, A. (2000). Characteristics and outcomes of older patients presenting to the emergency department after a fall: A retrospective analysis. Medical Journal of Australia, 173(4), 176-177. Bender, C. M., Engberg, S. J., Donovan, H., Cohen, S., Houze, L., Rosenzweig, M., et al. (2008). Symptom clusters in adults with chronic health problems and cancer as a comorbidity. Oncology Nursing Forum, 35(1), E1-11. Bennett, J. A., Winters, K. M., & Nail, L. (2007). Falls and characteristics of fallers among older breast cancer survivors. Disability & Rehabilitation, 29(20-21), 1651-1656. Beswick, A. D., Rees, K., Dieppe, P., Ayis, S., Gooberman-Hill, R., Horwood, J., et al. (2008). Complex interventions to improve physical function and maintain independent living in elderly people: A systematic review and meta-analysis. Lancet, 371(9614), 725-735. Blaauwbroek, R., Stant, A. D., Groenier, K. H., Kamps, W. A., Meyboom, B., & Postma, A. (2007). Health-related quality of life and adverse late effects in adult (very) long-term childhood cancer survivors. European Journal of Cancer, 43(1), 122-130. Bookchin, M. (1990). The philosophy of social ecology: Essays on dialectical naturalism. Montreal, Quebec: Black Rose Books. Bortz, W. M., II. (2002). A conceptual framework of frailty: A review. Journal of Gerontology and Biological Science, 57(5), M283-288. 172 Boult, C., Reider, L., Frey, K., Leff, B., Boyd, C. M., Wolff, J. L., et al. (2008). Early effects of "Guided Care" on the quality of health care for multimorbid older persons: A clusterrandomized controlled trial. Journal of Gerontology, 63A(3), 321-327. Bouman, A., van Rossum, E., Evers, S., Ambergen, T., Kempen, G. I. J., & Knipschild, P. (2008). Effects on health care use and associated cost of a home visiting program for older people with poor health status: A randomized clinical trial in the Netherlands. Journal of Gerontology, 63A(3), 291-297. Braun, I. M., Greenberg, D. B., & Pirl, W. F. (2008). Evidenced-based report on the occurrence of fatigue in long-term cancer survivors. Journal of the National Comprehensive Cancer Network, 6(4), 347-354. Bronfenbrenner, U. (1977). Toward an experimental ecology of human development. American Psychologist, 32, 513-531. Bronfenbrenner, U. (1979). The ecology of human development. Cambridge, MA: Harvard University Press. Browall, M. M., Ahlberg, K. M., Persson, L. O., Karlsson, P. O., & Danielson, E. B. (2008). The impact of age on health-related quality of life (HRQOL) and symptoms among postmenopausal women with breast cancer receiving adjuvant chemotherapy. Acta Oncologica, 47(2), 207-215. Brown, C. J., Gottschalk, M., Van Ness, P. H., Fortinsky, R. H., & Tinetti, M. E. (2005). Changes in physical therapy providers' use of fall prevention strategies following a multicomponent behavioral change intervention. Physical Therapy, 85(5), 394-403. Bylow, K., Dale, W., Mustian, K., Stadler, W. M., Rodin, M., Hall, W., et al. (2008). Falls and physical performance deficits in older patients with prostate cancer undergoing androgen deprivation therapy. Urology, 72(2), 422-427. Bylow, K., Mohile, S. G., Stadler, W. M., & Dale, W. (2007). Does androgen-deprivation therapy accelerate the development of frailty in older men with prostate cancer?: A conceptual review. Cancer, 110(12), 2604-2613. Cameron, I., Murray, G. R., Gillespie, L. D., Cumming, R. G., Robertson, M. C., K., H., [too many first initials?] et al. (2005). Interventions for preventing falls in older people in residential care facilities and hospitals (Protocol). Cochrane Database of Systematic Reviews, (3). Campbell, A. J., Borrie, M. J., Spears, G. F., Jackson, S. L., Brown, J. S., & Fitzgerald, J. L. (1990). Circumstances and consequences of fall experienced by a community population 70 years and over during a prospective study. Age & Ageing, 19, 136-141. 173 Centers for Disease Control and Prevention. (2007). A national action plan for cancer survivorship: Advancing public health strategies. Centers for Disease Control and Prevention. Centers for Disease Control and Prevention. (September 21, 2007). Fall Injury episodes among noninstitutionalized older adults: United States, 2001-2003, no. 292. Atlanta: Vital and Health Statistics. Centers for Disease Control and Prevention. (2006). Web-based injury statistics query and reporting system (WISQARS). National Center for Injury Prevention and Control, Centers for Disease Control and Prevention. Centers for Disease Control and Prevention. (2008). Preventing falls: How to develop community-based fall prevention programs for older adults. Atlanta, GA. National Center for Injury Prevention and Control, Centers for Disease Control and Prevention. Cesari, M., Landi, F., Torre, S., Onder, G., Lattanzio, F., & Bernabei, R. (2002). Prevalence and risk factors for falls in older community-dwelling population. Journals of Gerontology Series A: Biological Sciences & Medical Sciences, 57, M722-M726. Charlson, M. E., & Sax, F. L. (1987). The therapeutic efficacy of critical care units from two perspectives: A traditional cohort approach vs. a new case-control methodology. Journal of Chronic Diseases, 40(1), 31-39. Chelly, J. E., Conroy, L., Miller, G., Elliott, M. N., Horne, J. L., & Hudson, M. E. (2008). Risk factors and injury associated with falls in elderly hospitalized patients in a community hospital. Journal of Patient Safety, 4(3), 178-183. Chen, C. C., Kenefick, A. L., Tang, S. T., & McCorkle, R. (2004). Utilization of comprehensive geriatric assessment in cancer patients. Critical Review of Oncology and Hematology, 49(1), 53-67. Chen, J., Chan, D., Kiely, D. K., Morris, J. N., & Mitchell, S. L. (2007). Terminal trajectories of functional decline in the long-term care setting. Journals of Gerontology Series A: Biological Sciences & Medical Sciences, 62A(5), 531-536. Chen, J. S., Simpson, J. M., March, L. M., Cameron, I. D., Cumming, R. G., Lord, S. R., et al. (2008). Risk factors for fracture following a fall among older people in residential care facilities in Australia. Journal of the American Geriatric Society, 56(11), 1-7. Chen, Z., Maricic, M., Aragaki, A. K., Mouton, C., Arendell, L., Lopez, A. M., et al. (2008). Fracture risk increases after diagnosis of breast or other cancers in postmenopausal women: Results from the Women‘s Health Initiative. Osteoporosis International. Chen, Z., Maricic, M., Bassford, T. L., Pettinger, M., & Ritenbaugh, C. (2005). Fracture risk among breast cancer survivors: Results from the Women's Health Initiative Observational Study. Archives of Internal Medicine, 165(5), 552-558. 