University of Southern California Dissertations and Theses

Longitudinal change in active life expectancy: the longitudinal studies of aging 1984-2000

(USC Thesis Other)

Longitudinal change in active life expectancy: the longitudinal studies of aging 1984-2000

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

LONGITUDINAL CHANGE IN ACTIVE LIFE EXPECTANCY:
THE LONGITUDINAL STUDIES OF AGING 1984-2000

by

Aaron Timothy Hagedorn
________________________________________________________

A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(GERONTOLOGY)

May 2008

Copyright 2008 Aaron Timothy Hagedorn
ii
TABLE OF CONTENTS

List of Tables

iii
List of Figures

viii
Abstract

xiv
Chapter I: Introduction

1
Chapter II: Literature Review

10
Chapter III: Data Description and Methods

37
Chapter IV: Change in Disability-Free Life Expectancy for
Americans 70 Years Old and Over between 1984 and
2000
83

Chapter V: Changes in Life Expectancy and Active Life
Expectancy: Total and by Sex between 1984 and 2000

101
Chapter VI: Changes in Life Expectancy and Active Life
Expectancy by Race between 1984 and 2000

132
Chapter VII: Changes in Life Expectancy and Active Life
Expectancy by Education between 1984 and 2000

181
Chapter VIII: Conclusions and Discussion 230

Bibliography 245

iii
LIST OF TABLES

Table 2.1. Life Expectancy at Age 70 Across Recent Decades 21
Table 3.1 Description of the Similarities and Differences in the Two
LSOA Datasets
41
Table 3.2 Proportion Institutionalized in Later Waves, and the
Additional Weighting Factor Used to Match National
Long-term Care Survey Estimates.
46
Table 3.3 Basic Sample Frequencies at Baseline and Final Wave
Initial Wave
49
Table 3.4 Distribution of disability status variables in the LSOA I and
LSOA II
51
Table 3.5 Frequency of disability status variables in the LSOA I and
LSOA II
52
Table 3.6 Distribution of Missing Data in the LSOA I and LSOA II 54
Table 3.7 Cross Tabulation of Second and Third Interview with Vital
Status at Each Time Point
56
Table 3.8 Detailed View of Health Transitions Before and After
Imputation
59
Table 3.9 Distribution of Health Statuses in Each Survey 66
Table 3.10 Number of Health Transitions for Adults Aged 70 and
Older
69
Table 3.11 Total Life Expectancy Reported by NCHS for the United
States: Vital Statistics Comparison with Total Life
Expectancy Estimates at Two Time Points
75
Table 3.12 Vital Statistics Comparison with Total Life Expectancy
Estimates By Sex
76
Table 3.13 Vital Statistics Comparison with Total Life Expectancy
Estimates By Race
76
Table 3.14 Vital Statistics Comparison with Total Life Expectancy
Estimates: Males by Race
77
iv
Table 3.15 Vital Statistics Comparison with Total Life Expectancy
Estimates: Females by Race
77
Table 3.16 Comparison of Total Life Expectancy by Education level
with NCHS Education Life Tables
79
Table 3.17 Comparison of Active Life Expectancy by Education level
with NCHS Education Life Tables
80
Table 4.1 Disability Transition Equation Coefficients by Transition
Type
87
Table 4.2 Total, Disability-free, and Disabled Life Expectancy (with
confidence intervals): LSOA I and LSOA II
92
Table 4.3 Status Based Health Expectancies at Age 70 by Year and
Initial Health Status
96
Table 5.1 Distribution of Disability-state Transitions for Males and
Females Between Observation Intervals
107
Table 5.2 Disability Transition Equation Coefficients By Transition
Type
110
Table 5.3 Total, Active, and Disabled life expectancy (with
confidence intervals) for Males and Females: LSOA I and
LSOA II
124
Table 5.4 Status Based Health Expectancies at age 70 by year and
Initial Health Status
Males
128
Table 6.1 Disability Transition Equation Coefficients for Race and
Age
135
Table 6.2 Disability Transition Equation Coefficients for Race, Sex,
and Age
136
Table 6.3 Number of Health Transitions by Race for Adults Aged 70
and Older: Blacks
139
Table 6.4 Number of Health Transitions by Race for Adults Aged 70
and Older: Whites
139
v

