, .
.- meajupfia

«is»?

. 5.... a . . . . . . . V . «a .. V av.
muffin»... . . , . . .. . . «$.53 A .

. 5 . . . . . . . ., 22:3,”. ..
I L ‘. . . . . V. .V 014.5%

.134.) fig"

. . figufimniuw

 

”533...; I;

.i-Isr
25.....4mm
V 5“ n‘ . I
.r. gun: . . V . . :dvfhfl
1“.-!l . . . . . . A k b
.. . «:3 i:
itdr“;
huuuuunummmmmuw‘.
hnnu. x. . ax
‘ '9! E i. .
3.”... ..t L

'1... ll
Hagan.
...,‘
h»... .
. x;

 

.{i

 

VI. $4.57. v;
a 3!: 191
t‘lt.‘\i\x.._lsi\!.
II-‘Ilr’. 05".“
t.

. 3Y4
.191! '0 .h
IIIOD‘IKIA'I‘AAI‘
u‘tvl’IKIIO‘o
3'31!!!
$11.31

. . r
6%.... :1 14:55.4 It‘ll]. ’:
{311.63-3'317349‘2

. nth-5": slug still}, II.
’50 {$53.

.351: 19%:

 

 

In... .15.. . .

. .3 aye... . . . , . . . . . V
:m,Vw..§ .5, ,. . . .V . . f.
3.3:..- , . , , . . . .. m....A..p...u......$Vw

. . . V, V . .. . V Aunt. .v.1-1..-=.O.I.Ifi
w. . . . .

fig:

 

 

 

 

 

3‘ LIBRARY
10% 0 Michigan State
University
This is to certify that the
dissertation entitled

AN ECOLOGICAL BI-FOCAL MODEL FOR ELDER
PHYSICAL ABUSE BY ADULT CHILD: A STRUCTURAL
EQUATION MODELING OF RISK FACTORS
PREDICTING ELDER ABUSE IN THE UNITED STATES

presented by
LEVENTE VON HEYDRICH

has been accepted towards fulfillment
of the requlremente for the

Doctora degree in Family and Child col
Major Professor’s Signature \

R(u103

Date

.O-I-I-n-l-l-o-o—.-—..‘.-.-.-.-.-I--.-l-I-.-’-.-.-n-.— . -

MSU is on AMImetIve Action/Equal Opportunity Employer

 

 

 

 

PLACE IN RETURN BOX to remove this checkout from your record.
TO AVOID FINES return on or before date due.
MAY BE RECALLED with earlier due date if requested.

 

DATE DUE DATE DUE DATE DUE

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

5/08 K:IProj/Acc&Pres/ClRC/DateDuo.indd

 

AN ECOLOGICAL BI-FOCAL MODEL FOR ELDER PHYSICAL ABUSE BY
ADULT CHILD: A STRUCTURAL EQUATION MODELING OF RISK FACTORS
PREDICTING ELDER ABUSE IN THE UNITED STATES
By

Levente von Heydrich

A DISSERTATION
Submitted to
Michigan State University
In partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Family and Child Ecology

2009

ABSTRACT
AN ECOLOGICAL BI-FOCAL MODEL FOR ELDER PHYSICALABUSE BY

ADULT CHILD: A STRUCTURAL EQUATION MODELING OF RISK FACTORS
PREDICTING ELDER ABUSE ON THE UNITED STATES

By
Levente von Heydrich

The age structure of the US. population has been undergoing an unprecedented
shift toward a growing proportion of people aged 65 and older during the last two
decades. According to estimates projected by the US. Census Bureau the 65 and older
age group is expected to almost double between now and 2030, growing from 4.4 million
to 8.9 million. Simultaneously each year in the United States, 1 to 5 million adults older
than 65 are physically or sexually injured, exploited, or mistreated by their caregivers
(mainly adult child caregivers). Although elder abuse has reached epidemic proportions
in our society, there is a paucity of research in the field of marriage and family therapy on
theoretical frameworks, ecological risk factors associated with violence perpetrated on
the elderly and suggestions for successful therapy interventions. This study examines the
prevalence and risk factors involved in elder abuse, specifically physical abuse, in a
Random Digit Dial (RDD) sample of non-institutionalized and English speaking older
adult population in the United States. Using a subsample of 203 elderly participants from
Midlife Development in the United States (MIDUS H, 2004-2006), a study sponsored by
the MacArthur Foundation that examined age related differences in physical and mental
health due to individual and social factors (Ryff et al., 2006), relationships amongst
several individual, relational, and social context dimensions as they relate to elder abuse

were examined. Latent variable modeling was used to examine the causal pathways and

associations between individual elderly demographic characteristics, physical! emotional
health, and behavioral and contextual characteristics. The analysis of data patterns,
univariate calculations of data dimensions, and model syntaxes were written with the aid
of Mplus, PAXW, and SYSTAT statistical software packages.

The usefulness of this dissertation model is two pronged. First, the proposed
model will empirically examine the assumptions, tenets, and conclusions of the
Ecological Bi-Focal Model for Elder Abuse fiamework then discuss possible MFT
clinical implications to better understand the therapeutic needs of older adults and their
families.

Implications for the field of marriage and family therapy and future research directions

are discussed.

Key Words: Elder Physical Abuse, Ecological Risk Factors in Elder Physical Abuse,

LISREL Data Modeling, Social Relations

C0pyright by
LEVENTE VON HEYDRICH

2009

ACKNOWLEDGMENTS

The completion of this dissertation manuscript was the result of bi-directional and
synergistic interpersonal academic connections and the tireless help and encouragements
of several influential academic mentors / professors. I would like to express my deepest
gratitude to every wonderful professor who forgave my shortcomings and instead
encouraged, stood by me, and build me up during my most difficult trials.

To Dr. Lawrence Schiamberg, who graciously and humanly lifted me up when I
was down and scattered and through his kindness and belief in my abilities reenergized
and refocused my determination to complete the PhD process. During the last two years
you have spent countless hours (on a weekly basis) in helping, guiding, and mentoring
me in the field of gerontology and research. Your patience, kindness, empathy, and
hmnanity are well beyond anything I have experienced or even seen in this world. You
are more than a mentor and fiiend to me Dr. Schiamberg— I consider you a member of my
family. Thank you.

To Dr. Ruben Parra, Cardona, your class provided me with my first positive
research experience and started my interest and desire to learn even more about it. YoUr
kind, encouraging, and building up approach towards me greatly contributed to my
academic and personal growth. Thank you.

To Dr. Alexander von Eye, who opened my eyes to great, new, and wonderful
possibilities in the quantitative research field. Your class was the last and the greatest
impetus to change my research orientation from qualitative to quantitative research and to

my personal growth as a researcher. I loved the analogy of initial driver license in your

SEM class. It indeed opened my horizons and provided a feverish motivation to further
advance my knowledge in the field of quantitative research and measurement. Your
gracious approach to mentoring me in structural equations modeling and willingness to
patiently wait while I was attempting to grasp the deeper meanings of this statistical
approach meant the world for me. Thank you A

To Dr. Tenko Raykov, thank you for teaching and guiding me to a deeper
understanding of statistics in the domains of psychometrics, structural equation modeling
with ordinal and discrete data, and the Mplus statistical program. You graciously and
patiently helped me with the assessment of various statistical designs and models that
might fit my data, thus providing me with essential statistical scaffolding. I really
appreciate the fact that instead of giving me the final statistical answer (that might have
been easier on you) you allowed me to grow and come up with all the specific (and -
general) statistical solutions that my study needed. Thank you.

To Trang Heydrich, Demeny Istvan, and my two beautiful angels Eva and Noemi,

thank you. Without your support this dissertation would not have been possible.

TABLE OF CONTENTS

List of Tables ................................................................................ x
List of Figures ............................................................................... xi
CHAPTER I: BACKGROUND AND OVERVIEW .................................... 1
Introduction .......................................................................... 1
Statement of the Problem .......................................................... 6
Theoretical Foundations ............................................................ 9
Choice Theory ............................................................ l 1
Subjective Expected Utility Theory .................................... 13
Symbolic Interactionism ................................................ 14
Social learning Theory .................................................. 17
Interpersonal Theory ..................................................... 19
Effortful Control Theory ................................................ 21
Ecological Bi-Focal Model for Elder Abuse Theory ................ 23
Theoretical-Conceptual Study Map ................................... 26
Conclusions ............................................................... 28
Purpose of the Study .............................................................. 30
Significance of the Study ......................................................... 32
CHAPTER 11: REVIEW OF RESEARCH ............................................. 34
Prevalence/ Incidence of Elder Abuse .................. i ........................ 35
Issues Relating to Definition of Elder Abuse ........................ 35
Impact of Elder Abuse on the Victim and Physical Signs
of Abuse ................................................................... 37
Reporting and Detection of Elder Abuse .............................. 38
Risk Factors-of Elder Abuse ..................................................... 39
Elder Abuse Risk Factors at Microsystem Context: Adult Child/ Older
Adult Relationship ................................................................ 4O
Characteristics of the Adult Child that Enhance Aggression
Towards the Elder Parent ................................................ 43
Characteristics of the Elderly Victim .................................. 46
Contexts beyond the Immediate Elderly Parent / Adult Child Caregiver
Relationship ........................................................................ 53
Macrosystem Level Factors that Affect Elder Abuse ........................ 54
Study Research Questions ........................................................ 55
CHAPTER HI: METHODOLOGY ...................................................... 58
MIDUS I and MIDUS 11 Data Sets ..... . ........................................ 59
Measures .................................................................. 59
Instrument Development and Pretests ........................................ 59
vii

 

, .
' C
_.
§
3
;l
6‘.)
:5
"
.‘l.
9""
;‘_.,'
. I ,
l r
s .
We
I:
zit.
‘5
Ci
1::
n!
on!
.‘Tu‘.
..
5*
‘5"
£3-
...
H2
{$3
.4?
“E
iw
1%.
\v
.a
I

Data Collection Protocol ....... . ......................................... 6O

 

Interviewer Training ..... 61
Research Filed Procedures .............................................. 62
Survey Protocol ........................................................... 63
Sampling Procedures .................................................... 64
Proposed Study Sample ................... q .............................. 68
Human Subjects Protection ............................................. 71
Data Safety Monitoring Plan ........................................... 71
Measures ........................................................................... 71
Dependent Latent Variables and their Indicators ..................... 72
Independent Latent Variables and their Indicators .................. 74
Proposed Path Model ............................................................. 83
Specific Research Questions, Hypotheses, and Measures ................... 88
Data Analysis Plan ................................................................ 89
CHAPTER IV: RESULTS ................................................................ 98
Descriptive Statistics .............................................................. 98
Elderly Parent Descriptive Statistics ................................... 99
Adult Child Descriptive Statistics .................................... 104
Results of Missing Data Modeling ............................................ 109
Mathematical Equations Used to Compute Estimates ...................... 110
LISREL Data Modeling Results ............................................... 111
The Full Model ......................................................... 114
Measurement Dimensions of the Full Model ........................ 115
The Restricted (Mediational) Model ................................. 120
Measurement Dimensions of the Restricted Model ................ 121
Summary Results for Both Models ................................... 127
Hypotheses Testing. . .. ......................................................... 128
Hypothesis #1 ............................................................ 130 {;
Hypothesis #2 ............................................................ 131
Hypothesis #3 ............................................................ 131 l-
Hypothesis #4 ............................................................ 133 j,‘
Hypothesis #5 ............................................................ 134 :g
Power Assessment ............................................................... 134 2 -;‘
Additional Model Diagnostics ................................................. 136 it:
CHAPTER V: DISCUSSION ........................................................... 139
Sample Based Description of Elder Physical Abuse ........................ 139
Risk Dimensions of Elder Physical Abuse at Microsystem Level:
Immediate Older Adult/ Caregiver Context .................................. 140 §
Context beyond Focal Older Adult/ Adult Child Caregiver: ,
Ecological Model Discussion ................................................. 153 .3
viii Si

Theoretical, Research, and Practical Implications ........................... 159

Strengths, Limitations, and Directions for Future Research ............... 166
Concluding Thoughts ............................................ ' ................ 171
APPENDICES ............ . .............................................................. 173
Appendix A: MIDUS 11 Survey Protocol .................................... 173
Appendix B: MIDUS Sample ................................................. 175
REFERENCES ........................................................................... 17 7
ix

 

         

{Rf'finfinghtflflrWig-i{gill.mm‘ ‘ “f

M:

II.‘ '
I’ 1“- ' a. .

LIST OF TABLES

Table 3.1 Sample Sizes, Longitudinal Retention Rates, and Response Rates

for MIDUS II ..................................................................... 65
Table 3.2 SAQCompletion Rates for the MIDUS 11 Study ......................... 66
Table 3.3 Participation Rates. for the MIDUS II Cognitive Battery ................ 67
Table 3.4 Dissertation Sample Characteristics ........................................ 68
Table 4.1 Sample Characteristics ....................................................... 102
Table 4.2 Adult Child Characteristics .................................................. 105
Table 4.3 Univariate Properties of Manifest Dependent Variables ................ 109
Table 4.4 Six Factor Correlation Matrix ............................................... 112
Table 4.5 Full Model Factor Loadings ................................................ 116
Table 4.6 Restricted Model Factor Loadings .......................................... 122

Table 4.7 Structural Relationships amongst Latent Variables in the
Full Model ................................................... _ ............ 128

Table 4.8 Structural Relationships amongst Latent Variables in the
Restricted Model ........................................................ 129

LIST OF FIGURES

Figure 1.1 Circumplex Structure .......................................................... 21
Figure 1.2 Theoretical-Conceptual Map ................................................ 27
Figure 3.1 Full Model Path ................................................................. 84
Figure 3.1 Restricted Model Path ......................................................... 87
Figure 4.1 Gender Univariate Results ................................................... 99
Figure 4.2 Age Univariate Results ...................................................... 100
Figure 4.3 Physical, Psychological Health Sliced by Age ........................... 102
Figure 4.4 Elder Physical Abuse Univariate Results ................................. 107
Figure 4.5 Elder Sexual Abuse Univariate Results ................................... 108
Figure 4.6 P-P Plot ofStandardized Residuals........................... ............ 137
Figure 4.6. Structural Relationships in the Full Model .............................. 129
Figure 4.7. Structural Relationships in the Restricted Model ....................... 131

xi

 

CHAPTER I: BACKGROUND AND OVERVIEW
Introduction
The gravity of the problem of elder abuse can be understood, in part, as a
consequence of a drastically changing age structure of the US. population with growing
proportions of people aged 65 and older. The number of Americans aged 65 and older
grew fi'om 3.1 million in 1900 to 35.3 million in 2001 . This corresponded to a growth
fiom 4% of the population in 1990 to 12.4% in 2001. With baby boomers aging, the
number is expected to rise to 70.3 million by 2030, about 20% of the total projected US.
population (United States Adminishation On Aging, 2002). This trend was confirmed by
2007 census findings as well, which noted the two fastest-growing five-year age groups
in the United States since the last census as 50-54 year olds (55%), and 90—94 year olds
(45%) (U .8. Census Bureau, 2007). Furthermore, the age group of 75 and over increased
by more than 20%, constituting 20.6 million women and 14.4 million men age 65 and
older. The number of Americans aged 65 and older grew from 3.1 million in 1900 to
35.3 million in 2001. This corresponded to a growth from 4% of the population in 1990
to 12.4% in 2001. With baby boomers aging, the nmnber is expected to rise to 70.3
million by 2030, about 20% of the total projected US. population (United States
Administration On Aging, 2002). A similar trend was observed in the number of elder
abuse cases that have been steadily increasing over the last 20 years, according to
estimates by the National Center on Elder Abuse. Each year in the United States, 1 to 5
million adults older than 65 are physically or sexually injtued, exploited, or mistreated by
their caregivers (National Center on Elder Abuse, 2004). Furthermore, government

officials reported that cases of elder abuse increased nearly 20% fiorn 2000 and that 42.8

% were 80 years or older. Of the vast majority of elder abuse reports, 89.3% occurred
within community settings (family homes), and the most common relationship of victims
to alleged perpetrators was parent/child and other family members (National Center on
Elder Abuse, 2004). Pillemer and Finkelhor (1988) estimated that the rate of abuse for
persons aged 65 and over who were living with only a spouse was 41 per 1,000 couples
and that the perpetrators of this abuse were most likely to be spouses (58%) rather than
adult children (24%). Fmthermore, the authors reported that 32 out of every 1,000 elder
adults reported that they had experienced some form of maltreatment at least once since
reaching age 65.

As with most crimes, underreporting is the norm. Data on elder abuse in
commrmity settings suggest that only 1 in 14 incidents come to the attention of
authorities, and even fewer cases are resolved by prosecution (Pillemer & F inkelhor,
1988). More current studies on elder abuse estimate that the frequency and prevalence of
elder abuse range firm 2% to 10%, based on various sampling and survey methods and
case definitions (Lachs, Mark, & Pillemer, 2004). The high elder abuse incidence rate
and the continuous increase in allegations of elder abuse and neglect have contributed to
mormting scrutiny from politicians, non-profit commrmity—based organizations,
governmental oversight agencies, law enforcement officials, and the research/academia
community.

The most distmbing fact relating to elder abuse is that victims of abuse, neglect,
and exploitation are three times more likely to die at an earlier age than elders who were
not victims of abuse (Lachs et al., 2004). While the identification, apprehension, and

prosecution of perpetrators of elder abuse have generated considerable interest within our

society, these domains have lagged in relations to social needs. Although researchers
have sought to grasp the full scope and causes of elder maltreatment, studies are limited
both in quantity and scope, and the gaps in the literature are significant.

While there are a growing number of incidence and prevalence studies on elder
abuse in the field of gerontology, with few exceptions researchers based their results and
conclusions on small non-representative sample based data thus lacking sufficient
statistieal power to detect meaningful effects in the population therefore their findings
were non generalizable (Acierno, 2003). Seconding Acierno’s (2003) conclusions, the
National Research Council (NRC) on Elder Mistreatment (2003) as well documented that
researchers significantly lack lmowledge on elder abuse prevalence in the population.
Furthermore, the NRC representatives noted that most of these prevalence studies suffer
from methodological weaknesses in addition to using small convenience samples
therefore there is a consistmt lack of congruence between findings on elder abuse
prevalence and various risk factors influence on elder mistreatment. For example,
Pillemer and Finkelhor conducted a random sample survey among people aged 65+ and
reported a prevalence of 1.4% of verbal abuse and 0.4% of neglect occurring 210 times,
and 0.5% of any physical abuse over a year (Pillemer & Finkelhor, 1988). A few years
later, Paveza et al. (1992) concluded that the prevalence of elder physical abuse was at
least 5% amongst people aged 40+ with Alzheimer’s disease diagnosis. The incongruence
in prevalence numbers continued with Lachs et al. (1997), who found that elder physical
abuse incidence rates were closer to 1.6% in the population. Almost a decade later, in a
cross cultural study Cooper et al. (2006) indicated a physical abuse rate of 4.6% in a

probability sample of elderly patient (older than 65 years) receiving health and or

community services in 11 countries adult day health care programs. In addition to.
sampling and methodological weaknesses these studies differed in several other
dimensions as well. For instance, a variety of instruments were employed to measure the
construct of elder abuse with unknown or untested psychometric properties. Furthermore,
most of these studies did not go beyond a univariate statistical analysis of the data thus
they mainly reported frequencies and percentages found in the sample.

Therefore this study will address the gaps identified by the NRC Panel to Review
Risk and Prevalence of Elder Physical Abuse (2003) for studying elder abuse, including
physical abuse, by: (a) using a representative national random sample, (b) using a
standard definition of a specific type of elder abuse (6. g. physical abuse), (c) clearly
identified abuse measures based on these definitions, and (d) nploying effective
statistical models based upon sound mathematical principles that can yield generalizable
results. Inaddition, an ecological approach to elder abuse will be used principally
because the NRC (2003) review of elder abuse research suggests that models which
reflect the multiple causes of elder abuse ought to be utilized. One major theoretical
model identified by the NRC (2003) is the Ecological Bi-Focal Model (Schiamberg &
Gans, 2000), which indicated that elder abuse is not caused by any single agent or factor;
instead, it is the result of the interplay of multiple factors both within the immediate older
adult/ adult child caregiver context and beyond that context (Lamnan et al., 2002).

The paucity of research is even more striking in the field of marriage and family
thaapy. There is a scarcity of theoretical fiameworks and research literature on elder
abuse and geriatric mental/ emotional health and lifestyle correlates in the field of

marriage and family therapy (Ivey et al., 2004; Parra-Cardona et. al, 2007). To a

considerable degree the geriatric population has largely been neglected, specifically on
elder marriage and family therapy models and non-spousal domestic violence. Several
reasons have been provided for the paucity of research in the field of marriage and family
therapy including societal ageism (eg systematic stereotyping and discrimination) and
elders being uncomfortable when interacting with health care professionals including
family and marriage therapists (Ivey et al, 2004). Consequently specific ageist belief
systems such as that the elderly are rigid thus unlikely to change their family interactions
patterns, are mentally deteriorated, incompetent, or lack interest in most social interaction
limit the effectiveness of family and marriage therapist when interacting with elderly
persons or couples (Ivey et. al, 2004). The absence of family and marriage therapists with
specialized training in gerontology further elevates the level of misunderstandings when
responding to the needs of elderly couples and individuals. This author contends that
ageism and related biases exist in our society and this might inadvertently guide
contemporary couple and family therapy practice. Thus there is a need for a useful
theoretical framework and empirically validated clinical intervention models to guide

professionals in the field of marriage and family therapy.

Statement of the Problem

Aggression and physical violence directed toward the elderly have substantial
consequences for the health and well-being of the individual, and older adults exposed to
violence exhibit significant emotional and psychological distress (Comijs et al., 1999)
and face increased mortality rates compared to non-abused elderly (Lachs et al., 1998).
Although elder abuse is a major societal problem in the United States, few studies cover
in detail the various risk dimensions, their interplay, individuals involved, and cumulative
impact on the abused adult’s well-being. For instance, most elder abuse studies focus
either on the victim’s characteristics (e. g., cognitive factors such as Alzheimer’s,
Parkinsons, and dementia) as they correlate with abuse or on the abuser’s characteristics
(e.g., adult children of aging parents), as they increase the risk of violence directed
against older people (Cooney & Howard, 1995). The term elder has generally been used
to refer to people who are aged 65 or older. The Department of Health and Human
Services Administration on Aging seems to define older Americans as those 65 years and
over in some documents (Snapshot, 2003). Additionally, empirical research studies
examining the risk factors of elder abuse fi'om a bi-directional perspective (focusing
simultaneously on the adult caregiver and the aging elderly as a dyad) are virtually non-
existent. For the purposes of this study the term bi-focal model will be defined as a
simultaneous focus on both the adult child caregiver and the aging parent as a familial
dyad within a broader ecological context when examining risk factors of elder abuse.
The one and only theoretical model that adopted such an encompassing ecological
perspective was developed by Schiamberg and Gans that examined risk factor

dimensions at various systemic levels present dming the process of elder abuse in the

community and perpetrated by adult children (Schiamberg & Gans, 2000). Although this
Ecological Bi-Focal model of elder abuse has shown a great deal of promise as the
vanguard theoretical model of elder abuse, an empirically driven analysis of the model
has not been completed up to this date. Since the multidimensional approach to studying
violence against the elderly, as espoused by Schiamberg and Gans (2000) in the Bi-Focal
theoretical model, best explains risk factors involved in elder abuse; this author behaves
that we urgently need to undertake the challenge of empirically analyzing the tenants and
conclusions of this theoretical model. Moreover, if the assumptions and conclusions
reached by the authors of the Applied Ecological Bi-Focal model are validated via
empirical! statistical findings then their multi-level approach to preventing elder abuse
should immediately be employed at various societal levels to minimize violence
perpetrated on our most vulnerable population. In sum, given the enormous societal costs
(medical, legal, health, and etc.) that elder abuse exacts from all of us, there is an
immediate need for empirically validated and effective intervention strategies based on
sound research that could minimize if not altogether eliminate the health epidemic of
elder abuse.

Despite incongruent research findings and methodological difficulties mentioned
above, researchers reached consensus on a number of elder abuse types, definitions, and
related terms dming the past decades (Wilber & McNeilly, 2001). The usual types of
abuse includedwitbin most definitions are as follows:

1. Physical Abuse: Physical abuse is defined as the use of physical force that may

result in bodily injury, physical pain, or impairment (Wilber & McNeilly, 2001).

Examples of physical abuse are: a deliberate or inadvertert hitting, beating,
pushing, kicking, pinching, burning, and biting.

Sexual Abuse: Any non-consensual sexual contact with an elderly person such as
inappropriate touching, groping, forcing the individual to look at pornography,
promoting and facilitating unwanted sexual contact with a third party, or any
other unwelcomed sexual behavior is considered sexual abuse. Moreover, this
kind of abuse may also include acts such as sexual exhibition or coerced nudity.
. Emotional or Psychological Abuse: Behavior directed towards an older adult that
causes them to experience fear, mental anguish, or emotional pain or distress is
considered abusive (W ilber & McNeilly, 2001). Examples of this kind of abuse
includes name-calling, intimidation, insults, and threats; treating the older adult
like a child; and isolating the older adult fi'om family, friends, and social contact
by force, threats, or manipulation.

. Financial or material exploitation: Financial exploitation involves the illegal use
or improper use of an elder’s funds, property, or assets (W ilber & McNeilly,
2001). Examples of financial exploitations are actions such as denying older
adults access to their own fimds or home, taking their money under false
pretenses, forced property transfers, or even purchasing expensive items without
permission.

Neglect: Neglect can be defined as a refusal or failure to fulfill any part of care
giving obligations or duties to an older person under care (W ilber & McNeilly,
2001). Neglect can range from intentionally failing to meet the older adults’

physical, emotional, or social needs to withholding appropriate attention from the

older adult. It can include behaviors such as not food, water, clothing, medication,
or assistance with activities of daily living or personal hygiene (Wolf and Pillemer

.1989).

Theoretical Foundations

Over the last two decades numerous research studies were conducted on intimate
partner violence, psychological/ emotional impact of violence on women and children,
and profiling “the batterer” via the examination of multiple personalities and typologies
of batterers in the field of marriage and family therapy. Based upon sound research
findings, valid and reliable results, marriage and family therapy researchers identified
several major risk factors as well as protective factors present during the process of
physical abuse of intimate partners and elderly women. In addition, several theoretical
models explaining intimate partner violence were developed and consequently successful
couple and family therapy interventions were larmched. However, despite of these
promising successes in the field of intimate partner’s violence, MFI‘ researchers failed to
expand their study direction and include violence between other family members as well.
Consequently neither MFT theoretical frameworks nor intervention models were
developed to aid elderly patients abused by their own adult children. This conclusion was
reached after an intensive search of the PsychINFO, ProQuest, J STOR, and Wiley
InterScience Journals which resulted in fewer than five published MFT related research
studies. Out of these five studies, only Parra-Cardona et. al (2007) focused on building
an MFT elder abuse theoretical framework by describing precursors and dynamics

associated with elder abuse and neglect in Latino families. The authors used a qualitative

case study (one family) method to examine the tenets, assrmrptions, and conclusions of
the Ecological Bi-Focal Model for Elder Abuse framework as developed by Schiamberg
and Gans (2000). Given the inherent limitations (results cannot be replicated or
generalized) of a (case study) qualitative research approach, findings and their
applications are limited in scope.

Since this research study is only a first step towards an uncharted territory (elder
abuse in MFT) the approach taken here will have to combine innovations from the fields
of psychology, human development, gerontology, sociology, and family and marriage
therapy. The overarching theoretical foundations of this strrdy include both the
consideration of context of interaction and specific outcome measures and dimensions
that are derived from multiple social science theories. The principal focal theory utilized
in this study is the Ecological Bi-Focal Model for Elder Abuse (Schiamberg & Gans,
2000) and the human ecological perspective in general (Bronfenbrenner, 1979, 1986,
1997). Whereas the assumptions, tenets, and conclusions of the Bi—Focal model are
captured in this study with several abstract latart constructs at various systemic levels
(nricro, exo , and mesosystems), several specific indicator variables that directly measure
the model’s latent variables have their roots in other theoretical fiameworks such as: (a)
the Choice theory, (b) Subjective Expected Utility theory (c) Symbolic Interactionism,
(d) Social Learning theory, (e) Interpersonal theory, (f) the Effortful Control theory, and
(g) the Life Course Perspective. Since most theoretical models explaining human
behavior incorporate partly or to large extent previous theoretical fiameworks, the
development of a substantive theory for elder abuse requires an understanding of other

models as they relate to the behavior of elder abuse. In the following pages, various

lO

theoretical models and their contributions to understanding both the context and the

phenomenon of elder abuse will be illustrated.

Choice Theory

Psychiatrist William Glasser's (1998) suggested that human behavior is composed
of four simultaneous components: emotions, ideas, actions, and physiological states.
Furthermore, he espoused that humans choose the idea (e. g. helshe mistreated me) and
action (e. g. physically attacking an older adult) and that the associated emotions and
physiological states occur simultaneously and automatically; thus emotions do not arise
or ochn independently and without specific contexts. Glasser believed that previous
definitions of human behaviors were too narrow and exclusive so he coined the term
“total behavior” to distinguish his construct, a continuous loop of ideas, emotions, and
actions, M the general concept of human behavior. To accommodate his enhanced
construct of human behavior, Glasser used verbs to describe human emotional states. For
instance, he used the term “to depress” when describing the total behavior commonly
known as depression, which, to him, included depressing ideas, actions, notions, and
physiological states (Glasser, 1998). Consequently, the construct of elder abuse includes
not only the physical attack (the last step in a process) on the elderly person but also the
abuser’s mental picture and emotions aroused by those ideas in regards to the older adult.
Moreover, Glasser reasoned that internal choices (whether conscious or unconscious), as
opposed to external stimuli, cause the negative emotions in humans that will lead to the
final step of physical aggression against the older adult. To explain this, he suggested that

the individual’s cognitive appraisal of the environment (e. g. he/ she demeans me again)

11

filtered through the lens of past experiences was the motor behind her/ his emotional
processes which in turn would have a strong impact upon the caregiver’s subsequent
behavior (e. g. to physically assault the elderly or not). Glasser vehemently rejected the
idea that the external stimuli coupled with emotional knee jerk reaction were responsible
for the individual’s response behavior (Glasser, 1998). Consequently adult child
caregivers have a choice in whether or not physieally assault their elder parent and are
not mere drones that automatically react to certain stimuli. As such the adult child
caregiver first uses an inner soliloquy, part of which is reading the parent’s attitude,
which will lead to a decision based on past experiences and coupled with a certain
negative emotional state as a precursor to physical violence. In sum, a physical assault
on an older adult is a conscious and premeditated behavior option selected by an adult
child caregiver whether or not his/ her course of action makes sense to us.

Although Glasser’s Choice theory provides a well reasoned and detailed
explanation of a complex behavior such as elder abuse within the immediate individual
and family context, it does not acknowledge the impact of contexts beyond the dyads
control such as social isolation or societal attitudes towards elder abuse. In addition, this
framework minimizes the importance of external stimuli such as the victim’s behavioral
problems (e. g verbal or even physical assault on the adult child prior to the start of
confrontation) on the adult child’s choice to physically attack her/ his elderly parent.
Several empirical studies on elder abuse suggest a clear correlation between such an
external stimuli and the subsequent physical assault on an elderly person, (Shinoda-

Tagawa, et al. 2004; Schiamberg et al. 2009). In spite of these shortcomings Glasser has

12

Q

substantially contributed to the understanding of human behavior such as elder physical

abuse and its internal correlates.

Subjective Expected Utility (SE U) Theory

Although Glasser’s theory provides insight into the importance of cognitive and
psychological processes in explaining elder abuse, it has significant shortcomings. The
SEU theory filled these gaps. This theoretical framework is widely believed to be
applicable to all human decision making, whether or not is related to events or
experiences and regardless of time scale, the level of specificity under considerations, and
forms of cognitive processing involved (Mongin, 1997). The success of SEU model in
describing hummr cognition has to do with the model’s axiomatic foundations,
mathematical structure, and its balance between accruacy and parsimony. The SEU
theory assumes a mathematical and statistical (Bayesian roots) thus logical view of
human cognitive processes, including choice making.

The SEU fi'amework was intended both to explain and to predict human behaviors
by means of a small numba of concepts, e. g. intentions, attitudes, and beliefs.
Accordingly the caregiver (adult child) is assumed to have a consistent preference in
ordering of all possrble states of world and a prior distribution of exogenous events, (e. g.
past child abuse experience, current negative relationship, financial stress, etc.), (Simon,
1980). Thus the individual’s exogenous probability distribution coupled with action
strategy selected determines the probability of the expected outcome occurring (in our

case elder abuse). It follows that the decision maker (adult child) chooses between risky

13

L“ #
a

or uncertain prospects by comparing their expected utility values (personal benefits) and
the probability of negative or positive outcome.

To support this point further, an example is useful. Suppose an adult child
initiates a random ventrne with two possible outcomes, state “y” (verbal criticism) and
state “2” (physical aggression), where the first outcome is less regrettable than the
second. The outside observer has no idea of what the two states are composed of in the
cognitive world of the adult child. However, if the child selects verbal criticism of the
elderly parent rather than physically attacking her/ him, we can deduce that the caregiver
assigns a greater probability to state y relative to state 2. The fact that verbal criticism is
more desirable than physically attacking her/ his parent implies that this particular
behavior would be more consistent with her/ his beliefs. In addition to individual beliefs,
societal values, norms, and laws (prior distributional properties) will also impact the adult
child’s intentions and subsequent act or lack of it. Therefore, in essence individual
beliefs, attitudes, and intention precede the act of physical assault and their distributional

properties will either increase or decrease the probability of elder abuse happening.

Symbolic Interactionism

Whereas Glasser’s Choice Theory attributes behavior such as elder abuse to the
individual’s internal processes, the advocates of symbolic interactionsim suggest that
human behavior is best understood in relation to their environment and are dependent on
symbolic interactions for their existence (Dewey, 1922). Proponents of this theoretical
fiamework emphasize the importance of daily human interactions, relationships, and

associated experiences as they shape human behavior rather than the structures associated

l4

with individual psychological processes (micro system) or large scale thus rigid and
relatively fixed social forces and laws (meso, macro, and chrono systems). The principal
pillar of symbolic interactionism is that human life exists in a symbolic dimension.
Symbols provide the context by which reality is constructed and are based on shared
meanings that are created and maintained via social interaction. Thus an individual’s
reality is a social product and originates from a multitude of symbolic interactions.
Consequently the individual’s behavior (e.g. aggression) is a product of society, or more
specifically, social interaction rather than known or unknown internal individual
processes as advocated by Glasser. Symbolic Interactionism is based on three primary
premises: (a) human beings act towards things on the basis of the meanings those things
have for them, (b) these meanings arise out of the interaction of the individual with
others, and (c) that an interpretive process is used by the individual in each instance in
which she/ he must deal with things or people in be! his environment (Blurner, 1962). In
addition Blurner (1962) suggested that although people's selves are social products, these
selves are also purposive and creative and not a mere reflection of the society and
environment. As such the individual’s self emerges fiom the social interaction between
people in which the person takes on the role of the ”other” (objectification) and
internalizes the attitudes perceived in both real and imagined others (Blumer, 1962).
Hence the interaction between the individual’s self-concept (the "1") and the generalized
other, the perceived view that the others have of the individual (the “me”), are vital to
understanding people behavior (Morrione, 2004). Accordingly, the actions of the others

are instrumental in the formation of the individual’s beliefs in regard to any specific

15

object. Based on the symbolic interactionisms perspective the process of elder abuse
could be visualized as:

l. The perpetrator; “you stress me out too muc ”, “you never stop complaining
and demanding” or simply, “you physically attacked me” therefore “you are
evil” so I am justified in “setting you straight”.