174 Chen, Z., Maricic, M., Pettinger, M., Ritenbaugh, C., Lopez, A., Barad, D., et al. (2005). Osteoporosis and rate of bone loss among postmenopausal survivors of breast cancer. Cancer, 104(7), 1520-1530. Cheville, A. L., Troxel, A. B., Basford, J. R., & Kornblith, A. B. (2008). Prevalence and treatment patterns of physical impairments in patients with metastatic breast cancer. Journal of Clinical Oncology, 26(16), 2621-2629. Cimprich, B., Janz, N. K., Northouse, L., Wren, P. A., Given, B., & Given, C. W. (2005). Taking charge: A self-management program for women following breast cancer treatment. Psycho-oncology, 14(9), 704-717. Clay, C. A., Perera, S., Wagner, J. M., Miller, M. E., Nelson, J. B., & Greenspan, S. L. (2007). Physical function in men with prostate cancer on androgen deprivation therapy. Physical Therapy, 87(10), 1325-1333. Clough-Gorr, K. M., Erpen, T., Gillmann, G., von Renteln-Kruse, W., Iliffe, S., Beck, J. C., et al. (2008). Preclinical disability as a risk factor for falls in community-dwelling older adults. Journal of Gerontology, 63A(3), 314-320. CMS. (2008). Medicaid State Waiver Program Demonstration Projects - General Information. Retrieved Month XX, 2010, from the U.S. Department of Health and Human Services Web site: http://www.cms.hhs.gov/MedicaidStWaivProgDemoPGI Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, N.J.: Elbaum. Coleman, R. E., Banks, L. M., Girgis, S. I., Kilburn, L. S., Vrdoljak, E., & Fox, J. (2007). Skeletal effects of exemestane on bone-mineral density, bone biomarkers, and fracture incidence in postmenopausal women with early breast cancer participating in the Intergroup Exemestane Study (IES): A randomized control trial. The Lancet Oncology, 8(2), 119. Coons, S. J., Chongpison, Y., Wendel, C. S., Grant, M., & Krouse, R. S. (2007). Overall quality of life and difficulty paying for ostomy supplies in the Veterans Affairs Ostomy HealthRelated Quality of Life Study: An exploratory analysis. Medical Care, 45(9), 891-895. Cope, D. G., & Reb, A. M. (2006). An evidence-based approach to the treatment and care of the older adult with cancer. Pittsburgh: Oncology Nursing Society. Coussement, J., De Paepe, L., Schwendimann, R., Denhaerynck, K., Dejaeger, E., & Milisen, K. (2008). Interventions for preventing falls in acute- and chronic-care hospitals: A systematic review and meta-analysis. Journal of the American Geriatric Society, 56. Crimmins, E. M., Saito, Y., & Reynolds, S. L. (1997). Further evidence on recent trends in the prevalence and incidence of disability among older Americans from two sources: The 175 Longitudinal Study on Aging (LSOA) and the National Health Interview Survey (NHIS). Journals of Gerontology Series B: Psychological Sciences & Social Sciences, 52B(2), S59-71. Cronbach, L. J., Gleser, G. C., Nanda, H., & Rajaratnam, N. (1972). The dependability of behavioral measurements: Theory of generalizability for scores and profiles. New York, NY: John Wiley. Daubman, S. (2008). The missing link: Bringing survivorship to the continuum of cancer care. Oncology Nursing Forum, 35(3), 500. Davison, J., & Marrinan, S. (2007). Falls. Reviews in Clinical Gerontology, 17(2), 93-107. de Groot, V., Beckerman, H., Lankhorst, G. J., & Bouter, L. M. (2003). How to measure comorbidity: a critical review of available methods. Journal of Clinical Epidemiology, 56(3), 221-229. de Rekeneire, N., Visser, M., Peila, R., Nevitt, M. C., Cauley, J. A., Tylavsky, F. A., et al. (2003). Is a fall just a fall: Correlates of falling in healthy older persons. The health, aging, and body composition study. Journal of the American Geriatrics Society, 51(6), 841-846. 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. Deimling, G. T., Sterns, S., Bowman, K. F., & Kahana, B. (2007). Functioning and activity participation restrictions among older adult, long-term cancer survivors. Cancer Investigation, 25(2), 106-116. Delmas, P. D., & Fontana, A. (1998). Bone loss induced by cancer treatment and its management. European Journal of Cancer, 34(2), 260-262. DeSanto-Madeya, S., Bauer-Wu, S., & Gross, A. (2007). Activities of daily living in women with advanced breast cancer. Oncology Nursing Forum, 34(4), 841-846. Deyo, R. A., & Inui, T. S. (1984). Toward clinical applications of health status measures: Sensitivity of scales to clinically important changes. Health Services Research, 19(3), 275-289. Donald, I. P., & Bulpitt, C. J. (1999). The prognosis of falls in elderly people living at home. Age & Ageing, 32(1), 121-125. Dunn, J. E., Rudberg, M. A., Furner, S. E., & Cassel, C. K. (1992). Mortality, disability, and falls in older persons. The role of underlying disease and disability. American Journal of Public Health, 82, 395-400. 