Socioeconomic status and functional ability among older adults

MISSING IMAGE

Material Information

Title:
Socioeconomic status and functional ability among older adults
Physical Description:
xi, 129 leaves : ; 29 cm.
Language:
English
Creator:
McCawley, Dorothy Jean
Publication Date:

Subjects

Subjects / Keywords:
Sociology thesis, Ph.D   ( lcsh )
Dissertations, Academic -- Sociology -- UF   ( lcsh )
Genre:
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Thesis:
Thesis (Ph.D.)--University of Florida, 2000.
Bibliography:
Includes bibliographical references (leaves 118-127).
Statement of Responsibility:
by Dorothy Jean McCawley.
General Note:
Printout.
General Note:
Vita.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 025860019
oclc - 47109665
System ID:
AA00013601:00001


This item is only available as the following downloads:


Full Text










SOCIOECONOMIC STATUS AND FUNCTIONAL
ABILITY AMONG OLDER ADULTS
















By

DOROTHY JEAN MCCAWLEY


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF
DOCTOR OF PHILOSOPHY

UNIVERSITY OF FLORIDA



























Copyright 2000

by

Dorothy Jean McCawley














ACKNOWLEDGMENTS

None of my accomplishments have been due to my effort alone, nor would they

have been worthwhile if I accomplished them alone. I know that this dissertation might

have been possible as a lone effort, but it would not be the experience worth celebrating.

The most important influences in my life are my parents, Jean and Jim McCawley. My

mother and father taught me that I can accomplish anything and they imbued in me the

value of education. My life would not be the same without them. My brother and sister,

Mary Zinger and Rick McCawley, and their families have also been a great source of

support and joy. The admiration from my nieces and nephews gave me an additional

incentive to complete this process.

This is an academic endeavor and as such I owe a great debt of thanks to the

faculty and staff at the University of Florida. They are too many to name, but a few

deserve special mention. I was floundering for a time until my co-chairs, Drs. Barbara

Zsembik and Donna Berardo, took me under their wings and gave me a sense of

direction. Their mentoring inspired me and helped me maintain my enthusiasm

throughout the long process of creating this dissertation. In addition, for their assistance

and committee service during my graduate education, I gratefully acknowledge the

encouragement of Drs. Leonard Beeghley, Otto von Mering, and Chuck Peek. I

appreciate the time they have given to help me become a better scholar. The

administrative staff at the Department of Sociology saved me from the grief of

bureaucracy, especially Mary Robinson and Sheran Flowers.
iii








I also owe a great debt of gratitude to my support network of dear friends and co-

workers. Diana Hannah and Paula Coley persevered in friendship despite my single-

minded focus on my academic pursuit. The staff at Fringe Benefit Coordinators,

especially Beth Lege and George Zinger, kept the office running in my absence. They

gave me peace of mind and a continuing source of income over the past seven years.

My diploma should include the names of my family and friends, the faculty and

staff at the University of Florida and F.B.C. who all helped in their own way to make this

dissertation a reality. I apologize to those I have not mentioned by name. Know that I

gratefully acknowledge your encouragement and support in my graduate career and in the

completion of my doctoral dissertation.















TABLE OF CONTENTS

page

ACKNOWLEDGMENTS ........................................... ................ iii

LIST OF TABLES ................................. ....... ..................................... vii

L IST O F FIG U R E S ............................................................................ ......................... ix

A B ST R A C T ................. .... .................................................. ............................................x

CHAPTERS

1 HEALTH AMONG OLDER ADULTS ................................... ......................1

2 DETERMINANTS OF FUNCTIONAL STATUS....................... ..................... 5

Introduction ........................................... .................................................................... 5
M models of Functional Ability...................... ........... ......... ...................... 7
Socioeconom ic Status and Health................................................... ......................... 10
The Moderating Effect of Health Behaviors and Health Indicators........................... 18
C ovariates ...................................... ..... ................................................................ 29
Summary of Socioeconomic Status and Functional Ability in Older Adults............... 37
H ypotheses............................................. ....................................................................... 39

3 THE SURVEY OF ASSETS AND HEALTH DYNAMICS AMONG THE OLDEST
O L D ..................... ......................................................................................................... 4 1

Research Sample........................... .......................................................... 41
Measures .................................................................................... ................. 48
P rocedures.......................................... .............................................................. .... 59

4 DESCRIBING CHANGE AND STABILITY IN FUNCTIONAL PERFORMANCE 62

C correlation of M easures................................................................ .................... 67
Response to H ypotheses ............................................ ....................................... 80

5 PREDICTING CHANGE IN FUNCTIONAL PERFORMANCE..............................83

Multinomial Logistic Regression Modeling ................................ ... ............... .. 83
M odel C om parisons..................................................................................... ........... 84
v









M odels of C ausation ....................................................................... ........................ 94
Response to Hypotheses .................. ....................................................... 95

6 FOCUS ON HEALTH RECOVERY ........................................................ .......98

S um m ary ............................................................... ....................................... ......... 98
Economic Resources and Functional Limitations........... ........................................ 100
Social Policy Recommendations................................................. .................... 101
C ov ariates .................................................................................. .......................... 104
Why Separate Functional Measures?.......................................... ..... 105

APPENDICES

A SELECTED CODEBOOK SURVEY QUESTIONS............................................ 113

B MULTINOMIAL REGRESSION COEFFICIENTS FOR FUNCTIONS...................109

LIST OF REFEREN CES............................................................................. .......... 18

BIOGRAPHICAL SKETCH.................................................................................... 128





























vi















LIST OF TABLES


Table Page

1. Variable Description and Coding....................................................49

2. Decile Ranges for Income and Net Worth...................... .... ...........................53

3. Descriptive Statistics of Sample..........................................58

4. Transition Matrix of Walking Several Blocks............................................ 63

5. Transition M atrix of Climbing Stairs............................................................................. 64

6. Transition Matrix of Pushing/Pulling Large Objects...................... .........................65

7. Transition M atrix of Lifting 10 Pounds.......................................... .....................66

8. Transition M atrix of Picking Up a Dime................................... ...... .....................66

9. Correlation Matrix: Correlation with changes in Functional Performance Between
W aves 1 and 2..................... ........................ ................. ................ 68

10. Correlation Matrix: Correlation with SES Variables.................. ....................... 70

11. Correlation Matrix: Correlation with Health Behaviors ................... ...................74

12. Correlation Matrix: Correlation with Genetic Endowments...................................75

13. Correlation Matrix: Correlation with Covariates....................... .....................62

14. Correlation Matrix: Correlation between all variables ................... .......... ........... 77

15. The Odds Ratios of Being in a Stable with Limitations State...................................... 85

16. The Odds Ratios of Declining Functional Status............................ ......... .....89

17. The Odds Ratios of Recovery of Functional Status........................ .......................93

18. M odel C om parisons........................................................... .. ....................95








19. Net Worth Comparisons: Indirect Effects of SES through Health Behaviors.................97

20. Summary of Transition M atrices ........................................... ......... ...........100

21. Health Behaviors and Functional Abilities........................................................ 103

22. Predicting Health Recovery in Walking................................................... 113

23. Predicting Health Recovery in Climbing Stairs....................... ....... ............... 114

24. Predicting Health Recovery in Pushing/Pulling .................... ....................115

25. Predicting H health Recovery in Lifting............................................................................. 16

26. Predicting Health Recovery in Picking Up a Dime...................................................117














LIST OF FIGURES


Figure Page

1. A Model of the Disablement Process.............................................. 10

2. Conceptual Model of the Effects of SES and Health Behaviors on Functional Ability....38

3. Modification of the Model of the Disablement Process Showing Variables used in this
R research ........................ ...................... ................................................. 47














Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy

SOCIOECONOMIC STATUS AND FUNCTIONAL ABILITY AMONG OLDER
ADULTS

By

Dorothy Jean McCawley

December 2000

Chair: Dr. Barbara Zsembik
Major Department: Sociology

With increasing numbers of older adults in the population, research with a focus

on healthy recovery among older adults can have an impact on quality of life. Previous

research has examined the connection between socioeconomic status and health and

found that higher SES and better health are correlated.

This research examines the health-SES link specifically as regards the process of

disability in older adults as regards functional limitations, which include difficulty with

specific physical actions, which is the focus of this research. The Asset and Health

Dynamics Among the Oldest Old survey is a longitudinal data set that includes data about

difficulty walking, climbing stairs, lifting 10 pounds, pushing or pulling a large object

and picking up a dime from a table. It also includes detailed information about finances

and health behaviors. This data set offers the opportunity to analyze the specific

functional trajectories of the respondents through the first two waves of data collection.








Net worth proved a better indicator of health function over time than was annual

household income. Health behaviors such as exercising, not smoking, controlling weight,

and temperate drinking did moderate the relationship between net worth and functional

abilities. This group is mostly free of functional limitations, and between 18 40% of

older adults with functional limitations do recover from the first wave to the second.













CHAPTER 1
HEALTH AMONG OLDER ADULTS

Today the dialogue about aging and the changing expectations of our later years

has become an urgent topic. The U.S. Bureau of the Census (1996) has followed the

growth in the population of the over age 65 group. In 1994, the over-65 group

represented one-eighth of our total population, which is an 11-fold increase from 1990.

The Bureau expects another dramatic increase in the number of older adults after 2010

when the Baby Boomers reach age 65.

As we look at the aging process and its theoretical understanding, we see that

successful aging may be linked to productivity no longer a time of disengagement, but

of increasing activity. As the number of adults over 65 years of age becomes a larger

percentage of our population in the next century this debate becomes increasingly

significant. The continuing debate over whether older adults are healthy and capable of

recovering if disabled is likely to intensify.

In order to prevent disease and disability (or promote recovery), it is useful to

examine how social factors affect the aging process and find ways to facilitate a healthy

aging process. We all desire a good quality of life in our later years, which includes high

satisfaction, health, material security and happiness (Fry, 1996). The growth in the over-

65 portion of the population points to the need to understand the dynamics of health,

aging and socioeconomic status (SES), since poor health can undermine the quality of life

of older adults.








As we age, our bodies may experience disease, injury or the exacerbation of a

congenital condition that results in physical impairment. The "process" of becoming

disabled has been described various times in attempts to arrive at some standard model

for analyzing the dynamics of disability. Verbrugge and Jette (1994) developed a model

that has several steps. First, individuals experience a medical condition that leads to

physical impairment, functional limitations, and eventually disability. They called this

"the disablement process." Next, they added factors that impact the process, which they

label as "risk factors," "extra-individual factors," and "intra-individual factors."

This research uses an expanded version of this model, one that includes health

recovery from functional limitations as an alternative outcome to disability. For this

study, recovery is defined as a reduction in the difficulty of performing specific

functions. Recovery is a significant event when evaluating the process of disability. If

we can examine the connections between social factors and health recovery, we can help

elders maintain their independence and focus attention to preventing functional

limitations, which should reverse the process away from disability.

Studying the incidence of disease is useful for tracking the history of a particular

condition. Looking at the functional consequences is more useful for public policy and

for the lifestyle of elders and their families. If a disease results in physical impairment,

the adult experiences difficulty in performing activities that allow for independent living

and self-care (George, 1996). Dependence clearly has a strong negative effect on the

quality of life satisfaction of elders, but recovery is possible. Previous studies have

demonstrated that recovery from functional disability at some point occurs considerably

often (Myers, Juster and Suzman, 1997).








A focus on recovery from functional limitations can help us generate a picture of

what resources we need to provide that will allow for continued successful aging of

elders today and of successive cohorts of elders. Health recovery is worth study since

reversing or delaying the last step in the process to disability means that older adults can

maintain their independence longer. Promoting healthy aging contributes to the quality

of life and active life expectancy of elders. As Charmaz (1995) notes, recovery is the

goal of people with functional limitations. The initial reaction of people faced with

chronic illness is to plan to return to their prior level of ability, perhaps even to exceed

that level. She acknowledges that sufficient funds allow those who are functionally

limited flexibility in their recovery strategies. This research examines the connection

between SES and functional limitations in the context of the model of The Disablement

Process (Verbrugge and Jette, 1994).

House, Lepkowski, Kinney, Mero, Kessler, and Herzog (1994) note that

individuals in the lower SES strata experience greater levels of morbidity than do

individuals at higher levels of SES. We can view SES as a risk factor affecting the

functional levels of older adults. This has consequences for our public programs that

provide for treatment of illness, such as Medicare and Medicaid, and other entitlement

programs. As a result, we may realize that investing in prevention, screenings and a

focus on healthy aging will be socially beneficial.

Generally, one expects that in the retirement years economic stability is assured

through government programs and pension plans in place from previous work history.

However, most federal programs are designed to provide a floor of protection and often

do not provide coverage in the case of long-term illness (Estes, 1989; Wiener and Illston,








1996). In addition, if an older adult has had a history of poor health that began

significantly early in adulthood, he or she may not have had the opportunity to work in an

occupation that had fringe benefits such as a pension program. His or her health may

have affected the work history, and without consistent employment, older adults may not

have a strong pension program, if any.

One comprehensive source of information about the economic and functional

status of older adults is the Asset and Health Dynamics Among the Oldest Old (AHEAD)

survey. The AHEAD survey provides a longitudinal database for analysis of the

relationship between SES and health in the elderly population. The population under

study is adults aged 70 and older, despite the title of the survey as "Oldest Old," which is

generally considered to refer to individuals 85 and older.

This longitudinal data set offers the opportunity to analyze the specific functional

trajectories of the respondents through the first two waves of data collection. Since the

data set includes unusually detailed financial information about household income and

wealth accumulation, it provides a unique ability to describe and examine the SES/health

link among the older adults, the fastest growing portion of our population. Future public

policy direction needs such information to guide us into having a healthier aged cohort in

the next century.













CHAPTER 2
DETERMINANTS OF FUNCTIONAL STATUS

This chapter includes an overview of the model of functional ability used in this

research and of past research relating to the connection between health and

socioeconomic status. The focus is on the theory of cumulative advantage and

disadvantage. This is followed by a section that covers determinations of SES as well as

the role of education in determining SES and health. Next, the moderating effects of

health behaviors such as smoking, drinking alcohol, exercise, diet and nutrition,

preventive care and social networks are examined. This is followed by a review of

literature relating to the covariates. These are age, sex, race/ethnicity, insurance, medical

conditions, doctor visits, prescriptions and genetic endowments. The final section of this

chapter is a summary and statement of hypotheses.


Introduction

As we pass from one century to another, our population and our experiences

present new challenges and opportunities. First we have seen evidence of dramatic

increases in our life expectancy, accompanied with the decrease in acute, infectious

disease and concomitant increase in chronic, debilitating disease. The result of chronic

conditions can be poor physical functioning. Physical functioning, along with adaptation

to disease, influences recovery from disease and is also a significant marker for older

adults' quality of life. Understanding what social factors are related to the development

of poor functional status is informative to gain a full picture of health promotion among

older adults. (Guralnick and Lacroix, 1992)








Disability has gained prominence as an area of study because of changes in the

health of the population as a whole. For example, at the turn of the last century the

medical profession focused on acute, infectious diseases. At the turn of this century

chronic diseases have replaced acute conditions as the major cause of death, especially

for the elderly (Fried and Wallace, 1992). These conditions, while responsible for death

in the over age 65 age group, are also linked to limitations in functional ability, which can

be a precursor to disability.

The increasing incidence of chronic conditions leads us to study treatments, not

cures (Katz, 1999). Especially when it comes to disabling conditions, we are more

interested in increasing the length of time that one is stable or, if possible, the

improvement of a condition to avoid total disability. One consequential difference

between acute and chronic conditions is that an individual may live for a long time with

the limitations of a chronic condition. As Verbrugge and Jette (1994: 1) note, "People

mostly live with chronic conditions rather than die from them."

Older adults arrive in their post-retirement years with differing resources,

including wealth, health, knowledge, and social networks. The resources that older adults

have at this point of their lives are the result of their previous life experiences. Aging is,

after all, the accumulation of life experiences and biological changes across an entire life,

and a holistic approach recognizes the influences of the past on the present (Wallace,

1992).

Not all older adults arrive at advanced age with disability. Advanced age is not an

automatic indicator of functional limitations and resulting disability. Additionally, older

adults who do have functional limitations sometimes get better (Manton and Stallard,

1996). Knowing the dynamics of these processes of disability and recovery in








community dwelling older adults may help them maintain their independence. A study

such as this one also expands our understanding of the link between SES and health by

focusing on different health indicators, ones that are closely linked with independent

functioning (Wilkinson, 1996).

Arber and Evandrou (1993) point out that independence is perceived on three

dimensions. The first is physical independence or being independent in the domestic

sphere and being able to maintain one's own physical and personal care. The second is

autonomy, which implies the ability for self-direction free from interference. The third

aspect of independence is reciprocity or interdependence, which mitigates the negative

aspects of help if it is the result of a sense of mutuality. The point of this distinction is

that dependence and independence are endpoints of a spectrum and that we reach more

acceptable points on this spectrum during the life course. In addition, some loss of

independence is palatable if one can maintain a level of autonomy and a sense of

receiving that which is owed, rather than offered as "charity." Underlying all this is the

ability to function at a level sufficient to maintain self-esteem.