2. The victim; before taking provocative actions, the older person might reason as
‘you don’t tell me what to do,” “you can’t deny me taking as much prescription
meds as I wan ”, or “you are mean and abusive” therefore I can initiate
negative actions against you such as verbal abuse or even physical assault.

3. Although the abuser and victim rationalization processes originate fiorn
different sources and were differently constructed, they converge in the
initiation and consummation of the behavior of elder physical abuse.

For the proponents of symbolic interactionism perspective the social world is a
dynamic and dialectical web of situations and tmstable outcomes and individual
narratives are always in the process of shifting and transforming thus never fixed and
rigid. From this point of view there is no such thing as a solitary individual thus her/ his
internal processes are based on societal templates thus always understandable within the
“right” context (society). Therefore if the template fits than the adult child caregiver will
assault her/ his elder parent. The unfortunate shortcoming of this theoretical perspective
is that it removes or rationalizes the individuals’ choice during the process of elder abuse

and instead blames the society for the outcome.

16

Social Learning Theory

Social Learning Theory links Glasser’s Choice Theory and the Symbolic
Interactionsim framework by providing a learning- contextual framework for
rmderstanding the phenomenon of elder abuse. This fi'amework is based on the
assumption that most learning occurs in a social context thus people learn fiom one
another via observation,.irnitation, and modeling (Orrnrod, 1999). For instance, if an
adult child was subjected to child abuse during an early developmental stage she/ he
could have learned that violence is an acceptable avenue to settle disputes. Major tenets
of this theoretical framework as suggested by Bandura (1973) include:

1. People learn by watching displayed behaviors (most notably the significant
others’) and observe and weight both the rewards and punishments gained as a
direct result of that specific behavior.

2. Behavior does not necessarily have to change as a result of observing other’s
behavior, thus learning can happen without changes in behavior. Therefore
learning may or may not cause a change in one’s behavior.

3. Expectations of future punishments or reinforcements have a significant impact
on people’s displayed behavior. Thus the individual’s cognition affects her/ his
subsequent displayed behavior (Bandura, 1973).

Thus a frmdamental precept of this theoretical framework is that behavior
(aggressive) is acquired through learning, including observational learning, symbolic
modeling and direct experience (Athens, 1980, 1992). Athens’ theoretical model of
violent socialization, which includes but is not limited to child abuse or children’s

observational learning in families, describes how early developmental processes

17

influence individuals to resort to violence as a preferred method of handling disputes, or
“getting one’s way.” For example, being exposed to chronic and violent episodes of
child abuse early on, people learn to circumvent most non-violent social methods of
dispute resolution and resort to violent conflict resolutions. Athens attributes violence to
conscious, volitional, goal-directed activity (Athens, 1980; 1992). Athens's concluded
that there is empirical evidence to suggest that decisions to choose violence for settling
disputes almost always involve social conditioning to favor violent resolutions. He called
this process “violentization” and defined it as a process whereby individuals are injured
and trained to commit violence (Athens 1980; 1992).

Athens also suggested that victims often initiate the violent encounter (knowingly
or not) in which they subsequently get injured (Athens, 1992). For example, based on
analysis of clinical samples of Alzheimer’s patients Paveza et al. (1992) found that
symptoms of Alzheimer’s illness such as provocative physical and verbal behaviors often
lead to elder physical abuse. Such aggression elicitors during an elder abuse encounter
may include: (a) aversive treatment (e.g., physical, emotional, or financial withdrawal),
(b) anticipation of negative consequences in one person or, more often, in both the
aggressor and victim (e.g., fear would activate the fight or flight response in the
arnygdale), and (c) factors unique to the subjective perspective of either the aggressor or
the victim (e.g., tmfounded anger or negative thoughts attributed to the other)
(Williamson et al., 2005). In sum, aggression acts such as elder abuse is seen as a learned
behavior, and certain stimuli (in both the victim and perpetrator as well) may elicit or

inhibit the act of aggression.

18

Interpersonal Theory

Whereas Athens (1980; 1992) mainly focused on the internal cognitive processes
in an individual prior to a violent encormter, proponents of the Interpersonal fi‘amework
argue that emotional processes and personalities are as important as cognitions in
understanding elder abuse. The main assumption of the Interpersonal framework is that
everything individuals do in interaction with one another reflects an effort to achieve and
maintain self—esteem and to avoid anxiety (Leary, 1957). Hence, aggression and violence
could be conceptualized as a sum of interactions of interdependent individual dimensions
(cognitive, emotional, and personality) in a relationship and each action and/or reaction is
precipitated by a continuous chain of interactions (Day, 1995). Leary (1957) introduced
the notion of circular ordering of interpersonal variables known as the interpersonal
circurnplex (see Figure 1.1, a replica of Kiesler's 1983 circurnplex taken from Gurtman,
1997). Interpersonal theory rests upon the following three pillars: (a) the principle of
complementarity, (b) the principle of vector length, and (c) the principle of circurnplex
structure. The principle of complementarity suggests that in dyadic interactions
individuals negotiate their relationship through verbal and nonverbal cues. This
negotiation occurs along the affiliation axis (fiiendliness invites friendliness, and hostility
invites hostility), and reciprocity tends to occur on the power, vertical axis, (dominance
invites submission, and submission invites dominance) (Orford, 1986). This principle
specifies ways in which a person's displayed behavior (friendliness) evokes certain
behavior (fiiendliness) fiom an interactional partner, leading to a self-sustaining and
reinforcing system (Orford, 1986; Kiesler, 1983). However individual differences (the

overriding factor on the dominance- vertical— axis) often mediate complementarity thus

19

preexisting individual tendencies toward either submission or dominance will determine
aggressive behavior (Blulnn et al., 1990). For instance, if friendliness was unsuccessful,
anxiety might result which in turn causes the individual adult child caregiver to revert to
strategies that are less vulnerable to disruption, (e. g. hostility).

The principle of vector length suggests that personality types displayed on the
interpersonal circle and corresponding to vector lengths measure indices of
psychopathology that affects interpersonal relationships (Wiggins et al., 1989). For
example, older adults with adaptive and flexible personalities have fewer interpersonal
problems than individuals with inflexible rigid personalities. The concept behind the .
circurnplex structure is based upon the suggestion that variables measming interpersonal
relations are arranged arormd the (Figrne 1.1) circle in two-dimensional space (Leary,
1957). Interpersonal relations variables close to each other on the Circumplex are more
related than the variables that are further apart on the circle, with opposite variables being
negatively related and variables at right angles being unrelated to each other. For
example friendliness is closer associated with a warm and trusting personality than with a
mistrusting and cold personality. The interpersonal relations variables will determine how
the individual negotiates her/ his relations and whether or not aggressive acts such as
elder abuse will occur.

Therefore, the process of elder abuse could be conceptualized as a product of a
dysfrmctional interaction (behaviors driven by anxiety) between the elder and his or her
adult child caregiver (with opposite personality types) hence the bi-focal contribution to

elder abuse.

20

Figure 1.1 The Circumplex Structure

DOMJNANT
COMPETITIVE ASSURE]?

EDG‘IIBIONISTIC

   

RESTRUSTING

COLD '- _ socrABLE

HosmE , FRIENDLY
moor WARM
mmrrnn TRUSTING
UNASSU'RED pm
summsrvrz
Eflrorgfid Control Theory

Evolutionary psychologists suggest that most emotional/psychological adaptations
evolved as adaptive responses to recurrent environmental regularities over evolutionary
time. Furthermore, these evolutionary regularities influenced and primed various
affective states as a cue to action (Wilson, 1975). For instance, recurrent cues to danger
(e.g., caretaker raising his or her voice or assuming a threatening pose toward the elderly,
and social stimuli, such as strangers or mind-altering substances) will automatically
trigger the affective state of fear, in the older adult (Gray, 1987). These affective states
are the various emotional reflexes (fear, anger, etc.) that directly connect to the arnygdala
via the thalamic pathways and are implicit in emotion regulation (LeDoux, 2000).

Converging evidence in neuroscience and cognitive psychology indicate the

existence of two types of cognitive processing: implicit and explicit processing. Implicit

21

processing is automatic, relatively fast, and effortless and involves parallel processing of
large arnormts of information (Stanovich, 2004). Implicit processing responds
automatically to environment relevant information thus it fits the description of the
autonomous set of systems (Stanovich, 2004). Implicit processing is involved in the
perceptual interpretations of the other’s behavior and it has a priming effect on skills and
behaviors. Thus for example, if the elderly person via implicit processing perceives
danger may automatically assume the role of the aggressor and threaten the caretaker. In
turn, the caretaker might respond by physically attacking and injuring the older person.
Therefore when the knee-jerk reaction of implicit processing is involved in elder abuse it
i is much harder to prevent or deal with on a timely fashion.

Explicit processing on the other hand is a conscious and slower process thus
controllable by the individual. Such processing is characteristic of logical thought,
planning, and cognitive control processes (Stanovich, 2004). To use the scenario
provided above, instead of physically attacking the older adult, the caretaker may step
back think about the implications of his/ her actions after which decide on a de—escalating
approach instead of attacln'ng. Consequently, whether or not aggression toward the elder
occurs may also be influenced by the explicit evaluation of the societal context, such as
the possible costs and benefits of the aggressive act (e.g., probability of being caught,
penalties for the act, etc.). These explicitly calculated costs and benefits ratio will
definitely influence whether aggression will be perpetrated against the elderly by the

caretaker or vice versa.

22

Ecological Bi-Focal Model of Elder Abuse

Previous theoretical fi'arneworks conceptualized and operationalized elder abuse
by either focusing on the individual or the relational dimensions thus ignoring systemic
influences on this process. The Ecological Bi-Focal Model for Elder Abuse links the
individual, relational, and systemic processes as they affect elder abuse. This model
incorporates the tenets and assumptions of two broader theoretical fi'arneworks in their
explanations of elda abuse: the Bronfenbreener’s Eco-developmental model and the Live
Course Perspective (Bronfenbrenner, 1979, ; Bengston & Allen, 1993).

Bronfenbrenner and Evans (2000) espoused that individuals interact with different
contexts to form and guide their personal development and that these contexts are nested
within four levels of environment, or systems, in which the individual exists. These
mvironmental systems affecting the development of individuals are: (a) the microsystem,
which is the immediate context within which the human development takes place (e. g.
the individuals’ family), (b) the mesosystem, which is the relationship between family
and other principal settings in which the development occurs (e.g. formal or informal
support systems within the community), (c) the exosystem that includes external
environments to the deveIOping person, in which other family members participate (e. g.
the adult child’s workplace), ((1) the macrosystem that refers to major external influences
such as broad ideological values and norms, and (e) the chronosystem which refers to the
influence of temporal changes and continuities on the developing individual.

Hareven (1995) expanded on these assumptions by complementing them with
three major dimensions that are hinged upon timing. These three dimensions as they

apply to elder abuse are as follows: (a) individual timing of life transitions, (b) the

23

synchronization of individual life transitions with collective family transitions, and (c) the
impact of earlier life experiences on subsequent ones. These additional dimensions help
in explaining the relationship between the family’s demographic, social-cultural factors,
and intergenerational relations in the later years of life as they affect the developing
individual and her/ his behaviors (I-Iareven, 1995).

Instead of following the traditional models of elder abuse, that incorporate only one-
dimensional approach (the abuser or the victim characteristics as they relate to abuse) as
some of the previous theories reviewed, Schiamberg and Gans’s (2000) proposed an
ecological fi'amework for the study of elder abuse. They suggested that multiple systems
(microsystem through chronosystem) are involved with the etiology and processes of
elder abuse, thus the traditional models had to be expanded in order to better rmderstand
the phenomenon of elder abuse (Schiamberg & Gans, 2000). For instance, in addition to
victim/ abuser individual characteristics the authors proposed to investigate family
relational processes, multi-generational changes over time, and cultmal beliefs, values,
and norms they relate to and impact elder abuse. Furthermore, the reciprocal and bi-
directional relationship between the aging parents and their adult children are influenced
by the interpolations of multiple contexts such as biology, interpersonal relationships,
culture, and politieal-econornical dimensions, which, in turn, will impact the likelihood of
elder abuse and increase risk factors (Schiamberg & Gans, 2000). The four nested
enviromnental system that will impact elder abuse are:

1. Microsystem: comprised of elements in the individual’s immediate environment

and the bi—foeal interactions between the characteristics of elderiy victims and

their abuser such as age, gender, marital status, health (both physical and

24

psychological), and dependency factors such as living arrangements, history of
abuse, etc.). The microsystem is the most basic and immediate context in which
human development occurs

. Adult child- Mesosystem / aging parent-Exosystem: systems that indirectly
impact the older adult such as the employment status, financial resources, and in
general the outside stressors that might affect the caregiver adult child. In
addition, this level might also include the social isolation of either or both
individuals in the dyad and the existence of formal support systems for either or
both individuals or lack of thereof. For example, while the older parent is not
directly involved in her/ his child’s workplace activities, nonetheless the child’s
workplace policies such as family leave might directly impact the elderly.

. Macrosystem: influences such as cultm‘al values and norms, economics, and
national politics that have an overarching impact on the dynamics of elder abuse
and individuals involved in this process. Examples of macro influences are:
sexism, ageism, attitudes towards violence, and etc.

. Chronosystem: synchronizations of multiple time clocks to encompass the
evolving and interconnected dimensions of the individual and her/ his
environment over time. For example, important events such as the addition to the
household of an elderly parent in declining health and the loss of employment of
the adult child will create challenges that will modify existing relationships.
Caring for an elderly parent in ill health is more challenging for an adult child

without income and stability than for an adult caregiver who has no such

25

problems. Figure 1.2 provides an overview of elder physical abuse from an
Ecological Bin-Focal perspective (Schiamberg & Gans, 2000).

Finally, the “norm of reciprocity”, as a vital construct in elder abuse theories, needs
some focus. Schiamberg and Gans (2000) used this construct fiorn the social exchange
theory (Sabatelly & Shehan, 1993) and integrated into their model to explain why people
remain in relationships even if it is abusive. According Schiamberg and Gans (2000)
peOple will remain in relationships (exchanges) as long as the benefits are greater than the
costs and the level of relationship satisfaction is greater than the alternative (no

relationships).

Theoretical Conceptual Map

The theoretical conceptual map presented below (Figure 1.2) outlines anticipated
influences of the ecological factors on the bi-directional dimensions of elder physical
abuse. Ecological systems theory places the developing person within a complex system
of relationships affected by multiple dimensions of the surrounding environment,
(Bronfenbrenner, 1979, 1997). In this study, the environment of elder physical abuse is
envisioned as a series of nested structures, each of which has a powerfirl impact on the

developing person and the act of elder abuse.

26

Figure 1.2 Theoretical-Conceptual Map of the Dissertation study

       
 
   
  
   
    

CHRONOSYSTEM: e.g. Intergenerational Farn. Violence
\

\

MACROSYSTEMiwgctQI‘VaIues, Laws,

/ Ageism , . I \‘x ‘ \

/ ./"
g EXOSYSTEM: Social Relations, APS, \
t / .
\ [ Therapist
. « /_,
\ ;
MICROSYSTEM: Elderly

Parent/Adult Child dyad

27

Conclusions

The Applied Ecological Bi-Focal Model for elder abuse (Schiamberg & Gans,
2000), has been presented as the most inclusive elder abuse framework that best explains
the interplay between the four environmental contexts and elder abuse. Whereas certain
frameworks such as the Choice, Subjective Utility, Interpersonal, and Effortful Control
theories place the emphasis on the individual’s internal processes (e. g. emotional,
behavioral, and physical well-being) that might directly or indirectly provoke abuse,
others attempt to explain elder abuse exclusively through the prism of relational
interactions (Symbolic Interactionsim, and Social Learning Theory). This rigid either or
approach to understanding elder abuse obscures the connections uniting all human
behaviors and reduces complex human actions into simplistic categories. Nevertheless
the contribution these theoretical perspectives provide to understanding elder abuse
cannot be minimized. For instance, the Choice, Interpersonal, and Effortful Control
theories provide detailed explanations and information on basic human internal
processes, such as emotions and cognitions, which drive both the victim and the abuser’s
reactions to each other’s real or perceived intentions and subsequent behaviors. Although
numerous studies document the vital role of emotional and psychological problems in the
process of elder abuse (Kosberg, 1988; Wolf & Pillemer, 1989; Mortimer, 1995), they
lack specificity as to what exactly and how they influence the likelihood of elder abuse.
These three theories fill this gap in literature.

The Symbolic Interactionism and Social Learning‘fiameworks provide a detailed
explanation on how social interactions and relationships, both internal and external, will

influence the adult child caregiver’s decision to violently attack or not the elderly parent.

28

Athens explained this phenomenon by suggesting that perpetrators of violence first use an
inner soliloquy to define the situation at hand partly based on previous experiences and '
partly on reading the victim’s attitude and behaviors (Athens, 1997). Then based upon
this internal soliloquy between the “I” and “Me” (the generalized other) the individual
will decide if the situation and the victim’s attitude require violence or not. These models
complement Schiamberg and Gans (2000)’s explanation on the role of interaction
between aging parent and adult child (microsystem level) plays in the process of elder
abuse.

Although Bronfenbrenners’s Human Ecological Perspective and the Life Course
fiamework (Hareven , 1995, 1996) provide a more general understanding of elder abuse,
these explanatory models lack specificity and detail required to empirically test a
complex substantive theory of elder abuse. Consequently, elder abuse cannot be
adequately conceptualized, understood, operationalized, and mathematically/ statistically
analyzed with the aid of these models. Nevertheless these perspectives provide a general
framework that was incorporated into the Ecological Bi-Focal Model for Elder Abuse
thus they had to be discussed.

Schiamberg and Gans’s (2000) Ecological Bi-Focal Model for Elder Abuse
suggested that in addition to both unique individual and relational risk factors general
contextual influences also impact the phenomenon of elder abuse. Their model provides
us (and this study) with an all inclusive prism that takes into account the complexity of
individual/ relational processes in addition to understanding how ecological contexts

impact the outcome of elder abuse.

7.0

While this study mainly focuses on the bi-focal risk factors (nested mainly at the
micro, exo and meso systems level) that impact elder abuse, it is important to note that
macro and chronosystem dimensions will also have an important influence on violence
against the older adult. Given the limited scope of this study and limitations in the
MIDUS II data file macro and chronosysterns level influences could not be adequately

explored thus these aspects will be the main focus of the next study.

Purpose of the Study

The purpose of this study was three pronged: (1) to encourage the establishment
of an elder abuse quantitative research direction in the field of marriage and family
therapy, (2) empirically examine the assumptions, tenets, and conclusions of the
Ecological Bi-Focal Model for Elder Abuse theoretical framework, (3) raise marriage and
family therapists’ awareness to the needs and challenges the elderly population faces in
our society.

The Bi-Focal Model of Elder Abuse will be used as the principal guide in this
dissertation study. This author will examine whether empirical results support
conclusions reached by Dr. Schiamberg and Gans (2000) in the Ecological Bi-Focal
Model for Elder Abuse theoretical framework. The data base used for this study was
gathered by the MacArthur Foundation Midlife Research Network on Suecessfirl Midlife
Development (MIDMAC), lmder the auspices of the John D. and Catherine T. MacArthur
Formdation. Specifically, the National Survey of Midlife Development in the United

States (MIDUS) II data set will be used for this study purposes.

30

 

Bi-focal risk factors such as mental health issues (Schiamberg et al., 2009; Clark
& Winters, 2002); emotional/affective regulation difficulties influencing aggression
(Buss, 2005); behavioral problems (Schiamberg et al., 2009); and social isolation (Wolf,
1989, cited in Bennett & Kingston, 1993); caregiver abuse and/or dependence on alcohol
and drugs (Cohen, Teresi, & Holmes, 1988; Cohen, Onserud, & Monaco, 1993; Hurlburt,
Hough, & Wood, 1996); and the individual’s demographics and characteristics
(Schiamberg et al., 2009) will be used to establish casual pathways in this study. Since
the MIDUS H sample used in this study was drawn from a nationally representative
random-digit-dial (RDD) survey of non-institutionalized, English-speaking adults aged
65 and over in the United States (thus representative) and there is an existing theoretical
formdation (Schiamberg & Gans, 2000), this research is confirmatory in nature.

The present study proposes to empirically examine the causal bi-focal paths in the
familial dyw (elderly parent and adult child) as they related to risk factors in elder abuse
incidences. Another major goal for this study was to take a first step towards identifying
ecological risk factors at higher systemic levels and their individual and additive
contributions to the likelihood of elder physical abuse. Finally, given the ecological
approach of the Applied Bi-Focal Model, results of and conclusions reached in this study
can be used as a springboard for additional elder abuse and neglect research in the field of
marriage and family therapy in developing family interventions geared toward reducing

abuse in our society.

31

Significance of the Study

The significance of this study primarily involves the following outcomes: (1)
establishing the Ecological Bi-Focal Model for Elder Abuse as a useful basis fora more
complete and effective understanding of elder abuse in the family and community (2)
suggesting a quantitative research direction in the field of marriage and family therapy
upon which therapy outreach strategies could be based, and (3) developing a better
understanding of the multidirectional interactions of the individual] caretaker dyad and
context beyond that dyad in understanding the risk factors of elder abuse in the field of
marriage and family therapy.

As the first study that will empirically examine the assumptions, tenets, and
conclusions of the Ecological Bi-Focal Model for Elder Abuse (Schiamberg & Gans,
2000) will significantly contribute to the research literature on elder abuse. The current
study will specifically focus on the ecology of intergenerational relationships between
individual family members (e. g. elderly parent and adult child) over a life course
(ontogenic development) and then suggest useful and novel prevention and intervention
strategies to reduce elder abuse in the United States. Furthermore, the importance of the
elderly parent/ adult child relationship quality construct will be tested via the utilization
of a meditational mathematical data modeling design. In addition, results and findings
from this study will aid in developing a specific line of future research platform that will
further the knowledge in ecological risk factors contributing to elder abuse in the United
States. This study will also aid in launching a quantitative research platform focusing on-

elder abuse in the field of marriage and family therapy. Finally, the innovative research

32

methods and statistics used in this study could be used as a possible standard for future

methodologically sound studies in most areas of social sciences.

 

33

CHAPTER 11: REVIEW OF RESEARCH
Prevalence! Incidences of Elder Abuse

Annually, an estimated 2.1 million elder Americans are victims of physical,
psychological, or other forms of abuse and neglect (U .S. Department of Health and
Human Services, 2006). In the NatiOnal Elder Abuse Incidence Study, government
officials at the Administration on Aging documented, based on overwhelming evidence,
that the rate of abuse for individuals aged 65 and over living independently in a
community setting or semi-independently with family members was 41 per 1,000
families. The perpetrators of these abuses were most likely to be adult children or
grandchildren (77%) rather than spouses (23%) based on NCEA findings, (National
Center on Elder Abuse, 1998). The definition of elder abuse incorporates several types of
abusive behaviors such as physical, sexual, emotional! psychological, financial, and some
would prefer neglect in this category as well (Griffore et al, 2009; Schiamberg & Gans,
2000; Schiamberg et al., 2010 ). This study will only focus on elder physical abuse
(physical and sexual abuse) given that: (1) there is a lack of agreement between
researchers on whether psychological/ emotional and financial abuse measure the latent
vmiable elder physical abuse (serious concerns relating uni-dimensionality), (2)
investigators overwhelmingly exclude neglect fi'orn the family of variables measuring
physical abuse, and (3) there is an agreement in the academic and research communities
on physical abuse including sexual abuse as well (Schiamberg et. al, 2011; Grifi‘ore et al,

2009, Fulmer et al., 2004; and Meeks-Sjostrum, 2004).

34

Although emotional] psychological elder abuse and elder neglect have great
relevance to the field of marriage and family therapy, (Parra-Cardona et al., 2008), these
measures could not be included in this study for the following reasons:

1. As indicated above, there is a significant controversy in the academic and
research circles on whether emotional and psychological abuses measure the
latent variable physical abuse.

2. The MIDUS II data base had no direct indictors (manifest variables) to measure
elda' neglect and respondent emotional abuse, and

3. The principal focus of this dissertation study is elder physical abuse.

Issues relating to definition of elder abuse

Since elder abuse takes on a variety of forms, involves different relationships,
and occm's in several types of settings, it is important to first define what elder abuse is.
Elder abuse in all its types is a relatively common form of interpersonal violence
therefore researchers, academics, and government officials are in agreement on its
definitions (Hawes, 2002; Clarke & Pierson, 1999, Schiamberg & Gans, 2000). In
general, physical abuse can range from slapping or shoving to severe beatings, the use
inappropriate restraints (e.g. with ropes or chains), force feeding, or sexual abuse
(Schiamberg et. al, 2009). When a caregiver or other family members use enough force to
cause lmnecessary injury or pain, even if the reason is to help the elder, the behavior is
regarded as abusive. Elder physical abuse may only include the use of personal weapons

(hands, feet, other body parts) and performing abusive acts such as slapping, pushing,

35

kicking, or striking an older adult or may also include the use of objects such as sticks or
even baseball bats with the intent to inflict harm or pain.

In accordance with other researchers who include sexual abuse and non-
consensual sexual involvement (e.g., being forced, threatened, or deceived into sexual
activities ranging from looking or touching to intercourse or rape), the current study treats
sexual abuse and non-consensual sexual involvement in the category of physical abuse
(Hawes, 2002). Sexual abuse can range fiom rape to sexual exhibition. The act of sexual
abuse can include physically forcing the individual to look at pornography, inappropriate
touching, and imposing sexual contact with a third party, or any unwanted sexualized
behavior (Hawes, 2002). It also includes rape, sodomy, or coerced nudity. Sexual abuse is
a less often reported form of elder abuse (Hawes, 2002).

Although this study’s principal focus is on elder physical abuse (including sexual
abuse) and its correlates, it is important to define the otha types of elder abuse in order to
clarify and distinguish amongst the various forms of aggression directed towards older
adults. Emotional or psychological abuse can range fiom giving the "silent treatment" or
name-calling to verbally or physically intimidating and threatening the elderly individual.
As a rule of thumb, when a family member or a caregiver behaves in a way that causes
fear, distress, or emotional pain, that behavior can be regarded as abusive. Emotional and
psychological abuse can also include treating the older person like a child (e. g, denying
access to a phone) or isolating the elderly fiorn family, community members, and routine
activities (e.g., walking in a park or shopping at a grocery store), either through

manipulation, threats of force, or actual physical aggression (Bennett & Kingston, 1993).

36

The conscious and premeditated refusal or failme to meet the essential needs of
the older adult in order to punish her/ him or “teach a lesson” for various perceived
reasons is considered active neglect (Lachs & Pillemer, 1995). Certain care giving
behaviors, such as denial of food, health services, medication or outright abandonment of
the elderly, are considered elder neglect. In addition, the mintentional care taker
neglecting behavior stemming from developmental needs ignorance is also considered a
form of neglect (passive neglect).

The last form of elder abuse that will be defined in this study is financial
exploitation, which can range fi'om misuse of an elder’s funds to outright embezzlement
(Bennett & Kingston, 1993). Financial exploitation may include denying the older person
access to his or her personal funds, pm'chasing of items or services with the elderly
person’s money without her or his knowledge or permission, forced property transfers,
hand, or forgery (Bennett & Kingston, 1993). Given that older persons are more trusting
and often starved for companionship, they may fall prey to non-family members as well,

such as sales people and healflr, mortgage, and financial company representatives or any

other opportunistic acquaintance (Bennett & Kingston, 1993).

Impact of elder abuse on the victim and physical signs of abuse
Elder abuse seriously impacts the health and well-being of older adults on several
dimensions. Victims of elder abuse show increased levels of psychological, emotional,
and physical distress (Comijs et al., 1999) and experience increased mortality rates
compared to non-abused elderly (Lachs et al., 1998). Recent research studies found that

elders who have been abused tend to die earlier than those who are not abused, even in

37

the absence of chronic conditions and/or life threatening illnesses (Bonnie & Wallace,
2003). In addition, since older adults in United States are mainly covered by taxpayer
founded Medicare health insurance policy, medical treatment of victims imposes an
additional financial burden on the US society.

Physical signs that cannot be explained medically might suggest elder abuse and
need to be further investigated. For instance, bnrises or grip marks around the neck or
arms, welts on the wrists and/or ankles, rope marks, repeated “mysterious” injuries, and
more importantly, dismissive statements or attitudes by relatives about the elder’s injuries
(U .S. Department of Health and Human Services, Administration on Aging, 2004).
Unexplained vaginal or anal bleeding, bloody or ripped underwear, bruises mound the
breast or genitalia areas, and the presence of venereal diseases or vaginal infections might
indicate the possibility sexual abuse (U .S. Department of Health and Human Services,

Administration on Aging, 2004).

Reporting and detection of elder abuse

Since the majority of older peOple live on their own or with their spouses, adult
children, or other relatives, and not in institutional settings, elder abuse overwhelmingly
takes place at home (U .S. Department of Health and Human Services, Administration on
Aging, 2004). Therefore, when elder abuse happens, family, other household members,
or paid caregivers are the principal perpetrators of physical abuse against the older adult
Eldedy victims seldom report abuse for a variety of reasons, including concern for
themselves or for the abusive family member (Griffin & Williams, 1992). Other reasons

for concealing the abuse might include feelings of shame and grief, love for the abuser, or

38

fear of institutionalization. Unfortunately, the abused older adult will often go to great
lengths to protect their abusive children, thus keeping their own abuse a secret (Quinn &
Tomita, 1986). Even when confronted, the victimized elderly will often deny the abuse
altogether or mitigate the severity and seriousness of their adult child’s aggression, thus
creating a make belief illusion of normalcy (Kethineni 2004). In addition most abused
elderly would rather suffer in silence than break the family’s rmity and create discord
amongst members (Kethineni 2004). Finally, many abused elderly victims suffer from a
variety of cognitive, physical, emotional, and psychological illnesses hence they are often
isolated from the rest of society consequently, the abuse becomes more difficult to detect
(Kosberg, 1988). Despite these obstacles to reporting of incidence rates and prevalence of
elder abuse, studies suggest that child parent bi-directional violence is a real problem for
a significant number of elderly parents (Jackson 2003).

In addition to problems with reporting elder abuse, methodological issues such as
the use of rigorous sample surveys methods and employing non-uniform sampling and
research instruments used for data collection produce often non-replicable or even
contradictory results (Pillemer & Finkelhor, 1988). Therefore, not surprisingly, there is
an incongruity in research findings, due to discrepancies in research methods, designs,
and inaccurate reporting and/or underreporting of elder abuse (Hudson, 1986). All things
considered, the problems with underreporting and non-uniform theoretical perspectives

and research designs underscore the need for more studies in this field.

39

Risk Factors of Elder Abuse

Often nonprofessional caregivers (e. g. adult children, spouses, relatives and
friends) find taking care of an older adult relative to be an enriching and rewarding
experience. However the demands and responsibilities of the day to day elder care giving
can also be extremely stressful, especially when the older adult’s physical, cognitive and
psychological conditions deteriorate. The stress of elder care can lead to caregiver bin-n
out most notably in the areas of emotional, mental, physical, and physical health in
addition to the sometimes incurred financial costs associated with caring for the elderly.
Caregivers pushed beyond their coping threshold, financial or psychological resources
may not mean to yell at, strike, or ignore the needs of the elders in their care. There is no
single pattern of elder abuse or a specific road map of abuse that could be recognized by
most family members or social services, law enforcement, and court officials. Sometimes
the abuse is a continuation of long-standing patterns of physical or emotional abuse
within the family. In other instances, more commonly, the abuse is related to changes in
the older adults’ cognitive, psychological, and financial situation and her/ his dependence
on family members or relatives for meeting basic needs. Therefore the process of elder
abuse is a complex problem with numerous causes and risks and affected by the interplay
of at least one or more ecological systems. These risk factors might emerge at any
systemic levels or multiple levels simultaneously during the same time frames ranging

fiom the microsystem to the macrosystem level.

40

Elder Abuse Risk Factors at Microsystem Context: Adult Child/Older Adult
Relationship
History of relationship between the older adult parent and her/ his child:

The history of the relationship between the older adult parent and her/ his child
has a significant impact and often plays a determinant role (mediates) in whether the
behavior of elder physical abuse will happen or not (J afi‘e et al., 1990; Athens, 1992;
Heide, 1995). Dysfunctional family relationships implicated in the elder abuse process
include discord in the family created by the older person’s presence, a history and pattern
of violent interactions within the family (Kethineni, 2004; Athens, 1992), and prior child
abuse patterns accommodated by the family (Heide, 1995). Adult children or
grandchildren who have witnessed or have been victims of family violence are more
likely to resolve difficult life challenges with violent tactics they learned during their
childhood (Athens, 1992). For example, an adult child may take the opportunity to ”turn
the tables” on the formerly abusive parmt by forcefully “restraining” the parent. The use
of corporal plmishment in addition to a non-strict parenting approach and inappropriate
discipline in the child’s formative years appears to increase the likelihood of violence
from the adult child to her/ his elder parent (Kratcoski, 1984; Peek et al., 1985).
Furthermore, it has been suggested that in families where child abuse was occlu'ling,
violence was accepted as a legitimate and often successful means to an end and an
acceptable expression of anger and fi'ustration (J affe et al., 1990). Rhodes (1999)
suggested that this desensitization to violence is due to a dysfimctional family process
and a cultural phenomenon and not a supposed biological deviation of individuals. The

human subject is constructed by a life story. “Selves are not given. They are constructed.

41

They are built, modified, altered, refin'bished, even replaced over time” (Rhodes, 1999,
pp. 55). Rhodes (1999) describes this self creation process as children are first brutalized
into learning that they will not be protected by the adults responsible for them, that they
must brutalize others or be brutalized themselves, and finally, through the performance of
such brutalization they become violent individuals themselves.