Table 6.5 Number of Health Transitions by Race for Adults Aged 70
and Older: Blacks Males
139
Table 6.6 Number of Health Transitions by Race for Adults Aged 70
and Older: Blacks Females
140
Table 6.7 Number of Health Transitions by Race for Adults Aged 70
and Older: White Males
140
Table 6.8 Number of Health Transitions by Race for Adults Aged 70
and Older: White Females
140
Table 6.9 Life Expectancy by Decade by Race: Total, Active, and
Disabled: Black Population
163
Table 6.10 Life Expectancy by Decade by Race: Total, Active, and
Disabled: White Population
164
Table 6.11 Proportion of Remaining Expected Life in Active State by
Race, Sex and Decade
164
Table 6.12 Total Life expectancy by Sex and Race: Total, Active, and
Disabled
166
Table 6.13 Active Life expectancy by Sex and Race: Total, Active,
and Disabled
166
Table 6.14 Disabled Life expectancy by Sex and Race: Total, Active,
and Disabled
166
Table 6.15 Proportion of Remaining Expected Life in Active State by
Race, Sex and Decade
166
Table 6.16 Status Based Health Expectancies at age 70 by Year and
Initial Health Status among Blacks
170
Table 6.17 Status Based Health Expectancies at age 70 by Year and
Initial Health Status among Whites
170
Table 6.18 Status Based Health Expectancies at age 70 by Year and
Initial Health Status among Black Males
171
Table 6.19 Status Based Health Expectancies at age 70 by Year and
Initial Health Status among White Male
171
vi
Table 6.20 Status Based Health Expectancies at age 70 by Year and
Initial Health Status among Black Females
172
Table 6.21 Status Based Health Expectancies at age 70 by Year and
Initial Health Status among White Females
172
Table 7.1 Number of Health Transitions By Education: Transitions
among Those with Less Than 12 Years of Education
187
Table 7.2 Number of Health Transitions By Education: Transitions
among Those with 12 or More Years of Education
187
Table 7.3 Number of Health Transitions By Education: Low
Educated Males
187
Table 7.4 Number of Health Transitions By Education: High
Education Males
188
Table 7.5 Number of Health Transitions By Education: Low
Education Females
188
Table 7.6 Number of Health Transitions By Education: High
Educated Females
188
Table 7.7 Disability Transition Equation Coefficients for Education
and Age
190
Table 7.8 Disability Transition Equation Coefficients for Education,
Sex and Age
190
Table 7.9 Life Expectancy, Active Life Expectancy, Disabled Life
Expectancy, Among Those with <12 Years of Education
217
Table 7.10 Life Expectancy, Active Life Expectancy, Disabled Life
Expectancy, Among Those with 12 or More Years of
Education
217
Table 7.11 Proportion of Remaining Life Expected Active Among
Those with <12 years of Education
217
Table 7.12 Proportion of Remaining Life Expected Active Among
Those with 12 or more Years of Education
217
Table 7.13 Life Expectancy Among Males with <12 Years of
Education
219
vii
Table 7.14 Life Expectancy Among Females with <12 years of
Education
220
Table 7.15 Life Expectancy Among Males with 12 or More Years of
Education
220
Table 7.16 Life Expectancy Among Females with 12 or More Years of
Education
221
Table 7.17 Percent of Life Active Among Males with <12 Years of
Education
221
Table 7.18 Percent of Life Active Among Females with <12 Years of
Education
221
Table 7.19 Percent of Life Active Among Males with 12 or More
Years of Education
221
Table 7.20 Percent of Life Active Among Females with 12 or More
Years of Education
222
Table 7.21 Status Based Active and Disabled Life Expectancies:
Expected Able and Disabled Years by Decade among
Those Initially Able and <12 Years of Education
225
Table 7.22 Status Based Active and Disabled Life Expectancies:
Expected Able and Disabled Years by Decade among
Those Initially Disabled and <12 years of Education
225
Table 7.23 Status Based Active and Disabled Life Expectancies:
Expected Able and Disabled Years by Decade among
Those Initially Able and >=12 Years of Education
225
Table 7.24 Status Based Active and Disabled Life Expectancies:
Expected Able and Disabled Years by Decade among
Those Initially Disabled and >=12 Years of Education
226
Table 7.25 Expected Proportion of Remaining Years Able by Decade
among Those with <12 Years of Education
226
Table 7.26 Expected Proportion of Remaining Years Able by Decade
among Those with >=12 Years of Education
226
viii
LIST OF FIGURES