Models of Functional Ability

The combination of an aging population, increases in chronic conditions and life

expectancy, and widening economic inequality points to the significance of

understanding the mechanisms of functional limitations in old age. If we can clarify the

connections between resources and functional ability in older adults, it may be possible to

modify social policy to protect older adults.

There are a variety of measures available to determine physical morbidity.

Activities of Daily Living (ADLs) focus on activities such as bathing, eating, and

toileting. Instrumental Activities of Daily Living (IADLs) measure activities such as








shopping, doing laundry, and talking on the telephone. Functional abilities include basic

body functions such as walking a short distance, climbing stairs, picking up a dime from

a table, lifting 10 pounds, or pushing/pulling an object. ADLs and IADLs are associated

with "disability" in Verbrugge and Jette's disablement model (1994). This differs from

other schema of disability, including one set forth by Johnson and Wolinsky (1993).

Health is a multi-dimensional phenomenon, and as Crimmins (1996) points out,

conclusions about trends in disability/functional ability will vary depending on the

definition of functional limitations used.

In the Verbrugge and Jette (1994) model, functional limitations occur in the step

before disability. The distinction is made that functional limitations reflect difficulty

with a specific action, one that is situation-free, and thus not a disability. This differs

from other research on disability, which may use measures of functional ability that are

context specific. A disability in the Verbrugge and Jette (1994) model is situational and

impacts an individual's social role, which is what ADLs and IADLs measure. Thus, this

research focuses on measures of function such as the ability to walk several blocks, climb

stairs, push or pull a large object, lift 10 pounds, and pick a dime up off a table. These

are actions, not activities.

Other definitions of functional limitations include self rated health (poor, fair,

good, excellent) or diagnosis by a medical provider. Using the functional limitations

listed in the previous paragraph to examine the health SES connection is useful because

they have a direct relation to issues of independence and are less subjective than self-

rated health. Another benefit of using functional ability is that it links disease states with

environmental influences (Guralnik and Lacroix, 1992). This serves as a valuable tool in

describing the needs of the aging population as well as helps in understanding the








influences on health status. It also avoids some of the traps of ADL and IADL, which are

influenced by socially defined roles and the sociocultural environment (Freedman and

Martin, 1998). Functional ability is no less objective than physician's ratings of

respondents' health (Markides, Lee, Ray and Black, 1993) and is less subject to

complications due to limited access secondary to low SES. This research is focusing on

measures relating to physical functioning, especially those early in the disablement

process.

The model used in this research to analyze the connection between SES and

functional limitations is the model of the disablement process developed by Verbrugge

and Jette in 1994 (Figure 1). This model points to the connections between pathological

causes of functional limitations as well as the social factors that influence the

exacerbation or diminishment of limitations. Disability is a gradual process, and too

often we analyze the end result of the serious disability, but not at the impact of activity

limitations that affect participation in quality of life (or leisure) activities (Atchley, 1998).

This reflects the current thinking of aging as a social pathology (Arber and Ginn, 1991).

One additional "risk factor" that could possibly add to our understanding of the

disablement process that is not explicitly included in Verbrugge and Jette's model is the

accumulation of resources such as wealth, education, and income. These resources,

which determine one's socioeconomic status (SES), are not evenly distributed in our

society and previous research has linked SES and health (Deaton and Paxson 1998;

Feinstein 1993; Link and Phelan 1995; Nagi 1976; Preston and Taubman 1994). Risk

factors are considered pre-disposing because they occur before the disablement process

begins.









EXTRA-INDIVIDUAL FACTORS
Medical Care & Rehabilitation
Medications & Other Therapeutic Regimens
External Supports
Built, Physical & Social Environment




PATHOLOGY -- IMPAIRMENTS FUNCTIONAL DISABILITY
(diagooes of disease, injury, (dysfunction and stnrctul LIMITATIONS (diffulty doing
congital/ developmental abnormalities in specific body (r tions b physical and tie of daily life
condition) systems: musculoskeletal mental tio: ambuate, job. household
cardiovascular, nurologcal, etc.) s, cim stais, produce management, personal
"1, stairs, i 3 care, bobbins, active
intelligible speech, see standard cbion, tis

childcare, errands,
sleep, rips, ec.)


SK FACTORS INTRA-INDIVIDUAL FACTORS
Predisposing characteristics: Lifestyle & behavior changes
demographic, social, lifestyle, Psychosocial attributes & coping
behavioral, psychological, Activity Accommodations
environmental, biological


Figure 1. A Model of the Disablement Process (Verbrugge and Jette, 1994)




Socioeconomic Status and Health

Theoretical Implications

Older adults have accumulated advantageous resources in terms of savings,

accumulation of a pension, and access to governmental benefits due to the type of

employment they enjoyed in their younger years (Evandrou and Falkingham, 1993).

Their lack of accumulation may also be the result of disadvantages that they have faced

over their life course. As O'Rand and Henretta (1999: 10) point out, cumulativeie

advantage characterizes patterns of divergence or increased inequality over time."








Public programs reduce inequality in later years, but do not result in a total

leveling effect among older adults (O'Rand and Henretta, 1999). They cannot counteract

the accumulation of pension and wealth over the life course. Individuals with greater

income have the option of savings, which, over time, translate into greater wealth. The

resulting accumulation can ultimately produce a larger SES-health gap as the cohort ages

(Kubzansky, Berkman, Glass, and Seeman, 1998). The theory of cumulative

advantage/disadvantage points to consequent disadvantages, such as illness or functional

limitations, which can result in increases in inequality at older ages (O'Rand, 1996). In

other words, as individuals age, they have opportunities that make it possible for them to

accumulate resources. Alternatively, they may face barriers. The more the opportunities,

the greater the accumulated resources, the better health older adults will enjoy.

Another consequential aspect of this theory is that, even when negative events

occur, the accumulation of advantage may allow individuals with the benefit of resources

the ability to recover more quickly than individuals without resources. Aging is not a

story of inevitable decline. It is possible for some older adults to reach extreme old age

in good health with no functional limitations as well as to recover from health limitations

they experience. This research adds to this discussion by providing a description and

explanation for elders in the functional limitation stage of the disability process.

Disability is one of the most significant risk factors contributing to mortality among older

adults (Rogers, 1995). The goal of this research is to examine the relative contribution of

resources in the form of finances and in the form of health behaviors/health stock through

the mechanism of cumulative advantage/disadvantage theory.

This pattern of cumulative advantage intersects with issues of health and well-

being. Individuals who enjoy good health in their middle years have formed a basis for








good health in their older years. Individuals who have accumulated disadvantages in

their middle years may find their older years even more difficult and as a result

experience an increase in inequality over time. O'Rand (1996) outlines the process of

cumulative advantage and disadvantage of resources (wealth and health) over the course

of one's life that produces stratification of our elders. Initial inequalities in the

distribution of resources (economic or health-related) result in barriers to individuals in

disadvantaged groups. As a result, over the course of the life cycle, any disruptive social

changes result in multiplying the disadvantages one experiences (Uhlenberg and Miner,

1996). Conversely, individuals who have access to resources have access to opportunities

to accumulate advantages over the life course. The result is stratification of our elders

with inequality in SES and health which may be exacerbated by negative health events.

Mortality and morbidity rates for the upper social classes have dropped in recent

decades (Feinstein, 1993) which has resulted in a widening of the health gap between the

rich and poor. Elders in the upper social classes may have access to resources that allow

for prompt treatment of health and the purchase of goods and services that promote good

health and a lifestyle free of violence and negative environmental influences. This gap

between rich and poor is also a result of cumulative disadvantage and advantage as we

see inequalities increasing with age (Deaton and Paxson, 1998) and for each succeeding

cohort of elders (O'Rand, 1996).

Socioeconomic Status Income and Wealth

Previous research has found a consistent negative correlation between health and

socioeconomic status (Arber and Cooper, 1999; Deaton and Paxson, 1998; Feinstein,

1993; Link and Phelan, 1995; Kington and Smith, 1997; Preston and Taubman, 1994;

Smith, 1995;), although the exact nature of the causal relationship is unknown








(Cartwright, 1992). One longitudinal study using income found that a period of

economic deprivation does predict future functional limitations (Lynch, Kaplan, and

Shema, 1997). They did not have information on wealth, and they measured economic

deprivation as the number of times individuals were below 200% of the poverty level.

The mean age of their study population was about 65 years of age. Income, wealth, and

older ages were not the focus of that research. How does SES link to health? Which

comes first?

One argument is that poor health limits one's ability to work consistently at better

paying jobs, is associated with unhealthy living conditions and limits access to good

health care and preventive services (Adler, Boyce, Chesney, Cohen, Folkman, Kahn and

Syme, 1994). Smith (1995) points out that ill health also has associated medical expenses

that deplete savings and the ability to save. Health in older ages reflects the life history of

the individual.

The opposite argument is that living with limited opportunities can result in

poorer health. If one has wealth, the household is able to afford access to a preventive

lifestyle and better medical care (Smith, 1995), while individuals in limited financial

circumstances may engage in less healthful behaviors due to a lack of resources or

limited access, or lack of knowledge. Arber and Cooper (1999) found that social class (as

measured by last main occupation) was a better predictor of health status than was age,

for men and for women. "Access to money, and the desirable attributes that go with it,

such as car ownership and the power to buy services as they are needed, depends on life

history rather than any special circumstances related to old age" (Wilson, 1993: 63).

Researchers using cross-sectional data to analyze the relationship between SES

and health can determine that health and SES are linked, in that individuals with higher








SES have better health (Adler et al., 1994, House 1994). However, cross-sectional

studies cannot provide us with information about temporal ordering of SES and health. It

is evident that we need to extend prior research to analyze causal inferences between SES

and health at older ages (House, Lepkowski, Kinney, Mero, Kessler, and Herzog, 1994).

Some researchers look at the relationship between socioeconomic status and

health and assume that we can improve health by adding to individuals' income and

wealth (Smith and Kington, 1997). This would work if the assumption about the causal

connection is true. Older adults in the AHEAD have a certain "stock" of health that can

serve as a baseline for analysis, and with the longitudinal data available this research can

examine the direction of the SES effects on health recovery (Smith and Kington, 1977).

We can ask, does knowing SES in wave 1 of a longitudinal data set help us predict health

status in wave 2? This will not allow for conclusions about the causal association across

the lifespan, but does allow for uncovering some of the mechanisms underlying the

relationship between health and SES.

In considering the connection between SES and health, it is essential to determine

the composition of SES. The AHEAD data set includes information on all possible

sources of income for the individual respondent and for the household of the respondent.

Another determinant of SES is accumulated wealth.

Wealth, or net worth, includes all assets minus debt, including property

ownership, which reflects the "legacy of working life" (Wilson, 1993: 59). Income and

net worth are linked, especially for the upper percentiles of the wealthy. The disparity in

net worth is explained by looking at differences in income among the wealthier groups

(Smith, 1995). Smith (1995) also found that lower income people save less than higher

income people regardless of their race or ethnicity, which contributes to their lack of net








worth at older ages. In fact, the main source of net worth for the poor and the middle

income group is their Social Security benefit (Smith 1995).


Education Indicator of SES?

Education has been used as one of the factors determining SES. Education,

occupation and income tap into different features of SES position (Preston and Taubman,

1994). For older adults, education is stable and so the link between education, as a

measure of SES, and health diminishes by age (Beckett, 2000). In addition, education

may act in determining health status in other ways rather than directly as an indicator of

SES. As such, it may be linked to health behaviors more than SES at older ages.

A good education generally leads to favorable work and living conditions that can

have a direct effect of earnings potential and thus higher SES and better health. Education

can also indirectly influence health due to the connection between education and better

health habits and behaviors (Marmot, 1998). "Educational attainment is associated with

the availability of information and with cognitive skills" (Preston and Taubman, 1994, p.

282). As a result, individuals with a good education can obtain information about healthy

behaviors and comply with them.

Research by Beckett (2000) found that education and functional impairment are

significantly and negatively correlated for individuals over the age of 65. This may

reflect the fact that individuals in higher SES brackets and with better education live

longer, which may in part be due to behavioral factors related to education as an indicator

of self-efficacy. If education continues to hold the same relationship with mortality over

time, we can expect longer, healthier lives as better educated cohorts age (Elo and

Preston, 1996).








This is what Freedman and Martin (1998) found in their evaluation of functional

limitations of adults over the age of 50 evaluated in 1984 and 1993. They noted an

increase in educational attainment and a drop in functional limitations over that decade,

leading them to conclude that education operates in several ways to improve health. It is

associated with a beneficial lifestyle, but it may also be associated with the ability to

follow health care treatment plans and to modify the environment, whether that is the

physical environment or behaviors.

Manton (1996) also speculates that education gives older adults the motivation

they need to stay functional as it also helps them avoid risk factors such as smoking and

high-fat diets. The results of his research show that older adults in a low education group

were 5.73% more likely to be disabled. Manton's (1996) work was with adults who

experienced impairments measured by IADL and ADL, and he found that being a high

school graduate makes a difference in disability rates. If disability rates for high school

drop-outs were the same as for high school graduates, disability rates would be reduced

by thirteen percent.

Education was a measure of healthful behavior by Smith and Kington (1997b)

also. They argue that "education may... affect the way individuals can transform inputs

into good health" (Smith and Kington, 1997b: 108). Greater education gives households

access to and awareness of preventive behaviors, avoiding environmental risks, better

problem-solving skills, and effective self-care. Reynolds and Ross (1998) found that

education has power in predicting good health other than as a credential or that it leads to

high status employment. Rather, they found that education is significant, especially if

individuals are economically disadvantaged. Other research on older Black adults also

found that individuals with more education engaged in health promotion activities








(Ferraro, 1993). In this research, education is used as an indicator of health promoting

behavior.

Another view of education that is apart from SES and is an indication of human

capital, or the ability to use education to solve a variety of problems, not just as a work-

related skill enhancement (Mirowsky, 1998). In this sense, education gives one a sense

of control over one's lifestyle, leading to healthful behaviors and fewer functional

problems. However, this effect is present only for individuals who enjoy general

prosperity. Educated individuals know how to maximize the usefulness of economic

resources to reduce any sense of economic hardship (Mirowsky and Ross 1999).


Multiple and Interlocking Mechanisms

Straus (1999, 106) points out that we need to appreciate that "social and cultural

forces might be causing or complicating the patient's illness." This is true when we look

at the relationship between socioeconomic status and health in the elderly. Parental

income has an effect on children's access to education, which shapes employment and

earning history. Mirowsky and Hu (1996) found that education was linked to physical

impairment. They theorized that a lack of education is related to low income and as a

result individuals experience increased risk of physical impairment because they cannot

meet their basic physical need for food, clothing, shelter and care. Consequently their

limitations prevent them from improving their economic condition which results in

concentrating their economic hardship. Mirowsky and Hu (1996, 1091) found "a web of

reinforcing effects" in that lack of income, poor health, and lack of exercise work

together to compound problems over time.

Life span accumulations of resources may increase the differential between

classes experiencing chronic illness. Material resources include current annual income








from various sources for respondents and their spouses. Past employment history has an

effect on current levels of income and asset accumulation, and health shocks during

employment could adversely and cumulatively result in greater inequalities in older ages

(Deaton and Paxson, 1994).

Few individuals over the age of 70 are employed, but they have income from

public sources, such as Social Security, and from pension plans. In addition to current

income, material resources include assets such as home ownership and savings. Older

adults who were at a disadvantage during their work life and who suffered illness start

their retirement with fewer resources than those who have accumulated advantages.

Smith (1997) found, in his study using the AHEAD data set, that net worth is very

concentrated in this sample of elders, much more so than is income. Half the population

owns 99% of the net worth and the other half has the remaining 1%. The concentration

of net worth seems to be among families who are White. The average minority (Black or

Latino) household has no financial wealth at all.


The Moderating Effect of Health Behaviors and Health Indicators

Health Behaviors

Verbrugge and Jette's (1994) model of the disablement process points to the

existence of lifestyle and behavior factors that moderate the course of functional

limitations. They point out that wealth can allow for access to sources of information or

services that can alter the impact of medical conditions. Material resources make it

possible for the individual to purchase goods and services that can increase the odds of

remaining healthy. This includes more than purchasing health care services. For

example, owning an automobile allows for freedom of movement to take advantage of

preventive medical services or health-related activities such as health club or ajogging








park. Wealth is also connected to behavioral factors such as smoking, lifestyle and

exposure to violence (Feinstein, 1993). In this way, SES is linked to behavioral factors

that can also determine morbidity.

Individual responsibility for health, especially in the face of increasing incidence

of chronic disease, can make the difference between functional independence and

disability. Mechanic (1995) points out that as chronic disease problems increase in the

aging population, the mainstream medical diagnostic disease model is inadequate to

address the needs of disabled elderly. Their own intervention and assertiveness to engage

in healthful behaviors may be the best source of prevention and recovery. As a result, the

wealthier older adults may enjoy higher functional status and be able to recover quicker

from limitations.

The quality of health goods and services such as access to providers of care and

information about healthy behaviors are usually positively related to price. Generally

better quality care costs more, and so individuals with greater economic resources can

afford to purchase better quality "health." According to Feinstein (1993) researchers

have not reached a consensus about the relative contribution of SES and behavior to

health. This may be because SES and health behaviors are interrelated and that behavior

is a mediating factor between SES and health, although probably not as significant a

contributor as accumulated net worth and income (Wilkinson, 1996).