176 Elder, G. H. (1985). Life course dynamics: Trajectories and transitions, 1968-1980. London, England: Cornell University Press. Elixhauser, A., Steiner, C., Harris, D. R., & Coffey, R. M. (1998). Comorbidity measures for use with administrative data. Medical Care, 36(1), 8-27. Erfelder, E., Faul, F., & Buchner, A. (1996). GPower: A general power analysis program. Behavioral Research Methods, Instruments, & Computers, 28, 1-11. Extermann, M. (2000a). Measurement and impact of comorbidity in older cancer patients. Critical Review in Oncology and Hematology, 35, 181-200. Extermann, M. (2000b). Measurement and impact of comorbidity in older cancer patients. Critical Review of Oncology and Hematology, 35, 181-200. Extermann, M. (2000c). Measuring comorbidity in older cancer patients. European Journal of Cancer, 36, 453-471. Fauth, E., Zarit, S. H., Malmberg, B., & Johansson, B. (2007). Physical, cognitive, and psychosocial variables from the disablement process model predict patterns of independence and the transition into disability for the oldest old. The Gerontologist, 47(5), 613-624. Ferrans, C., Zerwic, J., Wilbur, J., & Larson, J. (2005). Conceptual model of health-related quality of life. Journal of Nursing Scholarship, 37(4), 223-337. Ferrucci, L., Baldasseroni, S., Bandinelli, S., De Alfieri, W., Cartei, A., Calvani, D., et al. (2000). Disease severity and health-related quality of life across different chronic conditions. Journal of the American Geriatrics Society, 48, 1490-1495. Ferrucci, L., Guralnik, J., & Cavazzini, C. (2003). The frailty syndrome: A critical issue in geriatric oncology. Critical Review of Oncology and Hematology, 46(2), 127-137. Flood, K. L., Carroll, M. B., Le, C. V., Ball, L., Esker, D. A., & Carr, D. B. (2006). Geriatric syndromes in elderly patients admitted to an oncology-acute care for elders unit. Journal of Clinical Oncology, 24(15), 2298-2303. Fogarty, L., Roter, D., Larson, S., Burke, J., Gillespie, J., & Levy, R. (2002). Patient adherence to HIV medication regimens: A review of published and abstract reports. Patient Education and Counseling, 46(2), 93-108. Forsyth, K. A. (2000). Using the minimum data set to measure change in functional status of nursing home residents. Unpublished Dissertation, University of Illinois at Chicago, Health Sciences Center. 177 Fortinsky, R. H., Baker, D., Gottschalk, M., King, M., Trella, P., & Tinetti, M. E. (2008). Extent of implementation of evidence-based fall prevention practices for older patients in home health care. Journal of the American Geriatrics Society, 56(4), 737-743. Fortinsky, R. H., Iannuzzi-Sucich, M., Baker, D. I., Gottschalk, M., King, M. B., Brown, C. J., et al. (2004). Fall-risk assessment and management in clinical practice: Views from health care providers. Journal of the American Geriatrics Society, 52(9), 1522-1526. Fox, S. W., & Lyon, D. (2007). Symptom clusters and quality of life in survivors of ovarian cancer. Cancer Nursing, 30(5), 354-361. Freedman, V. A., Martin, L. G., Schoeni, R. F., & Cornman, J. C. (2008). Declines in late-life disability: The role of early- and mid-life factors. Social Science & Medicine, 66(7), 1588-1602. French, D. D., Werner, D. C., Campbell, R. R., Powell-Cope, G. M., Nelson, A. L., Rubenstein, L. Z., et al. (2007). A multivariate fall risk assessment model for VHA nursing homes using the minimum data set. Journal of the American Medical Directors Association, 8(2), 115-122. Fried, L. F., Lee, J. S., Shlipak, M., Chertow, G. M., Green, C., Ding, J., et al. (2006). Chronic kidney disease and functional limitation in older people: Health, aging and body composition study. Journal of the American Geriatrics Society, 54(5), 750-756. Fried, L. P., Ferrucci, L., Darer, J., Williamson, J. D., & Anderson, G. (2004). Untangling the concepts of disability, frailty, and comorbidity: Implications for improved targeting of care. Journal of Gerontology, 4(59), 255-263. Fried, L. P., Tangen, C. M., & Walston, J. (2001). Frailty in older adults: Evidence for a phenotype. Journal of Gerontological Medical Science 56A, M146-M156. Friedlaender, G. E., Tross, R. B., & Doganis, A. C. (1984). Effects of chemotherapeutic agents on bone. Short-term methotrexate and doxorubicin (Adriamycin) treatment in a rat model. Journal of Bone Joint Surgery, 66, 602–607. Ganz, P. A., Casillas, J., & Hahn, E. E. (2008). Ensuring quality care for cancer survivors: implementing the survivorship care plan. Seminars in Oncology Nursing, 24(3), 208-217. Gardiner, J. C., Luo, Z., & Roman, L. A. (2009). Fixed effects, random effects, and GEE: What are the differences? Statistics in Medicine, 28. Gill, T. M., Robison, J. T., & Tinetti, M. E. (1998). Difficulty and dependence: Two components of the disability continuum among community-living older persons. Annals of Internal Medicine, 128(2), 96-101. 178 Gill, T. M., Williams, C. S., & Tinetti, M. E. (2000). Environmental hazards and the risk of nonsyncopal falls in the homes of community-living older persons. Medical Care, 38(12), 1174-1183. Gillespie, L. D., Gillespie, W. J., Robertson, M. C., et al. (2004 ). Interventions for preventing falls in elderly people. Cochrane Database Systematic Review, 6. Gillespie, L. D., Gillespie, W. J., Robertson, M. C., Lambert, S. D., Cummings, S. R., & Rowe, B. H. (2003). Interventions for preventing falls in elderly people. Cochrane Database Systematic Review, 4. Given, B., Given, C., Azzouz, F., & Stommel, M. (2001). Physical functioning of elderly cancer patients prior to diagnosis and following initial treatment. Nursing Research, 50(4), 222232. Given, B. A., Given, C. W., Sikorskii, A., & Hadar, N. (2007). Symptom clusters and physical function for patients receiving chemotherapy. Oncology Nursing Forum, 23(2), 121-126. Given, B. A., & Sherwood, P. R. (2005). Nursing-sensitive patient outcomes — a white paper. Oncology Nursing Forum, 32(4), 773-784. Given, C., Sikorskii, A. P., Spoelstra, S., & You, M. (2010). The burden of cancer among a longitudinal cohort of Medicaid patients in a home and community based waiver program. Paper presented at the National Cancer Institute 5th Biennial Cancer Survivorship Research Conference, Washington, D.C. Goodwin, J. A. (2007). Older adults' functional performance loss and adaptation during chemotherapy. Geriatric Nursing, 28(6), 370-376. Goodwin, P. J., Black, J., Bordeleau, L. J., & Gantz, P. (2003). Health-related quality of life measurement in randomized clinical trials in breast cancer: Taking stock. Journal of the National Cancer Institute, 95, 263-281. Greenspan, S. L., Bhattacharya, R. K., Sereika, S. M., Brufsky, A., & Vogel, V. G. (2007). Prevention of bone loss in survivors of breast cancer: A randomized, double-blind, placebo-controlled clinical trial. Journal of Clinical Endocrinology Metabolism, 92(1), 131-136. Guilley, E., Ghisletta, P., Armi, F., Berchtold, A., d'Epinay, C. L., Michel, J., et al. (2008). Dynamics of frailty and ADL dependence in a 5-year longitudinal study of octogenarians. Research on Aging, 30(3), 299-317. Gulluoglu, B. M., Cingi, A., Cakir, T., Gercek, A., Barlas, A., & Eti, Z. (2006). Factors related to post-treatment chronic pain in breast cancer survivors: The interference of pain with life functions. International Journal of Fertility & Womens Medicine, 51(2), 75-82. 