Figure 2.1 Mortality for the Older Population: 1900-2000 21
Figure 2.2 Mortality for Specified Age Groups: 1980-1996 22
Figure 3.1 Multi-state Table Representation of Transition between
Live States, and to Death
70
Figure 4.1 Probability of Becoming Disabled: LSOA I and LSOA II 88
Figure 4.2 Probability of Recovery from Disability: LSOA I and
LSOA II
89
Figure 4.3 Death Rates for the Non-Disabled and Disabled 90
Figure 4.4 Implied Prevalence of Disability 94
Figure 5.1 Probability of Becoming Disabled for the 1987 and 1997
Cohorts: Males and Females
111
Figure 5.2 Probability of Becoming Disabled for Males and
Females: LSOA I and LSOA II
113
Figure 5.3 Probability of Recovering from Disability for the 1987
and 1997 Cohorts: Males and Females
114
Figure 5.4 Probability of Recovery from Disability for Males and
Females: LSOA I and LSOA II
116
Figure 5.5 Probability of Death Separated by Survey Year for the
1987 and 1997 Cohorts: Males and Females: Probability
of Death among Non-Disabled in the 1987 Survey
118
Figure 5.6 Probability of Death among Non-Disabled in the 1997
Survey
118
Figure 5.7 Probability of Death among Disabled in the 1987 Survey 119
Figure 5.8 Probability of Death among Disabled in the 1997 Survey 119
Figure 5.9 Probability of Death Separated by Sex for Males and
Females from the Non-Disabled and Disabled States: A.
Probability of Death among Non-Disabled Males in Each
Survey
120
ix
Figure 5.10
Probability of Death among Non-Disabled Females in
Each Survey
121
Figure 5.11 Probability of Death among Disabled Males in Each
Survey
121
Figure 5.12 Probability of Death among Disabled Females in Each
Survey
122
Figure 5.13 Implied Prevalence of Disability by Sex 126
Figure 6.1 Probability of Transition to Disability in 24 Months by
Race: Race Differences in Disability Onset the 1987
cohort
143
Figure 6.2 Probability of Transition to Disability in 24 Months by
Race: Race Differences in Disability Onset the 1997
cohort
143
Figure 6.3 Probability of Transition to Disability in 24 Months by
Race: Change in the Probability of Disability Onset for
Blacks
144
Figure 6.4 Probability of Transition to Disability in 24 Months by
Race: Change in the Probability of Disability Onset for
Whites
144
Figure 6.5 Probability of Recovery from Disability in 24 Months by
Race: Race Differences in Disability Recovery for the
1987 cohort
146
Figure 6.6 Probability of Recovery from Disability in 24 Months by
Race: Race Differences in Disability Recovery for the
1997 cohort
146
Figure 6.7 Probability of Recovery from Disability in 24 Months by
Race: Change in the Probability of Disability Recovery
for Blacks
147
Figure 6.8 Probability of Recovery from Disability in 24 Months by
Race: Change in the Probability of Disability Recovery
for Whites
147
Figure 6.9 Probability of Transition to Death from the Non-disabled
State in 24 Months by Race: Differences in the
Probability of Death the 1987 cohort
149
x
Figure 6.10 Probability of Transition to Death from the Non-disabled
State in 24 Months by Race: Differences in the
Probability of Death the 1997 cohort
149
Figure 6.11 Probability of Transition to Death from the Non-disabled
State in 24 Months by Race: Change in the Probability
of the Probability of Death for Blacks
150
Figure 6.12 Probability of Transition to Death from the Non-disabled
State in 24 Months by Race: Change in the Probability
of Death for Whites
150
Figure 6.13 Probability of Transition to Death from the Disabled
State in 24 Months by Race: Differences in the
Probability of Death the 1997 Cohort
151
Figure 6.14 Probability of Transition to Death from the Disabled
State in 24 Months by Race: Differences in the
Probability of Death the 1997 Cohort
151
Figure 6.15 Probability of Transition to Death from the Disabled
State in 24 Months by Race: Change in the Probability
of Death for Blacks
152
Figure 6.16 Probability of Transition to Death from the Disabled
State in 24 Months by Race: Change in the Probability
of Death for Whites
152
Figure 6.17 Probability of Transition to Disability in 24 Months for
Females by Race
153
Figure 6.18 Probability of Transition to Disability in 24 Months for
Males by Race
154
Figure 6.19 Probability of Recovery from Disability in 24 Months for
Females by Race
155
Figure 6.20 Probability of Recovery from Disability in 24 Months for
Males by Race
156
Figure 6.21 Probability of Death from the Non-Disabled State:
Females by Race
158
Figure 6.22 Probability of Death from the Disabled State: Females
by Race
159
xi
Figure 6.23 Probability of Death from the Non-Disabled State:
Males by Race
160
Figure 6.24 Probability of Death from the Disabled State: Males by
Race
161
Figure 6.25 Implied Prevalence of Disability by Race and Year 174
Figure 6.26 Implied Prevalence of Disability by Race and Sex for
the 1987 Cohort
175
Figure 6.27 Implied Prevalence of Disability by Race and Sex for
the 1997 Cohort
175
Figure 6.28 Implied Prevalence of Disability by Race and Sex for
Blacks
176
Figure 6.29 Implied Prevalence of Disability by Race and Sex for
Whites
176
Figure 7.1 Probability of Disability Onset Among those with <12
years of Education
192
Figure 7.2 Probability of Disability onset among those with >=12
years of Education
192
Figure 7.3 Probability of Disability Onset Among Males by
Education Level
193
Figure 7.4 Disability Onset Among Females by Education Level 194
Figure 7.5 Disability Onset Among those with <12 Years of
Education
195
Figure 7.6 Disability Onset Among those with >=12 Years of
Education
196
Figure 7.7 Probability of Recovery from Disability Among those
with <12 Years of Education
198
Figure 7.8 Probability of Recovery from Disability Among Those
with >=12 Years of Education
199
Figure 7.9 Probability of Recovery from Disability Among Males by
Education Level
200
xii
Figure 7.10 Probability of Recovery from Disability Among Females
by Education Level
201
Figure 7.11 Probability of Recovery from Disability Among those
with <12 Years of Education
202
Figure 7.12 Probability of Recovery from Disability Among those
with >=12 Years of Education
203
Figure 7.13 Probability of Mortality Among Lower-Educated
(<12yrs) Previously Active
204
Figure 7.14 Probability of Mortality Among Lower-Educated
(<12yrs) Previously Disabled
205
Figure 7.15 Probability of Mortality Among Higher-Educated
(12+yrs) Previously Active
206
Figure 7.16 Probability of Mortality Among Higher-Educated
(12+yrs) Previously Disabled
207
Figure 7.17 Probability of Mortality Among Males Previously Active 208
Figure 7.18 Probability of Mortality Among Females Previously
Active
209
Figure 7.19 Probability of Mortality Among Males Previously
Disabled
210
Figure 7.20 Probability of Mortality Among Males Previously Active 211
Figure 7.21 Probability of Mortality Among those with <12 years of
education previously active
212
Figure 7.22 Probability of Mortality Among those with >=12 years of
education previously active
213
Figure 7.23 Probability of Mortality Among those with <12 years of
education previously disabled
214
Figure 7.24 Probability of Mortality Among those with >=2 years of
education previously disabled
215
Figure 7.25 Implied Prevalence of Disability by Education and
Survey Year
222
xiii
Figure 7.26 Implied Prevalence by Sex and Education: Males 223
Figure 7.27 Implied Prevalence by Sex and Education: Females 223