Individuals in higher SES levels may be able to recover from functional

limitations faster because they avoid risky health behaviors such as smoking, drinking

and eating less healthful foods. These individuals have access to educational information

that helps form their health behaviors. House et al. (1994) note that individuals in higher

SES levels are approaching the ideal in aging experiencing their older years free of








significant morbidity and functional limitations until the very last years of life. In

addition, they can avoid the stress of economic deprivation that influences the incidence

of health problems.


Smoking. Adler et al. (1994) note the strong link between SES components of

income, occupation, and education with smoking behavior. The connection is meaningful

due to the strong negative effect that smoking has on health. Wilkinson (1996) has noted

that smoking has become an identifier for socioeconomic stress. Smoking is closely

linked to cancer and cardiovascular disease, medical conditions that can cause functional

limitations or adversely affect recovery efforts from functional limitations.

Smoking behavior varies by race and education. Higher educated men and

women smoke less than do individuals who do not have a college degree (Berkman and

Mullen, 1997). Cessation rates are also higher for individuals with more education.

Berkman and Mullen (1997) also note that Black men are more likely to smoke than are

White men, but the opposite is true for women. Smoking is an indicator of risk-taking

behavior and of public health concern in light of its connection to chronic conditions and

subsequent functional limitations (Clark, Callahan, Mungai, Wolinsky, 1996).

Traditionally Mexican-Americans and Puerto Ricans have had lower rates of

smoking, but that rate has increased among males in the 1990s (Markides, Rudkin,

Angel, and Espino, 1997). Cuban-American males have the highest rate of smoking. As

a result, minority individuals (especially men) who are also poorer may have higher

incidences of chronic conditions relating to smoking and subsequent limitations.

Alcohol Use. Alcohol consumption follows the opposite pattern as smoking,

with individuals in the higher SES levels drinking more than individuals at lower levels

(Adler et al., 1994). Excessive alcohol consumption has negative consequences on








health, resulting in cirrhosis of the liver (Stoller, 1994). It can also exacerbate other

medical conditions, such as ulcers, respiratory disease, and heart conditions, and interact

with prescription medications (Council on Scientific Affairs, 1996). Alcohol is also

associated with a higher incidence of hip fractures. Not only does alcohol impair

balance, but heavier drinkers do not eat a balanced diet and are found to have a lower

bone density (Council on Scientific Affairs, 1996).

Previous research on elderly Danes found a U-shaped relation between alcohol

consumption and mortality. Both abstaining adults and heavy drinkers had a higher risk

of mortality than do light drinkers (Gronbaek, Deis, Becher, Hein, Schnohr, Jensen,

Borch, and Sorensen, 1998). The results of this study were the same for older adults as

for middle-aged adults and the mortality risk for women was 1.29 for abstaining women

as compared to light drinkers and 1.22 for abstaining men as compared to light drinkers.

These researchers concluded that light alcohol intake is associated with a lower mortality

than is abstaining or heavy drinking. The difference is the result of higher risk of

mortality from cardiovascular disease among abstaining adults. Light drinkers have

higher high-density lipoprotein (HDL cholesterol) and lower platelet aggregation

resulting in better cardiovascular health and lower mortality.

The Council on Scientific Affairs found similar patterns in research of older U.S.

adults (1996). Drinking behavior can be damaging to elders' health at any level if it

interacts with medications or if the individual is a smoker or has hypertension (Moore,

Morton, Beck, Hays, Oishi, Partridge, Genovese and Fink, 1999). Thus, alcohol

consumption is a critical variable to consider in an overall view of the health of older

adults.








Exercise. Physical activities are no longer integral to work or commuting and are

now part of our leisure activity. As a result, lack of regular physical activity has become a

health issue. Individuals who do not exercise suffer from loss of muscle mass and are

more susceptible to functional limitations. In addition to the direct effect that exercise

has in enhancing our health, it also has an indirect effect in that individuals who do not

exercise often suffer from obesity (Adler et al., 1994). Ruchlin and Lachs (1999) found

that less than half of older adults walk (the most popular exercise of this age group) and

when they did walk, most spent less than 15 minutes per walk. Individuals with higher

SES and more education exercise more than do individuals at lower SES (Mirowsky and

Hu, 1996; Ruchlin and Lachs, 1999) and the older an individual becomes, the less he or

she exercises (Bennett and Morgan, 1993). SES and education are indicators of lifestyle,

one that includes regular exercise, and helps hinder the development of functional

limitations. Exercise is seen as part of a high-status lifestyle and is reinforced by the

availability of economic resources in reducing the incidence of functional limitations.

These behavioral characteristics have an effect on an individual's health in a

cumulative fashion over the course of the life span. Individuals who participate in

regular exercise are less likely to suffer functional impairments, and thus are able to

pursue employment on a regular basis and accumulate resources to maintain good health.

Exercise can benefit older adults even if they do not exercise when they are younger.

Research has shown that exercise, even started after age 70, can reduce physical decline

and enhance functional abilities (Cress, Buchner, Questad, Esselman, deLateur, and

Schwartz, 1999). Older adults who exercise with weights to increase their endurance see

increases in their muscle strength as well as their aerobic capacity. Consequently, these








adults move more quickly and are able to carry more weight. Their physiologic reserve is

improved, and this helps maintain an independent lifestyle free of functional limitations.


Diet and Nutrition. Additional measures relating lifestyle to health and SES

include waist-to-hip ratios and body-mass index measures. The waist-to-hip ratio is a

measure of where an individual carries body fat. It can be around the waist, an "apple

shape," or around the hips, a "pear shape." The more weight carried around the waist the

greater the risk of cardiovascular disease because it indicates a higher concentration of

body fat around the heart with resulting atherosclerosis of the blood vessels.

Body-mass index (BMI) is a measure related to waist-to-hip ratio. The BMI is

determined by converting height and weight to metric measures and dividing weight by

height squared (McBride, 1992). This is a measure of body fat, and higher results are

associated with physical disability (Visser, Harris, Langlois, Hannan, Roubenoff, Felson,

Wilson and Kiel, 1998). High body fat is associated with chronic disease and low

physical activity. High BMI also indicates an increased physical burden on the body.

This burden places a strain on joints and muscles and limits the body's ability to move

easily. All of these factors increase the risk of functional limitations.

The risk of mortality also increases with greater BMI but is less of a risk at older

ages (Stevens, Cai, Pamuk, Williamson, Thun, Wood, 1998; Visser et al. 1998). A

moderate increase in BMI as one ages may provide physical resources in the event of

metabolic stress from disease. However, large increases in BMI are correlated with an

increased risk of disease, such diabetes, arthritis, hypertension and cardiovascular disease

(Clark, Callahan, Mungai, and Wolinsky, 1996).

The risk of cardiovascular disease is linked to social class, in that individuals in

the lower SES had worse indicators (Marmot et al. 1998). BMI is also correlated with








education, indicating that better educated older adults engage in healthier eating as well

as physical activity (Kubzansky et al., 1998; Stoller, 1994). The result is a lower risk of

functional limitations at older ages.

Similar to patterns in exercising, diet, the other component ofBMI, also reflects

social class and education. Upper and lower class individuals all prefer healthy diets, but

usually upper class individuals know that low fat and high fiber diets are good for their

health (Howarth, 1993). Eating may also be a response to stress. Wilkinson (1996)

points out that the poor may eat for comfort and as a source of relaxation. The

combination of eating out of stress as well as eating the wrong foods puts individuals in a

lower SES at greater risk for health consequences of poor eating habits.

Risk also varies by race and ethnicity. Clark et al. (1996) found that 40% of

Black women aged 51 to 61 have a BMI considered to be obese and that 25% of Black

men were also in the obese category. Obesity was related to difficulty in physical

functioning, a link in the path of disability. For example, about 50% of the women and

30% of the men had arthritis and the same percentage reported taking medication for

hypertension. Visser et al. (1998) found BMI more predictive of difficulties with physical

functioning than changes in muscle mass. They point to weight loss intervention in an

attempt to improve the functional performance of older adults.

We cannot overlook the influence of the "food" culture and social structure in the

United States in the connections between social class and obesity. McKinlay (1997)

argues that we are surrounded by an increasing corporatization of our diet for profit We

are encouraged to eat processed synthetic foods, not the basic, natural, nutritious foods.

As a consequence our health care system must focus on "downstream endeavors;" that is,

fixing our health problems after they are manifested in functional limitations (McKinlay,








1997: 520). Efforts at increasing healthy eating at earlier ages could result in a

cumulative health advantage for older adults.


Preventive Services. Checkups, immunizations, routine screenings, and

preventive care are shown to be health enhancing behaviors (Stoller, 1994). However, the

amount and type of preventive care an individual receives is dependent on factors related

to SES, specifically insurance coverage and access to medical services (Feinstein, 1993).

Generally, individuals who participate in preventive care have ongoing

relationships with their physicians. They visit their doctors regularly and the doctors are

familiar with their health status. As a result, medical conditions may be detected at an

earlier stage and treated more effectively before they become serious. Feinstein (1993)

notes that individuals from lower SES admitted through emergency rooms are usually

sicker than those from higher SES are. This may be the result of poor quality,

intermittent health care received by individuals in lower SES. They may wait to receive

care until after the condition is more serious because they do not have a regular provider

of care.

Another limitation of U.S. preventive care programs is their focus exclusively on

individual behavior. It is a "blaming the victim" type of approach in its underlying

ideology (McKinlay, 1997: 529). All this is linked to social structural issues such as

moral uniformity that may stem from Puritan roots and are reflected in middle-class

values. The medical establishment operates as an instrument of social control

condemning certain behaviors and encouraging conformity to other behaviors without

considering social context and the wider cultural values. For example, McKinlay (1997)

points to the health goal of relaxing to avoid heart disease while the economic system

demands hard working dedicated employees. The resulting expectation of behavior









change is unrealistic in light of these conflicting values. Focusing on individual behavior

also keeps us from recognizing the cultural and economic influences that encourage

unhealthful behavior.

While McKinlay (1997) wants to focus on the beginning point of the process of

disability, that is, the point at which we experience influences that increase our risk of

illness, this research will focus on the midpoint. This is the point at which at-risk

behaviors are identified and intervention used at the individual level to prevent the arrival

at the endpoint, or actual disability. With the AHEAD data set we can analyze the

midpoint and determine if individuals who engage in preventive behaviors have more

success at avoiding functional limitations or even recovering once they occur. Whether

or not older adults participate in preventive care gives an indication of preventive health

behaviors and their connection with functional ability. The AHEAD provides

information about respondents' participation in various health screenings, which will be

used in this research to indicate preventive care.


Social Networks. Social relationships have an influence of health and well-being

(Berkman, Oxman and Seeman, 1992) as well as functional ability (Harwood, Prince,

Mann, and Ebrahim, 1998). Berkman and Mullen (1997) note that supportive social

networks can reduce mortality risk and delay institutionalization. Part of the effect is the

result of improvement in the quality of life due to social connections. Additional factors

include monetary support and assistance with routine tasks. A study of women's health

found that membership in clubs and organization is especially significant as a predictor of

women's health (Moen, Dempster-McClain and Williams, 1992).

Unfortunately the AHEAD data set does not include information on the size,

structure, or perception of social relationships. In this case, the closest proxy is the








number of children of the respondent and the presence of a partner. These people can be

influential in the event of functional limitations, if the children live at home or nearby

(Morris, Sherwood, and Morris, 1996). Knowing only the number of children and

marital status does not give any sense of the strength of the network, nor its availability

or perceived adequacy.

Moen, Dempster-McClain and Williams (1992) found that number of children

had no significant contribution to predictions of duration of health for women, but other

measures of social support did have a positive relationship with good health. Individuals

in good health tend to accumulate roles as they remain active. So, we would expect older

adults with good functional abilities to have more children and a partner who can provide

the emotional and physical support for continued good health or recovery in the event of

a functional limitation.

Social support can be measured in several ways. The help pattern is a

determination of the source and types of help offered in through social networks.

Researchers have found variation in financial aid, types of services exchanged, and

generational linkages (Wilkinson, 1988). One focus of such research is the pattern of

help between daughters and their parents.

In the article by Wilkinson (1988) on mother-daughter bonds, she concludes that

the study of the help pattern between mother and daughter is complex and influenced by

the social environment, the structure of family life, and the changing roles of women.

She call for "a more thorough scrutiny of generational ties encompassed by the 'help

pattern'" (Wilkinson, 1988: 189). The AHEAD data set does not allow for this thorough

analysis, although such a help pattern could be essential in the recovery from functional

limitations.








Another tie that contributes to social networks is the marriage bond. The data set

includes marital status and the presence of a partner in the household. The marital status

of elderly men and women differ, as a result of women's greater longevity. Only after

the age of 85 a small majority (54%) of men are widowed, while most women live alone,

due primarily to widowhood, starting in their mid-seventies (Arber and Evandrou, 1993).

Thus, marital status may help in determining who among older adults is likely to recover

from functional limitations.

The divorced, separated, never married, and widowed have much less than one-

half the household net worth of married couples (Smith, 1995). Smith (1995) also found

that marriage and savings behavior are positively correlated. However, when measuring

marital status and its predictive ability for functional limitations, Arber and Cooper

(1999) found no statistical relationship.

Marital status has a strong relationship with mortality, with currently married

individuals having lowest death rates (Elo and Preston, 1996), benefiting men slightly

more than women. Elo and Preston (1996) suggest that this is because of a selection

effect, in that the never-married most likely suffer from some health problems. This

research looks at functional ability, but we expect to see a similar relationship due to the

selection effect as well. Goldman, Korenman, and Weinstein (1995) researched the

connection between marital status and disability and found that there may be a survival

effect for older adults, such that the differences in disability by marital status are not as

expected. In fact, single women in their study were in better health than the married

women were. This may indicate that single women are more likely to recover from

functional limitations. Recovery was not included in their study. Their work did not

include data on net worth nor all sources of income either, which they note as a weakness








of their study. This research will include measures of SES to gain insight into the process

of recovery from functional limitations considering marital status as part of a social

network. This allows for counting members of a social network, but does not provide

measures of its instrumental or emotional nature.

Covariates

Demographic Variables

Gender. Women and minorities face multiple disadvantages as they age. Older

women and minorities are two or more times likely to be poor than are white men

(Choudhury and Leonesio, 1997). Past racial or gender discrimination puts them in a

disadvantaged position regarding pension accumulation. Pensions tend to favor people

who have had advantages during their lifetime (Uhlenberg and Miner, 1996).

Women live longer, but have more reported illness than men do (Johnson and

Wolinsky, 1994). Women are also more likely to suffer functional limitations than men

are as they age (Arber and Cooper, 1999; Daltroy et al. 1999) and are less likely to

recover (Beckett, Brock, Lemke, Mendes de Leon, Guralnik, Gillenbaum, Branch, Wetle,

and Evans, 1996). "Above age 80, nearly 20% more women than men are functionally

disabled" (Arber and Ginn, 1993: 37).

It is difficult to determine the exact mechanisms at work here. It may be that

there is a link with SES, which was not accounted for in the research referenced above.

This could also be the result of a survivorship curve, that the men who have lived to older

ages are stronger and less susceptible to disability. In addition, the older men who are

disabled are possibly removed from the community-based population and living in care

facilities, while women may be more likely to remain in the community.








Additional factors which leave women particularly vulnerable to health and SES

inequalities are marital instability (SES is generally the result of husband's work history),

lack of opportunities for consistent employment and lack of pension accumulation

(Choudhury and Leonesio, 1997). Again, the intersection of SES and various

demographic factors put individuals at greater risk of functional impairment. If women

have higher levels of disability and longer life expectancy, then the issue of SES is even

more critical since women could be living with longer periods of functional limitations

and eventual disability than men do. Issues of recovery from functional limitations may

also differ by gender. This research will attempt to gain insight into the health-SES

connection for men and women.


Race and Ethnicity. A combination of factors place ethnic minorities at a

disadvantage when determining functional status. Issues of SES and health are even

more essential for ethnic and racial minorities as their population is expected to increase

substantially in the next 50 years (Martin and Soldo, 1997). Biology is not the issue

here, rather social and economic circumstances affect race and ethnic groups as social

entities (Berkman and Mullen, 1997; Kington and Smith, 1997). Especially at older ages,

there is less of a race-based difference in mortality and morbidity; social and economic

variables explain more of the differences in death rates (Elo and Preston, 1996) and

functional limitations (Kington and Smith, 1997).

Previous studies have found that Puerto Ricans and African-Americans suffer

from greater disability and functional limitations when over the age of 60 than non-

Hispanic White older adults (Jette, Crawford and Tenestedt, 1996). Older African-

Americans are concerned about their health and rate their health as worse than Whites

(Ferraro. 1993; Berkman and Mullen, 1997). They also suffer from more functional








limitations (women more so than men), even though they do not have a significantly

different number of chronic health conditions (Ferraro, 1993). Ferraro's research did not

include income and net worth variables, but did include education, which was a

significantly negatively correlated with functional limitations.