179 Haines, T. P., Hill, K., Walsh, W., & Osborn, R. (2007). Design-related bias in hospital fall risk screening tool predictive accuracy evaluations: Systematic review and meta-analysis. Journals of Gerontology Series A: Biological Sciences & Medical Sciences, 62A(6), 664672. Harrison, B., Booth, D., & Algase, D. (2001). Studying fall risk factors among nursing home residents who fell. Journal of Gerontological Nursing, 27(10), 26-34. Hawkins, M. M., Lancashire, E. R., Winter, D. L., Frobisher, C., Reulen, R. C., Taylor, A. J., et al. (2008). The British Childhood Cancer Survivor Study: Objectives, methods, population structure, response rates, and initial descriptive information. Pediatric Blood & Cancer, 50(5), 1018-1025. Hawley, A. H. (1950). Human ecology: A theory of community structure. New York, NY: Ronald Press. Hewitt, M., & Rowland, J. H. (2002). Mental health service use among adult cancer survivors: Analyses of the National Health Interview Survey. Journal of Clinical Oncology, 20(23), 4581-4590. Hewitt, M., Rowland, J. H., & Yancik, R. (2003). Cancer survivors in the United States: Age, health, and disability. Journals of Gerontology Series A: Biological Sciences & Medical Sciences, 58A(1), 82-91. Hewitt, M. E., Bamundo, A., Day, R., & Harvey, C. (2007). Perspectives on post-treatment cancer care: Qualitative research with survivors, nurses, and physicians. Journal of Clinical Oncology, 25(16), 2270-2273. Hirdes, J., Fries, B., Morris, J. N., Ikegami, N., Zimmerman, D., Dalby, D., et al. (2004). Home care quality indicators (HCQIs) based on the MDS-HC. The Gerontologist, 44(5), 665679. Hodgson, N. A. (2002). Epidemiological trends of cancer in older adults: Implications for gerontological nursing practice and research. Journal of Gerontological Nursing, 101(4), 34-23. Holen, J. C., Lydersen, S., Klepstad, P., Loge, J. H., & Kaasa, S. (2008). The Brief Pain Inventory: Pain's interference with functions is different in cancer pain compared with noncancer chronic pain. Clinical Journal of Pain, 24(3), 219-225. Holley, S. (2000). Cancer-related fatigue. Suffering a different fatigue. Cancer Practice, 8, 8795. Holley, S. (2002). A look at the problem of falls among people with cancer. Clinical Journal of Oncology Nursing, 6(4), 193. 180 Horning, S. J. (2008). Follow-up of adult cancer survivors: New paradigms for survivorship care planning. Hematology/Oncology Clinics of North America, 22(2), 201-210. Hurria, A., Lichtman, S. M., Gardes, J., Li, D., Limaye, S., Patil, S., et al. (2007). Identifying vulnerable older adults with cancer: Integrating geriatric assessment into oncology practice. Journal of the American Geriatrics Society, 55(10), 1604-1608. Huss, A., Stuck, A., Rubenstein, L. Z., Egger, M., & Glough-Gorr, K. M. (2008). Multidimensional preventive home visit programs for community-dwelling older adults: A systematic review and a meta-analysis of randomized control trials. Journal of Gerontology, 63A(3), 298-307. Iezzoni, L., & Freedman, V. A. (2008). Turning the disability tide: The importance of definitions. JAMA, 299(3), 332-334. Inouye, S. K., Studenski, S., Tinetti, M. E., & Kuchel, G. A. (2007). Geriatric syndromes: Clinical, research, and policy implications of a core geriatric concept. Journal of the American Geriatrics Society, 55(5), 780-791. Institute of Medicine. (2005). Cancer Survivorship Care Planning. Retrieved October 19, 2009, from http://www.iom.edu/~/media/Files/Report%20Files/2005/From-Cancer-Patient-toCancer-Survivor-Lost-in-Transition/factsheetcareplanning.ashx InterRai. (2008). Homepage. Retrieved October 26, 2008, from http://www.interrai.org/section/view/? Jager, T. E., Weiss, H. B., Coben, J. H., & Pepe, P. E. (2000). Traumatic brain injuries evaluated in U. S. emergency departments 1992 to 1994. Academic Emergency Medicine, 7(2), 134-140. Janz, N. K., Mujahid, M., Chung, L. K., Lantz, P. M., Hawley, S. T., Morrow, M., et al. (2007). Symptom experience and quality of life of women following breast cancer treatment. Journal of Women's Health (15409996), 16(9), 1348-1361. Johnson, R. W., & Wiener, J. M. (2006). A profile of frail older Americans and their caregivers. Changes in health care financing and organization initiative grant no. 049919. Robert Wood Johnson Foundation. Jones, H., Jones, M., Bayley, N., Macfarlane, J., & Honzik. (1971). The course of human development: Selected papers from the longitudinal studies, Institute of Human Development, The University of California, Berkley. Waltham, MA: Xerox College Publishing. Joreskog, K. G., & Sorbom, D. (2004). LISREL 8.7 for Windows. Lincolnwood, IL: Scientific Software International. 181 Kagan, S. H. (2004). Gero-oncology nursing research. Oncology Nursing Forum, 31(2), 293299. Kanis, J. A., McCloskey, E. V., Powles, T., Paterson, A. H. G., Ashley, S., & Spector, T. (1999). A high incidence of vertebral fracture in women with breast cancer. British Journal of Cancer, 79(7-8), 1179-1181. Keating, N. L., Landrum, M. B., Klabunde, C. N., Fletcher, R. H., Rogers, S. O., Doucette, W. R., et al. (2008). Adjuvant chemotherapy for stage III colon cancer: Do physicians agree about the importance of patient age and comorbidity? Journal of Clinical Oncology, 26(15), 2532-2540. Keating, N. L., Narredam, 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 Geriatrics Society, 53(12), 2145-2152. Kerber, C. S., Dyck, M. J., Culp, K. R., & Buckwalter, K. (2005). Comparing the Geriatric Depression Scale, Minimum Data Set, and primary care provider diagnosis for depression in rural nursing home residents. Journal of the American Psychiatric Nurses Association, 11(5), 269-275. Klabunde, C., Harlen, M., & 