Since its first official acknowledgement, child parent violence has remained an

often avoided or neglected subtype of family violence, with a scarcity of research to
document prevalence, severity, and factors influencing aggression against the elderly,
(Ulman & Straus, 2003). Pelletier and Coutu (1992) documented that 18% of two-parent
and 29% of single-parent families experience child parent violence. Although the
prevalence of child-parent violence is significant, these findings do not automatically
mean that adult children who were abused while growing up will turn the table on their
parents and abuse them when they are most vulnerable. The authors suggested that in
most cases the elder mother was injured (71.5%) and that most child parent violence
incidences did not result in injuries (injuries in less than 50% of the cases). Walsh and
Krienert (2007) also lmderscored the importance of studying elder abuse in an effort to
better understand victim, offender, and offense characteristics.
Therefore, an examination of the bi-directional process of violence would enable a better
grasp of the elder abuse phenomenon, especially since there is evidence to suggest an
association between child maltreatment and later elder abuse (Heide, 1995).

Kethineni’s (2004) concluded in his study that this form of violence is more

prevalent than both spousal abuse and child abuse (in the adult child’s formative years)

42

and his findings are consistent with estimates documented by the National Center on
Elder Abuse (1998) data

Given the significance of the relationship quality between the adult child and her/
his elder parmt in mediating or moderating elder abuse, this author proposes to construct
a latent variable that will measure this construct and examine its direct and meditational

influence on elder physical abuse.

Characteristics of the Adult Child that Increase Aggression towards the Elder Parent
A caregiver’s (such as an adult child) personal problems, such as the presence of

mental or ernotional'illnesses, addiction to alcohol or other drugs and financial pressmes
stress and therefore dependency on the elderly parent, might lead to elder abuse
(Schiamberg et al., 2009; Griffore et al., 2009; Schiamberg & Gans, 2000). Intrapersonal
theories of violence in the field of marriage and family therapy assume that certain male
demographic characteristics increase the risk of engaging in physical violence. Pagani et
al. (2004) found that women engage in slightly more physical aggression (9.7%) than
men (8.8%) within family relationship particularly among young adults. On the other
hand males engage more in “severe physical aggression” according to O’Leary et al.
(1994). In addition, younger males are significantly likelier to solve relational problems

via violence (O’Leary et al., 1994).

Gender

The gender predominance of the perpetrator of elder abuse varies across studies

depending on the methodology used. Conclusions reached as regards to gender

43

 

predominance in some studies suggest that sons are likelier to abuse their parents than
daughters (Cochran et al., 1994; Laurent & Deny, 1999). On the other hand other studies
suggest that females were the likelier gender to commit parent abuse (Pagani et al. 2004;
Nock & Kazdin, 2002). Yet others found no gender differences in reported parent abuse,
(McCloskey & Lichter, 2003). Walsh and Krienert (2007) concluded that teenage and
adult male children were more likely to commit aggavated assault or intimidation against
their elderly parents than female children, and female children more often used personal
weapons (eg fists, leg kicks, etc.) against their parents. Furthermore, female children .
were significantly more likely than male children to cause injury to their elder mother,
but no major differences in inflicting injm'y were formd with fathers. Given the
contradictory research findings on the adult child’s gender as they relate to likelihood of
him / her physically attacking the elderly parent, this variable won’t be used in this study.
Furthermore there is no direct measure of this in the MIDUS II data set as it relates to

abuse.

Chronological Age

The chronological age of adult child when she! he is more likely to perpetrate
violence on her/ his elderly parents is still in debate and often varies depending upon the
methodology used in specific studies and age inclusion parameters. Cornell and Gelles
(1990) suggested that adult male children carry out more assault on their elderly parents
as the get older (linear positive relationship) than females. Other researchers found that
the prevalence of assault from males decreases as they get older (Strauss & Gelles, 198 8).

Browne and Hamilton (1998) found no significant associations between age of offender

44

and physical violence perpeuated on parents CE. (1998). Results of Ulman and Straus’s
(2003) metaanalysis seems to indicate that age categories of children who perpetrate
violence on their parents vary dramatically and that there is no certainty about whether
age impacts the likelihood of elder abuse at all. In sum, the abusers age might not be a
good predictor of whether he/ she will abuse the elderly parent. In addition there is no

direct measure of this variable in the MIDUS II data set.

Psychological and emotional problems/ illnesses

Ongoing individual psychological and emotional problems (illnesses) in the
caretaker have been identified as predictors of elder abuse (Pillemer & Finkelhor, 1988;
Kosberg 1988). For instance, if a caregiver (in our case the adult child) suffers fiom
psychological/ emotional problems and who cares for a frail older adult with cognitive
impairment is likely to elicit resistant behavior fiom the old parent, which may lead the
caregiver to become physically violent (Kosberg 1988). Therefore the adult child
caregivers with severe emotional, cognitive, and psychological health problems will
likelier abuse an older adult in their care than a caregiver who has no such problems,
(Kosberg, 1988). Wolf and Pillemer (1989) also concluded in their study (based on
frequencies in their sample) that 38% of the abusers had a history of mental illness and
46% of abusers reported a recent decline in their mental health status. Cooney and
Mortimer (1995) also suggested that caregivers who admitted being physically abusive in
their sarrrple had notably higher rates of psychological and emotional health problems

than non abusive caregivers (Cooney and Mortimer, 1995). Pillemer (1989) as well

45

.rlhhr ",r'._.¢.,‘ "
li"'l"0~:l;lfl“.§ll‘3r‘l“!t ill"!

9
o

' .
‘K
'1
to
‘2‘.
‘re
'(
.5.-
s
;:e
.g‘
,s‘.
-.
':e
._».
n...
J

,x

‘

u

in

IL.

5:.

 

' T. -. r‘ "I "‘ ' " a” ‘
‘g'fll '4’ “5‘68 {Nagy-X31411

concurred with conclusions reached in others studies that suggest that psychological and

physical abuse appeared to be most related to the deterioration in health of the caregiver.

Alcohol and substance abuse addiction

Chemical dependency problems in adult caregivers have often been significantly
correlated with the process of elder abuse (Kratcoski, 1984). In a similar study Browne and
Hamilton (1998) conclusions concurred with findings that alcohol and drugs abuse by the
caregiver (adult child) was linearly (positive direction) associated with an increase in
violent encounters between the adult child and the elderly parent. The likely scenario
during which the adult child’s chemical dependence might play a siglificant role in a
violent encormter with the older adult would be when both the parent objected to the
adult child’s alcohol/drug use and when the adult child punished the elderly parent while

under the influence.

F inancial stress and burden

Caregiver’ financial stress (economic pressures and or lack of financial
resources) may also be a significant risk factor for elder abuse. Kosberg and Nahmias
(1996) found that adult children lacking in their financial well-being might be resentful of
their parents which in turn might lead to physical abuse. Steinmetz (1990) suggested that
financially overburdened and stressed out caregivers were likelier to commit (probably
repeatedly) the act of elder abuse than caregivers who could effectively cope with their

own stress.

Characteristics of the Elderly Victim

A general review of the elder abuse literature yields a number of victim risk factor
dimensions, interacting at various ecological systems level that might impact the
likelihood of abuse of older adults living in their own homes, with relatives, or in a
community setting (Pillemer & Moore, 1989). In addition, several research studies on
elder abuse over the last twenty years highlight individual victim characteristies that are
correlated with elderly abuse, such as; (a) suffering from diagnosed Alzheimer’s illness
(Paveza et al., 1992), (b) elderly in need of assisted home care services, (Fulmer et al.,
2005), (c) limitations in mobility, especially in regards to self-care (Fulmer, et al., 2005;
Pillemer & Finkelhor, 1988), ((1) social isolation (Fulmer et al., 2005), (e) cogritive
impairments such as Alzheimer and dementia (Lachs et al., 1997), (f) behavioral
problems (Coyne et al., 1993; Paveza et al., 1992), (g) serious chronic health problems
(Pillemer & Finkelhor, 1988). Based upon findings in studies (such as listed above)
conducted over several decades, it is evident that a relatively large number of late life
conditions associated with the last stages of human existence might negatively impact the
physical, psychological, and emotional safety of older adults. In addition, the literature on
elder abuse reveals that there are several unique factors that might also impact the

likelihood of abuse such as:

Gender
Much of the literatme on elder abuse suggests that mothers are the most likely
victims of child-parent violence. For instance, Nock and Kazdin (2002) suggested that in

88% of the time child violence was directed against the biological mother, followed by

47

the ad0ptive mother (5.4%), and finally only 2.7% was against the father. Several other
studies as well reached similar conclusions suggesting that elderly women are more
vulnerable to abuse than men (Pillmer & Finkelhor, 1988; Kosberg, 1988). Similarly,

Kethineni (2004) found in a sample of 83‘ juveniles arrested for violence against their

parents that biological mothers were the victims of child parent violence in 81% of cases.

Contrary Kosberg and Nahmias (1996) concluded in their study that there were no major
differences in the likelihood of abuse along the gender line. The apparently contradictory
results might be due to non-uniform research designs, sampling procedures, and

statistical calculations.

Chronological age

Kosberg (1988) suggested that the incremental increase in the elder’s

chronological age directly impacted (positive correlation) the likelihood of her/ his abuse.

This partially might be due to that advancenent in age was often linked to a decrease in
physical and cognitive health which logically would lead to an increase in psychological
and emotional problems that in turn directly correlates with abuse (Kosberg & Nahmias,
1996). Just as with gender, findings on the variable age’s effect on elder abuse are not

uniform. For instance, Schiamberg et al. (2009) suggested that age had a negative direct

effect on the likelihood of elder physical abuse in nursing homes.
Marital status

Although marital violence occms across the life span, its fiequency appears to

decline over time (Pillemer et al., 1990). In some instances one might argue that elder

48

 

C
.
i'
i "I
I .3
pt
. _ _
‘9‘-
'e
:f
,l
.
' .E'u
. .‘g
‘ I ‘—
. 'D
‘ .3 tr
--..
t?-
:11-
. ‘;_;
4.
r!"
1'}
3’3-
1:0».
.5}
..
Lin
5;;
11'}
~35
‘tt;
i a
it"
. (:3
. a
.7?
. .5
O
I:%
2".“

abuse is simply a continuation of abuse that has been occurring in the family over many
years. Evidence of spousal abuse suggests that it is proportionately (23%) less than abuse
perpetrated by adult children (National Center on Elder Abuse, 1998). Despite the
apparent decline in prevalence of abuse, a considerable number of elderly women
continue to be victims of spousal abuse. Contrary to this Pillemer and Finkelhor (1988)
found that a married older adult is likelier to be abused than a widowed or divorced one.
They further suggested that this might be due to living arrangements since an elderly
person living with someone else is at higher risk of abuse than someone living alone. In
other instances marital stress due to sharing a residence with gown up children or other
relatives might lead to marital domestic violence. Once again findings seem to be

contradictory and nebulous at best.

Chronic and serious physical health problems

Findings in many empirical studies suggest that physical stressors such as chronic
physical illnesses and ADLlimitations were significantly associated with the degee of
elder abuse (Bonnie & Wallace, 2003). Poor health and limited functional ability,
particularly in regard to self-care, were strongly correlated with elder abuse occurring in
community settings (Fulmer, et al., 2005; Pillemer & Finkelhor, 1988). In addition, older
adults with one or more physical impairments (health problems) are more vulnerable to
abuse because of their diminished ability to protect themselves, their dependence on the
caregiver, or seek help to end the violence (Lauman et al., 2008). These findings were
confirmed by a follow up study completed by Lauman et al. (2008). They confirmed that

chronic and serious health problems will definitely increase the dependence on the

49

caregiver, which strongly correlates with elder abuse (Lannnan et al., 2008). Additionally,
the elder’s physical impairments might increase the stress experienced by the caregiver,
which in turn increases the likelihood of a negative and aggessive interaction. When the
reverse is true, and the impaired older person is completely dependent on the caregiver,
the caregiver may experience resentment that leads to abusive behavior (Lauman et al.,
2008).
Presence of cognitive disorders

Pillemer (1992) suggested that older people with Alzheimer’s disease were
violent on occasion, which in turn, was robustly associated with caretaker physical
violence and violent feelings. Older adults who exhibit signs of Alzheimer’s may become
abusive as part of the disease progression process, and the object of the abuse more likely
is the caregiver who provides the day to day care for the impaired elder (Kosberg, 1988).
The abuse of the caregiver may manifest as various types of assaults resulting in minor
bnrises or cuts. The link between cognitive disorders (associated with aggressive and
provocative symptoms such as verbal/physical aggessiveness) and an increase in the risk
of elder abuse was documented by Cooney and Mortimer (1995) as well.

Williamson et al. (2005) evaluated the reactions of caregivers of cognitively
impaired elderly care recipients to of physically impaired elders without cognitive
impairment. Their results confirmed that caring for an elder with cognitive deficits can
mean not only providing more care but also dealing with more confused and delusional

behavior which in turn is the primary contributor to caregiver’s feelings of resentment

and hostility towards the older adult.

50

, ,. ,
. 3"
v.8
'.
:-
1. '
2.;
J?

.J
.1:
3%.,
." rt"
.32.

. , .
2315.
I,"
..-._-g~.
.241
..$(.
.' --
224'.

-r.-.
Lt ‘
a
.3:
. e‘

-0

...°'
’ r
:r'
.....
'.:¢
'
&e_
b 7 ‘l

 

Homer and Gilleard (1990) suggested that although there is no clear association
between the diagnosis of cognitive disorders such as Alzheimer or dementia and abuse
but evidence unequivocally indicated that violence (or threat of violence) by the person
with cognitive impairments seemed to lead to a violent response by the caregiver. Their
finding clearly indicates that it is the elder’s disturbed and disruptive behavior that was
likely to result in abuse by the caregiver, rather than the presence of the diagnosis of
cognitive impairment such as dementia. In the end the outcome (elder abuse) is the
important part rather than our inability to decide whether the diagnosis or its

manifestations are is the reason for the violence directed against the elderly.

Presence of mental and emotional illnesses

Psychological and emotional problems may be both a cause and an effect of elder
abuse in many circumstances. When elders behave in disturbing ways due to either
psychological or emotional problems (mostly combined effects), it becomes more
difficult to provide care for them. Higher levels of physical aggession in the elderly were
associated with depression, psychosis, impaired commurnication, and antipsychotic drug
use (T alerico et al., 2002). Furthermore these authors suggested that aggession in the
elderly may be a symptom of inadequately treated mental health disorders rather than a
conscious and volitional decision to antagonize the caregiver.

In a cross sectional study of 184 elderly patents with mental disorders (65 with
depression, 97 with dementia, arid 22 with anxiety disorders) Racic et al. (2006) found
that the prevalence of elder abuse amongst older adults suffering fiom depression and

anxiety disorders was high. Their findings were consistent with results fiom others

51

studies on this domain (Flannery, 2003; Coyne et al., 1993). Findings of Cooney and
Howard’s (1995) supported previous findings as well that the rate of abuse among older
people with mental health problems are higher than in the general population of older
people. Finally, co-morbid alcohol/ drug abuse problems are not infrequent companions
to psychiatric disorders, particularly depression (Gambert & Katsoyannis, 1995).
Unlike with most other victim! abuser characteristics, the literature is consistent
the correlation and probably cause and effect relationships between emotional and

psychological problems and elder abuse.

Substance abuse

A massive amount of data exist on prescription and over the counter (medication)
drug use by the elderly. Although the elderly populations makes-up only 12.4% of the
population, they consume 25% to 30% of all prescription drugs. They routinely consume
more prescription drugs than any other age goup and are more likely to consume
psychotropic medications with a potential abuse and addiction (Rigler, 2000). The
relationships between elder abuse and substance abuse/ drinking in late-life is
complicated given disparate and often contradictory findings on the health benefits of
moderate drinking at one end and the co-morbidity between psychiatric problems and
substance abuse at the other (T urvey et al., 2006). The negative effects of alcohol use and
abuse relate mostly to excessive drinking, whereas substance abuse at any level can have
potentially harmful effects on the health and welfare of the elderly (Turvey et al., 2006).
Alcohol abuse can lead to a plethora of physical health problems (cirrhosis, pancreatitis,

or even cardiomyopathy), major emotional and mental health disorders, and memory and

52

sleep problems, (Speer & Bates, 1992). Rigler (2000) also suggested that elderly
individuals with alcoholism problems are mairnly early onset drinkers and they
overwhelmingly have significant physical, psychological, and emotional health
complieations.

Given the high co-morbidity rate between the elderly victim’s characteristics such
as physical, psychological, and emotional problems and substance abuse, in addition to a
small sample size (203), this author made the decision no to include elder substance/
alcohol abuse in the final dissertation analysis. Furthermore, not only the literature
indicates very high correlations between the above cited victim’s characteristics and

substance abuse but also the data in form of collinearity.

Victim behavioral problems

Kosberg (1988) suggested that aggessive, overly demanding, or otherwise
unpleasant behavior of the elde'ly might contribute to the risk of abuse. Higher levels of
both verbal and physical aggession in the elderly were associated with cognitive
(Kosberg, 1988; Talerico et al., 2002), psychological and emotional health correlates
(Racic et al., 2006), substance abuse (Rigler, 2000), impaired communication, and higher
disorientation (Talerico et al., 2002). Given that elder behavior problems might only be
symptoms of one or more untreated physical, cognitive, mental, arid emotional health
problems, this study will be guided by a conceptual understanding that elderly aggession
may represent an expression of certain types of cognitive, psychological, and emotional

manifestations or met needs of the elderly.

53

Contexts beyond the Immediate Older Adult/ Adult Child Caregiver Relationship
Several studies on elder abuse suggest the influence of Exo and Mesosystem level
factors (outside of the immediate family) that might contribute to an increase in the
probability of violence, such as social relationships with people outside the family who
maltreat others, social deprivation, and beliefs about behavior and attitudes to education
(Paulson et al., 1990). Therefore, an examination of the bi-directional violence at these
levels would help us in better understanding and conceptualizing the process of elder

abuse.

Social isolation

Fulmer et al. (2005) documented that the strongest predictors of elder abuse in
the home were the social isolation of and to some degee the lack of a formal support
system for both the victim and the caregiver. Social isolation of the elderly can provide a
clue that a family may be in trouble, and it also can be a risk factor for abuse. Social
isolation can be a strategy for keeping abuse secret, or it can be a result of the stresses of
caring for a dependent older family member (Fulmer et al., 2005). Isolation also makes it
more difficult for outsiders to see and intervene in an abusive situation to protect the
elderly and to offer help to the abuser. When commnmity members are a part of the
elderly person’s life, tensions are less likely to reach unmanageable levels, and both the
victim and abuser are more likely to receive help. Furthermore, of there not sufficient
support pillars for the adult child to provide care, or if their own difficulties are not
recognized and assistance offered, as relevant, these difficulties may escalate into an

abusive situation (Wolf, 1989).

54

Financial resources/ employment status

Additional risk factors at higher systemic levels (the adult child! mesosystem —
elderly parent/ exosystem) were also identified in the literatme of elder abuse. Aspects as
stressors within the family system, such as the caregiver’s nmemployment thus financial
dependence on the older adult may also be implicated in the process of elder abuse
(Pillemer, 1986). Some research investigators have indicated that abusers may be
dependent on their victims for housing and accommodations (Pillemer, 1986). In
addition, apparently some caregivers are not suited to provide care, either physically or
psychologically, but who are nevertheless expected to slide into a caregiving role for their
parent.

Macrosystem Level Factors that Affect Elder Abuse

Beliefs and views of elderly people as being vulnerable and dependent may
contribute to negative perceptions and attitudes about older people. Ageism is a general
term that encompasses differential attitudes such as older people need assistance even if
they do not want it based on age designations and age discrimination is a displayed
behavior based on attitudes. In that sense, the cultural characterization of age goups
necessarily implies differences among them, and is a form of age discrimination. This
differentiation coupled with meanings and attitudes of ageism may affect the rights and
privileges of the elderly by placing heavy constraints on their life and potentially
devaluing them. For example, while working for the Family Independence Agency, State
of Michigan, I came across several cases where the caregiver nmduly withheld finances

firm the older adult when she/ he intended to purchase items that might have enhanced

55

her/ his quality of life. When asked for a reason, caregivers almost in unison replied that
their relatives (parents) mind (logical reasoning) was gone so what did they no what is

needed to adequately live their life.

Research Questions

Given the complexity of the process of elder abuse and the various ecological
systems at work interacting with and shaping several risk dimensions, the historical
approach to conceptualizing and explaining elder abuse is no longer adequate. The
different types of research studies conducted on elder abuse (case control, agency data,
studies of awareness and attitudes, and survey of professionals) often employed non-
uniform samples, research methodologies, and statistics that resulted in an often
contradictory knowledge base in the filed of elder abuse (Pillemer & Finkelhor, 1988;
Lachs et al., 1997). In contrast the Ecological Bi-Focal Model for elder abuse has been
presented as the most inclusive framework that best explains the complex processes
occurring before, during, and after the conclusion of the act of elder abuse.

To date, an understanding of family violence directed against an older adult
parent by the an adult child, from an Ecological Bi-Focal perspective has been limited
because of the paucity of research targeting adult child members involvement in elder
abuse and the fi'agnented nature of risk factors/correlates analyzed in previous studies
(Chermack et al., 2001). In addition, the correct incorporation of all tenets and
assnrrnptions of the theoretical Ecological Bi-Focal model of elder abuse requires a large
representative sample and a structural equation modeling that was not possible before.

Therefore complex research questions and related hypothesis could be advanced in areas

56

where research on elder abuse had significant gaps. In order to examine causal factors
relating to and predicting elder abuse from an Ecological Bi-Focal model for elder abuse

the aim of this study was to find answers to the following questions:

I. Prevalence
1. What is the prevalence of elder physical abuse committed by an adult child in the
farrnily/ community setting?
11. Immediate older adult/ caregiver context
2. Do victim characteristics influence the likelihood of elder physical abuse?
3. Do victim behavioral problems influence elder physical abuse?
4. Do victim and adult child relationship quality predict, mediate and or moderate

elder physical abuse?

III. Contexts beyond focal older adult/ caregiver (ecological model testing)

5. Does the victim’s social isolation influence elder physical abuse?
6. Do adult child characteristics influence elder physical abuse?

The temporal dimension (chronosystem)
The impact of ecological temporal transitions on elder physical abuse could not be
assessed given the natnue of MIDUS H data thus it is beyond the scope of this study to

integate chronosystem level variables.

57

CHAPTER HI: METHODOLOGY

In 1995-2006, the MacArthm' Midlife Research Network snn'veyed over 7,000
Americans aged 25 to 86 to collect data on the role of behavioral, psychological, and
social factors in understanding age-related differences in physical and mental health. Data ‘
was collected with the aid of baseline assessments (e.g., phone interview and self-
adrninistered questionnaire) and additional questions in selected areas (e.g., caregiving,
cognitive and emotional functioning, stressful life events, and health condition). The
guiding hypotheses for this innovative study were that behavioral and psychosocial
factors influence people’s physical and mental health. The MIDUS data file was used in
221 studies ranging fiom stress related health conditions to sexual experience across
midlife. Although the MIDUS data set was widely used to investigate various human
behavioral characteristics and health conditions, up to this date, it was not used to study
elder abuse in the United States. Given the innovative and the broad scientific scope of
the MIDUS study in addition to lack of previous research on elder abuse using this data
file, the MIDUS data set is optimal to empirically test the Ecological Bi-Focal Model for
Elda' Abuse (Schiamberg & Gans, 2000).

This chapter describes the dissertation study procedures in a two pronged format. The
first part of this chapter will be used to describe the methodology employed in the
MIDUS study to compile the data set. The second part will provide a detailed description
of the methodology used to empirically test the Ecological Bi-Focal Model for Elder

Abuse. This chapter will be organized based on the following format:

58

I. MIDUS I and MIDUS II data sets
Measures, instrument development and pretest, data collection protocol,
interviewer training, research field procedures, survey protocol, and sampling
procedures for the MIDUS I and MIDUS H studies.
II. Proposed Dissertation Study
1. Definitions of latent variables and their manifest indicators.
2. Overview of the measurement and structural models used in this study;
3. Overview specific hypotheses to be tested

4. Data analytic plans.

MIDUS I and MIDUS H Data Sets
Measures
In 2004-2006, the MacArthur Midlife Research Network developed, pre-tested,
and then used the following three survey instruments to measure several life comse
dimensions in their study. (1) MIDUS 11: Phone Questionnaire With Index (see Appendix
A), (2) MIDUSH Self-Administered Questionnaires (SAQ 1 & 2) - mailed to
participants, (see Appendix A), and (3) MIDUS II Cognitive TACI‘ instrument,

Telephone Assisted Cognitive Testing, (see Appendix A).
Instruments Development and Pretests

The development of the MIDUS 2 phone survey began at the University of

Wisconsin in January 2003. The survey instrument came in two waves, wave two was

59

based on wave one and included modifications (measuring instrument calibrations and
improvements).

The MIDUS one wave instruments (telephone and self-administered) were
pretested and calibrated by the fall of 2003. A random sample was obtained fiom Survey
Sampling Incorporated, State of Wisconsin, and 50 random numbers were fielded. Out of
this sample 38 respondents completed the phone survey and 27 (out of 38) mailed back
the SAQ’s. 32 respondents completed the Cognitive surveys. All of the data from the
pretests were delivered to the University of Wisconsin IOA staff during the late fall and
winter of 2003/2004. Based on result outcomes minor calibration changes were made to
all three instruments after pretesting process was complete. Final research instrmnents
were prepared for fielding and then utilized to obtain the MIDUS study sample.

The improved wave-two instruments then were used in the creating the final
instrument form and proan the Cases CAT'I computer instrument. MIDUS staff
decided to convert the 114 pages two self-administered questionnaires (SAQ’s) into PDF
files for ease of printing and mailing. Interview data obtained with the Cognitive phone
survey instrument (set of digital wave files) were digitally recorded for later computer
scoring and analysis. The end product of these recording processes was a set of computer
digital wave files that could be listened in, analyzed, and later scored by the research
team.

Data collection protocol

The MIDUS data base was generated by both self-completed individual surveys

and phone interviews by employing the using CATI (computer-assisted telephone

interviewing) protocol. The CATI system was developed and copyrighted by the

60

University of California- Berkeley's Computer-Assisted Survey Methods Progarn or
CSM. This system is a flexible data collection method since it allows for pre-coded
questions, open-ended questions, and combinations of the two. A second major advantage
of the CATI system is that during the interview process the text of the survey appears
question by question on a computer screen so the data collector can read it to the
respondent and enter data values in a timely fashion. The CATI system accepts only
valid responses (data values) and interviewer can’t move on to a different question before
correctly entered a value for the current one. In addition, this data collection system has
extra progam module added that creates and maintains an electronic cover sheet for each
participant that can be retrieved in an a timely and prompt manner thus eliminating the
cumbasome process of opening individual interview snn'veys for identifying information
such as the respondent’s phone number, address, interview status (complete versus

incomplete interviews) for ease of access during follow-up interviews.

Interviewer (data collector) training

MIDUS data collectors (interviewers) received 12 hours of general survey
training and four hours of specific MIDUS related issues and instruments. Specifically,
training was provided on: (a) general MIDUS research protocols, (b) phone callirng
guidelines and real life practice simulations of simultaneously talking with “research
respondents” and entering data into computer, (b) how surveys (cases) will be fielded and
classified, (c) quickly identifying respondents based on the electronic coversheets, and
((1) how to conduct mortality (closeout) interviews. In order to assure a quality, valid, and

reliable data set, MIDUS interviewers were required to practice the interview format on

61

at leas three different real respondents and complete entering three practice cases in the
computer database. Performance on these three practice simulations were directly
supa'vised and monitored by senior MIDUS research staff and feedback was provided on
performance and areas needing improvement to all prospective interviewers.

Close process monitoring and data quality assurance did not cease once
interviewers completed the MIDUS data collection training sessions. Supervision and
monitoring was extended into data collection as well so once the survey started,
interviewer data collecting interviews were regulme monitored using the blind
monitoring system. This system was comprised of: (a) randorrnly a senior research
supervisor was placed in a room not visible by the interviewers, (b) supervisors having
real life access to the interview conversation transpiring between the interviewer and
respondents, and (c) supervisors seeing the same computer screen as interviewers do thus
monitoring live data entry.

Interviewers’ work quality was reviewed at least on a monthly basis and they
. received a written feedback that contained constructive work performance critiques and
suggestions for improvement, as needed. Interviewers had to implement these
suggestions effective immediately and their supervisors continued monitoring how the

improvements helped their data collector in fulfilling the MIDUS research expectations.

Research field procedures
Several meetings were held with staff to discuss and implement procedmes on
how best to deal with a sample of this magnitude and develop a protocol on how to

document and completed the multiple contacts with the same respondent After staff

62

brainstorming and attending several meetings the following data processing and quality

control filed protocol was developed:

1. All confidential records to be kept in electronic format for easy access and follow-
up contact.

2. Establishment of a toll-fiee respondent telephone line: To maximize response
rates, a toll-free line for MIDUS was established and kept available only for the
MIDUS study purposes. In addition a telephone voicemail box was set up for the
events that phone room staff was not available. A voice recording instructed
callers to leave a message if they were calling about an interview. Study
participants were notified of the toll-fi‘ee number in advance via letters that were
sent to all MIDUS participants.

3. Each MIDUS participant received a personalized advance letter from staff one
week prior to being contacted. For the follow-study respondents were asked to
call the MIDUS toll-flee number to provide updates and future contact times. A
Brochure was also enclosed to reacquaint the respondent with the MIDUS survey.

4. Establishing an auto-scheduling system (from of electronic cover sheet) that -
allowed anyone accessing the case to read all previous contact notes and

information and for the tracing department to efficiently trace the case.

Survey Protocol
MIDUS II sm'vey protocol was a complicated process due to a multipronged
approach whereby the telephone surveys (also phone cognitive survey) and mail surveys

were taking place mosfly concurrently with different respondents at different points in the

63

survey at all times. To further complicate the process, some survey participants
completed their mail survey within days of finishing the phone survey and also mailed
them to MIDUS headquarters, whereas others procrastinated in completing their mail
survey for montlns and needed constant prompting to mail their surveys. These
complications were addressed by MIDUS senior staff by developing and implementing
survey protocol that was flexible enough to accommodate these challenges. For more

information on this protocol see Appendix A.

Sampling Procedures

The first wave of MIDUS national survey study (MIDUS I) of over 7,108
participants (aged 25 to 74) was carried out dming 1995-1996 by the MacArthm' Midlife
Research Network. The purpose of this study was to examine age related factors as they
correlate with the individual’s psychological, behavioral, and social factors. For optimal
sampling results MIDUS researchers used the random digit dialing survey method (RDD)
which is based on the largest database of working residential telephone exchanges and
working blocks in the United States. Thus, MIDUS researchers could tlneoretically draw
the sample fiorn any part of the United States (a real random sample), regardless how
large or small. Furthermore, this telephone survey method has the advantage of including
unlisted nmnbers as well that would be missed if the numbers were selected fiom a
regular phone book. The MIDUS I study was goundbreaking due to its broad scientific
scope (age related differences in mental and physical health due to psychological,

behavioral, and social factors) its large and diverse samples, and its creative use of both

phone and mail surveys. A more detailed description of the MIDUS I research study can
be found in Appendix B or at http://www.midus.wisc.edu.

The MIDUS II study sample, which is a longitudinal follow-up of the original
MIDUS I study, was conducted between 2004 and 2006. MIDUS researchers made
every possible attempt to contact and persuade all the original respondents of MIDUS I
study to participate in a follow up data collection effort (the MIDUS H study). To
increase participation of MIDUS I participants monetary incentives were raised fiom $20
(MIDUS I participants received) to $60. Of the 7,108 respondents in MIDUS 1 study,
interviewers successfully reached 4,963 participants who completed another 30 minutes
long phone interview.

Retention and response rates that were cross referenced and then readjusted based
on data received from the National Death Index (NDI), for the various MIDUS samples

are provided in Table 3.1.

Table 3.1 Sample Sizes, Longitudinal Retention Rates, and Response Rates for MIDUS II

 

 

 

Phone Interview Sample # MIDUS l # MIDUS 2 Longit. Ret. Adjusted for
4 Respondents Respondents Rates mortality

Main RDD 3,487 2,257 65% 71%

City Ovetsatnplw 757 489 65% 71%

Sibling 950 733 77% 83%

Twin 1,914 1,484 +78% 82%

Full Sample 7,108 4,963 70% 75%

 

 

 

 

 

*UWSC fielded 7,105 cases at MIDUS 2 because contact information was unavailable

for 3 cases.

65

 

** Note: From Ryff et al. (2207). Midlife DevelOpment in the United States
(MIDUS II), 2004-2006 [Computer file]. ICPSR04652-vl. Ann Arbor, MI: Inter-
university Consortium for Political and Social Research [distributor],

2007-03-22. doi:10.3886/ICPSR04652

Of the original MIDUS I participants 30% of respondents did not complete the
MIDUS 11 phone interview for the following reasons:

The MIDUS II study also included two self-administered questionnaires (SAQS),
each about 55 pages in length, which were mailed to participants, and when completed,
returned by mail. Response rates for the MIDUS II SAMPLE are summarized in Table

3.2 provided below.

Table 3.2 SAQs Completion Rates for the MIDUS II study.

 

 

 

 

 

 

Phone Interview Sample # of Complete Completion Rates
SAQS

Main RDD 3,487 2,257

City Oversamples 757 489

Sibling 950 733

Twin 1,914 1,484

Full Sample 7,108* 4,963

 

** Note: From Ryff et al. (2207). Midlife Development in the United States
(MIDUS II), 2004-2006 [Computer file]. lCPSR04652-vl. Ann Arbor, MI: Inter-
university Consortimn for Political and Social Research [distributor],

2007—03-22. doi:10.3886/ICPSR04652

66

In addition, to the initial 30 minutes phone interview, a second 20 minutes telephone
interview was conducted in order to collect cognitive battery data. Response rates for

completion of the cognitive battery across the above samples are provided in Table 3.3.

Table 3.3 Participation Rates for the MIDUS II Cognitive Battery.