Just as with any grouping by race and ethnicity, Latinos experience diversity

within subgroups (Whitfield and Baker-Thomas 1999). Generally Mexican-Americans,

Cubans and Puerto-Ricans experience differences in mortality and morbidity from each

other (Markides, Rudkin, Angel, Espino, 1997). Some of the differences are due to

location of birth (immigrants versus U.S.-born), but some is due to socioeconomic status.

Differences in education between racial and ethnic groups have a correlation with

the differences in health experiences. Older adults who are Black experience worse

health and generally have less formal education (Bound, Schoenbaum and Waidmann

1995). If they do have a health problem is it more likely to progress to disability than is a

similar health problem for a White older adult (Ferraro and Farmer, 1996). In a

longitudinal study using ADLs the researchers found that older Blacks had a health

disadvantage when compared to Whites but that considerations of SES mitigated the

difference somewhat (Mendes de Leon, Beckett, Fillenbaum, Brock, Branch, Evans, and

Berkman, 1997). This research used only three levels of income (low [<$5,000], middle

[$5,000 $10,000], income [>$10,000]) and did not include any measures of net worth.

Unfortunately, there is little consensus about how or why racial and ethnic groups

may differ in health due to age or SES as many studies do not examine these relationships

(Manton and Stallard, 1997). New data is being generated and this research will add to

knowledge about differences in SES, functional ability, and race and ethnicity.








Previous studies using the AHEAD data have found net worth gaps by racial and

ethnic groups using household wealth (Smith, 1995). This is similar to the results that

Brown (1996) found in that the ratio of median earnings of Black to Whites is .6 to .8.

The income gap is significantly less than the net worth gap, which may reflect the

accumulation of advantage over the life course of non-Hispanic White older adults.

Smith (1995) also found that household net worth was unequally distributed across

percentiles of household net worth with the upper 5% holding seven times ($655,000) the

net worth of the average household among non-Hispanic White households. Brown

(1996) found the ratio of median net worth of Blacks to Whites to be .2 to .3. This

research looks at differences in functional ability, however, and race is used as a control

variable to avoid confounding findings by SES (Ferraro, 1993). Additionally, this

research will explore the effect of health promoting behaviors on functional ability.


Age. The oldest old are especially vulnerable to disabling medical conditions.

Poverty rates increase with age, especially for women (Soldo, Hurd, Rodgers, and

Wallace, 1997). The oldest old are defined as individuals age 85 or older. For the

AHEAD data set, which includes interviews with individuals born 1923 and earlier and

age 70 at the time of the interview, the oldest old were born in 1908 or earlier. The oldest

old may have outlived their resources or they may be experiencing more ill health as they

age. This particular cohort was growing up during the Depression era and may be at

additional cumulative disadvantage due to the historical period of their work life, which

probably started in 1926 (Soldo et al. 1997).

Studies of older adults find that there is a correlation between age and declining

physical function, but some older adults, even at the oldest ages, actually recover from

disability (Beckett et al. 1996). This is again a point against the problemitizing of old age








and assuming that aging is a process of inevitable decline. Beckett et al. (1996) did find

that decline increased with increasing age, but there is variability by individuals. Their

study did not include measures of SES or health behaviors, which could explain the

variability. The purposes of this research is to address this gap. As a result we may be

able to determine how older adults who recover from functional limitations differ from

those who do not.


Genetic Influences

Our inherited characteristics are not health behaviors, but they do point to the

potential of our heath state, so for this study they are characterized as covariates and

grouped with health behaviors. Health status in older ages reflects health at younger ages,

even back to fetal status and genetic endowments (Smith and Kington, 1997). Thus,

some families are healthier than others with cumulative advantages and disadvantages

shared by family members. Genetic endowments promote health recovery as they

represent early childhood environments (even fetal environments) and the advantages a

good environment can contribute to good health in later years. Good health starts in utero

with some families being healthier than others and passing this benefit on to children

(Smith and Kington, 1997).

Smith and Kington (1997) used the AHEAD wave 1 data set and found several

measures within the data set that could be used as proxy measures for genetic

endowments. The data set includes age of death and education level of the respondent's

parents, which they used to measure the relative good health of the previous generation.

It also includes data on the number of surviving siblings and children. They used this all

information to create a proxy for genetic endowment and the promotion of health

behaviors. They found evidence of"intergenerational health transmission" in the positive








correlation of respondents' functional abilities and parents' age at death (Smith and

Kington, 1997, 165). Respondents with higher functional abilities also had parents who

lived long, even past the respondents' 70th birthday. Similarly, higher functioning

respondents had long-lived siblings, another indication of intergenerational health

transmission.


Medical Utilization

Existing Medical Conditions: Co-Morbidities. The first step in the

disablement process is the presence of chronic conditions or medical events that can

result in impairments and then functional limitations. Changes in our environment have

reduced the incidence and prevalence of acute, infectious conditions. Limitations on

physical activity can have social consequences with the resulting loss of independence

and social interaction. Individuals who maintain an active lifestyle are more likely to

maintain good physical functioning and be able to recover from them when they do

occur. They will have the physiologic resources to regain good functional status (Clark

et al. 1996).

Today chronic conditions are increasingly the cause of health problems. Previous

studies found that approximately 40% of adults over the age of 65 report activity

limitations due to chronic conditions (Fried and Wallace, 1992). Individuals with

functional limitations and co-morbidities are more likely to remain functionally disabled

than individuals with no other chronic conditions (Chirikos and Nickel, 1986).

Researchers using data from the Framingham Study (Guccione, Felson, Anderson,

Anthony, Zhang, Wilson, Kelly-Hayes, Wolf, Kreger, and Kannel, 1994) found that knee

osteoarthritis, heart disease, and stroke were conditions most attributable to functional

limitations. They also noted that chronic obstructive pulmonary disease and heart disease








made significant associations with functional limitations. In the AHEAD data set, we can

control for cancer, diabetes, emphysema, heart condition, stroke and arthritis as well as

test connections between SES and functional ability and potential recovery of good

functional status.


Insurance Coverage. Medicare coverage is available for almost all adults over

the age of 65. The U.S. Census (1996) reports that 99.4% of elderly had continuous

coverage between Medicare, Medicaid and military health care. However, such coverage

is not distributed equally among elders, leaving some groups at greater risk For

example, Mexican Americans have low levels of health insurance coverage and the

coverage they do have is minimal (Angel and Angel, 1996). Additionally, elders in poor

health or with functional limitations are less likely to have private health insurance

(Wilcox-Gok and Rubin 1994). Without this coverage, older adults may not seek care in a

timely fashion when it is possible to treat their medical conditions more effectively and

increase the probability of recovery from functional limitations. However, the presence

of publicly funded programs is no guarantee of adequate coverage for health care

expenses.

Unfortunately publicly funded programs such as Medicare and Medicaid do not

cover all health care costs. AARP estimates that the elderly will pay 43% of health care

costs out of their own pockets (Crystal, 1996). For some older adults this out of pocket

payment represents in excess of 16% of their annual income (Estes, 1989). SES status

has a direct effect on the amount and type of coverage that elders can afford. This may

affect their efforts to receive preventive care or treat chronic conditions that could result

in functional impairment and, thus, lower the odds of recovery.








So, the assumption that the elders of our population will successfully age even in

the face of chronic illness or disability since public programs will cover the cost of their

treatment may be erroneous. Without the full protection of public programs, individuals

must look to their own savings to help defray the expenses associated with treating

chronic illness and disability. Yet we know that financial assets are not evenly

distributed among the elderly (Crystal, 1996), and neither is public or private insurance

(U.S. Bureau of the Census, 1996). There is a direct link between the type of coverage

one enjoyed during employment years to the adequacy of coverage in the retirement

years, once again pointing to the importance of examining the connections between SES

and health (Angel and Angel, 1996).


Medical Services. Contact with medical providers and taking prescription

medications are indicators of possible pathology that leads to functional limitations.

Previous research has found that previous hospital stays are the biggest predictor of

continuing functional limitations (Chikiros and Nickels, 1986). Chikiros and Nickels

(1986) research did not include prescription medications as a variable, nor did they

include net worth as a SES variable (only income). Research using prescription

medications did find a strong statistical correlation between use of medication and lower

levels of functioning in older adults (Daltroy, Larson, Eaton, Phillips, Liang, 1999).

Another, subtle, economic factor may be at work here as well. Providers of health

care are socialized beings and susceptible to economic incentives and the perceived

opportunity cost of continuing disability. The subsequent patient care plan may vary

depending on the SES of the patient (Chikiros and Nickels, 1986).

Regular medical services are integral to ensuring good health and recovery from

functional imitations, since medical advances have the possibility of slowing or stopping








the progress of disease to disability (Crimmins, 1997). With increases in life expectancy,

researchers are asking if disability-free years are increasing, thus indicating a

"compression of morbidity" (Fries, 1989: 208). The concept of compressing morbidity

was re-introduced by Fries (1989) who analyzed the concept using data through the

1980s. He points out that individuals in lower SES are not enjoying the benefits of

reduced health risks and that their old age will be more expensive if they suffer more

disability over a longer period of time in their later years.


Summary of Socioeconomic Status and Functional Ability in Older Adults

Most older adults enjoy a relatively active life; however, about 14% of older

adults have some degree of activity limitation (Jette, 1996). This has a negative effect on

the quality of life they enjoy and the independence they are able to maintain. Jette (1996)

points out that disability prevention is possible and that recovery is a possible and

desirable goal for our elders as they experience longer life expectancy and increasing

numbers. At the very least, further decline may be stalled or delayed. Another

consideration in determining the possibility of health recovery, as well as delayed

deterioration, is understanding the mechanisms that increase or decrease the risk of

functional limitations at older ages.

Figure 2 is the conceptual model proposed for this research that incorporates the

correlates described above. Education is in the health behavior box. In this research

education will serve as an indication of self-efficacy regarding healthy behaviors.

Generally, individuals at lower SES suffer poorer health. They tend to adopt

more risky health behaviors (House et al. 1994) and have an accumulation of poor health

and a lack of resources such education, income and net worth. As House et al. (1994)

point out, the accumulated effect of psychosocial risk factors is harder on the physiology









of older adults due to biological declines that occur with age. The stratification that

accumulates over the course of a lifetime leaves elders particularly vulnerable.



Health Behaviors
*Smoking
Socioeconomic Status: Alcohol
Income Exercise
Net Worth Wecigcontrol
Preventive cam
Social Support (# children
& marital status)
Education
Genetic Endowments: age Education
at death for mother &
father, # living siblings.

Covariates (controls):Functional Ability
Age # walk several blocks
Sex climbing stairs
Race/Ethnicity pushing/pulling a large
Insurance coverage object
Specific medical conditions lifting 10 pounds
Doctor visits picking up a dime
Prescription Medications




Figure 2. Conceptual Model of the Effects of SES and Health Behaviors on Functional
Ability





There is a feedback loop between health and SES, and with cross-sectional data it

is difficult to see where the loop begins. Longitudinal data is useful in this endeavor to

determine the causal relationship between SES and functional status. Elders are a

heterogeneous population with great variation in their functional abilities, and studying

change in functional ability over time is a helpful tool in understanding the source of

variability and improving the ability to promote health recovery (Jette, 1996).








The disablement process provides a model for examining different outcomes of

functional limitations. This model is useful as a basis for determining the relationship of

SES and functional status as well as any moderating effects from preventive health

behaviors and the relative contribution of income versus net worth to the likelihood of

recovery from functional limitations.

Hypotheses

In light of previous research in this area, this research will focus on six

hypotheses, as follows:


1) A significant proportion of older adults are free of functional limitations and

some of those older adults who do suffer from functional limitation recover

from them within two years.

2) Individuals with greater economic resources have stronger functional status.

3) Individuals with greater economic resources are less likely to suffer a decline in

their functional status.

4) Individuals with greater economic resources are more likely to recover from

functional limitations when they occur.

5) Income and net worth will affect functional limitations differently depending on

the functional status of the individual:

a) Net worth is a better predictor of a stable state regarding functional limitations

across both waves of data. Individuals with a higher net worth will be more

likely to experience a stable state with no functional limitations, while the

opposite is likely to be true for individuals with lower net worth. This is due to





40


the accumulative advantage of a higher net worth or disadvantage of a lower

net worth.


b) Income is a better predictor of a transition state regarding functional limitations

across both waves of data. Individuals with a higher income will be more

likely to recover from functional limitations experienced in wave 1 by wave 2.

Individuals with a lower income will be more likely to suffer a decline between

Waves. This is due to the current onset of the disabled state and the need for

more accessible resources to improve the functional status.


6) Intervening health behaviors will modify the relationship between

socioeconomic status and health.













CHAPTER 3
THE SURVEY OF ASSETS AND HEALTH DYNAMICS AMONG THE OLDEST
OLD


Research Sample

Data

The Assets and Health Dynamics Among the Oldest Old (AHEAD) survey is a

national panel study designed to be used for analysis of older Americans and their

experiences with health, finances and families. This is an ongoing longitudinal survey of

community based individuals born in 1923 and earlier. The initial sample of 7,447

respondents were taken from the Health and Retirement Survey (HRS) screenings of area

probability household sample. Additional respondents aged 80 and older were taken

from the Medicare Master Enrollment File (HCFA) for a total of 8,221 respondents. The

dual sampling frame was used to test for bias in the selection criteria. Additionally,

Mexican-Americans, African-Americans, and Floridians were sampled at 1.8 times the

probability as the general population.

The data are organized into three waves with plans to accumulate additional

waves of information merged as part of the HRS. Wave 1 includes data that were

collected between October 1993 and July 1994 and wave 2 data that was completed in

May 1996. The survey was sponsored by the National Institute on Aging. The Institute

for Social Research at the University of Michigan oversees the data collection.

The designers of the AHEAD data set recognized the importance of evaluating the

interaction of health status and financial well being for setting public policy (Myers,

41








Juster and Suzman, 1997). As a result they have gathered detailed information at a

household level about health, finances and family relationships. This makes the AHEAD

an extremely useful data set for determining the direct effect of SES, and its indirect

effect through health behaviors, upon the health status of the elders of the United States.

The AHEAD data set is designed to examine the health and economic dynamics

of the oldest old. The data set includes detailed information on all sources of household

income and net worth, presence or absence of functional status, details of medical

diagnoses and health care services, participation in preventive health behaviors as well as

limited data on the respondents' parents, siblings, children and grandchildren. This data

set makes a unique contribution to research in its richness of detail regarding the health

and economic condition that is critical to understanding the SES/health relationship.

An additional benefit of the AHEAD data set is that is it longitudinal.

Longitudinal data provide a good picture of onset and desistance of medical conditions

that provides us with models of individual health experiences not available with cross-

sectional data. With these data researchers can determine the relationship between SES

and health for individual cases and categorize respondents by stability or change over

time. It is possible to determine the order in which changes occur in functional status and

the relationship between functional status, SES and health behaviors among the oldest

old. We can then analyze the benefit of knowing SES at wave 1 and functional ability at

wave 2 and determine if that knowledge gives us insight into the causative relationship

between SES and health over this time period. Through longitudinal data we gain the

ability to examine the chronological and developmental course of the relationship

between SES and health.








The concepts of "successful aging" and "compression of morbidity" need careful

study if we want to encourage them in our rapidly aging society. What social factors

influence the continuing good health of the oldest old? Ensuring quality of life for older

adults as well as freedom from disability and dependence saves time, money, and

emotional distress.

Another advantage of the AHEAD data set and this research is that the functional

status will be unbundled into particular activities. Prior research tends to create indices

or scales combining various measures ofADL, IADL or functional limitations. Each of

these measures of functional ability (walking, climbing, pushing/pulling, lifting and

picking up a dime) are influenced by different environmental, social, and physical

factors. By analyzing the SES/health connection for each of these independently we

eliminate the possibility of complicating upper body versus lower body issues. In this

way it is possible to see if the relationship between SES and health varies by specific

functional abilities, since different pathologies may influence each of the functional

abilities diversely.

The SES/health link is a consequential one to study in this age group for several

reasons. First, we will all hopefully be in this category at some point in our lives, and the

number of the U.S. population who are aging is increasing. To examine the factors

influencing a happy old age aids the oldest old as well as the following generations that

are responsible for their care and support. It is informative to see if the SES/health link is

the same for the oldest old as it is for other age groups. Secondly, this group is at a

unique stage. They have probably accumulated all the net worth and education possible








for them, so there is a limit to the question of "reverse causation" (that is, health causing

wealth) from this point forward.

The AHEAD data set also offers us information about health behaviors of the

oldest old and their impact on functional ability. Does a lower risk lifestyle have an

effect on functional ability at older ages? If so, it is not too late for even the oldest old to

make changes in health behaviors in order to improve their health status.

This group can also help us recognize the social influences on our health status.

They have experienced a lifetime of social forces and influences that have determined

their social standing and resulting good or bad health. Critically analyzing this through

measures of SES can help succeeding generations to choose a different course.