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., Legler, J. M., Warren, J. L., Baldwin, L. M., & Schrag, D. (2007). A refined comorbidity measurement algorithm for claims-based studies of breast, prostate, colorectal, and lung cancer patients. Annals of Epidemiology, 17(8), 584-590. Kline, R. B. (2005). Principles and Practice of Structural Equation Modeling. New York, NY: The Guilford Press. Knobf, M. T., Musanti, R., & Dorward, J. (2007). Exercise and quality of life outcomes in patients with cancer. Seminars in Oncology Nursing, 23(4), 285-296. Koch, M., Gottschalk, M., Baker, D. I., Palumbo, S., & Tinetti, M. E. (1994). An impairment and disability assessment and treatment protocol for community-living elderly persons. Physical Therapy, 74(4), 286-298. Koroukian, S., Murray, P., & Madigan, E. (2006). Comorbidity, disability, and geriatric syndromes in elderly cancer patients receiving home health care. Journal of Clinical Oncology, 24(15), 2304-2310. Kupper, L. L., Karon, J. M., Kleinbaum, D. G., Morgenstern, H., & Lewis, D. K. (1981). Matching in epidemiologic studies: Validity and efficiency considerations. Biometrics, 37(2), 271-291. 182 Lamb, S. E., Jorstad-Stein, E. C., Hauer, K., & Becker, C. (2005). Prevention of falls network Europe and Outcome consensus group. Development of a common outcome data set for fall injury prevention trials: The prevention of falls network Europe consensus. Journal of American Geriatric Society, 53, 1618-1622. Landi, F., Onder, G., Cesari, M., Barillaro, C., Russo, A., & Bernabei, R. (2005). Psychotropic medications and risk for falls among community-dwelling frail older people: An observational study. Journal of Gerontology: Medical Sciences, 60(5), 622-626. 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 & Symptom Management, 29(3), 255-262. Lee, J., & Rantz, M. (2008). Correlates of post-hospital physical function at 1 year in skilled nursing facility residents. Journal of Advanced Nursing, 62(4), 479-486. Lee, Y., Santacroce, S. J., & Sadler, L. (2008). Predictors of healthy behavior in long-term survivors of childhood cancer. Journal of Clinical Nursing, 16(11c), 285-295. Leipzig, R. M., Cumming, R. G., & Tinetti, M. E. (1999a). Drugs and falls in older people: A systematic review and meta-analysis: Psychotropic drugs. Journal of the American Geriatrics Society, 47(1), 30-39. Leipzig, R. M., Cumming, R. G., & Tinetti, M. E. (1999b). Drugs and falls in older people: A systematic review and meta-analysis: II. Cardiac and analgesic drugs. Journal of the American Geriatrics Society, 47(1), 40-50. Levy, M. E., Perera, S., van Londen, G. J., Nelson, J. B., Clay, C. A., & Greenspan, S. L. (2008). Physical function changes in prostate cancer patients on androgen deprivation therapy: A 2-year prospective study. Urology, 71(4), 735-739. Li, L. W. (2005). Trajectories of ADL disability among community-dwelling frail older persons. Research on Aging, 27(1), 56-79. Liu, S., Dixon, J. M., Qiu, G., Tian, Y., & McCorkle, R. (2009). Using generalized estimating equations to analyze longitudinal data in nursing research. Western Journal of Nursing Research, 31(7), 948-964. Lim, L. S., & Chutka, D. S. (2006). Preventive medicine beyond 65. Geriatrics & Gerontology International, 6(2), 73-81. Limburg, C. E. (2007). Screening, prevention, detection, and treatment of cancer therapyinduced bone loss in patients with breast cancer. Oncology Nursing Forum, 34(1), 55-61. 183 Luctkar-Flude, M. F., Groll, D. L., Tranmer, J. E., & Woodend, K. (2007). Fatigue and physical activity in older adults with cancer: A systematic review of the literature. Cancer Nursing, 30(5), E35-45. Mandelblatt, J. S., Lawrence, W. F., Cullen, J., Stanton, A. L., Krupnick, J. L., Kwan, L., et al. (2006). Patterns of care in early-stage breast cancer survivors in the first year after cessation of active treatment. Journal of Clinical Oncology, 24(1), 77-84. Mc Elroy, K. R., Bibeau, D., Steckler, A., & Glanz, K. (1988). An ecological perspective on health promotion programs. Health Education Quarterly, 15(4), 351-377. Mc Nab, P. (March 24, 2007). Discussion on types of patients serviced in home and communitybased waiver program in the State of Michigan with S. Spoelstra. East Lansing, Michigan. McCullagh, P., & Nelder, J. (1989). Generalized Linear Models. London, England: Chapman and Hall. Melis, R. J., Adand, E., Teerenstra, S., van Eijken, M. I., Wimo, A., van Achterberg, T., et al. (2008). Cost-effectiveness of a multidisciplinary intervention model for communitydwelling frail older people. Journal of Gerontology, 63A(3), 275-282. Melis, R. J., Van Eijken, M. I., Teerenstra, S., van Achterberg, T., Parker, S. G., Borm, G. F., et al. (2008). A randomized study of a multidisciplinary program to intervene on geriatric syndromes in vulnerable older people who live at home (Dutch EASYcare Study). Journal of Gerontology, 63A(3), 283-290. Michaud, L. B., & Goodin, S. (2006). Cancer-treatment-induced bone loss, part 1. American Journal of Health System Pharmacies, 63, 419-430. Michigan State University Institutional Review Boards (IRB). Human research protection program: Data and safety monitoring. Retrieved September 29, 2008, from http://www.humanresearch.msu.edu/hrp_manual/6-7_Data_and_Safety_Monitoring_5-608.pdf Mock, V., Frangakis, C., Davidson, N. E., Ropka, M. E., Pickett, M., Poniatowski, B., et al. (2005). Exercise manages fatigue during breast cancer treatment: A randomized controlled trial. Psychooncology, 14(6), 464-477. Mohile, S. G., Xian, Y., Dale, W., Fisher, S. G., Rodin, M., Morow, G. R., et al. (2009). Association of a cancer diagnosis with vulnerability and frailty in older Medicare beneficiaries. Journal of the National Cancer Institute, 101(17), 1206-1215. Moreh, E., Jacobs, J. M., & Stessman, J. (2010). Fatigue, function, and mortality in older adults. The Journals of Gerontology Series A: Biological Sciences and Medical Sciences, 65A(8), 887-895. 