 

 

 

# of Complete Completion Rates
Samrle Cognitive (% ofPhone
Batteries Participants)*
Main RDD 1,846 83%
City Ovensarnples 417 87%
Sibling 661 91%
Twin 1,262 86%
4,186 86%
Full Sample

 

 

 

 

* Completion rates do not include 71 cases who, after completing the telephone
interview, indicated they did not wish to participate in filture MIDUS research.
** Note: From Ryff et al. (2207). Midlife Development in the United States
(MIDUS II), 2004—2006 [Computer file]. ICPSR04652-v1. Ann Arbor, MI: Inter-
, university Consortium for Political and Social Research [distributor],

2007-03—22. doi:10.3 886/ICPSR04652

67

Proposed Study

Study Sample

For the pmposes of this dissertation study the final MIDUS II sample (participants
ranging in age fiom 35 to 86 old — a total of 4,963 cases) had to be reduced to a specific
subsample in order to comply with the criteria set forth in the Applied Ecological Bi-
Focal Model for Elder Abuse (Schiamberg & Gans, 2000) and proposed earlier in this
study. To be included in this study, participants had to meet the following criteria: (1)
their chronological age had to be at least 65 years old and (2) had to be living with an
adult child in the home who was providing care to the older adult. Out of 4,963 cases
1880 respondents indicated that they were at least 65 years old. Out of the total of 1880
participants only 203 met the criteria of residing with an adult child who provided the
care to their elderly parent. The dernogaphic characteristics of the remaining sample

(203 observations) used for this dissertation study is as follows:

68

Table 3.4. Dissertation Sample Characteristics

 

 

 

Demographics Frequencies Percentages
Age 65-70 years old: 95 46.8%
71-75 years old: 33 16.3%
76-80 years old: 37 18.2%
81-85 years old: 22 10.8%
86 and above: 16 7.9%
Gender Male: 75 36.9%
Female: 128 611%
Race Caucasian: 180 88.7%
Afiican American: 8 3.9%
Native American: 3 1.5%
Other: 11 54%
Marital Status Married: 74 36.5%
Not Married: 129
63.5%
Education Not high school: 34 16.7%
High school: 64 63%
BA/ Masters : 37 18.3 %
PhD or Equiv.: 4 2%
Total 203 100%

 

 

 

69

 

Human Subjects Protection

The primary institutional review board (IRB) for the participation of human
subjects in this project will be the Michigan State University Committee for Research
Involving Human Subjects (UCRIHS). It is IRB’s (MSU Institutional Review Board)
- policy that research that only involves secondary analysis of pre-approved public data
files, does not require IRB approval. The IRB must be informed of the conduct of the
research project howeva and certify the use of an approved database. Furthermore, the
data file is available for ICPSR (Inter University Consortium for Political and Social
Research) members and researchers (public) consumption and can be downloaded fiom
ha://www.igpsr.umich.edu/mydata?path=ICPSR. In addition, although the original data
file was gathered on identified subjects (to members of the McArthur Foundation), the
file available for public use was stripped of all direct and indirect identifiers thus any risk
of third party disclosure is non-existent. Before posting the data files both the McArthur
Foundation researchers and ICPSR representatives performed a formal disclosure
arnalysis to prevent any accidental identification of individual subjects by third party
research investigators.

In order to be in compliance with MSU IRB’s requirements, this researcher has
submitted the IRB form of Certification for Use of an Approved Public Data File and
subsequently obtained response! certification (see Appendix) to use the data file for this

dissertation study.

70

Data and Safety Monitoring Plan

Because this study is a dissertation project thus does not meet the criteria of a
NIH-defined phase HI clinical trial, a data and safety monitoring board (DSMB) will not
be established. However, the investigator will be responsible for complying with all the
policies, procedures, and recommendations indicated by UCRIHS. Rigorous control of
data will be implemented and all necessary provisions will be made according to
UCRIHS requirements. This investigator will be responsible for ensuring that an
adequate data and safety monitoring plan is designed and implemented. As such the data
file will be stored on this investigators password protected computer and only this author
will have access to the information stored in the data file.

This research investigator has been certified since 2006 in the use of human

subjects’ protections and the current training expiration data will be on 09/18/2010.

Measures
Dependent Variables and their Indicators (Latent Endogenous)

Physical abuse

The dependent latent variable (endogenous) used for this dissertation research is
Physical Abuse. Physical abuse is defined as any direct or indirect action that affects the
physical survival, welfare or health of elderly, causing pain, unnecessary suffering or
health deficiency (Hawes, 2002). Specific manifestations of physical abuse: (a) injury to
the body of the older adult, being hit, kicked, slapped, pushed, etc., (b) inappropriate
restriction of mobility, including application of physical and chemical restraints, and (0)

sexual abuse and nonconsensual sexual involvement: including being forced, threatened,

7l

or deceived into sexual activities ranging from looking or touching to intercourse or rape
(Hawes, 2002).

The reflective indicators (manifest variables) used to measure the main
endogenous latent variable (Physical Abuse) as suggested by Schiamberg et al. (2009)
are:

Indicator 1: Question: “Ever physically assaulted?”
Coding: Binary variable: 1=No (absence of abuse); 2=Yes (presence of
abuse)
Variable scaling: scales were constructed by calculating the mean across
each set of items. The variable’s coefficient alpha was over .70.
Indicator 2: Question: “Ever sexually assaulted?”
Coding: Binary variable: 1=No (absence of abuse); 2=Yes (presence of
abuse). Variable scaling: scales were constructed by calculating the mean

across each set of items. The variable’s coefficiernt alpha was over .70.

Victim Adult Child Relationship Quality

The victim / adult child relationship latent variable captures the quality of relationship
between the elderly parent and the adult child caregiver. Using this key factor serves
several purposes in this study: (1) measnues how the relationship quality between the
adult child and her/ his elderly parent directly influences the risk of elder abuse (the
factors flmctions as an exogenous variable), (2) whether this latent variable mediates and

or moderates elder physical abuse (factor functions as an endogenous variable), (3) to

72

propose a future elder physical abuse research direction that links child abuse/ neglect to
physical abuse of the oldest of the old.

This construct is measured via five proxies as listed below. In addition this latent
construct will act as a mediator between the four exogenous latent variables and the
Physical Abuse construct. Analytic considerations, model conceptualization, and testing
for mediating effects will follow guidelines set forth by Baron & Kenny (1986).
Manifest indicators measnning the victim-abuser relationship are as follows:

Indicator 1: Question- “Rate current relationship with child”
Scaling: were consu'ucted by calculating the sum across each set of responses.
Positive items were recodes so that higher scores reflect higher sense of social
well- being.
Coding: variables are continuous and based on 11 values where 0 stands for worst
possible relationship TO 10 representing best relationship ever.

Indicator II: Question- “Children have difficulties in getting along with family
Members”
Scaling: were constructed by calculating the sum across each set of responses.
Positive items were recodes so that higher scores reflect higher sense of social
well- being
Coding: this variable is discrete dichotomous where 1 stands for No and 2 for the
Yes category.

Indicator III: Question- “Rate control over relationship with your child”

73

Scaling: were constructed by calculating the sum across each set of responses.
Positive items were recodes so that higher scores reflect higher sense of social
well- being.

Coding: This manifest indicator is considered continuous and based on 11 values
where 0 stands for either no relationship TO 10 that represents best relationship.

Indicator VI: Question- “Rate thouglnt/effort put into relationship with your child”.
Scaling: were constructed by calculating the sum across each set of responses.
Positive items were recodes so that higher scores reflect higher sense of social
well- being.

Coding: This manifest variable is continuous and based on 11 values where 0
stands for either no thought effort put into relationship TO 10 represents the most
thoughts/ efforts put into relationship.

Indicator V: Question- “Family life with children more negative” (possibly abusive)
Scaling: were constructed by calculating the sum across each set of responses. ‘
Positive items were recodes so that higher scores reflect higher sense of social
well- being.

Coding: This manifest indicator is categorical and based on four values where 1

stands for not true at all TO four that represents extremely true.
Independent (exogenous) Latent Variables and their Indicators

The exogenous latent variables in the model are as follows: Victim Health, Victim

Behavioral Problems, Adult Child Characteristics, and Social Isolation.

74

Victim Health

This construct captures the physical and mental well-being of the elderly person]
victim. Older adults in poor health require a geat deal of care which increases the
demands, dependency, and stress on the caregiver (Lachs & Pillemer, 1995). Victim’s
emotional and mental health conditions increases the risk for elder abuse (Wang et al.,
2000). In addition to the presence of victim psychological disorders, this study uses

individual negative affect variables as suggested by Mroczek et al (1988).

Manifest indicators measuring the Victim 's Health construct
Indicator 1: Question- “Were you diagnosed with Parkinson’s illness?”
The variable is coded as: 1- No (absent of condition) and 2—Yes (condition
present). Variable scaling: scales were constructed by calculating the mean across
each set of items.
Coding: Coding: Binary variable: 1=No (absence of abuse); 2=Yes (presence of
abuse).
Scaling: Variable is constructed by taking the total number of “Yes” responses to
the items
Indicator 11: Question: “Do you have a history (were diagnosed with in the past)
neurological disorders?”
Coding: Coding: Binary variable: 1=No (absence of abuse); 2=Yes (presence of
abuse).
Scaling: Variable is constructed by taking the total number of “Yes” responses to

the items.

75

Indicator III: Question- “Are you having any chronic! health/ medical condition?”
Coding: Coding: Binary variable: 1=No (absence of abuse); 2=Yes (presence of
abuse).

Scaling: Variable is constructed by taking the total number of “Yes” responses to
the items.

Indicator IV: Question- “Were you diagnosed with Anxiety Disorder?”

Coding: Coding: Binary variable: 1=No (absence of abuse); 2=Yes (presence of
abuse).

Scaling. Variable is constructed by taking the total number of “Yes” responses to
the items.

Indicator V: Question- “Did you feel sad and or depressed during the last two weeks?”
Coding: Coding: Binary variable: 1=No (absence of abuse); 2=Yes (presence of
abuse).

Scaling: Variable is constructed by taking the total number of “Yes” responses to
the items.

Indicator VI: Question- “Mental and enotional / health self evaluated”.

Coding: The variable is coded on a five point Likert scale as: 1= Excellent; 2=
Very Good; 3=Good; 4= Fair; and 5= Poor.

Scaling: scales were constructed by calculating the mean across each set

of items. Higher scores indicate geater levels of amplification. Reverse coding.

Indicator VII: Question- “Frequency of upset feelings during the last 30 days?”

76

Coding: The variable is coded on a five point Likert scale as follows: 1= All of
the time; 2= Most of the time; 3= Some of the time; 4= A little of the time; 5:
None of the time. Note: The variable is reverse coded.
Scaling: scales were constructed by calculating the mean across each set of items.
Higher scores indicate geater levels of amplification.

Indicator VIII: Question- “Frequency of angy feelings during the last 30 days?”
Coding: The variable is coded on a five point Likert scale as follows: 1= All of
the time; 2= Most of the time; 3= Some of the time; 4= A little of the time; 5=
None of the time. Note: The variable is reverse coded.
Scaling: scales were constructed by calculating the mean across each set of items.
Lower scores indicate geater levels of amplification.

Indicator XI: Question— “More or less than usual negative feelings?”
Coding: The variable is coded on a seven points Likert scale as follows: 1= A lot
less negative than usual; 2= Somewhat less negative than usual; 3= A little less
negative than usual; 4= About the same time as usual; 5= A little more negative
than usual; 6= Somewhat more negative than usual; 7= A lot more negative than
usual. Note: The variable is revese coded.
Scaling: scales were constructed by calculating the mean across each set of items.
Lower scores indicate geater levels of amplification.

Indicator X: Question- “Ever attended emotional problems support goup?”
Coding: The variable is coded as: 1=Yes (condition present); 2= No (absence of

condition).

77

Scaling: Variable is constructed by taking the total number of “Yes” responses to
the itens.
Source and studies using these variables include: Wang et al., 2000; Kessler et a1. 1999;

Kessler et al, 1998.

Victim Behavioral Problems
Althougln there is no 1miversally, or even commonly, accepted definition of what
is a behavior problem, there is a general ageemernt to what it refers to: (a) behavior that
goes to an extreme — behavior that is not slightly different fiom the usual, (b) a problem
that is chronic - one that does not quickly disappear, and (c) behavior that is unacceptable
because of social or cultural expectations (McGraw-Hill, 2002). The reflective indicators
measuring this exogenous latent variable as developed by Patrick et al. (2002), are:
Indicator 1: Question- “Sometimes I just like to hit someone”.
Coding: 1= False; 2= Somewhat false; 3= Somewhat true; 4= True.
Scaling: All scales are constructed by calculating the sum of the values of the
items. All items except ones marked with (R) were reverse-coded so that high
scores reflect higher standing in each dimension.
Indicator 11: Question- “When people insult me, I try to get even.”
Coding: 1= False; 2= Somewhat false; 3= Somewhat true; 4= True.
Sealing: All scales are constructed by calculating the sum of the values of the
items. All items except ones marked with (R) were reverse-coded so that high
scores reflect hi glner standing in each dimension.

Indicator IH: “When angy I am ready to hit someone”.

78

Coding: 1= False; 2= Somewhat false; 3= Somewhat true; 4: True.
Scaling: All scales are constructed by calculating the sum of the values of the
items. All items except ones marked with (R) were reverse—coded so that high

scores reflect higher standing in each dimension.

Adult Child Characteristics
Although there is no universally, or even commonly, accepted definition of what
. is a behavior problen, thee is a general ageemernt to what it refers to: (a) behavior that
goes to an extreme - behavior that is not slightly different fiom the usual, (b) a problem
that is chronic - one that does not quickly disappear, and (c) behavior that is unacceptable
because of social or cultural expectations (McGraw—Hill, 2002). Latent indicator variable
of abuse (adult child) characteristics are as follows:
Indicator 1: Question- “Alcohol/substance problems during the last 12 months?”
Coding: The variable is coded as: 1=No (condition absent); 2= Yes (condition
present).
Scaling: All scales are constructed by calculating the sum of the values ofthe
items. All items except ones marked with (R) were reverse-coded so that high
scores reflect higher standing in each dimension.
Indicator 11: Question— “Emotional/ psychological problems dming the last 12
months?”
Coding: The variable is coded as: 1=No (condition absent); 2= Yes (condition

present).

79

Scaling: All scales are constructed by calculating the sum of the values of the
items. All items except ones marked with (R) were reverse-coded so that high
scores reflect higher standing in each dimension.
Indicator 111: Question- “Financial problems during the last 12 months”

Coding: 1 True of you; 2 Somewhat true; 3 Somewhat false; 4 False.

Scaling: All scales are constructed by calculating the sum of the values of the
items. All items except ones marked with (R) were reverse-coded so that high
scores reflect higher standing in each dimension.

Somce and studies using these variables include: Patrick et al. 2002.

(Victim ’s) Social Isolation

This concept captures the virtual absence of interaction with others, outside of the
contacts required to perform basic life functions—e. g. food shopping, transportation,
work, and entertainment (McGraw-Hill, 2002). Victim’s social isolation construct was
measured via the following manifest indicator variables, as suggested by Keyes (1995):
Indicator 1: Question- “I do not feel I belong to community?”

Coding: 1 Strongly agree; 2 Somewhat agree; 3 A little agree; 4 Don’t know;

5 A little disagree; 6 Somewhat disagree; 7 Strongly disagree.

Scaling: were constructed by calculating the sum across each set of responses.

Positive items were recodes so that higher scores reflect higher sense of social

well-being.

Indicator 11: Question- “Few close friends to share concerns with?”

80

Coding: 1 Strongly agree; 2 Somewhat agree; 3 A little agree; 4 Don’t know;
5 A little disagree; 6 Somewhat disagree; 7 Strongly disagree.
Scaling: were constructed by calculating the sum across each set of responses.
Positive items were recodes so that higher scores reflect higher sense of social
well-being.

Indicator III: Question- “I don’t fit in with people and community?”
Coding: l Strongly agree; 2 Somewhat agree; 3 A little agree; 4 Don’t know;
5 A little disagree; 6 Somewhat disagree; 7 Strongly disagree.
Scaling: were constructed by calculating the sum across each set of responses.
Positive items were recodes so that higher scores reflect higher sense of social
well-being.

Indicator VI: Question- “I feel close to others in community?”
Coding: l Strongly agree; 2 Somewhat agree; 3 A little agree; 4 Don’t know;
5 A little disagree; 6 Somewhat disagree; 7 Strongly disagree.
Scaling: were constructed by calculating the sum across each set of responses.
Positive items were recodes so that higher scores reflect higher sense of social
well-being.

Indicator V: Question-“Frequency of contact with fi'iends?”
Coding: l Strongly agree; 2 Somewhat agree; 3 A little agree; 4 Don’t know;
5 A little disagree; 6 Somewhat disagree; 7 Strongly disagree.
Scaling: were constructed by calculating the sum across each set of responses.
Positive items were recodes so that higher scores reflect higher sense of social

well-being.

81

Victim Background Independent Variables

Acknowledging and understanding victim demographic risk factors might prove
usefirl for family and marriage therapist when conceptualizing cases involving physical
aggression towards the elderly. Often used and tested background information in the field
of marriage and family therapy research are variables such as age, gender, and marital
status (Holtzworth et al., 2002). Individual victim demographic variables are available in
the MIDUS II data base so they will an integral part of this dissertation model. .
Unfortunately this data set lacks relevant demographic information on the abuser (adult
child) therefore these important backgrormd variables can’t be part of the proposed
model. In this study the following manifest indicators were used as model background
variables to introduce and estimate the influence of certain demographic characteristics
on the primary endogenous factor:
Variable 1: Victim’s chronological age

Coding: Continuous variable values ranging from 65 to 89_years of age.
Variable H: Victim’s gender.

Coding: Nominal level binary variable with values of 1 = Male and 2 =Female
Variable III: Victim’s marital status.

Coding: Nominal level binary variable with values of 1= Not married and 2 =

married.

82

Proposed Path model

The proposed path model diagram for this study is based on the Ecological Bi-
Focal model of elder abuse developed by Schiamberg and Gans (2000). This path
diagram is equivalent to a set of model equations and distributional assumptions (most)
and also permits a ready communication of models. Figure 3.1 (provided below) outlines
the anticipated causal relationships amongst the model’s 31 manifest variables with their
respective six latent variables and the influence of the four exogmous variables on the
mediator and the outcome endogenous latent (elder abuse) variables. The latent variables
(represent concepts of main interest and are not directly observable or measurable) are
designated with an oval in the model. The 31 observed manifest variables (directly
measured and recorded in the MIDUS II data file) are represented with squares.
Assumed explanatory relationships between the latent variables are expressed via a one
way arrow and the unspecified relationships (correlation) between models’ exogenous
latent variables are designated with a two way curved arrow. Each manifest variable has a
residual error hence all 31 will receive a short one way arrow. The model’s endogenous
variable (Elder Physical Abuse) receives five one way arrows from the exogenous

variables indicating explanatory (also causal) relationships.

83

Figure 3.1. Full Path Model Relationships

v v v v v_v
v V V. v v

5'

   

  
   
 
     

Problems

Victim] (:3 if
Adult Child l‘
Relationship )
‘ _,_~,w__... Quality '
J
J
J
'Victim ,Ji‘
Behavioral
J
_J‘

 

it}. 3 a; ' Social
1;” / Isolation
Adult Child

' J
J .9

According to the social learning framework, individuals’ (in cm study the elderly
parent and the adult child) ineffective attempts to deal with internal and external
challenges result in conflictual relationship which leads to escalating conflicts between
them. However, the individual’s inability to deal with internal and external stressors does

not automatically lead to a dysfunctional and abusive relationship (Kamey & Bradbury,

1995). Consequently the relationship quality between the elder parent and adult child
could be conceptualized as a mediator between the individual’s internal and external
challenges and the act of elder physical abuse.

Given the nested structure of MIDUS H data and the multilevel nature of the
study’s hypotheses, a two levels modeling approach had to be employed wherein the
model presented above (full study model) contains a restricted model or mediation
(model 2) which will allow the testing of mediating effects that Victim/ Adult Child
Relationship Quality latent variable has on the Physical Abuse endogenous latent
construct. The removal of direct loading parameters of Victim Health, Victim Behavioral
Problems, Adult Child Problems and Social Isolation on the Elder Physical Abuse latent
construct will yield the restricted model (Figure 3.2). The estimation of model 2
parameters will follow along similar dimensions as in the case of the study’s firll model.

Shrout and Bolger (2002) suggested that the power to detect causal relationships between
the exogenous and endogenous constructs (if) M and M-) 1]) through a proximal

relationship in a mediated model (nested in this study) is often stronger than is the power
to detect direct relationships between these variables. Accordingly, if the Victim/ Adult
Child Relationship Quality variable is the mediator in the study’s model than we should
expect more robust (larger estimate values, lower corresponding error terms, and

significant T- values) parameter estimates and better model fit of the nested model.

85

Figure 3 .2 Restricted W(Medition) Path Model Relationships

I; I! I! I! I]

Victim/ @
- .l Adult Child /)\
,r-«W’ Relationship
Quality

\f’ \/ Victim

 
  
    

i‘: If. Vlctlm

 
     

“a, Behavioral 4
‘\ Problems —.J‘
K __J
J/n.__.\\\\ Q _J‘
/ Social

, I l ' l I
i ‘ so atlon
I __J‘

.1

arms... 4 a a

new

86

Specific Research Questions, Hypotheses, and Measures
1. Prevalence
1. What is the prevalence of elder physical abuse (in the sample) committed by an
adult child in the family/ community setting?
Data analysis: Univariate and descriptive analyses of sample statistics.
II. Immediate Older Adult/ Caregiver Context
2. Do victim characteristics influence the likelihood of elder physical abuse?
Hypothesis #1 — Victim health factors (physical, psychological, and emotional,
health problems) are positively related to (more severe health problems will
results in higher likelihood of physical abuse) elder physical abuse (see Figure
3.1 ).
Data analysis: Latent variable modeling with SEM/ LISREL.
3. Do victim behavioral problems influence elder physical abuse?
Hypothesis #2 — Victim behavioral problems will positively correlate with
(increase) elder physical abuse, (see Figure 3.1)
Data analysis: latent variable modeling using SEM/ LISREL.
III Victim/ Adult Child Relationship Quality
4. Does victim/ adult child relationship quality predicts elder physical abuse when
allowed to mediate between the model exogenous and endogenous constructs?
Hypothesis #3: Victim/ Adult Child Relationship Quality latent variable is
positively correlated with elder physical abuse when allowed to mediate
between the model’s exogenous latent factors (Victim’s Health, Victim

Behavioral Problem, Adult Child Characteristics, and Victim’s Social Isolation) '

87

and the Elder Physical Abuse construct. Worst relationship will increase the
likelihood of physical abuse. In addition, more serious victim health problems
and victim social isolation will correlate with worst victim! adult child
relationship quality (negative relationship), (see Figure 3.2).
Data analysis: Latent variable modeling using SEM/ LISREL.
III. Contexts beyond focal older adult/ caregiver (ecological model testing)
5. Does the victim’s social isolation influence elder physical abuse?
Hypothesis # 4: Social isolation will positively co-vary with elder physical abuse
when mediated by Victim/ Adult Child Relationship Quality otherwise its
predictive effects will be non-significant.
Data analysis: Latent variable modeling using SEM/ LISREL techniques.
6. Do adult child characteristics influence elder physical abuse?
Hypothesis #5 Adult Child Characteristics (psychological/ emotional, alcohol and
substance abuse, and financial problems) will directly correlate (positive
relations) with elder physical abuse (see Figure 3.1)

Data analysis: Latent variable modeling using SEM/ LISREL techniques

Data Analysis Plan
The initial focus in the data analytic plan will be on obtaining sample descriptive
characteristics and a contingency cross-tabulation table of specific victim living
arrangements with elder physical abuse. The next step will be to closely examine the
MIDUS II— 65 and older, data base for systematic patterns in missing values. For this

study it is assumed that missing value indicators hide true values for subjects and

88

variables, which are meaningful for analysis (Little & Rubin, 2002). That is, missing
entries in the data set mask underlying (true) values. Therefore this dissertation study will
use neither pairwise nor listwise deletion approach instead all missing data values will be
imputed with Expectation Maximization (EM algorithm) used by Little’s MCAR test
statistic (SYSTAT software).

Since conventional regression techniques cannot be used to estimate latent
variables with multiple indicators, LISREL (Linear Structural Relationships) modeling
for categorical (discrete) data will be used to test hypotheses two, three, four, and five.
LISREL modeling will allow a closer examination of various paths (relationships)
between latent variables as well between manifest and latent variables while rejecting the
unrealistic assumption that our indicators will exhibit perfect validity and reliability
(absence of measurement and structural errors). Mplus statistical computer program will
be employed to complete the covariance structure analysis necessary to test whether the
hypothesized relationships amongst latent variables and manifest indicators and their
respective are real and significant (Muthen & Muthen, 2007). The specifications of
directional arrows in the model are based upon a substantive theory expressed in the
Applied Ecological Bi-Focal Model for Elder Abuse as a rich set of causal conjectures,
(Schiamberg & Gans, 2000). Thus published reports provided theoretical and causal
grounds for the presence or absence of every directed arc in this path model.

Since this study uses recursive models, with no closed cycles formed by directed
paths, the path equations of recursive models satisfy the conditions for regression
equations, thus interpretation of findings will be relatively straightforward. A sufficient

condition for the parameters of a measurement model to be identified (except for scaling)

89

is the condition that the factor loadings form independent clusters. This requires each
latent variable or common factor to have at least two pure indicators if the factors are
correlated and at least three if they are not (McDonald & Ho, 2002). The model’s latent
variables (both endogenous and exogenous) used in this dissertations study have at least
two pure (overwhelmingly three or more) and correlated indicators thus it meets
requirements as suggested by McDonald and Ho (2002).

The independence model (also called the null model), which assumes zero
population covariances among the observed variables or complete independence amongst
all variables, will be used as a baseline model to evaluate how well this doctoral
candidate’s model fits the MIDUS II- 65 and older-. data. Ifthe model accotmts for the
data well (that is for the observed covariances) and correctly estimates causal versus non-
causal aspects of observed correlations then it should be somewhere in between the
independence model and the saturated model (a model in which the number of estimated
parameters equals the number of covariances amongst observed variables). As proposed
by Muthen et al (2007), a set of goodness-of-fit indices (chi-square statistics, RMSEA,
TLI, and CFI) will be used to determine the degree to which this model as a whole is

consistent with the empirical data at hand (Diamantopoulos & Siguaw, 2006).

Data Analyses Rationale

The purpose of this study is to establish meaningful relationships between the
outcome variables (physical and sexual abuse) and their explanatory variables (e. g.
individually measurable health or behavioral problems variables) and to predict the

likelihood of elder physical abuse (study’s endogenous construct) given a set of

90

exogenous constructs based on elder abuse theories. Since the study proposes to predict
the likelihood of elder physical abuse some form of statistical regression analysis must be
used. The most general kind of structural equations model is a structural regression (SR)
model which is also called a LISREL model (Kline, 2005) is used in this study. The SR
model is a synthesis of structural and measurement models and it allows the testing of
direct and indirect causal effects between latent constructs, reciprocity, measurement
errors in both the dependent and independent variables, and interdependence (J oreskog &
Sorbom, 2001). This allows researchers to specify constructs related to a model and then
to simultaneously “test the plausibility of hypothetical assertions about potential
interrelationships among the constructs as well as their relationships to the indicators or
measures assessing them” (Raykov & Marcoulides, 2006, p. 1).

The assumptions made under the general LISREL framework, which are as
follows: (a) multivariate normality (normal data distribution, zero means on observed
variables or mean centered variables), (b) linearity (linear explanatory relationship
between exogenous and endogenous variables), (c) data should be free of outliers, (d)
non-spurious relationships (obse'ved covariance matrix must be true), (e) uncorrelated
error terms, (f) uncorrelated error terms with predictors in X (design matrix), and (3) LB

is invertible (Raykov & Marcoulides, 2006), will be applied to this study.

Mathematical definitions of the dissertation model
The general LISREL model for single samples continuous dependent variables is
defined by the following equations:

a)MeasurementpartintheY:Y= v + A n + r ,or

91

Y1 @hysical abuse) = 1 IF] + E],

Y2 (sexual abuse) = 12F1 + E2, and
b) Structuralpartzn=a+Bn+§where
a= intercept
A = factor loadings,
F = factors
11= random vector of latent dependent variables
C: vector of structural equations errors
a: vector of measurement errors in Y (including E1 & E2)
B = matrix of structural regression coefficients (q x q).
A = matrix of factor loadings (p x q)
Given the discrete natme (measurement levels) of both manifest dependent variables
(physical abuse and sexual abuse) used in this study, they may exhibit more than minor
violations of normality therefore the general LISREL modeling equation will not yield
accurate and reliable estimates. Past research studies routinely employed statistical
methods designed for continuous (at least ratio measurement level) data even if their
dependent variables were measured on categorical or nominal levels (Raykov &
Marcoulides, 2006). Ignoring data measurement levels yields potentially misleading and
serious biases in parameter estimates, incorrect standard errors (thus T-Values as well),
and serious biases in calculations of model fit and parameter restrictions (Raykov &
Marcoulides, 2006). Muthen and Muthen (1984) worked out an alternative statistical
modeling approach that is better suited for studies with dependent categorical (including

binary) variables as it is the case here. They postulated that with a large enough sample

92

size (above 200) and arbitrary distribution of Y’s the following model fit function would
yield accurate model fit and parameter estimates (Muthén and Muthén,1984):

F(ADF) = (s - 0(1))"W (8 - 0(7)),

where s and o (v ) are the stnmg-out vectors of non-redundant elements of S and 2( v ),

and W is a weight matrix (the inverse of the asymptotic covariance matrix of elements of
empirical covariance matrix S).

ADF = ‘asymptotically distribution-fiee’ method; the most general of all methods for
model fitting. The ADF function will be used to assess model fit and to estimate
parameters.

Since neither of the discrete manifest dependent indicators used in this study
posses metric scale properties the assmnption of linearity becomes an issue as well.
Therefore we have to use the corresponding underlying continuous variables instead to
build linear relationships. The use of the underlying metric variable is often both
plausible and meaningful in empirical research and this assumption is frequently invoked
within other analytic contexts outside of SEM (e.g., use of limited dependent or
categorical variables in medicine, psychology or economics) (Raykov & Marcoulides,
2006). Consequently this researcher will make the assumption that for each categorical
obse'ved dependent variable thee is an lmderlying latent continuous variable that has a
normal distribution. This assumption is often convenient and will not be limiting the
generalizability of results (Raykov, 2005; Raykov & Marcoulides, 2006).

In addition, the continuous latent variable (elder physical abuse) can only be
crudely measured with our instrumentation (MIDUS II) and will take on a value of 0 (no

abuse) or 1 (abuse). Hence the binary variables (manifest) physical abuse and sexual

93

abuse are the result of this measurenent and are related to underlying continuous variable
as:

Y = 1, if Y* > 1:, 0, otherwise , whee t is an unknown threshold.

Furthermore, this study model stipulates that the random variable vector Y is related to an
anothe random variable vector, X, as:

Y“ = 1: X + 5, (Raykov, 2005; Raykov & Marcoulides, 2006).

where 11: is an unknown parameter and 6 is a residual term (a random variable in its
own right, with a mean of 0 and lmrelated to X).

The probability of the outcome 1 (presence of abuse) on the Y variables given the
predictors could be expressed as:

P(Y = 1 IX = x) = F( a + B x) , (Raykov, 2005; Raykov & Marcoulides, 2006).
Where a and B are unknown constants (parameters) and F is either (a) the cumulative
distribution function (CDF) of the standard normal distribution, or (b) the CDF of the
standard logistic distribution.

The underlying continuous latent variable can be expressed as follows:
P(Y=1|X=x)=P(Y*> r |x)= 1 —P(Y* < 1.‘ Ix), (Raykov,2005; Raykov&
Marcoulides, 2006), where 1: is an tmknown threshold.

Finally, the underlying latent variable modeled as a linear function of predictors in this

study can be expressed as:
P(Y=1|X1=x1andX2=x2)= (P(a + B 1 x1 + 32x2+ B3x3...+328x28),
where (I) is the CDF (cumulative distribution function) of the standard normal

distribution, a and B ' s are our unknown parameters that will have to be estimated.

94

Dealing with possible normality violations

Given the discrete nature of both manifest dependent variables (physical abuse
and sexual abuse) used in this study, they may exhibit more than minor violations of
normality therefore this general LISREL equation will not yield accurate and reliable
estimates. For cases where normality does not hold yet the sample size is large enough
(over 200) there is an alternative method for parameter estimation and model testing,
(Raykov, 2005; Raykov & Marcoulides, 2006). With a large sample size and arbitrary
distributions in the Y’s the following fit firnction will yield accurate estimates and model

fit:
I"(MDF)=(S- 0(7))'W(S- 0(7)),
where s and o (v ) are the strung-out vectors of non-redundant elements of S and Z ( y ),

and W is a weight matrix (the inverse of the asymptotic covariance matrix of elements of
empirical covariance matrix S).
ADF = ‘asymptotically distribution-fiee’ method; the most general of all methods for

model fitting.

Relationship between latent variables and their predictors in the model

For the purposes of this study, the assumption that for each categorical observed
dependent variable there is an underlying latent continuous variable is made (Raykov,
2005; Raykov & Marcoulides, 2006). Therefore the binary manifest variables (physical
abuse and sexual abuse) are the result of this measurement and are related to lmderlying

continuous variable as:

Y = 1, if Y“ > r , 0, otherwise, where 1: is an unknown threshold.

95

The probability of the outcome 1 (presence of abuse) on the Y variables giVen the

predictors could be expressed as:

P(Y = 1 | X = x) = F( a + B x) , (Raykov, 2005; Raykov & Marcoulides, 2006).
Where a and B are lmknown constants (parameters) and F is either (a) the cumulative
distribution function (CDF) of the standard normal distribution, or (b) the CDF of the
standard logistic distribution.

The Imderlying continuous latent variable can be expressed as follows:
P(Y=1|X=x)=P(Y*> 1: |x)=1—P(Y*< 1: Ix),

(Raykov, 2005 ; Raykov & Marcoulides, 2006), whee r is an unknown threshold.

Finally, the underlying latent variable modeled as a linear function of predictors in this

study can be expressed as:

P(Y=l |X1=x1andX2=x2)= (P(a + B 1 x1 + 32x2+ 83x3... +828x28),
where (I) is the CDF (cumulative distribution function) of the standard normal

distribution, a and B ' s are unknown constants (parameters).

96

CHAPTER IV: RESULTS
Descriptive Statistics

This study empirically examined the structural relationships between four
hypothesized latent exogerous factors (Victim ’s Health, Victim ’s Behavioral Problems,
Victim 's Social Isolation, and the Adult Child ’s Characteristics) and the main
endogenous latent variable (Elder Physical Abuse). This approach is based on the
ecological theory of elder abuse in the community as proposed by Schiambeg and Gans
(2000) and the related recommendations for elder abuse theory and theory testing by the
NRC Panel, NRC,2003). LISREL data modeling with ordinal variables was used to test
hypothesized relationships between theoretical constructs at various ecological levels. A ‘
second covariance structure data model was employed to examine the mediating role of
the Victim/ Adult Child Relationship Quality factor in the process of elder abuse.

The first part of this chapter will describe the dissertation study sample characteristics
from a bi-focal perspective, including both eldely parent and adult child demographic
fiequercies. In addition, this subsection will contain results of sample based univariate
statistics on the manifest dependent variables and missing data analyses. The following
subsection will provide an overview of analyses results obtained with LISREL data
modeling. Results of model fit will be presented and discussed in terms of how well they
reproduce the observed data. In addition, the measurement parts of both models will be
examined and evaluated to detemine the validity and reliability of measures used.
Finally, the hypotheses testing results, powe assessments, and additional model

diagnostics will be presented at the end of this chapte.