Sample

The AHEAD wave 1 includes interviews with 8,221 non-institutionalized

individuals from 6,047 different households, with a response rate of 80.4% and a dual

sampling frame for respondents aged 80 and older. The dual sampling frame, by the

HRS and HCFA for the over 80 portion of the sample, was used to eliminate sampling

bias. The households surveyed contained at least one individual 70 or older (born prior to

1923) and his or her spouse. Some surveys were conducted face-to-face (especially if the

respondent was over age 80), and others took place over the phone depending on the

respondent's preferences. The study was designed to over-sample individuals of Black

and Hispanic race/ethnicity as well as people living in Florida. Analyses use weighted

data unless otherwise indicated.

This study sample will include all individuals who completed the financial, health

and behavior sections of the survey in waves 1 and 2. Only survivors are included since








this research is trying to determine the influences on transitions in functional ability

between waves. The combined sample of wave I and wave 2 respondents used in this

study is 6,237 individuals. The variables used are summarized in Table 1 on page 49 and

the descriptive statistics are summarized in Table 3 on page 58.

One limitation of using this data set is that only three waves of data are available

at this time and the waves are only two years apart (this research uses the first two

waves). This, however, does allow us a first glimpse at the transitions between functional

states among the oldest old, recognizing that recovery from functional limitations may

take more time than two years. There is a possibility of the opposite situation as well.

With these data we do not know the picture in the intervening time frame. Verbrugge,

Reoma and Gruber-Baldini (1994) found that post-hospital older adults improved for a

month or two, but then their health declined. They noted that functional ability is

variable for the first year following a hospital stay. Respondents could experience

functional limitations and recovery several times between waves of data collection.

Other research found that lower body limitations were less likely to resolve than

upper body limitations (Wolinsky, Stump, Callahan, and Johnson 1996). Since most of

the functional measures in the research are of the lower body, we may have consistency

across waves. Any patterns shorter than two years in duration are obscured by the

schedule of interviews. As a result, researchers may overestimate the stability of

functional status.

Functional status is the third step in the main pathway of Verbrugge and Jette's

disablement process model. It is a point in the process where we can observe how

physical dysfunctions operate in the lived experiences of the oldest old. This point is also








before complete disability, as would be measured by ADLs and IADLs. As such, it is

divorced from social definitions and role responsibilities, and any biases this might cause

in answers from the respondents. Additionally, it may explain a point in the process

before the need for nursing homes or other institutional aids and at which recovery is

possible. Jette (1999) called for more longitudinal studies isolating the steps of the

disablement process for a more thorough analysis of the critical risk factors at each point

in the pathway. This research attempts to do that.

The AHEAD study only peripherally addresses issues of access to health care and

barriers to adequate rehabilitation from physical impairments. Respondents are asked

about their visits to various providers, but are not asked if they did not see a provider due

to access limitations or other barriers to care. The data do not include measures of

convalescence nor measures of social networks nor social support. These would be

beneficial additions to the study as would questions related to difficulties seeing a

provider, whether they are related to transportation issues, cost of care, or lack of

availability of providers. The Institute for Social Research has gathered thorough

information on the health and wealth status of the respondents.

The data set is also limited by problems common to all longitudinal research, such

as attrition due to death and those lost to follow-up. This could be a problem when

analyzing health care issues, as individuals more frail may die between waves of data

collection. As a result, their experiences are not included in this analysis which may

affect our understanding of the trajectory of individuals with greater functional

limitations. For the second wave of interviews, 9% of respondents had died between

waves, and 11.1% were lost through attrition or did not respond for other unspecified









reasons. However, 88.9% of the baseline respondents have provided interviews at all

waves in which they were eligible.

Additionally there are problems with confounding cohort effects. Historically,

this cohort grew up before we had the medical knowledge we have today which guides

our health behaviors. For example, smoking was a popular pastime; cigarettes were

passed out on airplane flights. Today, we know the detrimental effects of cigarette

smoking and we cannot know what health behaviors we would have seen among this

cohort if they had the knowledge we have today. In addition, they lived through a time of

segregated school and health care systems. As a result, it is difficult to know if the study

results will be applicable to this cohort only or generalizable to the older population of

succeeding cohorts or if patterns observed are typical or normative aging. Other

measurement limitations will be addressed specifically for each set of variables.


EXTRA-INDIVIDUAL FACTORS
Number ofprescriptions, Doctor visits, hospital and nursing
Home stays, outpatient surgery, insurance, number of
Children for social support and assistance



PATHOLOGY -- IMPAIRMENTS FUNCTIONAL DISABILITY
(Number ofmedical LIMITATIONS
con st) (Walk one block, Climb
stairs, Lift 10pounds,
Push/Pul heavy object,
Pick up a Dime)


RISK FACTORS INTRA-INDIVIDUAL FACTORS
Net Worth, Years ofEducation, Sex, Race;
Genetic Factors: Father/Mother's age at death, number of Siblings;
Behavioral Factors: drink alcohol, smoking, weight, exercise,
social support network and preventive care


Figure 3. Modification of the Model of the Disablement Process (Verbrugge and Jette,
1994) Showing Variables used in this Research








Measures

The data from Waves 1 and 2 of the Assets and Health Dynamics Among the

Oldest Old (AHEAD) were used to analyze the relationship between socioeconomic

status (SES) and functional limitations. The wording for selected questions from the

AHEAD codebook is included in Appendix A. The model presented by Verbrugge and

Jette (1994) is used as the basis for the design of the study. Variables are shown in Table

1.



Functional Performance

The response variable is functional performance. The functional performance

indicators in both waves are: walking several blocks, climbing one flight of stairs,

pushing or pulling large objects, lifting weights over 10 pounds, and picking up a dime

from the table. Respondents were eliminated if they said they "Don't do" the activity,

but not because of health reasons. The individuals may not have the occasion to perform

these activities, such as not being around stairs, for example. As a result, the response

does not provide meaningful information on physical ability, just on environmental

limitations and opportunities (Guralnik and Lacroix, 1992). Using functional

performance to measure health status has several advantages. In this research, each

measure is modeled separately. Aggregating the scales in a meaningful fashion is

difficult and may mask difficulties with specific tasks and therefore underestimate the

exact level of difficulty with functioning (Guralnik and Lacroix, 1992).

Measures of functional performance have advantages over basing measures of

health on medical diagnoses because the existence of a medical diagnosis may be

compromised by access to care issues, which are related to SES. Analyzing









improvements in functional ability allows for a description of the elders who avoid

moving into complete disability.


Table 1. Variable Description and Coding


Variable Description
Functional Performance (O=no difficulty, l=difficulty)
Walking 'b difficulty walking several blocks
Climbing"b difficulty climbing one flight of stairs
Pushing/ Pulling 'b difficulty pushing/pulling a large object
Lifting'" difficulty lifting 10 pounds
Pick up a Dime 'b difficulty picking up a dime off a table


Socioeconomic Status
Income" Total annual household income
Net Worth" All assets minus all debts, including any debt for mortgage in total dollar
amount
Health Behaviors
Smoking" Never smoked, former smoker, current smoker
Alcohol Use" 0=abstainer; Light drinker=l-2 drinks; Drinker=3+ drinks/day
Exerciseb Vigorous activity three times a week over the past 12 months
O=non exerciser l=exerciser
Body Mass Index' Weight & Height ratio: Underweight (BMI <18.5); Normal weight (BMI
between 18.5 and 24.9); Overweight (BMI >24.9)
Screeningsb Number of health care screenings for breast cancer, prostate cancer,
cholesterol, cervical cancer, or had a flu shot or breast self-exam:
Women: 3* the number of screenings; Men: 5*the number of screenings.
Social networks" Number of children ever had
Marital Status: 0=no partner; married or co-habiting
Education' Years of education completed
Covariates
Genetic Endowments
Mother's age" Mother's age if living. If deceased, her age at death
Father's age" Father's age if living. If deceased, his age at death
Siblings' Number of siblings still alive
Sex' Sex: 0=male; --female
Age" Age at interview
Race" Race/ethnicity: Non-Latino White, African-American, or Other (Latino, Asian,
or Native American)
Medical Conditions' Total number present of the following: cancer, diabetes, emphysema, heart
condition, stroke, or arthritis
Health Insurance" Three variables: 1) Medicaid = presence or absence of Medicaid, 2)
Government Insurance = other Government sponsored ins (Medicare Part A,
CHAMPUS), or 3) private pay insurance = Medicare supplements or individual
coverage.
Doctor Visit" Respondent visited a doctor in the past 12 months
0=no visits; 1= at least one visit
Prescriptions' Number of prescription medications taken each month
Source: AHEAD
a Wave 1
b Wave 2








Studies have shown that these functional limitations are correlated with many

measures of health, such as self-reported health, work disability, IADLs, and ADLs

(Johnson and Wolinsky 1993; Waldron and Jacobs 1988). This measure is less subjective

than self-reported health, which is useful because it provides a specific level of

performance for comparison. It is easily observable to the respondent, compared to some

medical conditions, but has not progressed into disability, can be treated to maintain

community independence with less costly medical care.

The measures of functional performance include upper and lower body. Walking,

and climbing focus on the lower body and its functioning, while lifting 10 pounds and

pushing/pulling a heavy object is a measure of both upper and lower body, and picking

up a dime is upper body performance. Difficulty with any of these activities can lead to a

reduction in ADL or IADL performance. For example, if lifting is a problem, that could

reflect problems with reaching, which could mean the individual would eventually be

unable to dress or perhaps to grocery shop and unload groceries in the home.

For this research, we focus on the change in functional performance from wave 1

to wave 2. If an individual reports no problem with the measure of functional

performance in waves 1 and 2, then the variable is coded as stable with no limitations. If

a respondent has difficulty in waves 1 and 2, then the variable is coded as stable,

limitations. If an individual has difficulty in wave 1, but is recovered by wave 2, then the

variable is coded as recovered. If the respondent had no difficulty in wave 1, but

difficulty in wave 2 the variable is coded as declined.

Approximately one-third of respondents in wave 1 experienced some level of

difficulty with the functional performance measures, except the ability to pick up a dime








from a table (see Table 3). The most difficulty was experienced with walking several

blocks (37.5%) and pushing or pulling heavy objects (36.7%). The least difficulty was

experienced with picking up a dime from a table (8.3%). In wave 2 all measures

increased in the percentage of respondents reporting some level of difficulty with the

functional performance measures. Again, the most difficulty was reported with walking

several blocks (45.2%) and the least with picking up a dime (11.5%). The second most

reportedly difficult task was pushing or pulling a large object followed by lifting 10

pounds and finally climbing one flight of stairs.


Socioeconomic Status

Independent variables will include income and net worth for each household and

the educational level of older adults, all measured in wave 1. The mean household

income for the respondents in wave 1 was $25,191.94 and ranged from $0.00 to

$700,000. For this research income is divided into deciles as shown in Table 2.

Household income is measured with a variable constructed by AHEAD staff summing

across all household members the amount of income from the following sources:

Social Security Retirement pension income

Supplemental Security Stock income

Welfare Income from bonds

Veterans' benefits Income from dividends or
interest on savings or
Rental income checking accounts or CDs,
SBusiness bonds or treasury bills
Business
Income from work (including
Farm self-employment)

SIRA Any other sources of income.

Annuity income









Net worth includes all other forms of economic resources measured as the total

dollar value of all assets owned by household members minus any debts, including the

mortgage. This is also a derived variable using assets such as:

The value ofreal estate Treasur bills and certificate


holdings

* Business holdings

* IRA holdings

* Retirement pensions

* Stock ownership


of deposit

* Savings bonds

* Ownership of means of
transportation

* Value ofjewelry and
collections


Mutual fund ownership Debts owed to respondent

Checking and savings Rights in a trust or estate
accounts
Debts were subtracted from assets to derive the variable for net worth.

Mortgage Life insurance policy loans

Utilities Loans from relatives

Credit card balances Real estate tax

Medical debts Home insurance

Net worth was measured as a total of all assets minus all debt, including home

mortgages. Net worth had a larger range and mean than household income. The mean

net worth was $182,765 and it ranged from -$285,000 to $14,655,000. Both income and

net worth were divided into deciles for regression analysis (Table 2).

Health Behaviors

The AHEAD data set includes information on preventive care behaviors that

affect health. The data for smoking, drinking, and BMI are used from wave 1 to establish

a baseline. Also, the questions were changed for wave 2 and did not include useful


-e








information, such as status as a former smoker. Wave 2 questions included exercise

activity and participation in health screenings, which were not asked in wave 1.




Table 2. Decile Ranges for Income and Net Worth

Decile Income Range Net Worth Range

1 $0-5,900 -$285,000 +$570
2 6,000- 8,976 600- 13,200
3 9,000 11,988 13,500- 35,700
4 12,000- 14,940 35,712- 59,200
5 14,976- 17,820 59,500- 86,200
6 18,000 21,842 86,500- 120,900
7 22,000 26,960 121,000 167,000
8 27,000- 33,600 167,900- 247,000
9 34,000- 47,600 247,500- 422,500
10 48,000- 700,000 423,000- 14,655,000


The first health behavior is smoking. Respondents are coded into 1 of 3

categories: 1) non-smoker at wave 1, 2) former smoker at wave 1, and 3) current smoker

at wave 1. The referent category is non-smoker.

Another health behavior is alcohol consumption. Details about alcohol

consumption include the number of drinks per day. Respondents were coded according

to the number of drinks consumed each day. This resulted in three categories: 1) abstains

from alcohol (0 drinks per day), 2) light drinker (2 or less drinks per day), and 3) drinker

(3+ drinks per day). The referent category is abstaining.

Respondents were also asked about their participation in vigorous activity in wave

2, but the question was not included in wave 1. The question covers physical activity

including sports, heavy housework, or ajob that requires physical labor. Respondents








were asked whether they participated in physical activity three times a week or more over

the past 12 months.

An additional behavioral variable related to nutritional status and physical activity

is the imputed body mass index based on the height and weight of the respondents at both

waves. The following categories of body size were used: 1) underweight (BMI less than

18.5), 2) normal weight (BMI between 18.5 and 24.9), or 3) overweight (BMI over 24.9).

The referent category is normal weight

The respondents were asked about preventive care services in wave 2, but not in

wave 1. They were asked if they had completed or received any of the following: a) flu

shot, b) blood test for cholesterol, c) self-test for breast cancer, d) mammogram, e) pap

smear, and f) prostate cancer screen since wave 1? Items a, b, c, d, and e will be included

for women. Items a, b, and fwill be included for men. For each test completed, men will

receive 5 points and women will receive 3 points. The scoring calibrates a single scale.

The preventive care scale will range from 0 (no screens completed) to 15 (all screens

completed).

Education is measured as the number of years of formal schooling the respondents

had, with a maximum of 17+ years. The mean number of years of schooling is 10.8 with

a range of 0 (1.7%) to 17+ (5.6%) years. Approximately 11% of the respondents

completed eighth grade and 30% of the group completed high school and 27% had some

college or graduated from college.

The final measure used in this section is of social networks. Unfortunately neither

wave includes any information about social networks nor social support. The closest








proxy for this is the number of children, assuming that they interact with the respondents,

and marital status. The data used is number of children and presence of a partner.

Overall this group follows healthy behaviors, except for weight control. Only

9.7% of the respondents are current smokers with 48.2% never smoking and 42% quitting

before wave 1. Eighty-nine point two percent of the group abstains from alcohol or

consumes less than one alcoholic drink per day. Slightly less than one-third of the

respondents (29.3%), however, engaged in regular vigorous exercise. This could explain

the 44% of the respondents who remained overweight between Waves 1 and 2, with a

BMI over 24.9. Two point seven percent remained underweight and 11% lost weight

while nearly 5% gained weight. The remaining 38% maintained a BMI between 18.5 and

24.9, which is an appropriate weight for their height.

The AHEAD respondents indicated if they participated in preventive care, which

included screenings for cholesterol, prostate cancer, breast cancer, cervical cancer,

receiving a flu shot, or performing a breast self-exam. The mean score for preventive

screens was 9.47 overall. Men received an average score of 10.8, but because the value

applied to each screening was 5 points, this means they obtained approximately 2.2

screenings of the total 3 available (cholesterol, flu, prostate). The average score for the

women of 8.6, divided by the score per screening of 3, results in a slightly higher

participation in preventive screenings of 2.9 of the total 5 available (flu, cholesterol, pap

smear, breast self-exam, mammogram). The most popular screening for men was

prostate screen (47.5% of the men receiving preventive screens). For women, a

mammogram was the most used preventive screen (23.3% of women receiving

preventive screens).








The final health behavior used in this research is social networks. The data set

does not contain any overt measures of social support or social networks. As a substitute,

marital status and number of children ever had was used. The average number of

children reported by the respondents was 2.7 with a minimum of 0 and a maximum of 21.

Marital status categories include married or cohabiting and unmarried (divorced,

widowed, and never married). A slight majority of the respondents were married or

cohabiting (51.8%).


Control Variables

Demographic Information. Control variables include sex, race/ethnicity, age,

insurance benefit coverage, medical conditions, doctor visits and prescription

medications. All controls will be measured at wave 1 to form a baseline of variables.

Sixty-one point seven percent of the sample respondents were women, which is

representative of this age cohort. Age was measured as exact age. The oldest respondent

at the time of the first interview was 103 years old and the youngest was 69, and the

mean age was 77.25 years.