184 Moreland, J. D., Richardson, J. A., Goldsmith, C. H., & Classe, C. (2004). Muscle weakness and falls in older adults: A systemic review and meta-analysis. Journal of the American Geriatrics Society, 52(7), 1121-1129. Morris, J. N., Fries, B., Steel, K., Ikegami, N., Bernabei, R., Carpenter, G. I., et al. (1997). Comprehensive clinical assessment in community setting: Applicability of the MDS-HC. Journal of the American Geriatrics Society, 45(8), 1-13. Morse, J. M., Morse, R. M., & Tylko, S. (1989). Development of a scale to identify the fallprone patient. Canada Journal of Aging, 8, 366. Morse, J. M., Tylko, S. J., & Dixon, H. A. (1987). Characteristics of the fall-prone patient. The Gerontologist, 27(4), 516-522. Mortimer, J. T., & Shanahan, M. J. (2003). Handbook of the Life Course. New York, NY: Kluwer Academic Plenium Publishers. Myers, H. (2003). Hospital falls risk assessment tools: A critique of the literature. International Journal of Nursing Practice, 9, 233-235. Mudge, A. M., O'Rourke, P., & Denaro, C. P. (2010). Timing and risk factors for functional changes associated with medical hospitalization in older patients. The Journals of Gerontology Series A: Biological Sciences and Medical Sciences, 65A(8), 866-872. Muthen, L. K., & Muthen, B. O. (1998-2007). Mplus User’s Guide (5th ed.). Los Angeles, CA: Muthen & Muthen. National Cancer Institute. (2008). National Cancer Institute dictionary of cancer terms. Retrieved August 18, 2003, from http://www.cancer.gov/dictionary National Institute for Health and Clinical Excellence. (2004). Falls. The assessment and prevention of falls in older people. Retrieved February 22, 2009, from www.nice.org.uk/guidance/CG21/guidance/pdf/English National Institute on Aging. (2008). Cognitive impairment in older people. Retrieved August 18, 2008, from http://www.nia.nih.gov/Alzheimers/ResearchInformation/NewsReleases/Archives/PR200 1/PR20011112cognitiveimpairment.htm Nelder, J., & Wedderburn, R. (1972). Generalized Linear Models. Journal of the Royal Statistical Society. Series A (General), 135(3), 370-384. Nevitt, M. C., Cummings, S. R., Kidd, S., & Black, D. (1989). Risk factors for recurrent nonsyncopal falls. A prospective study. JAMA, 261(18), 2663-2668. 185 O'Connell, B., Wellman, D., Cockayne, M., & Baker, L. (2005). Fall risk factors and the nature of falls in inpatient oncology and palliative care settings. Contemporary Nurse: A Journal for the Australian Nursing Profession, 18(3), 247-257. O'Connell, B. O., Baker, L., Gaskin, C. J., & Hawkins, M. T. (2007). Risk items associated with patient falls in oncology and medical settings. Journal of Nursing Care Quality, 22(2), 130-137. Oliver, D. (2007). Preventing falls and falls-injuries in hospitals and long-term care facilities. Reviews in Clinical Gerontology, 17(2), 75-91. Oliver, D. (2008). Evidence for fall prevention in hospitals a response to Coussement, J., De Paepe, L., Schwendimann, R. et al. Interventions for preventing falls in acute and chronic care hospitals: A systematic review and meta-analysis. Journal of the American Geriatrics Society, 56(9), 1774-1775. Oliver, M. M., Daly, F., & Martin, F. (2004). Risk factors and risk assessment tools for falls in hospital inpatients. A systematic review. Age Ageing, 33, 122-130. Overcash, J. (2007). Journal club. Prediction of falls in older adults with cancer: A preliminary study. Oncology Nursing Forum, 34(2), 341-346. Overcash, J. (2008). Vitamin D in older patients with cancer. Clinical Journal of Oncology Nursing, 12(4), 655-659. 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(4), 286-293. Patrick, D. (1997). Finding health-related quality of life outcomes sensitive to health care organization and delivery. Medical Care, 35(11), NS49-NS57. Patrick, D., & Chiang, Y. (2000). Measurement of health outcomes in treatment effectiveness evaluations: Conceptual and methodological challenges. Medical Care, 38(9), 14-25. Pautex, S., Herrmann, F. R., & Zulian, G. B. (2008). Factors associated with falls in patients with cancer hospitalized for palliative care. Journal of Palliative Medicine, 11(6), 878-884. Pearce, T., & Ryan, S. (2008). Cancer and falls risk assessment. Australian Nursing Journal, 15(8), 37-39. Pearse, H., Nicholson, L., & Bennett, M. (2004). Falls in hospices: A cancer network observational study of fall rates and risk factors. Palliative Medicine, 18(5), 478-481. Peeters, G. M. E. E., Verweij, L. M., van Schoor, N. M., Pijnappels, M., Pluijm, S. M. F., 186 Visser, M., et al. (2010). Which types of activities are associated with risk of recurrent falling in older persons? The Journals of Gerontology Series A: Biological Sciences and Medical Sciences, 65A(7), 743-750. Phipps, E., Braitman, L. E., Stites, S., & Leighton, J. C. (2008). Quality of life and symptom attribution in long-term colon cancer survivors. Journal of Evaluation in Clinical Practice, 14(2), 254-258. Rehse, B., & Pukrop, R. (2003). Effects of psychosocial interventions on quality of life in adult cancer patients: Meta analysis of 37 published controlled outcome studies. Patient Education & Counseling, 50, 179-186. Reid, M. C., Williams, C. S., Concato, J., Tinetti, M. E., & Gill, T. M. (2003). Depressive symptoms as a risk factor for disabling back pain in community-dwelling older persons. Journal of the American Geriatrics Society, 51(12), 1710-1717. Ries, L. A. G., Melber, D., Krapcho, M., Mariotto, A., Miller, B. B., Feuer, E. J., et al. (2005). Seer cancer statistics review, 1975-2004. Retrieved October 19, 2009, from http://seer.cancer.gov/csr/1975_2004 Roche, K. B., Xue, Q. L., Ferrucci, L., Walston, J. D., Gure, T., Chaves, P. H. M., et al. (2006). Phenotype of frailty: Characterization in the women's health and aging studies. Journal of Gerontology, 61A(3), 262-266. Rowland, J. H., & Yancik, R. (2006). Cancer survivorship: The interface of aging, comorbidity, and quality care. Journal of the National Cancer Institute, 98(8), 504-505. Rubenstein, L. Z., Kenny, R. A., Eccles, M., Martin, F., & Tinetti, M. E. (2002). Evidence-based guideline for falls prevention: Summary of the bi-national panel. Generations, 26(4), 3841. Rubin, D. (1973). Matching to remove bias in observational studies. Biometrics, 29, 159-183. Saarto, T., Vehmanen, L., Elomaa, I., Valimaki, M., Makela, P., & Blomqvist, C. (2001). The effects of clodronate and antioestrogens on bone loss associated with estrogen withdrawal in postmenopausal women with breast cancer. British Journal of Cancer, 84(8), 10471051. Sawyer, P., Lillis, J. P., Bodner, E. V., & Allman, R. M. (2007). Substantial daily pain among nursing home residents. Journal of the American Medical Directors Association, 8(3), 158-165. Scaf-Klomp, W., Van Sonderen, E., Sanderman, R., Ormel, J., & Kempen, G. I. (2001). Recovery of physical function after limb injuries in independent older people living at home. Age & Ageing, 30(3), 213-219. 