97

Sample Based Descriptive Statistics

Older Adult Descriptive Statistics

This section will provide an answer to the research question: What is the
prevalence of elder physical abuse (in the sample) committed by an adult child in the
family/ community setting? For the purposes of this dissertation study 203 olde adults
respondents fi'om the MIDUS H survey meet the dual selection criteria of: (1) being 65
years or older at the time of interviews and (2) residing with an adult child caregive. Of
the 203 total sample of elderly respondents 63.1% (n = 128) participants were females and
36.9% (n = 75) were males, (see Figure 4.1). This gender value distribution is unique to
the sample used for this dissertation study because this variable had diffeent distributional

(53% females vesus 47% males) properties in the original MIDUS II sample (n=4,963).

98

Figure 4.1. Gender Univariate Result

 

 

 

Gender

Legend

‘50 Cat 1: MALE
Cat 2: FEMALE

120

90

75
Mean = 1.631

45 SD. = 0.484

30

15 —— - X
Bar Chart

The age manifest variable was considered continuous (normal distribution) with a
mean of 73.296 and standard deviation of 7.001 . Given the selection criteia, (variable
range between 65 to 90 years), the data distribution was larger on the left distribution tail,
(see Figure 4.2). These findings make sense since age is strongly correlated (positive
relationship) with the presence of chronic and debilitating illnesses that result in death
(whether natural or suicide) in the oldest of the old population (65 years or olde). In

addition, Figure 4.2 clearly indicates a terminal decline in health after reaching age 87.

Figure 4.2. Age Univariate Results
Age

70

63

Mean = 73296
N = 203

S.D. = 7.001 ‘

 

 

65.6 68.7 71.7 74.8 77.9 81.0 84.0 87.1 90.2
Histogram

The skewness (0.564) and kurtosis (-0.902) of the Age variable were within
expected limits given a sharp increase in mortality rates due to physical, psychological,
emotional illnesses after age 65.

Of the total 203 respondents 63.5% (n = 129) reported that they wee married and
the remaining 36.5% (n = 74) wee either never married (11 = 1), or separated (e.g.

divorced, separated, or widowed). It is important to note that in the dissetation sample a

100

high percentage of abused olde adults (63.5%) wee still married and resided with a
spouse. These pmportions are even higher (71% married versus 29% not married), as
expected, in the complete (N= 4,963) MIDUS 11 sample. These sample based statistics
appear to suggest that since a higher percentage of olde adults wee married they
probably still lived in their own home. Consequently more adult children lived in their
eldeiy parents’ home than older single adults at their children’s residence. These results
might be due to individual adult child characteristics (e.g. psychological/ emotional,
substance abuse, and financial problems). In addition, these results raise more questions
(e. g. whether only one, the respondert, or both parents wee abused by the adult child)
than we have answers at this stage. Statistieal data modeling, analyses, and structural
results presented in the second pet of this chapter might provide answers to these
questions.

Regarding the racial make up of the sample, 88.7% (n = 180) reported that they
were Caucasians, 3.9% (n = 8) wee Afiican American, 1.5% (n = 3), and 5.4% (n = 11)
indicated that their race was “other”. Denographic results wee similar in the unrestricted
MIDUS 11 sample whee 90% indicated that they were Caucasians, 4% Afiican
Americans, and 5% othe. This skewed population sample demographic results might be
due to the tmcertainty introduced by the MIDUS investigators because they used ove
five variables to measure this one dimension, thus possibly confusing respondents.
Consequently the race variable is not an accurate indicator to measure this very important
denographic characteristic in this study. This issue will be addressed in the research -

limitations section.

101

Regarding education level, the largest pecentage were high school graduates (n
= 64, 63%), while over 18.3% (n = 37) reported of having obtained eithe a Bachelors or
Mastes Degree. A significant pecentage, 16.6% (n = 34), of respondents never
completed the requirenents for high school graduation. 2% (n = 4) of respondents had a
PhD level degree. Regarding religious preference, 86.8% (n = 161) of respondents
identified themselves as belonging to a Christian denomination, 3.6% (n = 7) identified
themselves as Jewish, and 3.1% (n = 6) indicated that they wee Mormons. 8.7% (n = 17)
of respondents indicated no religious preference. Regarding employment status, 47.3% (n
= 96) reported that they were retired while 40.4% (n = 82) checked the employment or
self—enployed boxes. Since the victim’s psychological/ emotional and physical health are
important indicators of elder abuse, it is vital to provide frequencies in this are as well.
Respondents rated mostly their physical health as good to very good [mean = 7.21 on a
scale fiom 1 (poor) to 10 (excellent)] and their psychological/ emotional health as fair to

good [mean = 2.28, on a scale from 1(poor) to 5 (excellent)], (See Figure 4.3).

102

Figure 4.3. Physical, Psychological Health Sliced by Age

 

 

lbs ndont's .
IeiooIrzno goo Iaino
Iaino Inoo Iwoo Iuno
Iano Inna .8100 Iowa
Ioooo I75oo Iano Iddast
Ieno moo rim
Irono Inno Iano
Inna Inno Iaino

 

 

 

Phsslouoouls

Montallomotlonal health self-evaluated

 

Physical health self-evaluated

It can be concluded that in this sample olde adults in general had a more positive

peeeption of their physical health than their psychological! emotional well-being. For

additional dangraphicsinformation consult Table 4.1.

103

Table 4.1. Elderly Parent Characteristics

 

 

 

Demographic s Frequencies Percentages

Age 65-70 years old: 95 46.8%
71-75 years old: 33 16.3%
76-80 years old: 37 18.2%
81-85 years old: 22 10.8%
86 and above: 16 7.9%

Gender Male: 75 36.9%
Fenale: 128 63.1%

Race Caucasian: 180 88.7%
Afiican American: 8 3.9%
Native American: 3 1.5%
Other: 11 54%

Marital Status Married: 74 36.5%
Not Married: 129 635%

Education Not high school: 34 16.7%
High school: 64 63%
BA/ Masters: 37 18.3 %
PhD or Equiv.: 4 2%

 

 

 

104

 

Adult Child (Caregiver) Descriptive Statistics

Although adult children (caregivers) of olde adults did not complete the MIDUS
11 surveys, nevertheless information provided by their parents could be used to assess
unique characteristics such as substance abuse, or financial the stress and burden that
might play a role in the process of elde abuse, (See Table 4.2). In this dissertation
sample, 24.9% (n = 43) of olde adults reported that their adult children suffered fiom a
major chronic illness during the last 12 months and 27% (n = 47) had a fiequert minor
illnesses. In addition, 38.3% (n = 69) of adult child caregivers suffered fiom fiequent
psychological and emotional problems that could have a negative impact on their care
delivery approach and potential behavioral problems. A total of 21% (n = 38) of adult
child caregivers appeared to have had a substance/ alcohol abuse problem during the last
12 months and a significant proportion, 49.4% (n = 88), had financial difficulties for the
last 12 months. Regarding relationship difficulties, a total of 3 l .1% (n = 56) of adult care
giving children had marital problens in their own homes, 15.2% (n = 27) had work
related relationship issues, and 25.6% (n = 46) had difficulties in finding and or keeping
an employment. Moreover, 15.2% (n = 27) of adult caregives had significant legal and
systemic problems during the last 12 months. Out of 203 eldely participants, 12.9% (n =
23) reported that their children had behavioral problems, including problems associated
with others. As well, a total of 6.4% (n = 130) olde adults reported that family

inteaction with children at home terded to be negative, (See Table 4.2).

105

Table 4.2. Adult Child Characteristics

 

 

 

Adult Child Characteristics Frequencies Percentage
Suffering fiom Major Physical YES - 43 Yes - 24.9%
Illness NO -160 No -75.1%
Suffering fiom Minor Physical
Illness YES -47 YES - 27%
NO - 156 NO - 73%
Psychological/ Emotional YES - 69 YES - 38.3%
Illnesses NO - 34 NO - 61.7%
Substance Abuse Problens YES — 38 YES - 21%
NO - 165 NO - 79%
Marital Problems YES—56' YES-31.1%
NO- 147 NO - 68.9%
Financial Problems YES — 88 YES — 49.4%
NO - 115 NO - 50.6%
Employment Problems YES — 27 YES — 25 .6%
NO - 176 NO - 74.4%
Total 203 100%

 

 

 

106

 

Finally, 12.2% (n = 45) reported that it was difficult for them to maintain close
relationships, 7.2% (n = 28) indicated that they did not fit in their community, and 10.3%
(n = 35) felt that they had few friends with whom they could share their concens. In
addition, 9.8% (n = 37) of eldely participants reported that they had no warm and

trusting relationships with others in their family.

Manifest Dependent Variable Frequencies and Distributions: Prevalence of Physical
Abuse

Of 203 respondents, 5.9% (n = 12) reported that they wee physically abused and
9.4% (n = 19) indicated that they wee sexually assaulted. In total, 15% (n=31) of eldely
respondents indicated that they were victims of physical abuse. In the complete MIDUS
II sample 6% of respondents indicated that they wee physically abused at some point in
time and 7% indicated that they wee sexually abused. These results are in accordance
with conclusions reached in historical elde abuse prevalence where physical abuse
ranged fiom 2% to 10% (Lachs et a1, 2004).

These two dependent variables are discrete thus their distributional properties are

not considered norm] (See Figure 4.4 and 4.5).

107

Figure 4.4. Elde Physical Abuse Univariate Results

 

 

 

Physical Assault

1 Legend
220
198 CAT 1: NO
176 Cat 2: YES
154
132
110
88
66
44

» Mean = 1.059
22. n N = 203

0 1 2 S.D. =O.236
Bar Chart

108

Figure 4.5. Elder Sexual Abuse Univariate Results

 

 

 

Sexual Assault
Legend
210
. Cat 1: NO
189 Cat 2: YES
168
l
147
126
105
34
l
63
42
21 N = 203 §
IMean = 1.094
0 = ,
1 2 S.D. 0.292
Bar Chart

‘ Both manifest outcome variables have a large data value pile up on the left side of
distribution (e. g. response No). Since the assumption of normality will not be invoked in
the data analysis, dealing with data distribution skewness and kurtosis is not necessary.
Specific data related information on the outcome manifest indicators is included in Table.

4.3.

109

Table 4.3. Univariate propeties of manifest dependent variables.

 

 

 

Variables Mean Standard Skewness Kurtosis
Deviation
Elder Physical Abuse
1.094 0.236 3.767 9.310
Physical Abuse
1.097 0.292 2.811 5.963
Sexual Abuse

 

 

 

 

 

 

Given that both manifest dependent variables (physically and sexually abused) are '
discrete (dichotomous) with binary coding of (1 = No; 2 = Yes) and do not represent
measuremerts on an interval or ratio scales, they cannot be treated as variables with
interval scale propeties. Furthemore, computing the appropriate covariance matrix
(polychoric) and subsequently analyzing it with eithe the maximum likelihood (ML) or
the genealized least squares (GLS) method may yield distorted pararnete estimates,
incorrect standard erors and goodness of fit indices in the model '(J oreskog & Sorbom,
2001). Therefore the weighted least squares parameter estimator using a diagonal weight
matrix (full weight matrix) with standard errors, mean, and variance adjusted chi-square
test statistics was used, in addition to the delta (logit regression) pararneterization option

(Muthen & Mather, 2007).

Results of Missing Data Modeling
Larger data sets such as the MIDUS II routinely have missing values due to such
factors as equipment failure, fatigue of respondents, or attrition of cases. While few

missing values in a large data file such as the MIDUS 11 might be of little concern

110

however a substantial ntunbcr of missing observation pose a problems (Allison, 2001).
For the purposes of this study, values wee considered missing if they met any of the
following criteria: (1) no response provided by the participant, blank space left, (2) whee
available the no response option was marked, (3) if more than one response was provided
on a given item (Allison, 2001).

Missing data patterns wee statistically modeled to detemine whether data wee
missing at random or not (Little & Rubin, 2002). Based upon Little’s MCAR test results
(cg p-valuc large than 0.05 therein rejection of the null hypothesis (Ho: Thee is a
pattern to missing data) it was detemined that data wee missing at random in the
dissertation sample. In addition, missing data was within acceptable statistical toleances.
In this study no individual variable had more than 10% of missing values. Missing data
was within acceptable statistical toleances. In this study no individual variable had more
than 10% of missing values. 95.7% (n=6219) of response values were complete and only
4.3% (n=277) of values had missing data. The univariate analysis of missing variables
yielded a mean of 7.21 14 and standard deviation of 1.7273 statistics. Data ranges wee
checked for each variable enteed (n=203) to ensure that all data wee enteed within the
prescribed ranges. Of the ove 12,000 cells ranges examined, slightly over 4% of the
cases had at least one value outside the delineated variable ranges. These values wee
marked missing and treated as such. The statistical modeling of missing data pattens and
results of the Little's MCAR tests yielded highly non-significant chi-square values such
that the null hypothesis (H0: data missing at random) was not rejected. Estimated new
values were. imputed using the SYTSTAT statistical program. Results suggest that data

values wee missing at random with no identifiable pattern for missing data).

111

Mathematical equations used to compute estimates
The structural relationships between the latent factors modeled in this study will
be given by the following mathematical equations:
P(Y=1|X=x)=P(Y* > 'c |x)= l —P(Y* < r Ix), (Raykov, 2005; Raykov&
Marcoulides, 2006), whee 1: is an unknown threshold, Y* is the undelying latent
normal variable, Y is a random vector of criterion variables, and X is the random vector

of predictor variables. Therefore the Elder Physical Abuse endogenous latent variable,

modeled as a linear function of its predictors can be expressed as:

P(Y=] |X1=xl andX2=x2)= @(a + 81 x1+ 32x2+ B3x3...+828x28),

whee (I) is the CDF (cumulative distribution function) of the standard normal
distribution, a and B ' s are unknown constants (parameters).

An altenative method for formally presenting the dissertation model involved taking the
invese fiom of both sides of the above equation (note that the invese function of 4’

exists, since the CDF’s of the standard normal like of the logistic distribution are

I monotonically strictly increasing and invetible) and expressing the result as:

-1
(I) [P(Y=] |X1=xlandX2=x2)]= 011+ lel+ 32x2+83x3...+328x28

' LISREL Data Modeling Results: Full and Restricted Models
The reporting of results will conform to guidelines set forth by Raykov et a1
(1991) and McDonald ct al (2002). For example, the analysis of the MIDUS II covariance

matrix reflects the theoretical framework of the Ecological Bi-Focal Model for Elde

112

Abuse fiamcwork (Schiamberg & Gans, 2000), in accordance with the statistical/
mathenatical theories discussed above. Two LISREL data models, the full model and a
restricted model, wee designed to allow the enpirical examination of the assumptions
and conclusions of the Ecological Bi-Focal Model for Elde Abuse. The
conceptualization and specification of both the fill] and restricted models were similar,
with the exception that the latter model having four restrictions (e. g. direct paths fiom the
exogenous variables to tlnc endogenous factor restricted).

The restricted model, whee Ho also holds, was nested in the full fitted model. The
null hypothesis, Ho, states that “There exist (admissible) paramete values with which the
model reproduces perfectly the population covariance matrix.” That is, in the paramete
space (of all meaningful values of all parameters), thee exists a point y such that E =
2‘. ( y ); whee 2 denotes the population covariance matrix; H1: 2 > 0 (e.g., 2 is
positive definite). The input readings teminated normally and both models wee
identified and converged yielding admissible solutions. No Heywood cases (e. g.
parameter estimates with illogical values such as negative variance estimates and or
correlation between factor and indicator with an absolute value greate than one) wee
present in the covariance matrix. Weighted least square paramete estimator (WLSMV)
was used. Six first order factors were extracted fiom the covariance matrix and all had
Eigcn values over 2.000 and their position wee above the elbow on the see plot. The

correlation matrix of six factors is presented below (T able 4.4).

113

Table 4.4. Six Factors Correlation Matrix

 

Victim’s Health

Social Isolation

Adult Child Behav. Problems
Victim! Abuse Relationship
Victim Behavioral Problems

Elde Physical Abuse

 

1.000 .

0.403

0.381

0.567

0.331

0.693

1.000

0.187

0.538

0.489

0.153

1.000
0.405
0.001

0.566

1.000
0.204

0.468

1.000

0.227 1.000

 

Attributes Loadings in Both Models

Model attributes loading on the first factor, Victim ’s Health, were: “Parkinson

disease”, “History of neurological disordes”, “Chronic health! medical conditions”,

“Anxiety disorde”, “Felt sad/ depressed for over two weeks during the last year”,

“Mental/ enotional health self-evaluated”, “Felt upset frequency”, “F clt angry

frequency”, “More or less than usual negative feelings”, and “Eve attended emotional

problems group”. Attributes loadings on the Social Isolation factor were: ”Do not feel I

belong to community”, “Few close fiiends to share concern with”, “Don’t fit in with

people and community”, ”Feel close to others in community”, and “Frequency of contact

with fiiends”. Factorial loadings on the Victim Behavioral Problems wee: “Sometimes

just like to hit someone”, “When people insult me, try to get even”, and “When angry I

am ready to hit someone”. Attributes loadings on the Adult Child Characteristics wee:

“Children financial problens”, “Children alcohol / substance abuse problems”, and

“Children psychological/ enotional problems”. Attributes loadings on the Victim/ Abuser

114

 

Relationship were: “Ratccurrcnt relationship with your child”, “Child has difficulty
getting along with others”, “Rate control ove the relationship with your child”, “Rate
thought/ effort put into relationship with your child”, and “Family life with your child is
more negative”. Finally, both the physical abuse and sexual abuse variables loaded on the

dependent latent variable, Elder Physical Abuse.

The Full Model

The full dissetation model converged afie few iterations and yielded adnnissiblc
estimates. The overall full model fit was adequate (x2 = 83.117; with 69 degrees of
fi'eedom; p = 0.1182). Based upon the large p-value (p > 0.05) the model’s null
hypothesis [1-10: 2‘. = 2 (y )] is not rejected, implying that the model fits more than
- adequately in the population. Since the chi-square statistics is sensitive to sample size and
violations of the multivariate normality assumptions this index should be regarded as a
goodness (or more accurately badness) of fit measure rathe than a test statistics
(Diamantopoulos & Siquaw, 2006). A rule of thumb when evaluating this mm is that
a large x2 —value corresponds to bad fit and a small x2 —value to good fit when compared
to the corresponding degrees of fieedom. The full models chi-square value (83.835) is
consideably less than twice the degrees of freedom (2 times 69) thecforc it can be
concluded that the model as a whole adequately fits the population covariance matrix.

The comparative fit index (CF I) assesses the relative improvement of the
dissetation model compared with a baseline model. The baseline model assumes zeo
population covariances among the manifest variables. The baseline chi-square was

344.138, with 56 degrees of freedom, and a highly sigrnificant p value (p = O .000). If we

115

compare the baseline chi-square value (344.138) to the full models clni-squarc value
(95.329), it is evident that the full model is a significant improvement over the null
model. This yields a significant CFI of 0.953 which indicates a good fit.

The Tucker-Lewis index (T Ll) represents the dimension at which the model being
tested falls on a continuum ranging fiom an unconstrained null model to perfect fit of the
hypothesized model while correcting for model complexity (meaning that it favors a
parsimonious model). The full model had a TLI of .960, indicating a good model fit.
Literature on structural equation modeling recommends relying upon the CF I and TLI for
model assessment (Diamantopoulos & Siquaw, 2006). Since botln indices wee
significantly above 0.900, theeforc it could be concluded that the fill] model as a whole
is consistent with the empirical data at hand.

The Root Mean Square of Approximation (RMSEA) is a population based index

and it focuses on the discrepancy between 2 and 2 (y ). The RMSEA index indicates

how well the model, with unkrnown but Optimally chosen paramete values, would fit the
p0pulation covariance matrix if it wee available (Diamantopoulos & Siquaw, 2006).
For the fill] model, the RMSEA (0.032) indicates an adequate fit.

In addition, the RMSEA index can be used to assess the powe of the dissertation
model and answe the important question: If the fit is good for the population (RMSEA =
.032) then do we have a high probability with this sample size to reject a hypotlnesis that
is bad? Consulting the “Power of a test of poor fit and sample sizes needed for powers of
.80 and .90” table provided by (MacCallum ct al., 1996) it becomes evident that our
sample size (203) required for 67 degees of freedom exceeded the number of

observations needed for a powe of 0.80.

116

Measurement Dimension of the Full Model: Validity and Reliability

To evaluate the adequacy and theoretical significance of the hill model was first
necessary to examine the relationships between the latent variables and their indicators.
Unless we are convinced that our measmemcnts can be trusted, any assessment of
theoretical and substantive relations of interest will be problematic at best. Thecforc an
evaluation of the measurement model must precede the evaluation of structural part. The
aim bee is to detemine the validity and reliability of the measures used to represent the
constructs of interest The validity of manifest indicators can be ascetained by examining
the magnitude and significance of the model paths between the latent variables and their
manifest variables. In order to be a valid measure of a theoretical construct, the irndicator
must substantially load on its respective latent variable and its magnitude must
significantly diffe fiom zero. The fill] model factorial loadings on the respective latent

variables are provided below (Table 4. 5).

117

Table 4.5. Full Model Factor Loadings

 

 

 

. ' _ . Loadings T- Statistical
Mamfest Variables Measuring Exogenous Variables Estimates Values Significance
Victim 's
Health
History of Parkinsons 1.000 N/A N/A
History of Neurological Dis. 0.439 2.329 0.020
Clnronic Medical Conditions 0.710 2.247 0.025
Anxiety Disorde 0.926 3.886 0.000
Emotional/ Psychological Measures
Felt Depressed for 2+wks 0383 5.1 16 0.000
Mental Emotional Health ' -1.340* -5.396 0.000
Felt Upset Frequency 0.788 5.455 0.000
Felt Angry Frequency -0.502* -4.612 0.000
Negative Feelings -0.503* -4.151 0.000
Ever Attended Emotional Group -1.130* —6.125 0.000
Victim '5
Behavioral
Problems
1.000 N/A N/A
Sometimes Just Like to Hit Someone
1.192 4.587 0.000
When insulted ITryto Get Even
-0.319 -3 .249 0.001
When Angry I am Ready to Hit
Victim is
Social
Isolation
Do Not Feel I Belong to Community 1.00 N/A N/A

 

 

 

 

118

 

Table 4.5 Continued

 

 

 

Loadings T- Statistical
Manifest Variables Measuring Exogenous Variables Estimates Values Significance

Social

Isolation
Few Close Friend 1.505 4.349 0.000
Don’t Fit in with People and Community 1,213 4,815 0,000
Feel Close to Others in Community -0.489* -3.697 0.000
Frequency of Contac with Friends 0376 2,750 0.006

Adult

Child

Charater.
Children Emotional/Psych Problems 1.00 N/A N/A
Children Ach Substance Abuse Problems 0.864 13.797 0.000
Children Financial Problems 0.929 11.303 0.000

Victim/

Child

Relat.
Rate Relationship with your Child 1.00 N/A N/A
Child Difficult to Get Along With -O.247* -2.007 0.048
Rate the Thought/Effort you Put into Relationship 1.661 7.625 0000
Rate Control over Relationship w. Child 0354 1.966 0050
Family Life with Child is More Negative '0-308" '6-659 0-000

Physical

Abuse
Ever Physically Abused 1,00 N/A N/A
Ever Sexually Abused 1.154 14.412 0.000

 

 

 

 

* Reverse coding

119

 

Inspection of Table 4.5 reveals that all parameter estimates are significantly different
from zero in the population as indicated by their t-values (greater than the absolute value
of 1.96). Therefore all 28 manifest indicators appear to be valid measurements of their
specific latent constructs.

The manifest indicators of the Victim Health latent variable all loaded robustly (T
value over 1.96 and two tailed p value less than .05) and their sign was as expected. The
“Mental/ emotional health self-evaluated” manifest variable (coded from negative, low
numbers to positive, higher numbers) had the highest loading (1 = -1.340, t-value= -
5.396, p < 0.000). The presence of a diagnosis of neurological disorders manifest variable
had the weakest loading (1. = 0.439, t-value= 2.329, p < 0.013) on the Victim Health
latent variable. The negative signs before the four ernotional/ psychological health
variables are due to reverse coding (See Table 4.4).

The attributes loadings on the Social Isolation factor loaded significantly with
“Few Close Friends to Share Concerns with” having the highest values (X = 1.505,
t-value = 4.349, p < 0.000) and “Feel Close to Others in Community” had the lowest
value (A. = -0.489, t-value= - 3.697, p < 0.000). The negative sign for this loading is due
to reverse coding (See Table 4.4) and in accordance with the hypothesized relationship.

Factorial loadings on the Victim Behavioral Problems were high, ranging from the
lowest “When angry I am Ready to Hit Someone”, 0L: -0.319, t-value = -3.249,

p < 0.001) to the more robust “When People Insult Me I Try to Get Even” (7»: 1.192, T
value=4.587, p < 0.000). The negative sign for this indicator is due to reverse coding (See

Table 4.4)

120

Attribute loadings on the Adult Child Characteristics were robustly high
indicating good psychometric properties and unidimensionality. Loading values ranged
from the lowest, “Children financial problems”, (71.: 0.864, t-value= 13.7 97, p < 0.000) to
the highest, “Child Substance Abuse Problem” 0» = .929, t-value = 11.303, p < 0.000).
The signs for the loadings were as hypothesized. In fact, child substance abuse problems
were the strongest predictors of the likelihood elder abuse in the model.

Attribute loadings on the Victim/ Abuser Relationship Quality endogenous factor
yielded robust scores, ranging fiom the lowest, “Children Had Difficulties Getting Along
with Family Members” (3. = -.247, t— value = 2.007, p < 0.000) to the highest, “Rate
Control over Yom Relationship with Child” (1 = 1.192, t-value = 4.587, p < 0.000). All
loadings on this latent variable were different from zero in the population and were
statistically significant. The negative sign for the indicators “Children Difficult to get
Along with Family Members” and “Family Life with Children More Negative”, are due
to reverse coding therefore all signs were as hypothesized.

The reliability of the full model indicators can be examined by looking at their
squared multiple correlations. The model estimated individual R-squares indicate the
proportions of variance in an indicator that is explained by its underlying latent variable.

In the full model the most reliable indicators were as follows: (1) “Children

psychological/ emotional problems”, R2 " 0.915; (2) “Children alcohol/ substance abuse
problems”, R2 " 0.798; (3) “Ever sexually abused”, R2 " 0.711; and (4) “Children

financial problems”, R2 ' 0.696. The least reliable indicator in the model was the “Rate

121

tlrought/ effort put into relationship with your children” (R2 ' 0.028) variable. The most

reliable latent variable was the Elder Physical Abuse (R2 ' 0.725).

Restricted Model Mediator): Testing the Mediating Eflhcts of Victim/ Abuser
Relationship Factor

The restricted model converged after few iterations and yielded admissible
estimates. The overall model fit was adequate (x2 = 84.834; with 67 degrees of fi'eedom;
p = 0.0697). Although the fill] model’s fit is more robust, nevertheless results suggest that
the restricted model adequately fits in the population. Based upon the relatively high p-
value (p > 0.05) the restricted model’s null hypothesis [Ho: 2 = 2 ( 'y )] is not rejected,
implying that the there exist admissible parameter values with which the model
reproduces perfectly the population covariance matrix. In addition, the restricted model’s
chi-square value is considerably less than twice its degrees of freedom (2 times 67)
consequently it is safe to conclude that the model as a whole adequately fits the
population covariance matrix.

The next measure to consider in evaluating the restricted model’s fit is the
RMSEA, estimate which did not exceed 0.036, confirming that the restricted model fits
acceptably in the population. Consequently the discrepancy between 2 and 2 ( 'y ) is
close enough to yield accurate model parameter estimates.

In addition, the restricted model had a CFI of 0.941 and a TLI of 0.948, suggesting once
again a good model fit. The baseline model chi—square in this study came to 361.938,
with 59 degrees of fi'eedom, and a highly significant p value (p = 0.000). If we compare

this baseline chi—square value to the restricted model’s chi-square value (84.834.329), it is

122

evident that the restricted model is a significant improvement over the null model. In
summary, it can be concluded that the restricted models as a whole fits adequately in the

population.

Measurement Dimension of the Restricted Model: Validity and Reliability Results

There are no conceptualized and or practical differences between the full and the
restricted models in regards to the measm‘ement side psychometrics. Thus, the evaluation
of the measurement part of the model is fairly straightforward and results are not unlike
the ones discussed in previous subsection. The restricted model’s factorial loadings on

the respective latent variables are provided below (Table 4. 6).

123

Table 4.6. Restricted Model Factor Loadings

 

 

 

Manifest Variables Measuring Exogenous Variables Loadings T- Statistical
Estimates Values Significance
Victim '5
Health
History of Parkinsons 1.000 N/A N/A
History of Neurological Dis. 0,776 4,664 0.000
Chronic Medical Conditions 0,699 2,225 0.026
Anxiety Disorder 0.909 3.747 0.000.
Felt Depressed for 2+wks - 0.270 4.950 0.000
Mental Emotional Health -1 346* -5.282 0.000
Felt Upset Frequency 0.785 5.405 0.000
Felt Angry Frequency -0.488* . -5.363 0.000
Negative Feelings -0.694* -4.891 0.000
Ever Attended Emotional Group -l.153* -5.986 0.000
Victim '3
Behavioral
Problems
Sometimes Just Like to Hit Someone 1,000 N/A N/A
When insulted I Try to Get Even 1.366 3.639 0.000
When Angry I am Ready to Hit -0.6l7* -2.762 0.006
Victim ’5
Social
Isolation
I Do Not Feel I Belong to Community 1.00 N/A N/A

 

 

 

 

124

 

Table 4.6 Continued

 

 

 

 

 

 

 

Loadings T- Statistical
Manifest Variables Measuring Exogenous Variables Estimates Values Significance

Social

Isolation
Few Close Friend 1.621 4.159 0.000
Don’t Fit in with People and Community 1,278 4,838 0000
Feel Close to Others in Community .0,535* -3,979 0.000 .
Frequency of Contac with Friends 0,403 2,130 0,033

Adult

Child

Charater.
Children Emotional/Psych Problems 1.00 N/A N/A
Children Alc./ Substance Abuse Problems 0.864 13.981 0.000
Children Financial Problems 0.937 11.678 0.000

Victim/

Child

Relat.
Rate Relationship with your Child 1.00 N/A N/A
Child Difficult to Get Along With -0.480* -2.086 0.037
Rate the Thought/Effort you Put into Relationship 1.390 6950 0000
Rate Control over Relationship w. Child 0436 1.990 0.050
anily Life with Child is More Negative ’0390" '5-739 0000

Physical

Abuse
Ever Physically Abused 1,00 ' N/A NIA
Ever Sexually Abused 1,193 14.267 0.000
* Reverse coding

125

 

Inspection of Table 4.6 reveals that all of the parameter estimates are significantly
different fi'om zero (at p < 0.05 or better) as indicated by t-values well in excess of 1.96
in absolute terms. This provides validity evidence in favor of the 28 manifest indicators
used to represent latent variables of interest. The measurement error variances of these 28
indicators were within normal ranges (e. g zero/ close to zero measmernent errors)
suggesting no model specification problems. A closer look at the indicator loadings on
the Victim Health latent variable reveals that every directly measured manifest variable
loaded robustly (t-value greater than the absolute value of 1.96) and their sign were as
hypothesized at the model conceptualization stage. The unstandarized loading of
indicator “Ever Attended Emotional Support Group” had the greatest loading value (it = -
1.153, t-value = -5 .986, p < 0.000). “The presence of depressive thoughts for the last two
weeks” indicator had the weakest unstandardized loading value (it: 0.270, t-value =
4.950, p < 0.000) on the Victim Health factor. The negative signs before the four
emotional/ psychological health variables (see Table 4.6) are due to reverse coding, as
discussed in the firll model section. Inspections of standardized loading values reveal a
similar trend. The “Ever attended emotional support group indicatbr” had largest
standardized value (0.737) and the “Presence of depressive thoughts for the last two

weeks” had the smallest number (0.305). “Ever attended emotional problems group”
(R2= 0.543) variable was the most reliable indicator of the Victim Health factor and the
“Felt sad/ depressed for the last two weeks” was the least reliable (R2 ' 0.093).

Every manifest indicator of the Social Isolation factor loaded significantly and
appeared to measure a unidimensional construct. The “Contact with fiiends frequency”

had the lowest loading estimate (2. = 0 .403, t-value = 2.130, p < 0.000) and the “Few

126

fiiends to share concerns with” manifest indicator had the highest estimate value (7» =
1.621, t-value = 4.838, p < 0.000). The negative sign before the loading of “Feel close to
others in community” is due to reverse coding (See Table 4.6), thus in accordance with
the hypothesized relationship. Inspections of standardized loadings reveal a similar
picture. The “Few close fi'iends to share concerns with” indicator had a robust (0.836)
loading whereas the “Contact with fiiends frequency” had the weakest loading value

(0.184). In regards to reliability, the “Few close fiiends to share concerns with” variable
was the most reliable, (R2= 0.699), and the “Contact with fiiends” was the least, (R2=

0.034).

The unstandardized loading estimates on the Victim Behavioral Problems were all
robust ranging fi'om the lowest, “When angry I am ready to hit someone”, (l.= —0.6l7,
t-value = -2.762, p < 0.001) to the highest, “When people insult me I try to get even” (1. =
1.366, t-value = 3.639, p < 0.000). The negative sign before the former indicator is due to
reverse coding (Table 4.3). The standardized loadings follow a similar trend. The lowest
loading (0.369) belongs to “When angry I am ready to hit someone” and the highest

(0.437) to “When people insult me I try to get even”. The “When people insult me, I try

to get even indicator was the most reliable, (Rz= 0.191), while the “When angry I am

ready to hit someone” was the least, (R2: 0.136).

The Adult Child Characteristics factor had the greatest loading values in the
model. Although the Children financial problems indicator had the lowest value, (3. =
0.866, t-value = 13.981, p < 0.000) within this dimension, nevertheless the loading

estimate was only second to the “Children alcohol/ substance abuse problem” variable (7»

127

= .937, t-value = 11.678, p < 0.000) in this model. All loadings signs were as
hypothesized. Standardized estimates of these variables follow a similar trend. All three
indicators had a very high degree of reliability ranging from the lowest, “Children

financial problems”, (R2 = 0.696) to the highest, “Children psychological emotional
problems”, (R2 = 0.910).