Ethnic groups will consist of non-Latino Whites, non-Latino African -Americans,

and Other (includes Latino, Asian and American Indian). There are too few members of

other ethnic origins to form meaningful homogeneous groups with risking zero cells and

unstable estimates of coefficients and standard errors. The study sample is 80.3% non-

Latino White, 13.0% African-American, and 6.7% other.


Medical Utilization. A summary measure of the number of medical conditions

offers a measure of the prevalence of pathological conditions, the first phase in the

disablement process. The disability model focuses on the following conditions: cancer,








diabetes, emphysema, heart condition, stroke and arthritis. Respondents had an average

of .99 conditions. Categories of insurance coverage are: 1) Medicaid; 2) government

insurance (Medicare and CHAMPUS); and 3) private pay individual plan (basic, Medi-

gap, supplemental, etc.). Most of the respondents have some type of insurance coverage.

Only 10% have Medicaid, but 94.2% have government insurance and a private pay plan.

As far as accessing the medical system, respondents were asked if they had had a

doctor's visit in the past 12 months. Eighty-nine point two percent of the respondents

had had at least one visit. The respondents were also asked the number of prescription

medications they were taking. The number ranged from 0 to 20, with an average of 2.78

prescriptions. Other measures of accessing the medical system, such as hospital stays or

nursing home stays or outpatient surgery were not used in this research. It was believed

that these would be confounding variables, measuring the same functional limitations as

the dependent variable.


Genetic Endowments. The AHEAD data set does not include any detailed

information regarding genetic background, yet such information is meaningful because

our genetic inheritance provides a starting point for our health in older ages and it

represents our childhood environment. Long life of family members is serving as an

indication of positive health genetic endowments. This research follows Smith and

Kington's (1997) strategy to use substitute variables. The variables here are: 1) mother's

age (if alive, or age at death), 2) father's age (if alive, or age at death) and 3) number of

living siblings.











Table 3. Descriptive Statistics of Sample
MEAN (S.D.)
FUNCTIONAL PERFORMANCE
Walking difficulty
Stair Climbing difficulty
Pushing/Pulling difficulty
Lifting difficulty
Picking up a dime difficulty
SES
Household income (SO $700,000 $25,191 94 ($29
Net worth (-$285,000 14,655,000) $182,765 ($395,
HEALTH BEHAVIORS
Smoking:
Never smoked
Former smoker
Current smoker
Alcohol use:
Abstainer
Light Drinker
Drinker
Exercise.
Regular exerciser
Diet/Nutrition
BMI s18.5
BM >18.5 .24.9
BMIc24 9
Screenings (0 15) 9.47 (4.7)
Females (0-15) 8.6(4.4)
Males (0- 15) 10.8 (4 8)
Social networks
Number of children ever had (0 21) 2.73 (2.3)
Marred/Cohabitng
Education (I 17+) 10.84 (3.7)
COVARIATES
Women
Men
Non-Latino Whites
African Americans
Other (Latino, Asian, or Natave American)
Age (69-103) 77.25(5.68)
Unmarried
Genetic endowments
Father's age (20 107) 71.6(15.1)
Mother's age (18 109) 74.2 (17.1)
Number of surviving siblings (0 13) 2.07 (2,02)
Number of medical conditions (0 6) 0.99(0.97)
Government Insursace (Medicare, CHAMPUS)
Medicaid
Private pay insurance (Medi-gp, basic, supplemental)
Doctor visit
Prescription Medications (0 20) 2.78(2.1)
Source: AHEAD, Waves 1 and 2
a range in parentheses
b change between waves
c current age or age at death


PERCENTAGE
Wnvel Wave2
37.5% 45.2%
29.8% 33.1%
367% 44.3%
33.5% 40.1%
8.3% 11.5%

.983.13)
668.10)


293%


Until sociological research and biomedical research merge data, we will not have


accurate measures of genetic endowments. As a proxy, this data set includes the age of


parents of the respondents. Some of the respondents' parents were still alive, so their age








is reported. If the respondent's parents are deceased, their age at death is reported. In

addition, the number of surviving siblings may provide some indication of the positive

biological influences on health in old age. The average age for fathers of the respondents

was 71.6 and ranged from 20 to 107. Mother's average age was slightly higher at 74.2

with a range of 18 to 109. The mean number of surviving siblings was 2.07 with a

minimum of 1 and a maximum of 13.


Procedures

The data will be analyzed in two phases, first through a transition matrix for each

functional limitation and, second, using multinomial logistic regression to analyze

correlates of changes in functional status. For each of the five performance measures,

there are four states that can occur over the two waves. One pair of states is stability, or

no change in functional status, whether the respondent is stable with limitations or stable

with no functional limitations in both waves. The second pair of states is a transition

between limitations and no limitations. Recovery occurs when the respondent cannot

perform the activity at wave 1, but can perform it by wave 2. Functional decline occurs

when the respondent was able to perform the activity without difficulty at wave 1 but in

wave 2 completes only with difficulty or no longer performs it at all. A transition matrix

is an origin-destination contingency table and provides descriptive statistics of the stasis

and change states in functional status of the respondents. The resulting matrix will be a 2

by 2 table for each of the measures of functional ability. This analytic step permits the

description of th levels of change, decline and recovery in the sample.

In the second analytical phase, multivariate procedures provide insight into

correlates of decline, recovery and stasis. Specifically, the research will estimate the








effect of SES and health behaviors on functional performance using multinomial logistic

regression. This type of modeling is useful for describing the relationship between the

dependent and independent variables when the dependent variable is not continuous and

polytomous. In this case the dependent variable is arrayed across four categories. The

categories are: 1) decline; 2) recovery; 3) stable with functional limitations; and 4) stable

with no functional limitations. The basic model equation is as follows:

Ln[t (category 4, categoryl] = a + IXI + 02X2 + .. + kXk
Ln[I (category 4, category2] = a + piXi + X2 +... + pkX
Ln[ (category 4, category3] = a + 1pXi + P2X2 + + kXk

For each of the five functional performance measures (walking, climbing,

pushing, lifting, picking up a dime), the relevant transition matrix will be used to create a

four category dependent variable: maintenance of non-limited state, maintenance of

functionally limited state, recovery from limited state, decline into an limited state. The

reference category, category 4, is the maintenance of the non-impaired state. Three sets

of contrasts are enabled: 1) decline versus maintenance of non-impaired state (category 1,

category 4); 2) recovery versus maintenance of non-impaired state (category 2, category

4); and, 3) maintenance of impaired state versus maintenance of non-impaired state

(category 3, category 4).

This research will estimate two regression models for each functional

performance indicator. A positive coefficient indicates that a higher value of the

correlate is associated with either stasis in limitation, decline, or recovery compared to

maintenance of a non-impaired state. A negative coefficient indicates that a lower value

of the correlate is associated with either stasis in limitation, functional decline or

recovery, compared to maintenance of a non-impaired state.






61

The first model includes measures of SES, genetic endowments, and the

covariates. The second model adds the measures of health behaviors. Comparisons of

these two models describe the moderating effect of health promotion activities on the

relationship between SES and functional performance. Specifically, significant SES

coefficients describe a direct SES effect.













CHAPTER 4
DESCRIBING CHANGE AND STABILITY IN FUNCTIONAL PERFORMANCE

This chapter describes change and stability in functional performance. The first

section presents the transition tables for each of the five performance measures. The

second section evaluates the relationship of each of the correlates with the four states of

each functional performance measure.

Transition Tables

A transition matrix is useful for examining changes in functional ability over

time. First respondents are sorted into two categories with regards to each functional

measure using the wave 1 data. The two categories are no difficulty versus difficulty in

performing the function. This is the origin state. Next, it is determined if the respondents

continue to have difficulty or no difficulty in wave 2. This is the destination state.

A cross-tabulation is developed with the cells on the diagonal indicating stability.

That is, respondents who had no difficulty in wave 1 and no difficulty in wave 2 as well

as respondents who had difficulty in both waves 1 and 2.

The upper right hand cell holds information about respondents who deteriorated

over the two waves. The lower left-hand cells provide data about respondents who

actually improved between waves. The cell with the largest number of respondents is the

upper left hand cell indicating that the respondent had no difficulty with the functional

measure in wave 1 and in wave 2. Tables 4 through 8 contain the transition matrices for

the five measures of functional performance.











Table 4. Transition Matrix of Walking Several Blocks

Destination State
Origin State No Difficulty Difficulty Total
(0) (1)
No Difficulty 3100.8 854.14 3954.9
(0) 78.4 21.6 100.0

Difficulty (1) 402.62 1893.9 2296.6
17.5 82.5 100.0

Total 3503.37 2748.08 6251.46
56.04 43.96 100.00
Note: Analysis used weighted data. Each cell in the transition matrix
shows the cell size and the frequency percent. Source: AHEAD, Wave 1
and Wave 2


Walking several blocks. Table 4 shows that 79.9% of the respondents were

stable, either with or without limitations, between wave 1 and wave 2. Forty-nine point

six percent had no difficulty walking several blocks in wave 1 and in wave 2. They are

shown in the upper left cell, representing stability and no difficulty, which is the most

common of the cell states. The other stable state, continuing functional limitations, is

shown in the lower right cell. These respondents, about 30.2% of group, had difficulty in

wave 1 and continued to have difficulty in wave 2.

The upper right-hand and lower left-hand cells show the respondents who

experienced a transition in functional ability between waves. Interestingly enough, 17.5%

of respondents actually improved between waves 1 and 2, having difficulty performing

the function in wave 1 and no difficulty in wave 2. Twenty-one point six percent of the

respondents declined in functional ability between waves 1 and 2.









Climbing stairs. The transition matrix for stair climbing demonstrates similar

results with walking. Approximately 79% of the respondents were stable from one wave

to the next, with 75% of the stable group with no limitations, and 25% stable with

functional limitations. Among the groups experiencing transition in functional ability,

16.8% declined in ability and 24.4% improved between waves 1 and 2.



Table 5. Transition Matrix of Climbing Stairs

Destination State
Origin State No Difficulty Difficulty Total
(0) (1)
No Difficulty 3726.8 750.57 4477.4
(0) 83.2 16.8 100.0

Difficulty (1) 560.63 1213.4 2296.6
24.4 52.8 100.0

Total 4287.46 1964 6251.46
68.58 31.42 100.00
Note: Analysis used weighted data Each cell in the transition matrix
shows the cell size and the frequency percent Source: AHEAD, Wave 1
and Wave 2


Pushing/pulling large objects. This functional ability parallels the results of the

transition matrix for walking. Almost one-half of the group had no difficulty performing

this task in either wave 1 or wave 2. Eighty-six point five percent of respondents were

stable between the two waves of data, with the largest proportion of respondents being in

a state of stability, no difficulty with pushing/pulling large objects (47.09%). Once again,

a meaningful finding here is that almost 24.9% of respondents actually improved in

functional ability between wave 1 and wave 2.









Lifting ten pounds. More than one-half of the respondents (52.3%) reported no

difficulty in lifting 10 pounds in wave 1 and in wave 2. In addition, 24.8% of

respondents recovered by wave 2 from the difficulty they had in wave 1 in lifting 10

pounds. Consistent with walking and pushing/pulling, approximately 21.6% of the

respondents had a decline in functional ability and reported difficulty in lifting weight in

wave 2 when they had no difficulty in wave 1.



Table 6. Transition Matrix of Pushing/Pulling Large Objects

Destination State
Origin State No Difficulty Difficulty Total
(0) (1)
No Difficulty 2943.9 1024.5 3968.4
(0) 74.2 25.8 100.0

Difficulty (1) 568.13 1714.9 2283
24.9 75.1 100.0

Total 3512.08 2739.37 6251.46
56.18 43.82 100.00
Note: Analysis used weighted data. Each cell in the transition matrix
shows the cell size and the frequency percent. Source: AHEAD, Wave 1
and Wave 2



Picking up a dime. As noted previously, the responses to difficulty with this

functional ability differ from the other four. Here almost 85% of respondents had no

difficulty with this function in wave 1 and wave 2 and only 4% had difficulty in both

waves. Similarly to the other functional transition matrices a significant number of

respondents recovered from the functional limitation. Here one-half the number who

declined in ability recovered between waves of the data set.












Table 7. Transition Matrix of Lifting 10 Pounds

Destination State
Origin State No Difficulty Difficulty Total
(0) (1)
No Difficulty 3272.1 900.56 4172.7
(0) 78.4 21.6 100.0

Difficulty (1) 515.93 1562.9 2078.8
24.8 75.2 100.0

Total 3788.04 2463.41 6251.46
60.59 39.41 100.00
Note: Analysis used weighted data. Each cell in the transition matrix
shows the cell size and the frequency percent. Source: AHEAD, Wave 1
and Wave 2



Table 8. Transition Matrix of Picking Up a Dime

Destination State
Origin State No Difficulty Difficulty Total
(0) (1)
No Difficulty 5281.1 464.49 5745.6
(0) 91.9 8.1 100.0

Difficulty (1) 252.45 253.39 505.85
50 50 100.0

Total 5533.57 717.888 6251.46
88.52 11.48 100.00
Note: Analysis used weighted data. Each cell in the transition matrix
shows the cell size and the frequency percent. Source: AHEAD, Wave 1
and Wave 2


These transition matrices respond to the first hypothesis, that a significant

proportion of older adults are free of functional limitations and some of those older adults

who do suffer from functional limitations recover from them within two years. As









indicated in the matrices, half of the respondents have no limitations with walking,

climbing, lifting, or pushing/pulling. Over 80% have no difficulty with picking up a

dime, and 50% of those who have this limitation in wave 1 recovers the ability by wave

2. Respondents had the most difficulty with pushing/pulling and lifting. Both these

activities require both upper and lower body strength, and since this group is largely

female may reflect their lower upper body strength. Between 17% to 25% of those who

do have limitations with the other functional abilities recover between waves 1 and 2.


Correlation of Measures

Tables 9 through 13 contain the correlation coefficients of the measures used in

this study. The first table, Table 9, provides the correlations among the measures of

functional limitations. Functional limitations were measured in wave 1 and wave 2 and

the state in each wave is compared for the correlations. A "decline" state indicates that

the individual had no limitations in wave 1 but had limitations in wave 2. A "recover"

state indicates that the individual had limitations in wave 1 but had no limitations in wave

2. A "stable, limited" state indicates that the individual experienced functional

limitations in both waves. A "stable, no limits" state indicates that the respondent had no

functional limitations in either wave. Table 10 shows the correlations between changes

in functional limitations and SES measures and Table 11 between changes in functional

limitations and health behaviors. Table 12 summarizes the correlations between changes

in functional limitations and measures of genetic endowments and Table 13 correlations

with the covariates.










Table 9. Correlation Matrix: Correlation with changes in Functional Performance Between Waves 1 and 2
Pearson Correlation Coefficients/ N=6,237
Walking Climbing -Pushin /Pulling Picking up a Dime
(1) (2) (3) 1 (4) (1) (2) 1 (3) (4) (1) (2) (3) (4) (1) (2) (3) (4)
decline .22 -.03 .06 ** -.2**
recovery .05** -.21 -.09 .15 **

stable -.06 ** -.04 53 ** -.43
Limited **
stable no .07 -.07 -.52 .57**
limits ** **
decline 0.17 -.02 -004 -.11 .19 -.04 -.07 -.06
** ** ** ** **
recovery .02 -.13 -.03* .08 ** .04 .-11 -.01 -.05
S** ** ** **
stable -.04 -.006 47 ** -.41 .002 .09 .47 .44
limited ** ** ** ** **
stable no -08 ** -05 -.44 .49 ** -.12 -.12 -.38 .46
limits ** ** **
decline .17** .003 004 -.13 .2 ** -.03 -.03 -.09 .26 -.05 -.01 -.16
** ** ** ** **
recovery .01 -.12 .05 .1 ** .01 -.12 -.03* .09 .01 -.21 -.02 .13
** ** ** ** ** **
stable -.03 -.02 .5 ** .43 ** .006 .08 .47 -.44 -.04 -,008 .56 -.47
limited ** ** ** ** ** **
stable no -.09 ** -.05 -.5 ** .51** -.14 -.12 -.04 .5 ** -14 -.07 -.49 .59
limits ** ** ** ** ** ** a **
decline .06** -.04 .13** -14 09 -.03 .13 -.15 .06 -.03 .13 -.14 .07 -.03 -.15 -.16
S* ** ** ** ** ** ** ** ** ** **
Recovery .03* -.01 -.14 -.12 01 -.06 -.12 12 .04* -.04 -.13 .11 -.003 -.03 -.12 12
S_** *5 5* 5* *5 ** 5*
stable -02 -02 .18 ** -.15 -01 .01 18 -.15 -.03* .006 .16 -.13 -.005 -.002 .18 -15
Limited ** ** ** ** ** ** **
stable no -.02 .03 -.27 .24** -.05 -.17 -.26 .25 -.007 .0005 -.25 .23 -05 .008 -.27 .26
limits ** ** ** ** ** ** **
* p< .001 p< .05
Source. AHEAD, Waves I and 2








Functional Limitations

As shown in Table 9, the changes in five functional limitations vary in the

strength and significance of their correlation with one another. The strongest correlation

is between the measure of being disabled in both waves. The correlation ranged from .18

between picking up a dime and walking, climbing, and lifting (between picking up a dime

and pushing/pulling the correlation was .16) to .56 between pushing/pulling and lifting.