187 Schmitz, K. H., Cappola, A. R., Stricker, C. T., Sweeney, C., & Norman, S. A. (2007). The intersection of cancer and aging: Establishing the need for breast cancer rehabilitation. Cancer Epidemiology Biomarkers Prevention, 16(5), 866-872. Scott, V., Votova, K., & Scanlan, A. (2007). Multifactoral and functional mobility assessment tools for falls risk among older adults in the community, home-support, long-term care, and acute settings. Age & Ageing, 36, 130-140. Semba, R. D., Bartali, B., Zhou, J., Blaum, C., Ko, C., & Fried, L. P. (2006). Low serum micronutrient concentrations predict frailty among older women living in the community. Journals of Gerontology Series A: Biological Sciences & Medical Sciences, 61A(6), 594599. Shega, J. W., Weiner, D. K., Paice, J. A., Bilir, S. P., Rockwood, K., Herr, K., et al. (2010). The association between noncancer pain, cognitive impairment, and functional disability: An analysis of the Canadian Study of Health and Aging. The Journals of Gerontology Series A: Biological Sciences and Medical Sciences, 65A(8), 880-886. Sibbritt, D. W., Byles, J. E., & Regan, C. (2007). Factors associated with decline in physical functional health in a cohort of older women. Age & Ageing, 36(4), 382-388. Snyder, C. F., Garrett-Mayer, E., Brahmer, J. R., Carducci, M. A., Pili, R., Stearns, V., et al. (2008). Symptoms, supportive care needs, and function in cancer patients: How are they related? Quality of Life Research, 17(5), 665-677. Spoelstra, S. L., Given, B., & Given, C. (2010a). The influence of medications on falls in community dwelling elderly. Paper presented at the Sigma Theta Tau-Michigan State University. Spoelstra, S. L., Given, B., & Given, C. (2010b). The influence of medications on falls in community dwelling elderly. Paper presented at the Midwest Nursing Research Society. Spoelstra, S. L., Given, B., & Given, C. (2010). The influence of medications on falls in community dwelling elderly. Paper presented at the Sigma Theta Tau-Grand Valley State University. Spoelstra, S. L., Given, B., von Eye, A., & Given, C. (2009). Falls in the community dwelling elderly with a history of cancer. Paper presented at the Gerontological Society of American. Atlanta, GA. Spoelstra, S. L., Given, B., von Eye, A., & Given, C. (2010a). Fall risk in community dwelling elderly cancer survivors—a predictive model for gerontological nurses. Journal of Gerontological Nursing, 36(2), 110-118. Spoelstra, S. L., Given, B., von Eye, A., & Given, C. (2010b). Falls in the community dwelling 188 elderly with a history of cancer. Cancer Nursing, 33(2), 149-155. Stel, V. S., Pluijm, S. M. F., Deeg, D. J. H., Smit, J. H., Bouter, L. M., & Lips, P. (2003). A classification tree for predicting recurrent falling in community-dwelling older persons. Journal of the American Geriatrics Society, 51(10), 1356-1364. Stel, V. S., Smit, J. H., Pluijm, S. M. F., & Lips, P. (2004). Consequences of falling in older men and women and risk factors for health service use and functional decline. Age & Ageing, 33(1), 58-65. Sterling, D. A., O'Connor, J. A., & Bonadies, J. (2001). Geriatric falls: Injury severity is high and disproportionate to mechanism. Journal of Trauma-Injury, Infection and Critical Care, 50(1), 116-119. Stevens, J. A., Corso, P. S., Finkelstein, E. A., & Miller, T. R. (2006). The cost of fatal and nonfatal falls among older adults. Injury Prevention, 12, 290-295. Stevens, J. A., & Sogolow, E. D. (2005). Gender differences for non-fatal unintentional fallrelated injuries among older adults. Injury Prevention, 11, 115-119. Stommel, M., & Wills, C. E. (2004). Clinical research: Concepts and principles for Advanced Practice Nurses. Philadelphia: Lippincott, Williams, and Wilkins. Stovall, E. (2008). National Coalition for Cancer Survivorship: Advocacy for quality cancer care. Journal of Oncology Practice, 4(3), 145-149. Sweeney, C., Schmitz, K. H., Lazovich, D., Virnig, B. A., Wallace, R. B., & Folsom, A. R. (2006). Functional limitations in elderly female cancer survivors. Journal of the National Cancer Institute, 98(8), 521-529. Szabo, S. M., Jannsen, P. A., Khan, K., Potter, M. J., & Lord, S. R. (2008). Older women with age-related macular degeneration have a greater risk of falls: A physiological profile assessment study. Journal of the American Geriatrics Society, 56(5), 800-807. The Joint Commission. (2009). 2009 National Patient Safety Goals Ambulatory Care Program. Retrieved October 19, 2009, from http://www.jointcommission.org/PatientSafety/NationalPatientSafetyGoals/09_amb_npsg s.htm Tinetti, M. E. (2003). Clinical practice. Preventing falls in elderly persons. New England Journal of Medicine, 348(1), 42-49. Tinetti, M. E. (2008). Multifactoral fall-prevention strategies: Time to retreat or advance. Journal of the American Geriatrics Society, 56(8), 1563-1565. 