Attribute loadings on the Victim/ Abuser Relationship endogenous construct
yielded unidimensional and relatively high estimate values, ranging from the lowest,
“Family life with children more negative” (1. = -0.390, t-value = 5.789, p < 0.000) to the
highest, “Rate control over your relationship with child” 0t. = 1.890, t-value = 6.950,

p < 0.000). All loadings on this latent variable were definitely different fi'orn zero in the
population and were statistically significant. The negative sign before indicators
“Children difficult to get along with family members” and “Family life with children
more negative” are due to reverse coding therefore all signs were as hypothesized. The
standardized estimate of “Rate control over your relationship with child” had the highest
value, (0.656), for the relationship construct and the “Family life with children more
negative” had the lowest value, (0.240). The most reliable indicator of the relationship
quality was the variable “Children difficult to get along with family members”,

(R2= 0.605).

Finally, both the Ever physically abused” and “Ever sexually abused” manifest
variables loaded highly on the dependent latent variable, Elder Physical Abuse. The
loading of me physical abuse variable was set equal to one (requirement of factor

modeling). The sexual abuse manifest indicator robustly loaded on its latent factor

128

(l. = 1.193, t-value = 4.267, p < O .000) and the sign was as hypothesized. The most

reliable measure of Elder Physical Abuse factor was the “Ever sexually abused” manifest

indicator, (R2: 0.732).

Summary Results for Both Models

Having evaluated the fit of both models, including individual factor measurement
dimensions and their squared multiple correlations, we can safely conclude that: (1) the
relationships between the latent variables and their respective indicators are robust and
statistically significant (e.g. all indicator loadings are significant at p < 0.05 and t-values
in excess of absolute value of 1.96), (2) the validity of measures used to represent
constructs of interest are within acceptable and safe limits (e. g. t-values larger than
. absolute value of 1.96), (3) the reliability of indicators are high (e. g. the high squared
multiple correlations), and (4) there is no excess of residuals on either tail of the normal
distribution (e.g. no measurement errors with values of zero or few SE over 2.58). In
conclusion, since both models adequately fit the population matrix and the measurement
dimensions are robust, the evaluation of the structural models (hypothesis testing) can

safely proceed. Finally, the squared multiple correlations for the two endogenous

variables are quite impressive; for the Elder Physical Abuse (R2= 0.725) and for Victim/

Adult Child Relationship Quality (R2= 0.797).

129

Hypothesis Testing

The purpose of this section is to evaluate the structural parts of the full and
restricted models by focusing on the substantive relationship of interests as described in
the Ecological Bi—Focal Model for Elder Abuse and hypothesized in the methods section
of this study. Three relevant factors for this analysis will de addressed: (1) the signs of
parameters representing the paths between the latent variables are as hypothesized, (2)
magnitudes of estimated parameters reflect the strength of hypothesized relationships,
and (3) the amount of variance in the Elder Physical Abuse factors is accounted for by the
exogenous latent variables as indicated by the squared multiple correlations.

The Elder Physical Abuse latent endogenous variable regressed on the Victim/
Adult Child Relationship Quality secondary endogenous factor and on the four
exogenous variables in the full model yielded the following substantive results (See Table
4.7).

130

Table 4.7 Structural Relationships amongst Latent Variables in the Full Model

 

 

 

Latent Variables Factor T-Values Statistical
Estimates Significance
Elder Abuse ON:
Victim’s Health 0.822 2.759 0006*
Victim’s Behavioral Problems .0472 -1 ,058 0.290
Victim’s Social Isolation -0.343 -1529 0.126
Adult Child Characteristics 0,259 2,704 0007*
Victim/ Adult Child Relationship .0124 .0858 0.290
Victim’s Age 0.000 -0.006 0.995
Victim’s Gender 0.246 0.780 0.436
Victim’s Marital Status -os72 -1.711 0.087

 

 

 

 

* Statistical significance

131

 

Figure 4.6. Structural Relationships in the Full Model

' '~ i Q g -' i

h Victim/

. Adult Child
Relationship )
Quality .

         
  
   
     
 

 

Victim
5‘ Behavioral
._ Problems

     

 

. Adult Child
Characteristic

Thai-J

Note: Statistically significant paths signified by wide arrows

132

In the restricted model, structural loadings of the four exogenous variables on the

Elder Physical Abuse factor were mediated by the Victim/ Adult Child Relationship

Quality secondary endogenous variable. Therefore both the magnitude of estimates and

their signs were different, as expected, from previous findings (See Table 4.8).

Table 4.8 Structural Relationships amon

st Latent Variables in the Restricted Model

 

 

 

 

 

 

Latent Variables Factor T-Values Statistical
Estimates sgnificance
Victim! Adult Child Relationship ON:
Victim’s Health -O.738 -3072 0002*
Victim’s Behavioral Problems .054] .0320 0.446
Victim’s Social Isolation -0329 -2157 0041*
Adult Child Characteristics 0270 -3.422 0001*
Elder Abuse 0N Victim/Abuser Rel. -0331 -3903 0.0001:
Demographic
Variables
Victim’s Age 0014 -1357 0.175
Victim’s Gender 0.358 1.825 0.068
Marital Status 0.229 1.244 0.213

 

* Statistical significance.

133

 

Figure 4. 7. Structural Relatio his in the Restricted (Mediational) Model

‘i‘V‘VVI

Victim/ » I
i, Adult Child

- " Relationship
4: y’ Quality /\

I .0.
n‘ \ ,

.’\ ;
< Victim
Health

 
 
      
    
 
    
 

  
  

Behavioral
Problems

 

Social
Isolation

» Adult Child

Characteristic

Note: Statistically significant paths signified by wide arrows

134

Immediate Elderly Parent/ Adult Child Context! (microsystem level/ bi-focal directional
approach)
Hypothesis #1:

Victim health factors (physical, psychological, and emotional, health problems)
are positively related to (more severe health problems will results in higher likelihood of
physical abuse) elder physical abuse (Figure 3.1). Hypothesis one predicted a positive
relationship between Victim Health factors (physical and psychological/ emotional
health) and Elder Physical Abuse in the full model and, as shown in Table 4.7, was
supported (7= 0.822; t-value = 2.759; p = 0.006). As hypothesized, the Victim Health
construct was significantly associated with Elder Physical Abuse in the restricted model
as well (7= -0.738; t-value = -3.072; p = 0.002). The Opposite sign before the Health
constructs in the two models is due to the meditational effects of Victim/ Adult Child
Relationship Quality factor. The scale on the meditational variable measures variable
dimension fiom worst relationship to best. As such, findings are logical given that more
health problems appear to be associated with worst relationship quality between the
elderly parent and her/ his adult child. Results confirm that the presence of various
physical, psychological, and emotional illnesses in the victim will directly increase her/
his risk of physical abuse regardless the meditational presence of the relationship
construct. The measurement part of this model suggest that emotional issues such as
feeling upset, irritable, angry, or anxious have a greater impact on the likelihood of elder

physical abuse than the presence of certain medical diagnoses.

135

Hypothesis #2:

Victim Behavioral Problems positively correlate with (increase) Elder Physical
Abuse, (see Figure 3.1). Hypothesis two posits that victim behavioral problems will
directly predict elder physical abuse. Analyses results do not support this hypothesis. In
the full model the loading estimate of the Victim Behavioral Problems to Elder Physical
Abuse path came to (Y= -0.472; t-value = 4.058; p = 0.290) whereas in the restricted
results yielded an even worst loadings (7'=-0.541; t-value = -0.820; p = 0.446). Although
it is uncertain 'why the loading signs were negative in both models, the fact is that the
effects of victim behavioral problems on elder physical abuse were not different fi'om
zero in the population. This could be due to several reasons such as a lack of precise
definition of what constitute elder behavioral problems. For example, it is not clear what
Kosberg (1988) and Lachs and Pillemer (1995) meant with ungrateful, demanding, and
otherwise unpleasant behaviors. Vague concepts do not yield precise operationalization
and consequently clear (e. g. unidimensional) measurements. It could be that thus victim
behavioral problem dimension was captured instead by the victim’s health latent factor

(e.g. behavior exhibited due to feelings such as anger, anxiety, and panic).

Hypothesis #3 :

Victim/ Adult Child Relationship Quality latent variable is positively correlated
with elder physical abuse when allowed to mediate between the model’s exogenous latent
factors (Victim Health, Victim Behavioral Problem, Adult Child Characteristics, and
Victim ’3 Social Isolation) and the Elder Physical Abuse construct. Worst relationship will

increase the likelihood of physical abuse. In addition, more serious victim health

136

problems. and victim social isolation will correlate with worst victim/ adult child
relationship quality (negative relationship), (see Figure 3.2).

Thus, hypothesis three predicted tlmt the Victim/ Adult Child Relationship Quality
construct will predict elder physical abuse if and only if it is allowed to mediate between
the model’s exogenous latent constructs and the main endogenous latent variable,
otherwise its impact will be insignificant. Accordingly when the Victim/ Adult Child

Relationship Quality factor was not allowed to mediate (e.g. full model) then its path to

Elder Physical Abuse was not diflemnt from zero in the population (7= -0.124; t-value =

-O.858; p = 0.290). However when the relationship construct was allowed to mediate
between the four exogenous "variables and the elder physical abuse factor then the path
hem Victim/ Adult Child Relationship Quality to Elder Physical Abuse became
statistically significant (7: -0.831; t-value = -3.908; p = 0.000). The sign before the
structin'al estimate is negative, suggesting that worst (negative) relationship quality
corresponds to an increase in elder physical abuse. In addition, the exogenous constructs;
Social Isolation, (7= -0.329; t-value = -2.157; p = 0.041), Victim Health problems, (7:-
0.738; t-value = -3.072; p = 0.002), Adult Child Characteristics, (7: -0.270; t-value = -
3.422; p = 0.001) changed signs direction. Specifically, a negative relationship (worst
relationship quality) between the elderly parent and her/ his adult child will increase in
the older adult’s health problems, social isolation, adult child problem, and consequently
in elder physical abuse. In addition, the strength of all exogenous paths mediated via the
relationship construct to elder physical abuse has increased in magnitude. Final

conclusion: results support hypothesis three in entirety (all parts).

137

“l

Contexts beyond focal older adult/ caregiver (ecological model testing)
H}pothesis # 4:

Social isolation will positively co-vary with elder physical abuse when mediated
by Victim/ Adult Child Relationship Quality otherwise its predictive effects will be non-
significant. The term social relations, (Antonucci, 2001) describes a broad array of
factors and interpersonal interactions, including the relationship between the elderly
parent and adult child in addition to the victim’s social isolation , that characterizes social
interactions among people. As such, the victim’s social isolation construct together with
the relationship factor (e. g. social relations) represent an exosystem level influence on
elder physical abuse. Consequently it is reasonable to assume that the' mediation model
tests elder physical abuse from an ecological perspective. When not mediated (full
model), results indicate that this hypothesis was not support by data (Y= -0.343; t-value =
-l .529; p = 0.126), as expected. However when the path to elder abuse was mediated by
the Victim/ Adult Child Relationship Quality factor (restricted model), social isolation of
the elderly parent was a strong predictor of elder abuse, (7= -O.329; t-value = -2.157; p =
0.041). The negative sign before the structural estimate indicate that worst elder parent/
adult child relationship is associated with greater victim isolation and a greater likelihood

of elder abuse.

Hypothesis #5:
Adult Child Characteristics (psychological/ emotional, alcohol and substance
abuse, and financial problems) will directly correlate (positive relations) with elder

physical abuse. Although hypothesis five appears to operate at the immediate ontogenic

138

1r

level (microsystem), a closer examination suggests that it has distal repercussions as well.
It is evident that the adult child’s psychological, emotional, and substance abuse
problems are microsystem level factors that directly influence the risk of elder abuse.
However, we could consider the adult child’s financial problems (e.g. lack of income-no
employment, real estate foreclosure, vehicle repossessions) an ecological factor (adult
child-mesosytcrn/ aging parent exosystem) that influences elder physical abuse. For
example, the adult child’s home foreclosure does not contain the elderly parent,
nevertheless will affect elder abuse in the immediate setting. A similar argument could be
made regarding the adult child’s substance abuse problems.

Hypothesis five was supported by empirical evidence in both the full model (7=
0.259; t-value = 2.704; p = 0.007) and the restricted model (7:0.270; t-value = -3 .422; p
= 0.001). A closer look reveals that when the Adult Child Characteristics factor was
mediated by the relationship latent variable both the sign (fiom positive to negative) and
the magnitude (increased) of the gamma loading has changed. When mediated by the
relationship variable both the magnitude and sign before the Adult Child Characteristic
construct changes. Consequently the bi-focal factors (victim health construct and adult
child health constructs) do predict elder physical abuse at the microsystem level with

distal contexts repercussion.

Power Assessment
An important part in the model(s) evaluation process is the examination of the
statistical power associated with testing that model. In the previous sections, both the

main and nested models were evaluated based upon their specific chi-square tests,

1319

whether to reject or fail to reject the null hypothesis, and in both cases the null
hypotheses (model fits correctly in the population) were retained. These results suggest
that the probability of incorrectly retaining the null hypothesis (perfect fit) in both
instances were very low (less than .05). To accormt for a Type II error a power analysis
must be undertaken for the following reason: if the power is low it is hard to tell whether
the low value of test statistics reflects to correctness of the models or the lack of
sensitivity to specification error. Two options are available to assess the powers (both
models) associated with testing for exact fit: (1) consult a useful power table compiled by
McCallum et al, (1996) or (2) use the SAS program routine developed by these authors
for analyzing power. Since I have no access to SAS statistical program, I will use option
the first option which is to get the respective power estimates for both models fi‘om Table
2 of MacCallum, Browne, and Sugawara (1996) .

For a model (main model) with 69 degrees of fieedom and a sample size of 203,
the power estimate for the test of exact fit is 0.801 and that that for close fit is 0.877.
Thus both power estimates indicate that the analysis in the main model is sufficiently
powerful.

For a model (nested model) with 67 degrees of fi'eedorn and a sample size of 203,
the power estimate for the test of exact fit is 0.827 and that that for close fit is 0.854.
Thus, both power estimates indicate that the analysis in the nested model is sufficiently
powerful. In sum, the likelihood of rejecting the hypothesis of close fit when the true fit is
mediocre is in excess .08. Thus it is closed to certainty that serious misspecifications

would be detected if they were in either of the models.

140

if .

Additional Model Diagnostics

Further model diagnostic information can be obtained by: (1) examining the data
correlation matrix for values outside the range from - 1 to +1, (2) negative variance
estimates and (3) standardized residuals with absolute values greater than 2.58.

Both models yielded admissible solutions meaning that covariances did not
contain negative variance estimates or a correlations larger than 1. Visual inspection of
variance estimates and correlation estimates supported the admissible solutions Mplus
message.

If the distribution of standardized residuals is approximately normal then about
95% of the standardized residuals should be between -2 and +2 and 99% of residuals
between -2.58 to +2.58. 'Ihe dissertation model had 0.06% of cases (cases #: 2;3; 19; 20;
36; 48; 69; 108; 121; 158; 163; 171; aird 195) with standardized residuals outside of the
normal range (absolute value 2). The largest standardized residual value came to 4.064
(case 20). Even if the model fits well, we expect to see about 6% of the cases with
standardized residuals greater that 2.56 in absolute value. In addition, a normal P-P plot
(Figure 4.5) reveals a normal residual distribution (data points arrayed along the 45
degrees line in the plot). In addition the distribution of residuals appears decidedly linear
and steeper than or 45 degrees angle (See Figure 4.6). In summary, the model
diagnostics provide evidence that all statistical assumptions have been reasonable meet
which in turn inspire confidence in the fitted models. This confirms the conclusions

drawn from both the examination of models fit and the residuals analyses.

141

 

Figure 4.8. P—P Plot of Standardized Residuals

Normal P-P Plot of Regression Standardized Residual

Dependent Variable: Physical Abuse

 

1.0-

a 0
e 0.8" 8
o.

E

s 0.5-

0

e

3 _

o 0.4

o

o.

x

"J 0.2-

 

 

 

T l I l l j
0.0 0.2 0.4 0.6 0.8 1.0

Observed Cum Prob

142

CHAPTER V: DISCUSSION

Guided by the Ecological Bi-Focal Model for Elder Abuse theoretical fi'amework
(Schiamberg & Gans, 2000), the purpose of this investigation was to empirically examine
variOus risk factor dimensions and their interactions during the process of elder abuse at
various ecological systemic levels. Although past research advanced the understanding of
risk factors in elder abuse, there are important limitations in these studies as documented
by officials at the National Research Council (NRC, 2003; Acierno, 2003).

First, previous studies on elder abuse prevalence relied on small and often
unrepresentative samples that limited the generalizability of findings (Cooper et al., 2006;
Lauman et al., 2005; Acierno, 2003; NRC, 2003). In contrast, the data for this study came
fiom a representative subsample of adults surveyed in a nationwide study on Midlife in
the United States (MIDUS). Second, previous empirical studies of elder abuse have
typically examined a limited source of variability in certain risk factors such as the
elderly victim’s health problems (e. g. dementia, Alzheimer, etc) as it increase the risk of
abuse. This dissertation study, in contrast, uses data on a wide array of health variables,
such as physical, psychological, and emotional indicators, that capture the
unidimensional latent construct of elderly victim’s health. Third, previous studies
typically have relied on self-administered checklists of elder abuse risk factors that only
indicate whether or not a given stressor has occurred. This study used both structured and
semi structured telephone and personal sm'veys to measure several aspects of risk factors
dimensions, including their objective characteristics as rated by expert coders (e.g.,

content, severity) and their subjective appraisals by study participants. Finally, previous

143

studies mainly relied on simple data analysis methods (e. g. using descriptives only,
simple regression analyses) (Acierno, 2003). In contrast this study has employed a
complex and data relevant mathematical modeling that captured associations at various
levels and in various directions simultaneously. In conclusion, this study was innovative
due to its advanced scientific scope, its diverse and representative large sample base,

creative use of in-depth- survey assessments in key areas, LISREL data modeling, and

ecological perspective.

Discussion of Results
Sample Based Description (Prevalence) of Elder Physical Abuse
Overall, the proportion of elderly Americans that reported physical or sexual
abuse experiences in the MIDUS II sample was not extensive and did not exceed 15%
(n=31). Although this statistics does not appear striking at first glance, the problem of
elder maltreatment becomes critical if we consider that by 2030 the number of elderly
individuals will have reached over 70 millions in the United States, (United States
Adrninislration On Aging, 2002). Accordingly, in a population of 70 million over 10
million elderly persons would be at risk of physical abuse. This public health epidemic
would impact almost every aspect of the US society and radically transform it. For
example, the astronomical financial cost of providing physical and psychological
treatment for 10 million physically abused older adults could push the Medicare program
closer to bankruptcy. Furthermore, bankrupt Medicare/ Medicaid programs could usher in
serious economic, social, and political instabilities into our society. Consequently the

epidemic of elder physical abuse warrants our utmost attention.

144

Background Variables

For the purposes of this study three victim demographic background variables
were examined as they influenced elder physical abuse and the relationship between the
elderly parmt and her his adult child. These variables had no significant influence on the
endogenous latent variables in either of the models. Although the Gender background
variable had the strongest influence on both latent factors its effects were not
significantly different from zero in the population. Previous studies in the field of elder
abuse have documented either no effects or contradictory findings in regards to these
very important victim background variables. Historical study findings imply that both
men and women could be victims of elder abuse and that the issue of gender is not that
important when attanpting to predict the phenomenon of elder abuse (Pillemer &
Finkelhor, 1988; Kosberg & Nahmias, 1996; Schiamberg et al., 2009). Similarly, findings
on chronological age as a predictor of elder abuse are contradictory and its effects
obfuscated. Some researchers formd a direct positive linkage between age and physical
abuse (Kosberg, 1988; Kosberg & Nahmias, 1996) whereas others suggest that the
influence is negative (Schiamberg et al., 2009). The issue appears to be similar with
marital status as well. This uncertainty and contradiction regarding the effects of these
three background variables on elder physical abuse is mirrored by this study’s results as
well (See Tables 4.7 and 4.8). LISREL modeling found no relationships between these
three background variables and elder physical abuse in the full model (See Table 4.7) and

in the restricted models (See Table 4.8).

145

Risk Dimensions of Elder Abuse at Microsystem Level: Immediate Older Adult/
Caregiver Context (microsystem level/ bi-focal directional approach)

In this study three hypotheses (#1; #2; and #3) examined elder physical abuse risk
factor dimensions fi'om a bi-focal perspective at the immediate/ ontogenic level. The
ontogenic level (microsystem) consist of activities and interaction patterns in the older
adult’s immediate surrounding that affects the process of elder physical abuse (Berk,
2008, Schiamberg & Gans 2000). To understand risk factors at this level we must
consider that all relationships are bi-directional (Schaimberg & Gans, 2000; ~
Bronfenbrenner, 1979, 1986, 1997). In addition, activities and interaction patterns at this
immediate ecological level might result either in an increase (risk of elder abuse) or a
decrease (protective factors) in the likelihood of elder physical abuse. Although the focus
in this study was on the risk factors, inference to protective factors can be discussed in
the context of absence or minimal levels of negative risk dimensions.

Four latent exogenous constructs measured via 21 manifest indicators captured
relevant risk factor dimensions operating and impacting elder physical abuse at
microsystem level in both (full and restricted) models in this study. These risk factor
characteristics and dimensions were all identified and discussed in the “Elder Abuse by
Adult Children: An Applied Ecological Framework for Understanding Contextual Risk
Factors and'the Intergenerational Character of Quality of Life” by Schiamberg and Gans
2004). The risk dimensions (captured by latent variables in this study) operating at the

microsystem level and identified as affecting elder physical abuse were:

146

 

 

Victim ’s Health

Hypothesis one tested the victim’s health risk dimension and its affects on elder
physical abuse. The Victim Health construct includes: (1) the presence of certain physical
illnesses (mainly neurological in this study) and (2) the accompanying negative
psychological/ emotional states. The logic behind creating such an overarching risk
construct instead of separating physical illnesses from the psychological/ emotional part
was: (1) the existence of physical illnesses robustly increases the presence of negative
emotional states (e. g. feelings and moods) (Hysing et al., 2007, Bolger et al., 1989) and
(2) separating the physical dimension fi'om psychological/ emotional could be impossible
or even counterproductive. In accordance with the literature, victim’s health related
problems significantly increased the risk of elder physical abuse, [e. g. the latent variable
Victim Health robustly predicted (J’= - 0.738; t-value = -3.072, and p < 0.000) elder
physical abuse both when mediated by the Victim/ Adult Child Relationship Quality
factor and when not mediated (7= 0 .822; t-value = 2.759, and p = 0.006)]. Therefore it
is safe to conclude that health related problem dimensions directly increase the likelihood
of elder physical abuse.

Studies often report this relationship from a one-dimensional and linear
perspective and do not examine the possibility of a bi-directional aspect and
consequence. Study results suggest that interactional effects between victim’s health
dimensions and physical abuse are circular and simultaneous. Once this interaction began
it is hard to predict the direction of cause and effect and possible outcome results. Health
problems increase in the presence of physical abuse (6. g. broken bones, depression, self-

neglect) whereas an increase in illnesses may result in an increase in physical abuse

147

likelihood. This bi—directional interaction between health problems dimension and
physical abuse could potentially result in an avalanche of unintended consequences. For
instance, co-existing biological and psychological/ emotional conditions can potentially
lead to critical outcomes for older adults who are abused. The elderly have a decreased
ability to heal after injury and may experience greater trauma fiom physical injuries than

do younger adults. Consequently their tissue and internal organs can be lacerated,

 

bruised, and damaged faster with minimal trauma and their bones easier fractured. This

could lead to an increase in the elderly mortality rates. Although it is not immediately

evident the psychological/ emotional trauma can be as devastating as any physical trauma
for the elderly. Physically abused older adults may oflen engage in self-destructive acts
(e. g. refusing to eat or take prescribed medication) and hasten their own death (Kennedy
& Tannebaum, 2000). In addition, as health problems worsen and feelings of
hopelessness, helplessness, and fear of being assaulted deepen elder active suicide rate
will increase (Kennedy & Tannebaum, 2000).

The bi-directional influence between the victim health factors and physical abuse
could be further clarified by examining the influence of individual physical and
psychological illnesses on the likelihood of elder physical abuse. Results from this study
linked the presence of various neurological disorders (e.g. Alzheimer disease, Aphasia,
Parkinson illness) to an increased risk of physical abuse (1‘; 0.776; t-value = 4.446; p =
0.000). The literature on elder physical abuse contains several studies. that docmnent
various individual neurological disorders as a component factor in the process of physical
abuse, (Schiamberg et al., 2009; Suitor & Pillemer, 1993). The etiology of neurological

disorders suggest three types of clinical syndromes: (1) diffuse and multifocal brain

148

 

disorders affecting cognition and behavior (e. g. dementia), (2) neuro-behavioral
syndromes linked with focal brain lesions (e. g. aphasia, amnesia, apraxia), and (3)
neuropsychiatric manifestations of neurological disorders (e. g. depression, anxiety,
psychoses), (American Psychiatric Association, 2005). Symptoms of various neurological
disorders gradually lead to behavior and personality changes, a decline in cognitive
capabilities (e. g. language skills or sound decision-making), dificulties in recognizing
family and friends, and aggressive and disruptive behaviors in the elderly (American
Psychiatric Association, 2005). Talerico et al. (2002) formd a strong correlation between
impaired commrmication and difficulties in recognizing relatives with elder aggression.
Stressed out and often threatened caregivers may find the task of caring for an aggressive
and dependent older person too difficult and in some situations the process of physical
abuse can develop and take place. In addition, if the relationship quality between the
elderly parent and the adult child has been negative for several years or even decades the
often aggressive and threatening behavior of the older parent might elicit a response of
“in-kind” from the adult child thus sparking a violent confrontation in which the older
adult could be hurt.

This study confirmed that the presence of chronic health conditions (e.g.
rheumatoid arthritis, multiple sclerosis, diabetes) will significantly increase 0. = 0.699; t-
value= 2.225, p < 0.000) the risk for elder physical abuse (See Tables 4.5 and 4.6).
Elderly people in poor health due to one or more chronic health related conditions may
require around the clock help and extensive coordination of care, including symptom
management, disability, and help with mobility related challenges (Kosberg & Nahmias,

1996). In the face of these increasing responsibilities and challenges adult child

149

caregivers often feel tired, overwhelmed, and probably isolated as well. In addition,
caring for a disabled parent may require missing workdays, often taking personal days, or
even quitting employment to provide care. This in turn may increase the levels of
physical, emotional, and financial stress in the caretaking adult child. If in addition the
elderly parent displays an aggressive behavior due to other medical condition (e.g.
neurological disorders) that might overwhelm the already stressed out caretaker’s coping
mechanism which in turn will increase the potential of elder abuse.

The presence of recent depressive droughts and feelings (neuropsychiatric
manifestations of neurological disorders) in the elderly parent substantially increased
(7t= .270; t-value = 4.495, p < 0.000) the risk of potential elder physical abuse (See
Tables 4.5 and 4.6). These findings are in accordance with the existing literature on risk
factors for elder physical abuse, (Kossberg & Nahmias, 1996; Cooney & Mortimer,
1995). For instance, Talerico et al. (2002) found that higher levels of elder aggression
were often associated with depression, impaired communication, low dose of .
antipsychotic drug use (inadequately treated depression). Given the common
neurotransmitter basis for both aggression and depression, untreated depression will
definitely increase the aggression level in older adults (Ryden et al., 1999). Displayed
irritable and aggressive behavior may again overwhelm the adult child’s capacity to
manage stress which in turn could lead to violent behaviors.

In this study the presence of anxiety disorder (neuropsychiatric manifestations of
neurological disorders) in the older adult directly and very robustly increased the
likelihood of elder physical abuse Oe= 0.909; t-value= 3.747, p < 0.000). Although

several psychiatric disorders, including depression, post traumatic stress disorder, and

150

personality disorders, mimic the diagnoses on the anxiety continuum, none is as closely
linked to aggression as anxiety disorder (Zohar, 2003; Hoffman, 2005). Anxiety is a
mood state and is regarded as a warning signal that alerts an individual to impending
danger thereby enabling the person to deal with that threat or harm of threat (Zide el al.,
2001). Although a certain level of anxiety might'be beneficial to an elderly individual, it
becomes problematic when it is not within normal parameters. The perceived impending
doom will activate the flight or fight response mechanism in people (Zide & Gray, 2001).
Since all too often the “threaten ” elderly perceives no way out due to physical
limitations, social isolation, as well as other biological conditions they might respond
aggressively or even violently. Aggression exhibited by the elderly parent might evoke
aggression in the adult child as well, especially when their past relationship was
characterized by negativity or violence.

The elderly victim’s self rating of mental and emotional well being was also
closely linked with the probability of elder physical abuse Ot= 1.346; t-value = 5.282,
p = 0.000) (See Tables 4.5 and 4.6). The more and higher ratings of mental and
emotional health problems corresponded to an increase in the likelihood of physical
abuse. The often perceived and felt negative moods, emotions, and thoughts, including
the anger and fear dimensions, created a generally stressful and negative environment for
the caret providing adult child. This stressful and often hostile environmentcould
overwhelm the caretaker’s coping skills to a degree where aggression would have been
seen as the only plausible management technique. In addition, if the relationship history
between the parent and child was already checkered with aggression and violence then

elder physical abuse could be very likely.

151

In contrast to the above described manifest indicators that measured various
mental health diagnoses that included an array of negative feelings (e. g. fear and or
anger), the following two measrnes capture individual feelings in the category of anger
that might be responsible for elder physical abuse. These two feelings were: (a) being
upset or (b) the more intense anger.

The victim’s feelings of being fiequently upset were closely associated with elder
physical abuse (X= 0.785; t-value = 5.405, p = 0.000). In the field of human
psychopathology there is an agreement about that fact that human emotions involve both
a positive and negative valence. Generally speaking feelings of being upset or angry
(both members of the anger category) are considered negative. In addition, the presence
of negative feelings (e.g. anger) does not preclude the co—existence of more subdued
positive (less than negative) feelings during the same circumstance. The victim’s negative
feeling state of being upset with either the adult child or with something else might be a
precrnsor to the initiation of a violent confrontation between the adult child and her/ his
elderly parent.

The fiequency of victim’s anger was also associated, as hypothesized, with the act
of elder physical abuse OF 0.488; t-value= 5.363, p = 0.000). Interestingly at first glance
it does not appear to make any sense why this variable had a less robust loading than the
previously analyzed feeling (being upset). Common sense would tell us that since anger
is a more intense feeling than being upset and it is more likely to precipitate physical
hostility between the elderly parent and her/ his adult child. A closer examination of these
two estimates values (5.405 versus 5.363) reveals that in the population the effects of

these two feelings on elder abuse vary only a little (less than 0.042). One explanation of

152

why angry feelings loaded with almost half strength (.488 versus .7 85- compared to upset
feelings) on elder physical abuse could be the social stigma associated with confessing to
anger rather than to being upset. Regardless the final result is the same. Both individual
feelings in the elderly victim can be a precursor to physical abuse. In closing, discussing
these two feelings individually and their impact on elder physical abuse is very important
for clinical reasons that will be discussed later in this chapter.

The existence of negative feelings (variable: more negative feelings than usual) in
the elderly victim yields similar results as the existence of individual negative feelings
discussed above. The presence of negative feelings (more than one at the same time) is
strongly correlated with elder physical abuse Oe= 0.694; t-value= 4.891, p = 0.000).
Therefore if one or more negative feelings (possible fiom the anger family) are present in
the elderly during an interaction with her/ his adult child the likelihood of a violent
encounter is more pronounced than otherwise.

Logically the variable ever attended an emotional support group is closely related
to having been diagnosed with a psychopathological disorder or the existence of one or
more negative feelings that appear to be precursor to elder physical abuse. The rationale
behind this reasoning is that if the elderly victim had no emotional problems as measured
by the lack of intense negative feelings, that were either symptoms of more serious
underlying neurological disorders or not, she/ he did not need help fiom an emotional
support group. Consequently if the elderly parent indicated that she/ be has attarded an
emotional support group than the presence of negative feelings preceding the act of elder
physical abuse is justified. Therefore we expect that the variable attending an emotional

support group directly influence elder physical abuse. In this study it was found that this

153

 

variable had a strong impact (1= 1.153; t-value = 5.986, p = 0.000) on the outcome
variable. Moreover this estimate loading was only second to the variable rate your mental
and emotional health in strength which implies that if the help of an emotional support
group was sought than the intensity of the existing negative emotions as a precursor to
elder physical abuse were also detected by the elderly individual. Once again, this

conclusion has serious clinical implications which will be discussed later in this chapter.

Victim Behavioral Problems

Another risk dimension often associated with elder physical abuse was the victim
behavioral problem (Kosberg, 1988, Lachs & Pillemer, 1995). For the purposes of this
study the victim behavioral problem construct was defined as overly demanding,
combative, 1mgrateful, and otherwise unpleasant behavior of the aging parent, (Kosberg,
1988; Lachs & Pillemner, 1995). Before moving to the examination of this risk
dimension, it should be noted that there is no universally accepted definition (e. g.
theoretical or operational) of elder behavioral problems. The absence of acceptable
definitions is due to several factors: (a) competing definitions of what constitutes elder
behavior problem and non-problematic behavior characteristics; (b) significant variations

among conceptual models and theories regarding behavior problems (e. g. Kosberg’s

model compared to Pillemer and Suitor, 1992, 1993); (c) measurement related issues (e. g.

how do you measure an “ungrateful” behavior, and; (d) close relationship between
behavioral problems and other psychological/ emotional impairments (e. g. anxiety
disorders, personality disorders), (Pliszka et al.2000 ). Moreover, the correlation of co-

occurring neurological disorders (e. g. depressive) and problem behavior among

154

adolescents is well established in the literature (Pliszka et al., 2000; Vermeiren et al.,
2002). Accordingly the so called elder behavior problem could be the outward
manifestation (the action) of negative internal processes such as afflictive feelings and
mood (e. g. fear, anger, or rage). This appears to be confirmed by recent research studies
that examined how an individual’s mental health status moderates the exhibited behavior,
(Nebbitt et al., 2008). Consequently it is difficult to separate the so called behavior
problems from the existence of biological (e. g. neurological disorders) psychological/
emotional problens, thus a case for construct unidimensionality (e. g. biological,
psychological, emotional, and behavioral factors measuring the same dimension —
individual’s health) could be made. This complicated issue will be clarified when we
examine the loadings on the Victim Behavioral Problems factor.

Although every indicator loading on the Victim Behavioral Problems latent
variable was robust [e. g. “When people insult me I try to get even”- (1: 1.366; t-value =
3.639, p = 0.000) and “When angry I am ready to hit someone”—(7t= 0.617; t-value =
2.762, p = 0.000)], it did not appear to impact the Elder Physical Abuse endogenous
variable either in the full model, (7=- .472; T-value = -l .058, and p = 0.290) or in the
restricted model, (7: 0.541; t-value = -0.820, and p = 0.446). These results also
confirm previous findings indicating that there is no direct association between the
perceived elder unpleasant and or demanding behavior and the risk of suffering abuse.
Consequently it can be concluded that the elderly parent’s perceived behavioral problems

did not directly or indirectly influence the likelihood of elder physical abuse.