This may indicate that respondents suffer from several functional limitations at one time

and that they follow a similar trajectory.

The correlations of the same functional state for each measure of functional

performance are, for the most part, strongly and significantly correlated with each other,

except for picking up a dime. This functional performance may be fundamentally

different from the other measures of functional ability. The other all involve the lower

body, and pushing/pulling and lifting involves the upper and lower body with grosser

muscle movements. Picking up a dime is the only measure that requires upper body

mobility solely and depends on dexterity and fine motor skills. It requires more discrete

movement and may indicate a more serious problem than the other measures of

functional performance.

We gain additional insight by looking at each of the functional abilities on its own

rather than as combined in an index. The functions have differences and shared features

that provide useful information about the respondents. For example, the stable state with

limitations is strongly correlated between walking and climbing (.53) and between lifting

and pushing/pulling (.56). These two functional abilities share characteristics in that they

use similar parts of the body. It appears that when a respondent has limitations with one

functional measure they may also have difficulty with another. Limitations with walking









are also strongly correlated with limitations in lifting (.5). This indicates that difficulty

with walking is connected with difficulties in other functions. Additionally, the

correlations demonstrate that while respondents have a tendency to have more than one

functional limitation at a time, the other strong correlations are between the functional

abilities in the stable state without limitations. The correlation there range from .59 to a

low of.23 with picking up a dime.


Table 10. Correlation Matrix: Correlation with SES Variables
Pearson Correlation Coefficients N = 6237


Total Household
Income


-.044 **
.018
-.145 **
.173 **
-.055 **
.071 **
-.168 **
.216 **
-.036 *
.021
-.141 **
.165 **
-.064 **
.042 **
-.136 **
.187 **
-.048 **
.054 **
-.026 *
.078 **


Net Worth
(assets-debt)
-.046 **
.025*
-.209 **
.238 **
-.064 **
.088 **
-.231 **
.284**
-.046 **
.016
-.193 **
.217**
-.061 **
.053 **
-.228 **
.271**
-.072 **
.093 **
-.049 **
.130 **
Source: AHEAD, Waves 1 and 2


Socioeconomic Status


In Table 10, the correlations between SES and changes in functional limitations,

show a significant, if modest, relationship between income, net worth and the measures


decline
recovery
stable limited
stable no limits
decline
recovery
stable limited
stable no limits
decline
recovery
stable limited
stable no limits
decline
recovery
stable limited
stable no limits
decline
recovery
stable limited
stable no limits
p < 05


I
U


Sp< 001


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









of change in functional limitations. The correlation coefficients indicate that functional

abilities are less likely to be limited with increasing income and net worth. Maintenance

of non-limited state is correlated with higher SES. Similarly, stasis in a limited state is

associated with lower values of the SES measures. Transitions are more weakly

correlated with SES than static states. SES correlates more frequently with functional

decline than recovery, based on the number of significant correlations. Functional

declines are concentrated among lower values of SES. In contrast, higher values of

income and net worth are associated with recovery.


Health Behaviors

The measures for health behavior used in this study are smoking status in wave 1

(former, current, or non-smoker), number of alcoholic drinks consumed in an average day

at wave 1 (abstains, less than 1 2 is a light drinker, and 3 or more is a drinker), body

mass index (BMI) at wave 1, participation in vigorous physical exercise three times a

week in wave 2, and participating in preventive care (measured by number of preventive

health care screenings for cholesterol, breast, cervical or prostate cancer, high blood

pressure or a flu shot) in wave 2, social networks (number of children, marital status) and

years of education. The correlations are shown in Table 11.


Stable States. The strongest and significant correlations are between the stable

groups (no change in functional limitations between wave 1 and wave 2) and education,

drinking, exercise and participating in preventive screens. The stable states seem more

affected by health behaviors than are the transition states. Drinking behavior is measured

in wave 1, while exercise and preventive behavior were first asked of respondents in









wave 2. Abstaining and heavy drinking respondents are more likely to experience

functional limitations than are light drinkers, those who exercise vigorously, are educated

and those who engage in preventive behaviors. This is consistent with previous research

that pointed to a U-shaped relationship between drinking behavior and disability

(Gronbaek et al. 1998). In other words, light drinking is less correlated with health

problems than is abstaining or heavy drinking.

Additional health behaviors that are statistically correlated with the stable states

of most of the functional measures are weight and smoking. Being underweight or

overweight is associated with increases in functional limitations, especially with walking,

while normal weight is positively correlated with remaining stable with no functional

limitations over both waves. Being a former smoker is positively correlated with

remaining stable without functional limitations, but being a current non-smoker is not

Smoking is correlated with difficulties in walking while being a former or non smoker is

more highly correlated with difficulties in climbing stairs.


Transition States: Decline and Recovery. Exercise is significantly correlated

with all states of functional limitations. The association differs depending on functional

ability. Thus, exercise and recovering from functional limitations go hand-in-hand, as

does lack of exercise and decline in functional ability. Drinking alcohol is significantly

correlated with these functional states as well. Just as with the stable group, respondents

who abstain from alcohol are more likely to experience functional limitations, while

those who drink fewer than three drinks per day are less likely to experience limitations.









Covariates

The covariates used in the study are sex, age, race/ethnicity, number of medical

conditions, insurance coverage, doctor visits, number of prescriptions and genetic

endowments. The positive, significant correlations shown in Table 13 for the stable

status with functional limitations indicate this state is more likely for respondents who are

female, Black, older, without insurance, have many medical conditions and visits to

doctors, and are taking more medications. The opposite trends are noted for the stable

status with no functional limitations.

Strong correlations are seen for stable states of walking between age and number

of medical conditions. The same is true for climbing except that prescription use is also

strongly correlated. Pushing/pulling stable states are strongly correlated with the

respondent's sex, number of medical conditions and number of prescriptions. This may

reflect women's lesser upper body strength, as lifting follows a similar pattern, with the

addition of strong correlations with age. The strongest correlations among the stable

states of picking up a dime and the covariates are number of medical conditions and

number of prescriptions. The decline and recovery status groups follow the stable, with

limitations group, but the coefficients are not as large.


Genetic Endowments. The measures here are proxies for biomedical markers.

For this research the age of parents and the number of surviving siblings is used to

indicate the positive health benefits passed on from one generation to the next. Parent's

age is measured as the age at death for deceased parents or the current age of living

parents.
















Table 11. Correlation Matrix: Correlation with Health Behaviors
Pearson Correlation Coefficients N=6,237 unless noted otherwise
Educa- Non Former Smoker Abstain Light Drinker Normal Over Under Exercise Social Sooial Preventive
tion Smoker Smoker Drinker Weight weight weight N=6,235 network: network: Screens
marital children N=6,043
.. ...... ..._~ _status
D e c l in e .0 3 5 .0 0 4 .0 0 9 .. ,' ,. ,: ..." '
Recovery .017 -.01 .006 .007 -.007 .01 -007 .02 -.02 -01 .03' -.03* -.00001 .002
Stable limited -. 169*" .02 -03' .02 .17" 16" -.04* -07" 04* .06" -.3'* .31" .05" -12"*
Stableno .189* -.02 .04* -.03* -.18* .17"* .02 07" -.05" -.05 35"* -.32* -.07"* .1*
limitations
Decline -054" -.01 -.01 04" .04* -04" .01 .003 -.003 .006 1* 03' -.002 -.03'
S Recovery .082" -03* .03* .008 -07** .06** .02 .02 -.02 -006 .058" -.07' -.03* .023
Stable limited -.196 .04" -.04* .0004 18" -.17"* -03' -.04" .01 .08* -.22* .27"* .07"* -13"
Stable no .244" -04* .06* -.02 -2"* .2* .03* 05** -.02 -07** 29** -28* -.08* 14**
limitations
Decline -.045" .006 -02 .02 .04* -.04' -.02 0005 -008 .02 -097"* -.005 -.008 -011
Recovery .019 009 -.007 -005 -.04* .04* 003 .008 -.005 -.008 .04* -.07" -.006 -.016
Stable limited -.126" .1" -.1" -.004 17" -15"* -.06" .01 -.04* .08" -.24* .27" .06" 11**
Stableno .157 -.09' 1"* -.01 -.2* .18"* 07"* -.006 .04" -.09"* 3"* -27" -.05" .11'
limitations
Decline -.027' -.009 -004 .01 03 -.03' -002 03* -.03' -.004 .09"* 004 .007 -006
SRecovery .043* .001 -006 .007 -05" .05" 01 02 -.008 -.03' .02 -.06 -02 -.009
Stable limited -.141* .09* -.09* .008 .15" -.14" -05" .01 -.04' 07" -.24* .26" .05"* -09"*
Stableno 166" -.07* 08" -.01 -19* .17" 05" -.02 .05* -.08" .29" -.27" -06" .08"
limitations
Decline -.035- .03* -.03 .005 .05 -.05** -.006 -.01 -.003 .04* -.07** .08** -.01 -.004
SRecovery 082* -.01 .01 .006 -05** .05" .01 .02 -02 -001 .056"* -09** -.04"* .032*
Stable limited -.043* -.005 .002 .002 .04"* -.04 -.019 01 -.02 .02 -.05" .14" .03* -.02
Stableno .094" -.02" .03' -001 -.09" .08** .02 .01 004 .02 .11" -18'* -03* .03*
limitations
*p<.001 *p .05
Source: AHEAD, Wave 1












Table 12. Correlation Matrix: Correlation with Genetic Endowments
Pearson Correlation Coefficients


decline
recover
stable limited
stable no limits
S decline
S recovery
stable limited
U stable no limits
S decline
S recovery
stable limited
stable no limits
decline
.B recovery
3 stable limited
stable no limits
decline
recovery
stable limited
S stable no limits
**p<.001 *p<.05
Source: AHEAD, Wave 1


Mother's Age Father's Age
N=5,823 N=5,657
-.01 .007
.005 -.004
-.06** -.03*
.07** .02*
-.005 .03*
-.005 .03*
-.07** -.008
.06** .009
-.002 .01
.01 .03*
-.04* -.02
.04* .02
-.04** -.01
-.009 .01
-.03* -.01
.05** .03*
-.04* -.001
.04* -.008
-.0001 .013
.05** -.01


Mother's age and number of siblings are negatively correlated with the measures

of functional limitation (Table 12). The strongest relationship is between walking,

climbing and pushing/pulling in the stable states. These tasks all require large muscle

groups. Father's age is not as strongly related and is significant for the stable states in

walking function and the decline/recovery in climbing stairs. All the significant

correlations indicate that less functional limitations and recovery are related to increasing

age in parents and higher numbers of surviving siblings.


# Siblings
N=6,220
-.006
.02
-.04*
.05**
-.005
.0006
-.04**
.04**
.02
.002
-.05**
.03*
-.02
-.002
-.05**
.05**
-.02*
-.003
.002
.02


--











Table 13. Correlation Matrix: Correlation with Covariates
Pearson Correlation Coefficients Number ofObservations=6,237, Prescriptions N=5,262

Respond- Non-Latino Black Other Age Medical Insurance Doctor Prescrip-
ent's Sex White Conditions Plan(s) Visit tions
decline -.007 -04" .04"* .01 .03* .03* -.02 -.004 -.002
recover -.033' .003 .004 -.01 -002 -.03* .004 -02* -02
S stable limited 13"* -.06** .07 .01 .24** .31" -.05 .08"* .3"
stable no limit -.13" .09** -.09** -.03* -.24"* -.32* 06** -.08** -.03**
decline -.009 -03 .02 .008 .07* .03* -.05* -.014 .02
S recover .-06" .06* -.05" -.04* -.03* -07"* .005 -.02 -.06"
0 stable limited .14*" -. 11" .07" .09** .23" .27** -.07* .06" .25"*
stable no limit -.14"* .15"* -.1"* .1" -.25** -.28"* .09"* -.05" -.25"
decline .03* -.03* 03* .002 .04* .004 -.04** .003 .004
recover -.05" .03* -.01 -.02 -.009 -.06"* .01 -.03* -.04*
stable limited .23** -.05** .04 .02* .17" .26** -.03* .08* .24**
stable no limit -.25"* 08"* -.07" -.04* -.18"* -.27** .06** -.09" -.25**
decline 07** -04* .06" .009 .05** -.005 -02 .007 -01
recover -.05** .03* -.008 -.04* -.02 -.07" -.008 -.02 -.06
stable limited .25** -.06" .06" .01 .22* .27 -.05" .08** .26"
stable no limit -.28" .1"* .1 -.03* -.24" -.27* .06** -.08** -.26'
S decline .02 -.02 .03* -.004 .07* .08" -.02* .01 05**
S recover -.02 .06" -.05** -.03' -.07" -.09"* .07* -.03* -.06"*
stable limited -.005 -.03* .04* -.004 .08** .14" -.02 .03* .11"
stable no limit -.02 .07" .07* -.013 -.13" -.18"* 07" -.04** -.14**
*p<.001 *p<.05
Source: AHEAD. Wave I and Wave 2















Table 14. Correlation Matrix: Correlation between all variables


(1) Income
(2) Net Worth
(3) Education
(4) Non Smoker
(5) Former Smoker
(6) Smoker
(7) Abstainer
(8) Light Drinker
(9) Drinker
(10) Exercise
(11) Screens
(12) Normal Weight
(13) Underweight
(14) Overweight
(15) Mother's Age
(16) Father's Age
(17) # of Siblings
(18) # of Children
(19) Respondent's Sex
(20) White
(21) Black
(22) Other
(23) Age
(24) Married/Partner
(25)# Medical Cond.
(26)# Medications
(27) Gov. Insurance
(28) Medicaid
(29) Private Insurance
(30) Doctor Visit
*p<.01 "p<.001


.5" 1.0
.4" .5" 1.0
-.06" -.06" -.03 1.0
.09" .09" .05*" -.82" 1.0
-.04" -.05" -.02* -.32" -.28"1 10
-.22" -.27" -.26" .20" -.17" -.07" 1.0
.21" .26" .26" -.17" 15" 03' -.96" 1.0
05" .03* .02 -.11" .04" .11" -16" -.13" 1.0
11" .17" .14"* -.02 .04" -.04" -.1" .11" -009 1.0
.19" 23" .18" -.09" .13" -.08" -.12" .12" -.009 .11" 1.0
04" .06" 1" .02 -.06" 06" -.02 .02 .002 .05" -05" 1.0
-.05" -.07" -.05" .01 -.05" .06" .05" -.05" -.01 -.05" -.08" -.18" 1.0
-06' -03* -.08" -.02 07"- -.08" -.0004 .001 .002 -.03" .08" -9" -.21" 1.0
.07" .09" .09" .004 -.0004 -.007 -.03' .03* .01 .04* .04* .02 .009 -.03* 1.0
-.01 .002 .008 .003 .003 -.009 02 -.01 -.03* .009 -.004 -.001 .01 -.003 .08" 1.0
-.05" -.05" -17" .02 -.03' 02 .06" -.05" -.02' 05" .03* -03* -.02 .04' .03* .09"
-08" -.15" -.24" -.03* -.02 -.006 .09" -.08" -005 -.01 -.03' -09" -.006 .09" -.03' .01
-17" -.17" -.02 .38" .35" -.06"* .17" -.13" -.11"* -.12" -23" .05" .09" -.09" -.04" -.02'
.22" .35" .39" -.03' .04" -.02 -.17" .17" .004 .1" 09" .12" -.03* -.1" .07" -01
-.17" -.27" -.24" .04' -04" .01 .14" -.15" 007 -.07" -.06" -.1" .008 .09" -.06" .01
-.13" -.19" -.3" .002 -.007 .009 .08" -.08" -.02 -.07" -.05" -05" .03* .04* -.03' .003
-.15" -.18" -.15" .16" -.09" -12" .12" -.1" -.07" -.16" -.21" .11" .07" -14" -.05" .02
32" .35" .14" -.14" .15" -.01 -.13" 12" .04' ." .19" -.02 -.05" 04" 04* -.006
-08" -.13" -.11" -.06" .06" -.006 .12" -.11" -.02 -14'" 09" -.06" .008 .05** -.03 -.007
-03* -.08" -.03' .004 03* -.06 .11" -.1" .04' -.18" .14" -.06" .004 .06" -.06" -.03'
05" 03* .03' -.02 .008 .03* -.008 .005 .012 .007 .03* .005 .003 -.006 .03* .01
-.26" -.37"* -.35" .02* -.03' .02 .16" -.16" -.02 -10" -.08" -.07" .06" 05" -.05" .009
.13" 18" .19" -.008 .03' -.04' -.08" .09" -.007 .05" .1" .03* -.03' -02 .02 -.01
04* 04" .04* -.005 .07" -.11" -.004 .011 -02 -.04' .25" -.03' -.02 04' -05" -.005


)I )- (u )r )o (lv )r (lJ )1* (r~ ) ( )


--


) 2( (3 4 (5 6 7


8P 9 I\ ri i\ ri\ rr I\ /r












Table 14. Correlation Matrix (continued)


(1) Income
(2) Net Worth
(3) Education
(4) Non Smoker
(5) Former Smoker
(6) Smoker
(7) Abstainer
(8) Light Drinker
(9) Drinker
(10) Exercise
(11) Screens
(12) Normal Weight
(13) Underweight
(14) Overweight
(15) Mother's Age
(16) Father's Age
(17) # of Siblings
(18) # of Children
(19) Respondent's Sex
(20) White
(21) Black
(22) Other
(23) Age
(24) Married /Partner
(25) # Medical Cond.
(26) # Medications
(27) Gov. Insurance
(28) Medicaid
(29) Private Insurance
(30) Doctor Visit
*p<.Ol *p<.OO1


(17) (18) (19) (20) (21) (22) (23) (24) (25) (26) (27) (28) (29) (0)


.08" 1.0
-.03* -.009 1.0
-.07" -.19"* -03* 1.0
.01 08* 04" 8" 10
09* .2" -.01 -.54" -.1"* 1.0
-17" -07* .08" .007 0003 -01 1.0
01 -.006 -.38* .14'" -14"* -.03' -23"* 1.0
-.03 .06" -.03' -05" .04* .03' .04* -05 1 0
-.05* .02* .07** .01 -.009 -.007 .04* -.03' .47** 1.0
.02 -.02 -06" -.01 .02 -.007 -04' .04" .01 .00006 1.0
.01 22" 09** -.32'* 19'* .25* .07"* -.17"* .12"* .09"* -.06" 1.0
-.03* -.09" .001 .22" -.16"* -.14" -.01 .09" .003 .04' -05" 2* 10
-.03* .02 .02 -.004 .007 -.004 .009 .02 .2** .27** -.0006 02 .09" 1,0








Overall Correlations

Table 14 shows the correlations between the variables used in this research. First

we note that some multicollinearity exists between certain variables, such as between

income, net worth and education. The resulting models may be less precise and have

larger standard errors. To offset this difficulty, regression models will be estimated and

tested in a step-wise fashion to determine the effects on the significance of the estimates

and the standard errors. As a result, the regression modeling will use either income or net

worth as an indicator of SES.