xiv
ABSTRACT
The purpose of this dissertation was to examine trends in active life
expectancy between the mid 1980s and the late 1990s, and to examine
changes in disability onset, recovery, and mortality in different subgroups of
the population. Two nationally representative longitudinal datasets of the
population ages 70 and older were used to examine an incidence-based
measure of disability in a multi-state life table program known as IMaCh.
The results suggest that there was a slight increase in active life
expectancy between age-matched individuals in the 1990s relative to their
counterparts in the 1980s, with gains unequal across subgroups. For example,
males in the 1990s showed clear gains in years expected active relative to
their 1987 counterparts. No change over time was observed for females.
Larger gains in active life expectancy were observed for the white population
than the black population over the same time period. There are some signs,
however, that differences in disability onset by race may be gradually
declining, as rates on disability onset appear to be falling slightly in the black
population. There are no differences between blacks and whites in the
probability of recovery from disability, and only slight differences in the
probability of mortality.
Clear differences in disability onset and total and active life expectancy
were observed between those with less than 12 years of education compared
to those with 12 or more years of education. Individuals in the higher
xv
education group had life expectancies about 2 years longer than those of the
same age with less than 12 years of education. There appears to be no
convergence over time between the education groups, as neither the lower nor
higher education group showed any change in active life expectancy over the
time period.
The results suggest that overall there may be a slight compression of
disability in the overall population of the U.S. between 1984 and 2000,
however, most of the improvements are observed for males and the white
population. Continued improvements may be observed if minorities and lower
educated groups can gradually reduce the gap in disability incidence and
mortality.