155

The Victim/ Adult Child Relationship Quality

The percentage of middle aged adult children living and caring for their elderly
parents have increased fi‘orn 10% in the 1990’s to 50% during the last decade in United
States, (U .S. Census Bureau, 2006b). Furthermore about 20% of these adult children
provide care for an aging parent with at least one chronic physical, neurological, or
psychological illness (V anier Institute of the Family, 2004b). As these illnesses progress
both internal processes (e. g. negative emotions, moods and diagnosed psychological
conditions) and behavioral changes (e. g. actions driven by internal processes) will occur
(e. g. fiequent and angry outbursts, physically attacking) that will increase in the
difficulties of caring for an elderly parent. Consequently a worsening in the victim’s
health dimension will affect the relationship quality between the elderly parent and her/
his adult child. Furthermore, a strained relationship might increase stress and burden on
the adult child as well. Studies consistently suggest that stressed out and overburdened
adult children are more likely to abuse their elderly parents than those who are able to
adequately cope with increased stress levels (Steinmetz, 1990). Moreover if the
relationship between the elderly parent and adult child is primarily characterized by
conflict then there will be an increase in elder abuse risk dimensions (Korbin et al., 1995;
Kosberg & Nahmias, 1996; Browne & Hamilton, 1998).

In addition, there may also be a history of long-term difficulties in the relationship
between the elderly parent and adult child. Several studies document a relationship
between child abuse and parent abuse (Agnew & Huguley, 1989; Peek et al., 1985). The
Fourth National Incidence Study of Child Abuse and Neglect (2006) for instance,

estimated that between 20% and 35% of children identified as maltreated suffered a

156

serious injury, defined as "long-term impairment of physical, mental, or emotional
capacities or requiring professional treatment aimed at preventing such long-term
impairment. Children who have been maltreated may begin to show violence in the place
where it was first encountered, namely in the family home (Agnew & Huguley, 1989;
Peek et al., 1985). Consequently the relationship between the elder parent and adult child
might be primed for physical violmce given a history of violence in the family.
Unfortunately, there was no direct manifest indicator available in MIDUS H data to test
the connections between child abuse and subsequent elder abuse in this study. It would be
a useful direction for future research.

Finally, previous research studies suggest that the relationship quality between the
aging parent and her/ his adult child might be the most important factor in trying to
explain why some adult children resort to violence against their parents and others not
(Browne & Hamilton, 1998; Jaffe et al., 1990). Based upon these findings, it is safe to
state that the relationship between the elderly parent and adult child will mediate and or
moderate the likelihood of elder physical abuse. Worst relationship will increase the
likelihood of physical abuse. In addition, more serious victim health problems will
correlate with worst victim/ adult child relationship quality.

When mediating the four exogenous latent variables the Victim/ Adult Child
Relationship factor robustly predicted Elder Physical Abuser, (7: -.831; t-value = -
3.908 and p = 0.000). However when this factor was not mediating than its predictive
power of Elder Physical Abuse was not different from zero in the population, (7: -
.124; t-value = -0.858; p = 0.391). Findings underscore the vital importance that the

relationship quality between elderly parent and adult child either triggers or dampens

157

 

aggressive feelings and actions towards the older adult. In addition, the model
suggests a strong correlation between the Victim Health and the relationship quality
between the dyad.

A lack of victim’s control over the relationship is the most predictive of a
negative and conflictual interaction between the elderly parent and the adult child ,
031.890; t-value = 6.950, p = 0.000). In fact this lack of control over the relationship
appears to generate feelings of alienation, helplessness, and hopelessness in the older
adult and it is the most robust predictor in the model of a negative relationship and
increased physical abuse.

Based upon model results it is evident that the elderly parent clearly intends to
work on and improve the relationship with her/ his child, Oe= 0.436; t-value = 1.990,

p =.050). The more the elderly parent works on improving the quality of their relationship
the more it improves thus there is a reduction in the possibility of aggressive and violent
interaction. However it should be noted that not every relationship, especially those that
were characterized by long-standing family violence, can be effectively improved (Straus
et al., 1980; Jaffe et al., 1990; Browne & Hamilton, 1998). It appears conclusive that
family violence teaches children to use violence themselves, both in childhood and
adulthood. Therefore if a relationship has been characterized by serious violence it might
not leave place for much improvement regardless of the elderly parent’s efforts.
Consequently the reduction and elimination of family violence (e. g. child abuse) early on
will reduce elder abuse. There appears to be a clear link between child abuse and elder

abuse.

158

 

The last two manifest indicators of the Victim/ Adult Child Relationship construct
exmnined the adult child’s negativity towards family members in general. The variable
“Family life with the child more negative” was directly linked to worsening the
relationship between the adult child and elderly parent, (2.: -0.390; t-value = -5.789,

p =.000). Similarly, the indicator “Children have difficulties getting along with family
members” negatively correlated (towards worst possible relationship) with the
relationship construct, (2.: -0.480; t-value = -2.086, p =.037). If the elderly parent’s
perception about her/ his caregiver was not biased, the adult child’s behavior towards the
family was negative or even confrontational. This conclusion is logical if we consider the
adult child’s likely emotional, psychological, or substance abuse problems.

Finally, the Victim/ Adult Child Relationship factor did not appear to be

influenced by the elderly parent’s age, gender or marital status.

Contexts beyond Focal Older Adult/ Caregiver (Ecological Model Discussion)
Victim ’3 Social Isolation
Antonucci’s (2001) social relations construct includes an array of factors and

interpersonal interactions that characterizes social exchanges among people. The social
relations concept includes: (1) social networks, (2) social support, and (3) an individual’s
sense of control over relationships. In addition, social support can be defined as an
interpersonal transaction among people that includes: aid, affect, and affirmation,
(Antonucci, 2001). In this study, indictors measuring the victim’s Social Isolation (the
opposite of social support) construct include these dimensions (e. g. feel close to others in

community -) affect; do not fit in with people or cornmunity-) lack of affirmation)

159

 

consequently they can be used to describe the older adult’s lack of social relations
(isolation). Furthermore, the social isolation construct interacts with elder abuse in a
context beyond the immediate elderly parent/ adult child dimension. For example, the
risk of elder abuse could be minimized by close relationships the older adult has within
her/ his community (e. g. the elderly could tell about the negative relationship with her/
his adult child to the family physician). They in turn must report (since they are mandated
by law) the suspicion of elder abuse to government authorities such as the adult
protective services (APS). APS professionals will investigate and protect the older adult
if abuse is suspected. This social relations context (or lack of it in our study-9 isolation)
is beyond the focal elderly parent/ adult child interaction.

Research on elder abuse consistently documents the phenomenon that the victim’s
social isolation (the lack of social relations) positively covaries with increase in elder
physical abuse. Garre-Olmo et al. (2009) reached similar conclusions and, in addition,
they suggested that social relationships and networks are protective factors against elder
abuse and the social isolation of the elderly increases their risk of abuse. Fulmer et al.
(2005) defined social isolation as the lack of both informal (e. g. relative and fiiendship
network) and formal (e. g. adequate community based outreach programs, government
programs, and health care providers) support systems for both the victim and also the
caregiver. '

Social isolation of the elderly parent significantly covaried with the relationship
quality between the dyad, (7= - 0.329; t-value = -2.157, p = 0.041) as they jointly
influenced the main outcome construct, Elder Physical Abuse, ( 7= - 0.831; t-value = -

3.908, and p = 0.000). Consequently, the elderly parent became gradually more isolated

160

as the relationship quality worsened which in turn increased the risk of physical abuse.
These findings imply that the adult child has an active role in the elderly parent’s social
isolation. Ifthe relationship between the older adult and her/ his adult child is
characterized by strain and abuse, it appears that the child is responsible for the isolation
of her/ his parent. Fear of legal consequences, family and social ostracism, and or loss of
good reputation could motivate the adult child to actively control access to her / his
parent.

The victim’s Social Isolation path to Elder Physical Abuse was not different fi‘orn
zero in the p0pulation when the Social Isolation factor was not mediated by the Adult
Victim/ Adult Child Relationship Quality latent variable (1’ = - 0.343; t-value = -1.529,
and p =.126). If we did not take into account the relationship quality between the elderly
parent and her/ his adult child then the victim’s social isolation did not increase the risk
of physical abuse.

The manifest indicator “I have few fiiends to share concerns with” loaded the
strongest on the Victim 's Social Isolation construct, (X = l .621; t-value = 4.159,
pl= 0.000). The loss of fiiends could be due to several reasons including the denying
access to the elderly parent, older adults’ physical fiailty or fiiends deceased or moved
away. However it is likelier that the adult child denies access to her/ his parent when their
relationship is characterized by strain and abuse. The above reached conclusions are
further supported upon the examination of the last manifest indicator. The variable “The
frequency of contact with fiien ” also loaded significantly on the victim’s Social

Isolation construct, 0s = 0.403; t-value = 2.130, p =.033).

161

 

 

The next two direct measm‘es of the Social Isolation construct, “Don’t fit in with
people or community”, (it = 1.28; t-value = 4.838, p = 0.000) and “Feel close to others in
community”, (2. =-.535; t-value = -3.979, p = 0.000) (See Table 4.6), not only measure
the elderly parent’s social isolation but also offer a sense of how “life alienated” the older
adult became. Research consistently linked social isolation to higher risks of suicide and
an increase in mental/ emotional difficulties in the older population (Antonucci, 2001;
Berk, 2008). Consequently, severe social isolation will directly influence (worsen) the
older adult’s physical, mental, and emotional health and possibly lead to an early death,
due to either suicide or worsened health related complications, even if the psychical

abuse experienced was not directly fatal.

Adult Child ’s Characteristics

The Ecological Bi-Focal Theory for Elder Abuse posits that acute stressors (e. g.
within and outside of ontogenic relationship) experienced by the adult child lead to elder
physical abuse (Schiamberg & Gans, 2000; Steinmetz, 1999). Examples of such stressors
might be psychological/ emotional health problems (e. g. depression), caregiver substance
abuse dependence (e.g. cocaine habit) and ecological financial problems (e. g. real estate
repossession). In addition the stress of caring for an elderly parent can lead to worsening
existing mental and physical health problems that make caregivers burned out, impatient,
and unable to keep fiom physically lashing out against their parents. The adult child
characteristic construct has both ontogenic (e. g. psychological/ emotional health and
substance abuse) and distal repercussions (e. g home foreclosure) on elder abuse. In this

study, the latent variable Adult Child Characteristics tests risk factors in elder abuse at

162

the following ecological dimensions: (1) at the microsystem level (adult child’s
psychological well being and substance abuse problems), and (2) at the mesosystem
(adult child’s financial problems), and exosystem (aging parent) levels.

In accordance with previous research literature, this study found strong
associations between the adult child’s personal problems and an increase in the risk of
elder physical abuse. In fact, the relationship was so robust that in both models the impact
of the Adult Child Characteristics on the Elder Physical Abuse factor could be
docmnented. In the full model the relationship between the exogenous and endogenous
factors was as follows: 7: 0.259; t-value = 2.704 and p =.007 (See Table 4.7). In the
relationship mediated model the finding was as follows: V=- 0.270; t-value = 3.422 and
p =.001 (See Table 4.8). An examination of the above relationships brings up two
observations: (1) the different directions of the structural estimates in the two models and
(2) an increase in the strength of the gamma loading in the mediated model. In the full
model in increase in adult child’s problems corresponds to an increase in the risk of elder
physical abuse (positive direct relationship). In the mediated model the increase in adult
child’s problems is directly related to the worsening of the relationship between the
elderly parent and her/ his child which in turn increases the likelihood of physical abuse.
All three individual indicators of adult child problems robustly loaded on the Adult Child
Characteristics factor.

Caring for a chronically ill and dependent parent can be highly stressful and add
to the adult child’s psychological and emotional strain. It can lead to role overload, an
increase in depression, anxiety and feelings of hostility towards the aging parent, (Killian

et al., 2005). Unfortunately the loading on this manifest indicator had to be fixed to one

163

(LISREL requirements) consequently we can’t directly examine the influence of this
variable on elder physical abuse. Although the models did not directly produce useful
estimates to directly assess this risk dimension and its impact on elder abuse, several
studies document higher rates of psychological/ emotional problems (e.g. depression and
anxiety) in those adult children who engage in violent behaviors (e. g. elder physical
(e.g., Baskin-Sommers, 2006). Given prior conclusive findings on the relations between
psychological/ emotional problems and aggression in addition to the models’ LISREL
estimates (e. g. admissible solution and convergence) it can safely be stated that the
adult’s child psychological/ emotional problems directly increases the risk factors of
elder physical abuse.

The caretaker’s ongoing substance abuse problem has been identified as a
characteristic of some elder abusers and the level of addiction directly linked to an
increase in elder physical abuse, (Kratcoski, 1984; Kosberg & Nahmias, 1996; Hamilton,
1998). The role of controlled substances and alcohol abuse in aggressive behavior (6. g.
elder physical abuse) has been researched extensively. Studies have consistently found
relationships between physical aggression, alcohol, and cocaine use (see MacDonald et
al., 2003). Alcoholism was also associated with family violence against the elderly,
(Anetzberger et al., 1994). In this study, the adult child’s substance abuse problem was
robustly linked to the deterioration of the relationship between the older parent and her/
his child and to a strong increase in the risk of physical abuse, (1 = 0.937; t-value =
11.678, p = 0.000) (See Table 4.6).

Caregiver’s financial problems, including financial dependence on the older adult

(ontogenic level), has exosystern/ macrosystems repercussions as well. Accordingly,

164

repossession of the adult child’s car or home foreclosure can intensify her/ his stressors
including a worsening of psychological/emotional internal processes, which in turn can
increase the risk of aggression against the elderly parent.

In this study the adult child’s financial problems were directly related to the
worsening of relationship between the victim and her/ his child and a direct increase in
risk of elder physical abuse, (71.= 0.866; t-value = 13.981, p =0.000) (See Table 4.6). As
the elderly parent’s physical and mental health increasingly gets worse the care giving
tasks and their costs are likely to escalate. Conflicts are likely to arise over excessive
financial burdens on the adult child which in turn might create an unstable and dangerous
living environment for the elderly parent. Findings in this study are in accordance with
conclusions reached by other researchers in the field of elder physical abuse (Pillemer,

1986; Kosberg & Nahmias, 1996).

Theoretical, Research, and Clinical Implications

The overarching goals of this study were to: (l) empirically examine the
assumptions, tenets, and conclusions of the Ecological Bi-Focal Model for Elder Abuse
as developed by Schiamberg and Gans (2000) and (2) offer usefirl clinical suggestions on
risk and protective factors involved in elder physical abuse. Study findings will
contribute to the scholarly literature on elder abuse by establishing the Ecological Bi-
Focal Model for Elder Abuse as a preeminent framework that both theoretically and
empirically explain the dynamics and various risk dimensions of elder physical abuse.
In addition, the study’s findings have important practical implications as well that can

guide clinicians working with older adults and their families.

165

Theoretical Implications

NRC Panel to Review Risk and Prevalence of Elder Physical Abuse (2003) called
for the development of theoretical elder abuse models that: (1) are based upon a
representative national random sample, (2) use standard definition of a specific type of
elder abuse, (3) clearly identified abuse measures based onthese definitions, and (d) the
utilization of research methods and statistical modeling that is based upon sound
mathematical principles that can yield generalizable results. For the purposes of this
study a nationally representative random sample was used with cases in excess of 200
(large sample). Furthermore, all variable definitions and their measures in this study were
based upon standard and generally accepted definitions and guidelines. Both research
methods and statistical analyses used in this study were based on sound and robust
mathematical modeling which resulted in generalizable results and conclusions. Finally,
this study results reflected and confirmed previous elder abuse prevalence research
findings and in addition expanded the theoretical fiamework in this field by the following
contributions:

1. The established the usefidness of the Ecological Model for Elder Abuse:
Empirical results of this study indicate that the ecological perspective for elder
abuse framework offers a valid and useful theoretical explanation for elder abuse.
Risk factors of elder abuse by adult children were examined fi'om a bi-focal
perspective (e. g. focusing simultaneously on the aging parent and adult child as a
dyad) embedded within a larger social context. Study results suggest that elder

physical abuse risk factors are nested in three levels of environment or contexst of

166

deve10pment. At the ontogenic level the victim’s health factors, the relationship
quality of the dyad and, to some extent, certain abuser characteristics appear to
play a significant role. Social relations and certain abuser characteristics such as
financial problems (e. g. lack of employment, home foreclosure, etc.) robustly
impact elder physical abuse at higher systemic levels (e. g. mesosytem/ ‘
exosystems). In our study the construct social relations included: ( 1) the victim’s
social isolation (lack of social networks), (2) social support (e. g. study construct
elderly parent/ adult child relationship quality), and (3) sense of control over the
relationship, (6. g. manifest indicator control over the relationship). When the
social relations construct included all three dimensions, the construct had a robust
mediating effect between the immediate context’s risk factor dimensions and
elder physical abuse. The abuser"s characteristics construct both at the immediate
level and at context beyond the focal relationship, robustly impacted physical
abuse of the older adult. Study limitations: given the characteristics of the
MIDUS II data set, constructs at chronosystem level could not be exarrrined.

. Ofi'ered a holistic approach to conceptualizing the victim ’s health dimension: A
novel approach was followed when the victim’s'health construct was
conceptualized and then operationalized for data analyses purposes. This latent
factor combined the presence of diagnosed physical illnesses with the existence of
negative emotional/ psychological problems. I contend that chronic and
debilitating physical illness factors cannot be separated from the accompanying
(intertwined) psychological and emotional dimensions. This assumption was

based upon both clinical mental health experience and newly emerging research

167

findings. Chronic and serious physical illnesses are among the strongest risk
factors for late-life depression and other related emotional/ psychological
problems (Geerlings et al., 2002). This study results and findings clearly indicate
that examined chronic physical illnesses and psychological/ emotional indicators
measure one and only one factor dimension (in this case the Victim ’s Health latent
factor) and their presence robustly increase the likelihood of elder physical abuse.
Moreover the stronger loadings of psychological and emotional health indicators
suggest that the presence of these dimensions is the main and more robust risk
factors associated with physical abuse of the older adult. In addition, mental and
psychological health problems in the elderly population are strongly correlated _
with elder suicide (Heisel, 2006), thus elder abuse might also be robustly related
to elder suicide. More ecologically focused studies are needed in this area in order
to better rmderstand the effects of physical abuse on the elderly and society.

. Elder behavioral problems and physical abuse: a problematic factor:

Literature on elder abuse contains definitional and conceptual contradictions in
regards to the concept of “elder behavioral problem”. The first issue with the
Elder Behavior Problems construct is that the there is no unique and generally
accepted definition for it Thus, it is hard to operationalize the construct in order
to measure it. For example, ungrateful behavior could mean different things to
different people. It is unclear how ungrateful a behavior has to be in order to elicit
physical abuse. Second, this definition (behavior problems) was often
operationalized into one manifest indicator that yielded mixed results in previous

studies. In this study, the victim behavioral problem’s effect on elder physical

168

abuse was not different fiom zero in the population. At this point it appears that
what was conceptualized in previous research studies as victim behavioral
problem is not different from emotional factors (e. g. upset, angry, hostile, scared)
present during the process of elder physical abuse. In sum, the behavior problem
variable and psychological! emotional indicators are a one dimensional construct
and do not measure separate factors.

4. Social relations (victim ’3 social isolation, dyad ’s relationship quality, and the
elderly parent ’s control over therelationship) as a mediating factor: Study results
indicate that a robust one-dimensional relationship exists between the Victim ’3
Social Isolation and Elderly Parent/ Adult Child Relationship Quality constructs.
In addition, when separated neither construct impacts the elder abuse latent
variable in any significant way. Consequently, these two factors will have to be
combined into one overarching construct (e.g. Social Relations, Antonucci, 2002)
to gain the desired explanatory influence in future research. Study findings also
suggest that the victim’s social isolation is due to the adult’s child’s effort to
prevent the elderly parent fiom maintaining close community based relationships

in order to conceal various forms of abuse, neglect, and financial exploitation.

Clinical Implications for the Marriage and Family Therapist
Results of this study and the identification of the ecological etiology of elder
abuse might be useful for clinicians in the field of marriage and family therapy for

several reasons:

169

1.

In accordance with previous research findings this study also documented that
relationship factors (e. g. relationship quality, coping habits, social networks) were
strongly correlated with late-life physical and psychological well-being (V aillant
& Mukarnal, 2001), that at turn were associated with change in elder abuse
dimensions (Zide et al., 2001; Zohar, 2003; Hoffman, 2005). In addition, an
improved relationship between the dyad might mean a reduction in the older
adult’s social isolation thus more social oversight and better protection of the
older adult. Given the contextual, systemic, and relationship oriented approach of
marriage and family therapists, they could be at the forefront in providing clinical
treatment to older adults and their families. However, although family and
marriage therapists have well established therapy intervention methods to help
yormg and middle aged couples and families in improving their relations and
balancing the power and control in relationships, there is paucity in ecological
therapy models that could guide clinical work with physically abused older adults
and their families. Consequently this study’s results and findings could be useful
in developing comprehensive clinical assessment, intervention, and follow up
methods that would guide clinicians in their work with elderly clients and their
families.

To reduce elder physical abuse and suicides, clinicians must assess for several
risk factor dimensions at patient intake. Since there is a strong correlation
between elder abuse and control over relationship, the elderly client must be
interviewed alone. For various reasons (e. g. loss of control, fear of retaliation,

shame) the older client might not be willing to disclose abuse in the presence of

170

her/ his adult child. Undisclosed ongoing physical abuse might have fatal
consequences (e. g. active or passive suicide) for the older adult. Therefore
clinicians are ethically responsible to assess the relationship for abusive practices
and this assessment must take place in the absence of the adult child. Even though
there are no specific elder physical abuse measurement instruments, clinicians .
could use the Conflict Tactics Scale (CTS; Strauss, 1979) or similar scales until
such an ecological physical abuse instrument would be developed. In addition,
most existing aggression/ violence related self-report measures have a number of
shortcomings such as assessing only a limited set of violent behaviors (pushing,
hitting, or kicking) or missing the ecological context of violence (e. g. events
preceding the violent act, intentions dming the incidence, or consequences of
violence). Therefore I would recommend the use of measurements that assess
Social Relations, Victim ’s Health (both physical and psychological) and Adult
Child Characteristics (e. g. psychopathology, substance abuse, and financial
difficulties) in addition to the Conflict Tactics Scale as well. Untreated physical,
psychological, and emotional health conditions in the elderly increase the risk for .
elder abuse (Kossberg & Nahmias, 1996; Cooney & Mortimer, 1995).
Consequently information should be obtained in regards to the older adult’s
medical/ psychological background history and treatment interventions received
(e.g. physical and psychotropic medications regimens). In addition, signed
permission must be obtained from the patient to initiate and maintain ongoing
contact with the elder’s medical and psychiatric physicians. Failing to gather a

complete health related history might prevent the detection of abuse and suicidal

171

thoughts and ideations. The victim’s fear for her/ his immediate and long term
safety should be assessed in addition to the adult child’s willingness to
acknowledge that she/ he has anng physical aggression problems and take
responsibility for her/ his violent actions. If the clinician determines that ongoing
violence between the dyad could not be eliminated or lethality is present, the
clinical focus should shift to safety planning and resomce mobilization. As part of
the safety planning protocol the case must be referred to Adult Protective Services
and or local law enforcement agencies in order to ensure the safety of the older
adult. Furthermore, information on local resources and support networks should
be provided to the older adult in addition to obtaining permission and referring the
abused individual to appropriate community based agencies. Ifthe marriage and
family therapist determines that the safety of the older adult is not in jeopardy and
dyad appears willing to work on their relationship then providing conjoint therapy
is appropriate and beneficial to parties.

. p The primary goal of a conjoint therapy should be to build solutions that would
successfully stop every form of violence. in the dyad’s relationship. The ideal
clinical intervention method to attain this goal (e. g. eliminating elder physical
abuse) would be a family centered therapy approach that focuses on the dyad as
the target of intervention rather than the adult child or the older adult separately.
Strengthening the existing elderly parent - adult child bond is the most effective
and least intrusive way to protect the older adult. After the assessment stage is
completed, clear treatment goals must be set, sound implementation plans should

be jointly formulated, and a termination plan outlined. To accomplish the above

172

mentioned goals in a relatively short time fi'ame we could consider the use of the
brief solution-focused family therapy model as developed by de Shazer and Insoo
Kim Berg (1997) and expanded on by Hoyt (2000). The logic for using the brief
solution focused therapy approach is threefold: (a) constructing solutions are
faster than dissolving problems (especially when child abuse was an issue while
adult child was growing up) and success based upon brief therapy is of essence
for older adults (given the possibility of fast deteriorating medical conditions and
the approach of the dead drop stage), (b) it is preferable to repeat already
successful behavior patterns and build on them to improve relationships rather
than to teach and practice new models given the short timefiames available for
therapy, and (c) activities that center arormd finding solutions (e. g mainly to
protect the older adult) are different from problem solving activities where the
focus is to build safety for the elderly. Since the focus of this study was not to
design new therapy interventions, for more information on how to provide brief
solution focused therapy the works of de Shazer, Insoo Kim Berg, and Hoyt
should be considered.

. Ifeither the elderly parent and or the adult child exhibit symptoms of
psychopathology, the marriage and family therapy clinician should consider
referring the individual(s) to a psychiatrist or mental health professional for those
problems. The issue is not whether the marriage and family therapist can provide
individual mental health therapy rather splitting up the dyad and providing them
with individual sessions while conjoint therapy is in progress could present

unforeseen difficulties (e. g. redefining the problem and focus on individual

173

mental health issues instead of building solutions). A team approach (e. g.
coordination between marriage and family therapist, psychiatrist, family
physician, etc.) could help and support the difficult decisions that clinicians must
continually make to ensure the safety of the older adult as well as to improve the
dyads relationship. Being up to date with information on the client’s medications,
“recent meltdowns”, or physical pain and suffering is not optional but a must
when dealing with older adults. Since human systems are fluid, changing, and
continually evolving, the therapist needs the opinions and information obtained
fiom other team members to continuously adapt the therapy process to fit the
dyad’s situation and to achieve optimal results. Finally, since the focus of this
study was not on how to establish and maintain profitable social networking,
professional information on how to create such networks should be consulted.

. Recent trends in US. health care systems indicate that older adults with long-
term, chronic health problems are increasingly being cared for at home by family
members such as an adult child (Lim & Zebrack, 2004). Elder abuse prevalence
studies suggest that caring for a chronically ill family member impacts multiple
dimensions of caregivers' lives such as physical, psychological, social, and
financial aspects that might result in increased stress, psychopathology, and drain
in finances (Steinmetz, 1990). Because of these increasing challenges and
responsibilities, caregivers may feel overwhelmed, tired, isolated, and concerned
for their own financial survival. The financial burden (e.g. missing work, taking
personal days, and quitting or retiring early to provide care) due to caring for an

elderly parent with chronic and severe physical and psychological illnesses could

l74

be crippling to most middle class families. US news media consistently reported
over the last few years that an increasing number of elderly suffering fi'om
chronic and severe disabilities were abandoned (“granny dumping”) in addition to
possibly being abused by family members usually at hospital emergency rooms
(Berk, 2008). Marriage and family therapist should possess up to date and
accurate knowledge about the enormous toll (e. g. physical, psychological, and
financial) it takes to care for a severely disabled older adult and refiain fiom
judging or blaming an adult child caregiver who might be experiencing anger,
anxiety, and distress. Instead the client must be compassionately helped. Helping
might take the form of providing information on community resources,
completing referrals, and actively advocating on behalf of the client.

. Finally, clinicians should known the physical (visible) signs of elder abuse and
the elderly patient must be reassessed for the potential of physical abuse if signs
are present, at any stage of therapy. Michigan laws (Public Act 519 of 1982,
commonly called the APS law) require that physicians, therapists, and certain
public officials report all suspected cases of elder abuse to proper law
enforcement authorities. It is within the jurisdiction of Michigan APS to decide

whether elder abuse has taken place and not the role of clinicians.

175

Strengths, Limitations, and Directions for Future Research

There are several strengths in this study worth mentioning First, the measurement
instnnnent used to collect elder abuse relevant information was validated and calibrated
in a pilot study and based upon a large sample. By employing the random digit dialing
sampling method the study sample was as close to a nationally representative random
sample as possible. To add refinements to MIDUS II and create a relatively proportional
sample, males and certain ethnic groups were oversampled. Although most cases did not
meet this study’s eligibility requirements (65 years or older and living with an adult
child) and, nevertheless a large sample (over 200) was available for statistical analyses.
The latent variables used in this study were well defined (exiting and accepted definitions
used) and substantively more meaningful in their own right and their interrelationships
than simple manifest indicators overwhelmingly used in previous studies. Both the
measurement and structural parts of the models were based on the substantive
relationships of interest as set forth by Schiamberg and Gans (2000) in their “Ecological
Bi-Focal Model for Elder Abuse”. Data analysis was based upon a large national random
sample, sound mathematical equations, and carried out with the aid of LISREL modeling.
Therefore results obtained are generalizable. The empirical evidence provided by this
study, including the moderate or non-significant estimate loadings and paths, suggest that
the Ecological Bi-Focal Model for Elder Abuse framework was soundly conceptualized,
(Schiamberg & Gans, 2000) and provide a useful explanation of ecological risk and
protective factors involved in the process of elder physical abuse. In addition, our
findings have successfully broadened the way we conceptualize specific latent risk

factors, such as the Victim Health, and clarified how exactly it influences and or predicts

176

elder physical abuse. Finally, the importance of the relationship quality between the
elderly parent and adult child was examined and findings suggest that it definitely
frmctions as a mediator and to some extent even as a moderator (e. g. increased the
magnitude of estimates and changed their sings). In sum, empirical evidence confirmed
the accuracy and relevance of the Ecological Bi-Focal Model for Elder Abuse theoretical
framework in describing the ecological risk dimensions of elder physical abuse
(Schiamberg and Gans, 2000).

This study has several limitations that highlight opporttmities for future research
directions. The MIDUS II surveys obtained answers for the risk dimensions and elder
physical abuse prevalence from one source, the physically abused older adult. Therefore
the veracity of answers could not be verified. As such, intervention methods (e. g. crisis
and long term therapy) developed based upon this study have a limited scope and might
not be applicable until firrther research confirmed findings.

This study has several limitations that highlight opportunities for futme research
directions. The MIDUS II surveys obtained answers for the risk dimensions and elder
physical abuse prevalence from one source, the physically abused older adult. Therefore
the veracity of answers could not be verified. As such, intervention methods (e. g.
including crisis and long term therapy) developed based upon this study have a limited
scope and might not be applicable until findings are confirmed by future research.

Since the aim of the MIDUS research was not specifically directed towards
measuring elder physical abuse, several limitations attributed to the MIDUS H research
design have to be considered. First, for this study archival and legal (e. g. police reports)

files were not available to confirm the accuracy and veracity of respondents indication of

177

physical abuse in the self-report instrument. Second, the use of self-report in the MIDUS
II questionnaires might have limitations such as over-reporting or rmder-reporting along I
ethnic and gender lines (e. g. due to overrepresentation of the female gender and
Caucasian race in the dissertation sample the physical abuse distribution might be
skewed) (Griffin & Williams, 1992; Shaughnessy et al., 2006). Third, the direct manifest
outcome variables (e. g. ever physically abused and ever sexually abused) did not
specifically measure or refer to physical abuse occmring after age 65 and committed by
adult children. Finally, shame, dependency, and unwillingness to admit abuse due to
concerns for themselves as well as the abusive adult child might have resulted in serious
under-reporting (Griffin & Williams, 1992; Quinn & Tomita, 1986) of physical abuse.

The MIDUS H project was designed to examine the role of behavioral,
psychological, and social factors in rmderstanding age-related differences in physical,
emotional, and mental health, (Ryff et al., 2007). Although the MIDUS H data collecting
instruments were pilot tested for reliability and validity, it was not developed to
specifically measure ecological risk factors dming the process of elder physical abuse.
Therefore the next step in establishing the Ecological Bi-Focal Model as the preeminent
theoretical fi'amework for elder physical abuse would be the creation of a specific
measm'ement instrument that would capture and measure several ecological dimensions
of elder physical abuse. This instrument could be made available to researchers, APS
workers, physicians, and clinicians treating physically abused elderly patients and their
families.

Another limitation was the fact that both manifest dependent indicators of the

elder physical abuse dimension were discrete and dichotomous with very unique

178

distribution properties. Consequently these measurements might not have captured as
many dimensions of the elder abuse construct as a continuous variable. It would be
advisable to develop a continuous measure for the elder physical abuse manifest
indicators and re-nm the LISREL models on the MIDUS II data set. Consequently the
elder physical abuse measurement instrument will have continuous outcome direct
indicators with normal distributions properties.

The existing literature on elder physical abuse lacks models that tested the
mediating/ moderating role of the relationship between the elderly parent and her/ his
adult child. Therefore this study appears to be the first that examined the influential
aspects of this relationship construct in the process of elder abuse. Findings emerged
fiom this study suggest the Victim/ Adult Child Relationship Quality construct might
constitute a partial dimension of the Social Relations, (Antonucci, 2001) latent ecological
factor. As such the Social Isolation and thim/ Adult Child Relationship Quality latent
variables should have been combined into one factor (Social Relations) and the models
tested again. Future research could use the Social Relations ecological construct to test
correlations between ecological risk dimensions and elder physical abuse.

This study had no available measures to test how societal ageism (macrosystem)
and life transitional phases (chronosystem) would influence elder abuse and its risk
dimensions. A consideration of ageism as it affects elder abuse is important because it
facilitates a contextual understanding of economical, political, and social dimensions of
society within which elder abuse occurs. Devaluing the elderly (e. g. greedy senior
citizens) covertly justifies certain discriminatory behaviors as well as abusive activities

(e.g. financial exploitation and elder abuse). In addition, transitional phases in life (e. g.

179

terminal decline) and its effects on elder abuse are also missing from this study due to
unavailable measures. A complete ecological model on elder abuse will have to include
these two dimensions as well. The testing of these two systemic levels as they influence
elder abuse would be a very important contribution to scholarly literature hence a future
research direction in the field of elder abuse.

Finally, to examine the extent to which this study would replicate in a non
MIDUS II sample, a model cross-validation might be required in order to select the
“best” model among a set of alternatives to avoid the possibility of the accepted model
only fitting the data-base for which it was developed Given the limited scope this study,

models cross-validation was not possible at this time.