189 Tinetti, M. E., Allore, H., Araujo, K. L. B., & Seeman, T. (2005). Modifiable impairments predict progressive disability among older persons. Journal of Aging & Health, 17(2), 239-256. Tinetti, M. E., Baker, D. I., King, M., Gottschalk, M., Murphy, T. E., Acampora, D., et al. (2008). Effect of dissemination of evidence in reducing injuries from falls. New England Journal of Medicine, 359(3), 252-261. Tinetti, M. E., Baker, D. I., & McAvay, G. (1994). A multifactoral intervention to reduce the risk of falling among elderly people living in the community. New England Journal of Medicine, 331, 821-827. Tinetti, M. E., McAvay, G. J., Fried, T. R., Allore, H. G., Salmon, J. C., Foody, J. M., et al. (2008). Health outcome priorities among competing cardiovascular, fall injury, and medication-related symptom outcomes. Journal of the American Geriatrics Society, 56(8), 1409-1416. Tinetti, M. E., Speeckley, M., & Ginter, S. F. (1988). Risk factors for falls among elderly persons living in the community. New England Journal of Medicine, 3(19), 1701-1707. Tinetti, M. E., & Williams, C. S. (1998). The effect of falls and fall injuries on functioning in community-dwelling older persons. Journals of Gerontology Series A: Biological Sciences & Medical Sciences, 53A(2), M112-119. Tinetti, M. E., Williams, C. S., & Gill, T. M. (2000). Health, functional, and psychological outcomes among older persons with chronic dizziness. Journal of the American Geriatrics Society, 48(4), 417-421. Travis, L. B., & Yahalom, J. (2008). Cancer survivorship: Facing forward. Hematology/Oncology Clinics of North America, 22(2), 365-371. Twisk, J. W. R. (2003). Applied longitudinal data analysis for epidemiology: A practical guide. West Nyack, NY: Cambridge University Press. U.S. Census Bureau. ( 2008). Statistical Abstract of the United States, Washington, D.C. U.S. Department of Health and Human Services. (2005). United States Cancer Statistics: 19992002 incidence and mortality Web-based report. Retrieved Month XX, 2010, from Centers for Disease Control Web site: www.cdc.gov/cancer/npcr/uscs van Helden, S., van Geel, A. C. M., Geusens, P. P., Kessels, A., Kruseman, A. C. N., & Brink, P. R. G. (2008). Bone and fall-related fracture risks in women and men with a recent clinical fracture. Journal of Bone & Joint Surgery, 90A(2), 241-248. 190 Van Voorhees, B. W., Walters, A. E., Prochaska, M., & Quinn, M. T. (2007). Reducing health disparities in depressive disorders outcomes between non-Hispanic whites and ethnic minorities. Medical Care Research & Review, 64(5), 157S-194. Vandenberg, R. J. (2006). Software Review: Mplus 3.0. Organizational Research Methods, 9(3), 408-412. Vellas, B. J., Wayne, S. J., Romero, L., Baumgartner, R. N., & Garry, P. J. (1997). Fear of falling and restriction of mobility in elderly fallers. Age & Ageing, 26, 189-193. Visovsky, C. (2006). The effects of neuromuscular alterations in elders with cancer. Seminars in Oncology Nursing, 22(1), 36-42. Walke, L. M., Gallo, W. T., Tinetti, M. E., & Fried, T. R. (2004). The burden of symptoms among community-dwelling older persons with advanced chronic disease. Archives of Internal Medicine, 164(21), 2321-2324. Walston, J., Hadley, E. C., Ferrucci, L., Guralnik, J. M., Newman, A. B., Studenski, S. A., et al. (2006). Research agenda for frailty in older adults: Toward a better understanding of physiology and etiology: Summary from the American Geriatrics Society/National Institute on Aging Research Conference on Frailty in Older Adults. Journal of the American Geriatrics Society, 54(6), 991-1001. Waltman, N., Gross, G., Lindsey, A., Ott, C., Twiss, J., & Berg, K. (2006). Prevention of falls and osteoporotic fractures in postmenopausal breast cancer survivors. Oncology Nursing Society 31st Annual Congress podium and poster abstracts. Oncology Nursing Forum, 33(2), 402. Waltman, N., Ott, C., Twiss, J., Gross, G., Lindsey, A., & Berg, K. (2007). Predicting likelihood of multiple falls in postmenopausal breast cancer survivors (BCSs) with low bone mineral density. Oncology Nursing Forum, 34(1), 181-188. Wedding, U., Rohrig, B., Klippstein, A., Brix, C., Pientka, L., & Hoffken, K. (2007). Comorbidity and functional deficits independently contribute to quality of life before chemotherapy in elderly cancer patients. Supportive Care in Cancer, 15(9), 1097-1104. Weiner, M., Fan, M., Johnson, B. A., Kasper, J. D., Anderson, G. F., & Fried, L. P. (2003). Predictors of health resource use by disabled older female Medicare beneficiaries living in the community. Journal of the American Geriatrics Society, 51(3), 371-379. Wenger, N. S., Solomon, D. H., Roth, C. P., MacLean, C. H., Saliba, D., & Kamberg, C. J. (2003). The quality of medical care provided to vulnerable community-living older patients. Annals of Internal Medicine, 139, 740-747. 191 Werley, H. H., & Lang, N. (1988). The consensually derived Nursing Minimum Data Set: Elements and definitions. In H. Werley and N. Lang (Eds), Identification of the Nursing Minimum Data Set (pp. 402-411). New York, NY: Springer. Wilson, I. B., & Cleary, P. D. (1995). Linking clinical variables with health-related quality of life. A conceptual model of patient outcomes. JAMA, 273(1), 59-65. Wolfinger, R., & O'Connell, M. (1993). Generalized linear models: A pseudo-likelihood approach. Communications in Statistics-Simulations and Computation, 22, 1079-1106. Yabroff, K. R., McNeel, T. S., Waldron, W. R., Davis, W. W., Brown, M. L., Clauser, S., et al. (2007). Health limitations and quality of life associated with cancer and other chronic diseases by phase of care. Medical Care, 45(7), 629-637. Yancik, R. (1997). Epidemiology of cancer in the elderly: Current status and projections for the future. Rays, 22(Supplement 1), 3-9. Yancik, R., Havlik, R. J., & Wesley, M. N. (1996). Cancer and comorbidity in older patients: A descriptive profile. Annals of Epidemiology, 6(5), 399-412. Yancik, R., & Ries, L. A. (2000). Aging and cancer in America. Demographic and epidemiological perspectives. Hematology/Oncology Clinics of North America, 14, 1723. Yancik, R., Wesley, M. M., & Ries, L. A. (2001). Effect of age and comorbidity in postmenopausal breast cancer patients aged 55 years and older. JAMA, 285, 885-892. Zeger, S. L., & Liang, K. Y. (1986). Longitudinal data analysis for discrete and continuous outcomes. Biometrics, 42, 121-130. Zeger, S. L., Liang, K. Y., & Albert, P. S. (1988). Models for longitudinal data: A generalized estimating equation approach. Biometrics, 44, 1049-1060. Zheng, B. (2000). Summarizing the goodness of fit on generalized linear models for longitudinal data. Statistics in Medicine, 19, 1265-1275. 192