Education is used as a health behavior, as it indicates self-efficacy. Other

correlations with income and net worth show a positive correlation with respondents who

are White, male, married or co-habiting and of younger ages. Negative correlations

among health behaviors are seen among respondents who abstain from drinking, are

overweight or underweight and with increasing number of children. Among the

covariates, income and net worth are negatively correlated with Medicaid, number of

siblings, higher numbers of medical conditions, and higher numbers of prescriptions.

Other examples of multicollinearity exist between number of medical conditions

and number of medications and net worth, education and Medicaid coverage. However,

each measure adds to the theoretical basis of the model. These variables are included for

control purposes and so the research is not hampered by more conservative tests of

significance that may result from the multicollinearity

Indicators of the connection between education and positive health behaviors are

seen here with the significant positive correlation between education and normal weight,

light drinking, exercise, participation in preventive screens and former smoker, as well as

the negative correlation between education and number of medical conditions. This is an








additional indication that education can be used as an indicator of self-efficacy and

motivation to engage in preventive health behaviors.


Response to Hypotheses

The data indicate that the first hypothesis, that a significant proportion of older

adults are free of functional limitations and do experience recovery from limitations when

they do exist, is confirmed by the transition matrices. For walking, pushing/pulling, and

lifting, just under one-half of the respondents are free of functional limitations in both

waves of the AHEAD data set and between 17% to 25%, depending on the measure of

functional ability, experienced improvement in their functional status between waves.

Respondents had more difficulty with pushing/pulling and lifting and much less difficulty

with picking up a dime.

The second, third, and fourth hypotheses, that individuals with greater economic

resources: 2) have stronger functional status; 3) are less likely to suffer decline; and 4)

are more likely to recover, are also supported by the correlation coefficients. The

coefficients indicate the individuals with more economic resources are less likely to have

limitations in either wave. The largest coefficients are for climbing, lifting, and walking.

The signs of the coefficients also indicate that greater income and net worth are

associated with better functional ability, less likelihood of decline, and greater likelihood

of recovery.

Likewise, the fifth hypothesis is partially supported by the bivariate data. This

hypothesis is that income is a better predictor of recovery than is net worth for the

transition states of decline and recovery and net worth is a better predictor of stasis states.

According to the correlation coefficients, net worth is more strongly correlated with all









functional states than is income. The coefficients for the correlations are larger for net

worth than for income, indicating that net worth has more influence on functional

limitations than does income.

The sixth hypothesis, that intervening health behaviors will modify the

relationship between SES and health, will be addressed through multinomial logistic

regression. However the correlation coefficients in Table 14 show us the health

behaviors do vary by income and net worth. Education is strongly and positively

correlated with both income and net worth. Exercising, being a light drinker and

participating in preventive screens are also strongly, positively correlated with the

measures of SES. The measure of health behavior with a strong negative correlation

with SES is abstaining from alcohol. Other negative correlations are number of children,

being underweight or overweight, being a current smoker or a non-smoker. Being a

former smoker is positively correlated with SES.

Health behaviors also have influence on functional status, especially in the stable

states. Education, light drinking (versus abstaining), exercise and participating in

preventive screens are all positively correlated with remaining free of functional

limitations in both waves, and negatively correlated with the stable state with functional

limitations, especially for walking and climbing. Education, exercise and light drinking

are also strongly correlated with lifting and pushing/pulling, but not as strongly correlated

with picking up a dime. The presence of a social network is significantly correlated with

all the stasis states, but the correlation coefficient is smaller than for other measures of

health behavior. The same is true of the BMI and smoking variables. There are






82


indications even at the bivariate level that health behaviors have strong influence on

functional limitations.













CHAPTER 5
PREDICTING CHANGE IN FUNCTIONAL PERFORMANCE


Multinomial Logistic Regression Modeling

The correlation coefficients all indicate that recovery is more likely with greater

economic resources, as stated in the second hypothesis. In order to determine the exact

relationship as well as any intervening influences of health behaviors (the fourth

hypothesis), it is necessary to look at multiple variable modeling. In this case,

multinomial logistic regression modeling was used. The purpose of the modeling is to

determine the maximum likelihood of obtaining the particular data that are actually

observed (Power and Xie, 2000). This type of modeling is the comparison of

membership in the four categories established. As a result, membership in each category

is compared to the probability of membership in the reference category, in this case,

being stable in each wave with no functional limitations.

Through testing the models and looking at the correlation matrices, it appeared

that household income, net worth, and years of education were too correlated to use

together in the logistic regression model. First, education is kept in the models because,

while it is associated with SES, it is also a predictor that indicates motivation to engage in

preventive behavior, which may have a different affect on health recovery. It is a

measure of self-efficacy in this research and, thus, part of the set of variables measuring

health behaviors. Net worth is a stronger predictor of functional ability than is household

income in these models (p<.001). Therefore, the modeling was done using net worth as

the indicator of SES and years of education as a health behavior variable.
83








There are two regression models. The first model tests regressing functional

ability on net worth and the control variables. The second model adds preventive health

behaviors, including education. The results of comparing the first and second models

show that the second, saturated model is more predictive of functional limitations

(p<.001). Therefore, the following description is of the complete model that includes the

preventive health behavior variables.


Model Comparisons

Stable with Functional Limitations

Comparing the models of functional abilities illustrates the benefits of analyzing

each functional measure separately. Although walking and climbing are similar, there are

differences between them as among picking up a dime, lifting and pushing/pulling. This

section presents comparisons of both stable states (see Table 15). The only variables

statistically significant for all five functional abilities are age, number of medical

conditions, and number of prescriptions, which are all control variables.

By comparing odds ratios (coefficients are reported in Appendix B), we can

determine the risk of falling into the stable category with no limitations versus the stable

category with limitations. A positive coefficient and odds ratio grater than 1 signals

higher odds of be in the stable category with limitations A negative coefficient and odds

ratio less than one indicates higher odds of being stable with no functional limitations.


Socioeconomic Status. The measure of SES, net worth, is significant for all

functional limitations except for picking up a dime. The odds ratio indicates that the odds

of a stable state with functional limitations is higher among the older adults with a lower









net worth, especially for climbing stairs and lifting (.88). Walking and pushing/pulling

odds are slightly better at .93/.92.




Table 15. The Odds Ratios of Being in a Stable with Limitations State.
Walking Climbing Lifting Pushing/ Picking up a
Stairs Pulling dime
SES
Net Worth .93** .88** .88** .92** .98

Health Behaviors (ref. categories: Smoking = non-smoker; Drinking abstains; Weight= normal wt.)
Education .97* .97* 1.00 .98 1.01
Smoker 2.79** 1.69* 1.64* 1.64* 1.19
Former Smoker 1.47** 1.26* 1.22* 1.05 .99
Light Drinker .67* .60** .67** .69** .90
Drinker .58 .41* .28* .36* .30
Preventive Screens .95** .95** .98 .99 .99
Exercise .17** .22** .27** .27** .78
# Children 1.03 1.04* 1.05* 1.04 1.02
Married/Cohabiting 1.05 1.11 1.16 1.25* .78
Underweight 1.38 1.63* 2.12** 1.64* 1.20
Overweight 1.67** 1.37** .83* .83* .87
COVARIATES
Genetic Endowments
Father's Age .99 1.00 1.00 1.00 1.01
Mother'sAge 1.00 1.00 1.00 1.00 1.00
# Siblings 1.03 1.01 1.00 .98 1.10*
Respondent's Demographics (referent category: Race White)
Sex (1=female) 1.72** 1.96** 6.48** 4.69** .71
Black .86 .95 .90 .75 1.27
Other (Latino, Asian, .61* 1.65* .59* .85 .40
or Native American)
Age 1.12** 1.11** 1.10** 1.06** 1.07**
Medical Conditions 2.09** 2.01** 1.91** 2.07** 1.63**
Utilization Issues (referent category: Insurance = Government Insurance (Medicare/CHAMPUS))
Medicaid 1.22 1.21 1.85** 1.21 1.55
Private Pay .75 .71 .67* .63* .65
Doctor Visit .99 1.08 1.08 .96 .99
Prescriptions 1.41** 1.28** 1.28** 1.31** 1.16**
Intercept Coefficient
Constant -9.16 -9.81 -9.39 -6.33 -9.57
Model Log Likelihood 3689.853 3417.529 3365.676 3635.678 1430.555
N=4,977
*p<.05 **p<.001
Source: AHEAD, Waves I and 2.








Health Behaviors. Exercise is the most frequent significant health behavior

predictor. The odds ratio for exercise is approximately .2 for walking, climbing, lifting

and pushing/pulling, indicating that those who do not exercise are 80% more likely to

have functional limitations. Unfortunately the question regarding exercise was not asked

in wave 1, so it is difficult to determine if exercisers are more likely to be non-limited or

if the non-limited are more likely to exercise. However, this is an indication that

engaging in vigorous physical activity is beneficial to maintaining functional ability.

All the preventive health behaviors are significant for the functional ability of

climbing stairs, except for marital status. The highest odds ration of 1.69 is for smokers

as compared to non-smokers. This indicates that smokers are 69% more likely to report

difficulty in climbing stairs than non-smokers are. This is similar to the odds ratios for

being underweight as compared to normal weight. There, individuals who are

underweight are 63% more likely to report difficulty with climbing stairs than normal-

weight older adults are. This may reflect a loss of muscle when losing weight as one

ages. Similar results are seen with lifting and pushing/pulling, which may also reflect a

loss of muscle mass. Social networks is another health behavior that is positively

correlated with climbing and lifting limitations.

Overweight individuals have similar limitations with walking and smokers have

nearly 200% more difficulty walking than non-smokers do. Walking does not require as

much muscle, and lack of lung capacity may be more influential here reflected in the

odds ratios for smokers, even former smokers, and overweight older adults. Persons who

regularly complete preventive health screens are less likely to report walking and

climbing limitations; otherwise, preventive screens do not distinguish between stability in

limitations of upper body functions.








Drinking is also a significant predictor of stability in functional ability for all

functions, except for picking up a dime. None of the health behaviors' odds ratios are

statistically significant for picking up a dime. The odds ratios for drinking indicate that

abstainers are 30 to 40% more likely to suffer functional limitations than light drinkers.

These results mirror the correlation coefficients that indicate that abstaining has more

negative consequences. This is also true of older adults who consume more than 3 drinks

a day, which is a very small proportion of the respondents, for climbing, lifting and

pushing/pulling.


Covariates. Measures of genetic predisposition included parent's age at death, or

current age if still alive, and number of siblings. None of these measures are statistically

significant except for number of siblings with picking up a dime. The odds ratios here

are 1.10, indicating that persons who have more siblings have more difficulty picking up

a dime. This is counter to the expectations that a strong genetic heritage as measured by

many siblings would result in better health at older ages. This is also a change from the

bivariate analysis in which a number of the correlation coefficients, especially those

revealing the effects of mother's age at death, were significantly correlated with the

measures of functional limitations.

Unlike the correlation coefficients in the bivariate analysis there are no significant

differences between whites and African Americans. The race/ethnicity category of

"Other," which includes Latino, Asian American, and Native Americans, is statistically

significant, and indicate that these respondents are less likely to have functional

limitations in walking and lifting than the white respondents, and more likely to have

climbing limitations than white respondents. The other demographic variables that are

significantly predictive are age and sex. Women are more likely than men to have








walking, lifting and pushing/pulling limitations. For all functional limitations, each

additional year of age raised the risk of having limitations in waves 1 and 2 by

approximately 10%.

Forms of insurance are also statistically significant for certain measures of

functional limitations. Limitations in walking, climbing or picking up a dime is unrelated

to type of insurance. Respondents with Medicaid are more likely to suffer limitations as

compared to the reference group, which has Medicare or CHAMPUS, for climbing (19%)

and lifting (79%). Individuals with private pay supplemental coverage are less likely to

have functional limitations with lifting (34%) and with pushing/pulling (37%) than

individuals with Medicare or CHAMPUS only. It may be that this is not significant for

all measures because so many of this group have some form of coverage. The presence

or absence of coverage does not vary much.

From the odds ratios, it appears that the extra- individual factors of the

disablement process model (Verbrugge and Jette, 1993), or the health care access and

utilization factors, are the most predictive of functional limitations. Each additional

prescription raises the odds by roughly 30% of remaining with functional limitations

(16% for picking up a dime), and each additional medical conditions doubles the risk,

holding all other variables constant.


Decline in Functional Status

This section describes the contrast between functional decline and stable with no

limitations respondents (see Table 16).









Table 16. The Odds Ratios of Declining Functional Status
Walking Climbing Lifting Pushing/ Picking up a
Stairs Pulling dime
SES
Net Worth .93** .90** .90** .96* .93**

Health Behaviors (ref categories: Smoking = non-smoker; Drinking = abstains; Weight= normal wt.)
Education .97 .95** 1.01 .96* 1.01
Smoker 1.68* 1.65* 1.7** 1.48* 1.00
Former Smoker 1.17 1.06 1.31* 1.19 .82
Light Drinker .81* .76* .76* .83* .82
Drinker 1.47 .78 .38* .86 .64
Preventive Screens 1.01 .98 1.02 1.01 1.04*
Exercise .32** .39** .39** .43** .55**
# Children 1.05* .97 1.02 1.02 .96
Married/Cohabiting 1.19 1.18 1.19 1.22* .92
Underweight .74 1.52 1.69* 1.28 1.41
Overweight 1.19 1.14 .91 .81* 1.08
COVARIATES
Genetic Endowments
Father's Age 1.00 1.00 1.00 1.00 1.00
Mother's Age 1.00 1.00 1.00 1.00 1.00
# Siblings 1.00 1.00 .99 1.02 1.00
Respondent'csDemographics (referent category: Race = White)
Sex (1=female) 1.32* 1.15 3.05** 2.45** 1.02
Black 1.05 1.04 1.16 .98 .95
Other (Latino, Asian, .74 1.15 .62 .66 .64
or Native American)
Age 1.06** 1.07** 1.06** 1.04** 1.04**
Medical Conditions 1.53** 1.38** 1.28** 1.37** 1.35**
Utilization Issues (referent category: Insurance = Government Insurance (Medicare/CHAMPUS))
Medicaid .85 .73 1.02 .80 1.35
Private Pay .82 .69 .68 .56* .71
Doctor Visit .85 1.18 1.16 1.17 .82
Prescriptions 1.18** 1.14** 1.09** 1.14** 1.02
Intercept Coefficient
Constant -5.89 -6.07 -5.71 -4.06 -5.01
Model Log Likelihood 3689.853 3417.529 3365.676 3635.678 1430.555
N=4,977
*p<.05 **p<.001
Source: AHEAD, Waves 1 and 2.


Socioeconomic Status. The odds ratios for net worth are significant for all five

measures of functional ability. Relative to the referent group, the risk of decline decreases




Full Text
xml version 1.0 encoding UTF-8
REPORT xmlns http:www.fcla.edudlsmddaitss xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.fcla.edudlsmddaitssdaitssReport.xsd
INGEST IEID E130G09UG_J98SKJ INGEST_TIME 2013-03-25T12:30:15Z PACKAGE AA00013601_00001
AGREEMENT_INFO ACCOUNT UF PROJECT UFDC
FILES