Concluding Thoughts

This study provided a window into the interactions of several ecological risk
factor dimensions and their individual and interactional effects on elder physical abuse.
The novel approach to conceptualizing the Victim Health construct offered a holistic view
of how various sub dimensions (e. g. physical illness and psychological/ emotional
problems) of this construct interact with each other and influence in turn elder physical
abuse. Findings suggest that negative hmnan emotions such as anger, anxiety, and fear
are the strongest predictors of elder abuse and not the diagnoses attached to the presence
of certain illnesses such as Alzheimer. In addition, this study’s results confirm the vital
significance the relationship quality between the elderly parent and adult child plays in

the process of elder abuse. Finally, this study proved to be a useful first step in

180

confirming the assumptions and conclusions of the Ecological Bi-Focal Model for Elder
Abuse as developed by Schiamberg and Gans (2000).

While far from comprehensive, this study provides important and novel
contributions to the theoretical literature on elder abuse because: (1) there is a paucity of
research, especially with a quantitative designs, on elder abuse in the field of marriage
and family therapy, (2) it is the first study that empirically examined the Ecological Bi-
Focal Model for Elder Abuse, (Schiamberg & Gans, 2000), (3) provided empirical data
that confirmed the robust significance of victim/ abuser relationship as a mediator
between risk factors and elder abuse, and (4) suggested useful marriage and family
therapy implications when working with the elderly, (5) recommended new areas of

research programs in the field of elder abuse.

181

APPENDIX A

Survey Protocol
. An advance letter was sent to all respondent with all information necessary to
know about the MIDUS research.
. A week later respondents were called and the phone survey process completed.
When the telephone interview was completed, a check for $25.00 was sent out to
the respondent within one week
. A week after checks were sent out a packet was mailed to participants containing
the followings: (a) a letter explaining how to complete the mail sm'veys, (b) two
SAW booklets, (c) a tape measure to be used in providing body measurements,
(d) a survey reply envelope with postage paid, (e) and a pre-incentive of ten
dollars for completing the survey.

. One week after the SAQ surveys were mailed out, a postcard was sent to
encourage participants to complete and mail back the surveys. If a respondent did
not return the SAQ within five weeks of receiving it, another packet was mailed
out — without the ten dollars incentive.
. If a respondent did not return the SAQ within the next three weeks of receiving it,
another packet was mailed out — without the ten dollars incentive.

. If three months have elapsed fiorn the initial mail (first packet), a phone call was
made to the respondent to encourage them to mail back the survey, and to provide
information about the cognitive survey.
. When the completed SAQ’s were received by MIDUS staff participants were sent

a check for 25 dollars.

182

9.

10.

11.

12.

One week afier the payment for the mail survey was sent, cases were fielded for
the Cognitive telephone survey.

Before the fielding of Cognitive surveys another packet was sent to participants
contacting: (a) a package of post it note with the MIDUS logo, (b) a letter of
request for participation, and (c) a check for 25 dollars.

Two weeks later the fielding of Cognitive surveys started even if the respondent
had not mailed back their SAQ’s.

Before concluding the survey process a telephone version of the SAQ was
developed by the MIDUS researchers which contained approximately 25% of the
items in the original SAQ. These questions reflected the areas that investigators
believed to be the most valuable questions fiom the SAQ, but were limited to this
number to create a viable telephone survey length for what came to be called the

SAQ By Phone effort.

13. Respondents that failed to ma] back the SAQ’s were fielded for this effort in the

summer of 2006, and nearly half of them completed the telephone version of the

SAQ surveys.

183

APPENDIX B
MIDUS Sample
The initial MIDUS sample was composed of the four specific subsamples: (a) a

national RDD (random digit dialing) sample of (N=3,487), (b) oversamples from five
metropolitan areas in the US. (N=757), (c) siblings of individuals fiom the RDD sample
(n=950), and a national RDD sample of twin pairs (N=l ,914). Theoretically every adult,
who was non-institutionalized-and English language speaking, within the perimeters of
the United States were eligible to participate in the MIDUS research project.
Oversampling of elderly and of men was achieved by altering the likelihood of
completing the interview as frmction of sex and age and then randomly selecting from
sample acquired. The purpose of this type of oversampling was to advance targeted
research agendas. If these targeted respondents did not complete the interview the
household was dropped from eligibility. Selected respondents were invited to complete a
phone interview of approximately 30 minutes in length followed by two mailed self—
administered questionnaires (SAQs), each approximately 45 pages in length. Once the
two SAQ questionnaires were completed, respondents had to mail them back in a self
addressed and stamped envelope. Financial incentives were used to maximize
participation and respondents who completed all phases of data collection received $20.
Of respondents who reported having one or more siblings, 529 individuals were
randomly selected. Only siblings that shared the same biological mother and father were
selected for sampling purposes. This method yielded 1, 480 eligible participants. Out of
this sample 950 respondents successfirlly completed the phone interview, some of whom

(678) came from the same family. In order to recruit twin pairs for this study a two

184

 

pronged sampling design was employed. First a national sample of 50,000 households
was screened for the presence of a twin. Then 14.8% of respondents, that identified a
twin, were asked for permission to contact their sibling to solicit their participation as
well. 60% of the eligible recruits gave permission to contact their twin sibling In
addition, the twin subsample was administered a short screener to assess zygosity and
other twin-specific information. In conclusion, this study was groundbreaking due to its
broad scientific scope (age related differences in mental and physical health due to
psychological, behavioral, and social factors) its large and diverse samples, and its

creative use of both phone and mail surveys.

185

REFERENCES

Acierno, R. (2003). Elder mistreatment: Epidemiological assessment IN RJ. Bonnie &
R.B. Wallace (Eds). In Elder mistreatment: Abuse, neglect and exploitation in an
(Acierno, 2003)aging America (pp. 261-302). Washington, DC: National
Academic Press.

Administration for Children and Families, US. Department of Health and Human
Services, (2006). NIS-4 Rockville: MD.

Agnew, R., & Huguley, S. (1989). Adolescent violence towards parents. Journal of
Marriage and Family, (51), 699-711.

Allison, PD. (2003). Missing data techniques for structural equation modeling. Journal
of Abnormal Psychology, 112, 545-557.

Anetzberger, G.J., Korbin, J.E., & Austin, C. (1994). Alcoholism and elder abuse.
Journal of Interpersonal Violence, 9(2), 184-193.

Antonucci, T.C. (2001). Social Relations: An examination of Social Networks, Social
Support, and Sense of Control. In Handbook of the Psychology of Aging (pp. 427-
453). Academic Press.

Apter, M.J. (1982). The Experience of Motivation: A Theory of Psychology Reversals.
Chichester and New York: NY: Academic Press. '

Athens, L. (1980). Violent Criminal Acts and Actors: A symbolic Interactionist Study.
London: Routledge, Kegan & Paul.

Athens, L. (1992). The Creation of Dangerous Violent Criminals. Urbana: University of
Illinois Press.

Athens, L. (1998). Types of Violent Persons: Toward the development of a Symbolic
Interactionist Theory of Violent Criminal Behavior. In N.K Denzin (Ed), Studies
in Symbolic Interaction (V 01. 78, pp. 259-268).

Bandura, A. (1973). Agression: A social learning analysis. NJ : Englewood Cliffs:
Prentice-Hall.

Baron, R.M., & Kenny, DA. (1986). The Moderator-Mediator Variable Distinction in
Social Psychological Research: Conceptual, Strategic, and Statistical
Considerations. Journal of Personality and Social Psychology, 51(6), 1173-1182.

Baskin-Sommers, A. & Sommers, I. (2006). Methamphetamine use and violence. Journal
of Criminal Justice, 34, 661-674.

186

Bengston, V.L.& Allen KR. (1993). The life course perspective applied to families over
time. In In P.G. Boss, W.J. Doherty, R. LaRosse, WR. Schram, & SIC Steinmetz
(Eds), Sourcebook of families, theories, and methods: A contextual approach (pp.
469-499). New York: Plenum.

Bennet, G.C. & Kingston, RA. (1993). Elder Abuse: Theories, Concepts, and
Interventions. Boston: Chapman & Hall.

Beresford, T.P. & Lucey, MR. (1999). Ethanol metabolism and intoxication in the
elderly. In In T.P Beresford & E.S. Gomberg (Eds), Alcohol and aging. (pp. 117-
127). New York: Oxford University Press.

Berk, LE. (2008). Exploring Lifespan Development (lst ed.). Pearson Education, Inc.
Boston: MA.

Bluhrn, C., Widiger, T.A., & Miele, G.M. (1990). Interpersonal complementarity and
individual differences. Journal of Personality and Social Psychology, 58, 464-
47 1.

Blumer, H. (1962). Society as symbolic interaction. Human Behavior and Socail
Processes. Boston: Houghton-Mifflin.

Bolger, N., DeLongis, A., Kessler,& RC. (1989). Effects of Daily Stress on Negative
Mood. Journal of Personality and Social Psychologr, 5 7(5), 808-818.

Bonnie, R.J. & Wallace, RB. (2003). Elder mistreatment: Abuse, neglect, and
exploitation in an aging America. Washington DC: National Academic Press.

Bronfenbrenner, U. (1979). The ecology of human development. MA: Cambridge:
Harvard University Press.

Bronfenbrenner, U. & Evans, G.W. (2000). Developmental science in the 21st century.
(Bronfenbrenner & Evans, 2000)Emerging theoretical moels, research designs,
and empirical findings. Social Development, 9, 115-125.

Browne, K.D. & Hamilton, CE. (1998). Physical Violence Between Young Adults and
Their Parents: Associations with a History of Child Maltreatment. Journal of
Family Violence, 13(1), 59-79.

Bubolz, M & Sontag, S. (1993). Human Ecology Theory. In P.G. Boss, W.J. Doherty,
R.LaRosa, W.R Schram, & SJC Steinmetz (Eds) Source book of families, theories,
and methods: A contextual approach. (pp. 419-448). New York: Plenum.

Buss, D.M. (2005). The murderer next door: Why the mind is designed to kill. New York:
Penguin.

187

Butler, RN. (1969). Age-ism: Another form of bigotry. The Gerontologist, 9, 243-246.
Census USBot (2003). Statistical Abstract of the United States. Washington DC.

Chermack, S.T., Walton, M.A., Fuller, B.E.& Blow, EC. (2001). Correlates of expressed
and received violence across relationship types among men and women substance
abusers. Psychology of Addictive Behavior, 15, 140-151.

Cochran, D., Brown, M. E., Adams, S. L., & Doherty, D. (1994). Young adolescent
batterers: A profile of restraining order defendants in Massachusetts (Office of
the Commissioned on Probation). Boston: Office of the Commissioned on '
Probation,

Cohen, C.I., Onserud, H., & Monaco, C. (1993). Outcomes for the mentally ill in a
program for older homeless persons. Hospital and Community Psychiatry, 44,
650-656.

Cohen, C.I., Tereesi, 1., & Holmes, D. The mental health of old homeless men. Journal of
the American Geriatrics Society, 36, 492-501.

Cooney, C. & Mortimer, A. (1995). “Elder Abuse and Dementia”. International Journal
of Social Psychiatry, 41(4), 276-283.

Cooney, C., & Howard, R. (1995). Abuse of patients with dementia by their caretakers:
out of sight but not out of mind. Internationsl Journal of Geriatric Psychiatry, 10,
73 5-741. ~

Cooper C., Katona C., Finne-Soveri, H., Topinkova, E. Carpenter, G.I., & Livingston, G.
(2006). Indicators of elder abuse: A cross-national comparison of psychiatric
morbidity and other determinants in the Ad-HOC study. American Journal of
Geriatric Psychiatry, 14, 489-497.

Comijs, H.C., Penninx, B.W.J.H, Knipscheer, K.P.M., & van Tilburg, W. (1999).
Psychological distress in victims of elder mistreatment: The effects ofsocial

support and coping. Journal of Gerontology: Psychological Sciences, 543, 240-
2.45.

Cottrell, B. (2001). Parent abuse: The abuse of parents by their teenage children.
Retrieved March, 10, 2009.

Coyne, A. C., Reichman, W. E. & Berbig, L. J. (1993). The relationship between
dementia and elder abuse. American J orn‘nal of Psychiatry, 150(4) 643 — 646.

Day, RD. (1995). Family systems theory. In R.Day, K.R. Gilbert, B.H. Settles, & W.R.

Burr (Eds). Research and theory in family science. (pp. 91-101). Pacific Grove,
CA: Brooks Cole.

188

de Shazer, S. & Berg, I.K. (1997). "What works?": Remarks on research aspects of
solution-focused brief therapy. Journal of Family Therapy,(19), 121-124.

Dewey, J. (1922). An Introduction to Social Psychology. New York: Henry Holt and
Company.

Diamantopoulos, A. & Siguaw, J .A. (2006). Introducing LISREL. Thousands Oaks: CA:
Sage Publications Ltd.

Dooley. W.K., & Schulz, R. (2005). Caregiver Resentment: Explaining Why Care
Recipients Exhibit Problem Behavior. Rehabilitation Psychology, 50(3), 215—223.

Elder Mistreatrnent: Abuse, Neglect, and Exploitation. Washington, DC: National
Research Council Panel to Review Risk and Prevalence of Elder Abuse and
Neglect. National Center on Elder Abuse.

Flannery, RB. (2003). Domestic violence and elderly dementia sufferers. American
Journal of Alzheimer's Disease and other Dementias, 18, 21-23.

Fullmer, T., Guadagno, L., Bitondo, Dyer, C., & Connolly, M.T. (2004). Progress in
elder abuse screening and assessment instruments. Journal of the American
Geriatrics, 52 (2) 297-304.

Fulmer, T., Paveza G., , VandeWeerd, C., F airchild, S., Guadagno, L., Bolton-Blatt, M.,
& Norman, N. (2005). Dyadic vulnerability and risk profiling for elder neglect.
The Gerentologist, 45(4) 525-534.

Gambert, S.R. & Katsoyannis, KK (1995). Alcohol-related medical disorders of older
heavy drinkers. In T.P. Beresford & E. Gomberg (Eds). Alcohol and aging. (pp.
70-81). New (Hudson, 1986)York: Oxford University Press.

Garre-Olmo, J., MD, Planas-Pujol, X., Lopez-Pousa, S., Juvinya, D., Vila, A., & ,
Vilalta-Franch, J. (2009). Prevalence and Risk Factors of Suspected Elder Abuse
Subtypes in People Aged 75 and Older. Journal of American Geriatrics Society,
5 7, 815—822.

Geerlings, S.W., Beekman, A.T.F, Deeg, D.J., Twisk, J.W.R., & van Tilburg, W. (2001).
The longitudinal effect of depression on functional limitations and disability in
older adults: An eight-wave propspective community-based study. Psychological
Medicine, 31, 1362-1371.Hoffman, DA. (2005). Tiagabine for Rage, Aggression,
and Anxiety. American Psychiatric Press, Inc, 1 7, 252.

Gelles, R. J., & Cornell, C. P. (1990). Intimate Violence in Families (2nd ed.). Beverly
Hills, CA: Sage.

189

Glasser, W. (1998). Choice Theory: A New Psychology of Personal Freedom. Harper
Collins.

Gray, J .A. (1987). The psychology of fear and stress. Cambridge, England: Cambridge
University Press.

Griffin, L.W. & Williams, 0.]. (1992). Abuse among Afiican-American Elderly. Journal
of Family Violence, 7(1), 19-35.

Griffore, R.J., Braboza, G.E., Mastin, T., Oehmke, J., Schiamberg, L.B., & Post, LA.
(2009). Family Members Reports of Abuse in Michigan Nursing Homes. Journal
of Elder Abuse and Neglect, 21, 105-1 14.

Gurtrnan, M. B. ( 1997). The interpersonal Circumplex. Retrieved May 15, 2009 from
hip://www.uwp.edu/academic/psychology/faculg/netcirchtm

 

Hareven, T.K. (1978). Introduction: The historical study of the life cornse. In T.K
Hareven (Eds. ), Transitions: The family and the life course perspective (pp. 1-16).
New York: Academic Press.

Hareven, T.K. (1995). Historical perspectives on the family and aging. In In R Blieszner
& V.H. Bedford (Eds. ), Handbook of aging and the family, Westport, CT:
Greenwood Press.

Heide, KM. (1995). Why Kids Kill Parents: Child Abuse and Adolescent Homicide.
Beverly Hills: Sage.

Heisel, M.J. (2006). Suicide and its prevention among older adults. Canadian Journal of
Psychiatry, 51, 143-154.

Holtzworth-Mrmroe, A., Meehan, J.C., Rehman, U., & Marshall AD. (2002). Intimate
Partner Violence: An Introducation for Couple Therapists. In Clinical Handbook
of Couple Therapy (3rd ed., pp. 441-465). New York and London: The Guilford
Press.

Homer, A. & Gilleard, C. (1990). Abuse of Elderly People by their Carers. British
Medical Journal, 301, 1359-1362.

Hoyt, M.F. (2000). Some stories are better than others: Doing what works in brief
therapy and managed care. Philadelphia: anner/ Maze].

Hudson, M.F. (1986). Elder mistreatment. Current Research. In KA Pillemer & R.S.

Wolf (Eds. ), Elder abuse: Conflict in the family (pp. 125-16). Dover, MA: Auburn
House Publishing Company.

190

 

Hurlburt, M.S., Hough, R.L., & Wood, P.A. (1996). Effects of substance abuse on
housing stability of homeless mentally ill persons in supported housing.
Psychiatric Services, 4 7, 721-726.

Hysing, M., Elgen, 1., Gillberg, Ch., Lie, Stein, A. & Lundervold, A. J. (2007). Chronic
Physical Illness and Mental Health in Children: Results from a Large-Scale
Population Study. Journal of Child Psychology and Psychiatry, 48(8), 785-792.

Ivey, D.C., Wieling, E., & Harris, S.M. (2004). Save the Young—the Elderly Have Lived
Their Lives: Ageism in Marriage and Family Therapy. Family Process, 39(2),
1 63-1 75.

Jackson, D. (2003). Broadening constructions of family violence: Mother's perspectives
of aggression from their children. Child and Family Social Work, 8, 321-329.

J oreskog, K & Sorbom, D. (2001). LISREL 8: User's Reference Quide. LincolnwoodzlL:
Scientific Software International.

Karney, B.R., & Bradbury, T. N. (1995). The longitudial course of marital quality and
stability. A review of theory, method, and research. Psychological Bulletin, 118,
3-34.

Kennedy, G.E., & Tanenbarnn, S. (2000). Suicide and aging: International Perspective.
Psychiatric Quarterly, (71), 345-362.

Killian, T., Turner,J., & Cain, R (2005). Depressive symptoms of caregiving women in
midlife: The role of physical health. Journal of Women and Aging, 1 7, 115-127.

Kline, RB. (2005). Principles and Practice of Structural Equation Modeling (2nd ed.).
New York, NY: The Guilford Press.

Kleinschmidt, KC. (1997). Elder abuse: A review. Annals of Emergency Medicine, 30,
463-472.Kessler, R C., Andrews, A., Mroczek, D., Ustun, B., & Wittchen, H. U.
(1998). The World Health Organization Composite International Diagnostic
Interview Short-Form (CIDI-SF). International Journal of Methods in Psychiatric
Research, 7, 171-185.

Kessler, R. C., Mickelson, K D., & Williams, D. R. (1999). The prevalence, distribution,
and mental health correlates of perceived discrimination in the United States.
Journal of Health and Social Behavior, 40, 208-230.

Kethineni, S. (2004). Youtlr-on-parent violence in a central Illinois county. Youth
violence and Juvenile Justice, 2(4), 374-394.

Kiesler, D. J. (1983). The 1982 interpersonal circle: A taxonomy for complementarity in
human transactions. Psychological Review, 90, 185-214.

191

 

Korrapati, M.R. & Vestal, RE. (1995). Alcohol and medications in the elderly: Complex
interactions. In T.P. Beresford & E. Gomberg (Eds. ), Alcohol and aging. (pp. 42-
69). New York: Oxford University Press.

Kosberg, J.I. Preventing elder abuse. Identification of high risk factors prior to placement
decisions. The Gerontologist, 28(1), 43-50.

Kratcoski, P. C. (1984). Perspectives on intrafarnily violence. Human Relations, 37, 443-
453.

Lachs, M.S., Williams, C.S., O'Brien, 8., Pillemer, K.A., & Charlson, ME. (1998). The
mortality of elder mistreatment. Journal of the American medical Association,
280, 428-432.

Lachs, M.S., Mark, K., & Pillemer, KA. (2004). Elder Abuse. The Lancet, 364, 1192-
1263.

Laurent, A., & Derry, A. (1999). Violence of French adolescents toward their parents.
Journal of Adolescent Health, 25(1), 21-26.

Lauman, E.O., Leitsch, S.A., & Waite, LJ. (2008). Elder Mistreatrnent in the United
States: Prevalence Estimates From a National Representative Study. The Journal
of Gerontology, 633(4), 248-254.

Leary, T. (1957). Interpersonal diagnosis of personality. New York: Ronald
LeDoux, J. (2000). Emotion circuits in the brain. Annual Review of Neuroscience, 23,
1 55-1 84.

Little, RJ. & Rubin, D.B. Statistical analysis with missing data. New York: Wiley.
Ormrod, LE. (1999). Human Learning (3 ed.). Upper Saddle River, NJ: Prentice Hall.

MacCallmn, R.C., Browne, M.W., & Sugawara, HM. (1996). Power analysis and
determination of sample size for covariance structure modeling. Psychological
Methods, 1, 130-149.

McCloskey, L.A. & Lichter, EL. (2003). The contribution of marital violence to
adolescent aggression across different relationships. Journal of Interpersonal
Violence, 18(4), 390-412.

McDonald, R.P., & Ho, M.H.R. (2002). Principles and Practice in Reporting Structural
Equation Analyses. Psychological Methods, 7(1), 64-82.Midlife Development in
the United States (MIDUS H), 2004-2006 [Computer file. ICPSR04652-v1.].
(2007). Ann Arbor, MI: Inter- University Consortium for Political and Social
Research [Producer and Distributor].

192

 

Macdonald, S., Angling-Bodrug K, Mann, R.E., Erickson, P., Hathaway, A., Chipman,
M. A., & Rylett, M. (2003). Injury risk associated with cannabis. Drug Alcohol
Dependence, 72, 99-115.

McGraw & Hill. (2002). Behavior Problems. In McGraw-Hill Concise Dictionary of
Modern Medicine, The McGraw-Hill Companies, Inc.

Meeks-Sjostrum, D. (2004). A comparison of three measures of elder abuse. Journal of
Nursing Scholarship, 3 6, 247-250).

Mongin, P. (1997). Expected Utility Theory. In J. Davis, W. Hands, and U. Maki (Eds. ),
Handbook of Economic Methodology (pp. 342—350). London: Edward Elgar.

Morrione, T. (2004). Herbert Blumer. Edited and Introduced by Thomas Morrione.
George Herbert Mead and Human Conduct. Walnut Creek, CA: AltaMira Press.

Muthen, B. (1984). A general structural equation model with dichotomous, ordered
categorical, and continuous latent variable indicators. Psychometrika, 49, 115-
1 32.

Muthen, L.K., & Muthen, B.O. (1998-2007). Mplus User's Guide. (5 ed.). Los Angeles,
CA: Muthen & Muthen.

Mroczek, D.K & Kolarz, CM. (1998). The effect of age on positive and negative affect:
A developmental perspective on happiness. Journal of Personality and Social
Psychology 75(5), 1333-1349. ‘

National Research Council Panel to Review Riskand Prevalence of Elder Abuse and
Neglect (2003). Elder maltreatment: Abuse, neglect, and exploitation in an aging
America. Washington DC.

Nebbitt, V., Lombe, M., & Williams, J. (2008). Assessing the moderating effects of
anxiety sensitivity on antisocial behavior among urban Afiican American youth.
Journal of Health Care for the Poor and Underserved, 19, 277-293.

Nock, M.K. & Kazdin, A.E. (2002). Parent-directed physical aggression by clinic-
referred youths. Journal of Clinical Child Psychology, 31(2), 193-205.

O'Leary, K.D., Malone, J., & Tyree, A. (1994). Physical agression in early marriage: Pre-
relationship and relationship effects. Journal of Consulting and Clinical
Psychology, 62, 594-602.

Pagani, L.S., Tremblay, R.E., Nagin, D., Zoccolillo, M., Vitaro, F., & McDuff, P. (2004).

Risk factors models for adolescent verbal and physical aggression toward
mothers. Journal of Behavioral Development, 28(6), 528-537.

193

Parra-Cardona, J .R., Meyer, E., Schiamberg, L., & Post, L. (2007). Elder Abuse and
Neglect in Latino Families: An Ecological and Culturally Relevant Theoretical
Framework for Clinical Practice. Family Process, 46(4), 451-470.

Patrick, C.J., Curtin, J.J. 8c Tellegen, A. (2002). Development and validation of a brief
form of the multidimensional personality questionnaire. Psychological
Assessment, 14(2), 1 50—1 63 .

Paulson, M.J., Cooms, R.H.,& Landsvert, J. (1990). Youth who physically assault their
parents. Journal of Family Violence, 5, 121-133.

Paveza, G.J., Cohen, D., Eisdorher, C.E., Freels, S., Shram, T.S., Ashford, W., Gorelick,
P., Hirschman, R., Lunchis, D., & Levy, P. (1992). Severe family violence and
Alzheimer's disease: Prevalence and risk factors. The Gerontologist, 32(4), 493-
497.

Peek, C., Fisher, J. L., & Kidwell, J. S. (1985). Teenage violence towards parents: A
neglected dimension of family violence. Journal of Marriage and Family, (47),
l 05 l -1058.

Pelletier, D. & Coutu, S. (1992). Substance abuse and family violence in adolescents.
Canada's Mental Health, 40, 6-12.

Pillemer, K 8!. Finkelhor, D. (1988). The prevalence of eldr abuse: A random sample
survey. The Gerontologist, 28, 51 -57.

Pillemer, K.A. & Wolf, KS. (1988). Helping Elderly Victims. New York: Columbia.

Pleck, A. (1987). Domestic Tyranny. The making of social policy against family violence
fiom colonial times to the present. New York: Oxford University Press.

Pliszka, S., Sherman, J ., Barrow, M. (2000). Affective disorder in juvenile offenders: A
preliminary study. American Journal of Psychiatry, 15 7, 130-132.

Quinn, MJ. & Tomita, SK (1986). Elder abuse and neglect: Causes, diagnosis, and
intervention strategies. New York: Springer Publishing Company.

Racic, M., Kusmuk, S., Kozomara, L., Debelnogic, B., & Tepic, R. (2006). The

prevalence of mistreatment among the elderly with mental disorders in primary
health care settings. The Journal of Adult Protection, 8(4), 20-24.

Raykov, T., Tomer, A. & Nesselroade, J .R. ( 1991). Reporting Structural Equation
Modeling Results in Psychology and Aging: Some Proposed Guidelines.
Psychology and Aging, 6(4), 499-503.

Raykov, T. (2005). Bias-corrected estimation of noncentrality parameters of covariance

194

 

structure models. Structural Equation Modeling, 12(1), 120-129.

Raykov, T. & Marcoulides, GA. (2006). A first course in structural equation modeling
(2 ed.). Mahwah, NJ: Lawrence Erlbaum Associates.

Rigler, S. K (2000). Alcoholism in the elderly. American Family Physician, 61, 1710-
1716.

Robinson, P.W., Davidson, L.J. & Drebot, ME. (2004). Parent abuse on the rise: A
historical review. American Association of Behavioral Science (online journal),
58-67.

Rhodes, R. (1999). Why They Kill: The Discoveries of a Maverick Criminologist. Vintage
Books, Division of Random House.

Ryff, C.D., Magee, W.J., Kling, KC., 8c Wing, EH. (1999). Forging macro-micro
linkages in the study of psychological well-being. In C.D. Ryfld’t V. W. Marshall
(Eds. ), The self and society in aging processes. (pp. 247-278). New York:
Springer Publishing.

Ryff, C., Almeida, D. M., Ayanian J .S., Carr D.S., Cleary, P. D., Coe, R. , Davidson R.,
Krueger,R.F., Lachman, M.E., Marks, N.F., Mroczek, D.K, Tersa Seeman, T.,
Seltzer, M.M., Singer,B.H., Sloan, R.P., Tun, P.A., , Weinstein,M., and
Williams, D., (2007). Midlife Development in the United States (MIDUS II),
2004-2006 [Computer file]. ICPSR04652-vl . Ann Arbor, MI: Inter-university
Consortium for Political and Social Research [distributor], 2007-03-22.
doi:10.3886/ICPSR04652.

Schiamberg, L.B. & Gans, D. (2000). Elder Abuse by Adult Children: An Applied
Ecological Framework for Understanding Contextual Risk Factors and the
Intergenerationsl Character of Quality of Life. The International Journal of Aging
and Human Development; A Journal of Psychosocial Gerontology, 50(4), 329-
359.

Schiamberg, L.B., Barboza, G., Oehmke, J ., Zhang, Z., Griffore, R.J., Weatherhill, R.
Post, L.& von Heydrich, L. (2010). Elder Abuse in Nursing Homes: An
Ecological Perspective Journal of Elder Abuse and Negelct, 23 (2).

Schiamberg, L, von Heydrich, L., Oehmke, J ., Zhang, Z., Barboza, G., Grifi‘ore, R. Post,
L.Weatherill, R., Masten, T., (2010). Physical Abuse of Older Adults in Nursing
Homes: A Random Sample Survey. Accepted for publication by the Journal of
Family Violence.

Shaughnessy, J .J ., Zechmesiter, E.B., & Zechmeister, J .S. (2006). Research Methods in
Psychology (7th ed.). New York: NY: McGraw-Hill.

195

Shinoda-Tagawa, T., Leonard, R., Pontikas, J ., McDonough, (Meddaugh, 1993)J.E.,
Allem,D., & Dreyer,D. (2004). Resident -to- resident violent incidents in nursing
homes. Journal of the American Medical Association, 291(5), 591-598.

Shrout, P.E., & Bolger, N. (2002). Mediation in experimentaland non-experimental

studies: New procedures and recommendations. Psychological Methods, 7, 422-
445.

Simon, H.A. (1980). The meaning of causal ordering. In In RK. Merton, J.S. Coleman,
and RH. Rossi (Eds), Qualitative and quantitative social research (pp. 65-81).
New York, NY: The Free Press.Snapshot. (2003). A statistical profile of older
Americans aged 65R. US Department of Health and Human Services
Administration on Aging. Retrieved September 28, 2006, fiom
hgp://www.flgov/press/fact/fact.a_sp#Snapshots.

Stanovich, K E. (2004). The robot ’s rebellion: Finding meaning in the age of Darwin.
Chicago: The University of Chicago Press.

Steinmetz, SK. (1990). Myth and reality. In TH. Brubaker (Eds. ), Family relationships
in later life. (2 ed., pp. 193-211). Newburry ParszA: Sage Publications.

Straus, M. A., & Gelles, R. J. (1988). How violent are American families? Estimates
from the National Family Violence Resrn'vey and other studies. In Hotaling, G.
F inkelhor, D., Kirkpatrick, J. & Strauss, M. (eds), Family Abuse and its
Consequences: New Directions in Research (pp. 14-37). Beverly Hills: Sage.

Talerico, K.A., Evans, L.K., 8L Strumpf, NE. (2002). Mental Health Correlates of
Aggression in Nursing Home Residents with Dementia. The Gerontologist, 42(2),
169-177.

Turvey, C.L., Schultz, S.K, & Klein, MD. (2006). Alcohol use and health outcomes in
the oldest old. Substance Abuse Treatment, Prevention, and Policy, 1(8), 1-8.

Ulman, A. & Strauss, M.A. (2003). Violence by children against mothers in relation to
violence between parents and corporal pubishment by parents. Journal of
Comparative Family Studies, 34(1), 41-60.

US. Department of Health and Human Services, Administration on Aging. (2006).
Snapshot: A statistical profile of older Americans aged 65 +. Retrieved 7 March,
2009 from ht_tp://www.aoa.gov/PRESS/fact/pdf/Attachment_l304.pdf

US. Census Bureau (2006b). Statistical Abstract of the United States (125th ed.)

Washington: DC: US. Government Printing Office.("Profiling Canada's families
111,", 2004b).

196

 

Vaillant, (3.13. & Mukarnal, K. (2001). Successilll aging. Journal ofPsychiatry, 158, 839-
847.

Vanier Institute of the Family (2004b). Profiling Canada's families 11. Retn'ved 30
September 2009 from www.vifamily.ca/profiling[3d. htrn

Vermeiren, R., Deboutte, D., Ruchkin, V., et al. (2002). Antisocial behavior and mental
health: Findings from three communities. European Child Adolescent Psychology,
1 I , 168—175.

Walsh, J .W. & Krienert, J .L. (2007). Child-Parent Violence: An Empirical Analysis of
Offender, Victim, and Event Characteristics in an National Sample of Reported
Incidents. Journal of Family Violence, 22, 41-60.

Wang, P.S., Berglund, P. & Kessler, RC. (2000). Recent care of common mental
disorder in the United States: Prevalence and conformance with evidence-based
recommendations. Journal of General Internal Medicine, 15, 284-292.

Wiggins, J. S., & Trobst, K. K (1997). When is a circurnplex an ”interpersonal
Circumplex"? The case of supportive actions. In In R Plutchik & H. R. Conte
(Eds. ), Circumplex models of personality and emotions @p. 57-80), Washington,
DC: American Psychological Assciation.

Williamson, G.M., Martin-Cook, K., Weiner, M.F., Svetlik, D.A., Saine, K, Hynan, L.S.
W. Keith Dooley. W.K., & Schulz, R. (2005). Caregiver Resentment: Explaining
Why Care Recipients Exhibit Problem Behavior. Rehabilitation Psychology,
50(3), 215-223.

Wilson, ED. (1975). Sociobiology: The new synthesis. Cambridge: Harvard University
Press.

Wolf, R.S., Godkin, M.A. & Pillemer, K (1984). Elder abuse and neglect. Clinical
Gerontology, 2, 269-276.

Wolf, R.S. (1989). Testimony before the Subcommittee of Human Services. Select
Committee on Aging; US. House of Representatives Hearings on Elder Abuse. In
Bennett, G. & Kingston, P. (I 993) Elder Abuse: Concepts, theories, and
interventions, London: Chapman & Hall.

Wolf, R.S. & Pillemer, KA. (1989). Helping Elderly Victims: The reality of elder abuse.
New York: Columbia University Press.

Zarit, S.H. & Reid,J.D. (1994). Family cargiving and the older family. In In C.B. Fisher

& Lerner, RM. (Eds), Applied developmental psychology (pp. 23 7-264)., New
Yok: McGraw-Hill.

197

 

 

Zide, M.R. & Gray, S.W. (2001). Psychopathology: A Competency-Based Assessment
Model for Social Workers. BelmontzCA: Wadsworth.

Zohar, J. (2003). Posttraumatic stress disorder: a hidden epidemic. CNS Spectrums, 8(9),
/634.

198