1
Chapter 1
Introduction
One of the defining features of the twentieth century was an incredible
increase in life expectancy, accompanied by rapid population aging.
Understanding both the present and the future impact of population aging on
society requires consideration not just of life expectancy alone, but also the
interplay of health, disability, and mortality. Active life expectancy is a
measure of remaining years of life without disability, and can measure whether
increases in life expectancy are accompanied by improvements in health, or
less time spent in disability, or extensions of life spent with morbidity.
A variety of theories have been proposed related to future health
trends, with predictions ranging from an expansion of morbidity (Gruenberg,
1977, 1980; Kramer, 1980) to a compression of morbidity (Fries, 1980; Fries &
Crapo, 1981), as well as the dynamic equilibrium theory (Manton, 1982), which
predicts parallel changes in morbidity and mortality.
Changes in active life expectancy over time are greatly affected by
changes in mortality rates. Old-age mortality rates have declined throughout
the twentieth century (Riley, 2003; White 2002), particularly between 1970 and
the present (Bone et al. 1995; Crimmins, 2004). Age-adjusted death rates for
all causes of death among people age 65 and over declined by 12 percent in
the United States (U.S.) between 1981 and 2001 (NCHS Data Warehouse,
2004). However, while it is evident that expected length of life for the elderly
2
has increased over the last 30 years, it is less clear whether these additional
years are spent in good or poor health. This last consideration is of the utmost
importance due to the vast financial and practical considerations involved as
the aged population of the U.S. becomes both larger (in both numbers and in
proportion relative to younger age groups) and potentially less healthy.
In addition to questions of the interrelationship between mortality and
morbidity are issues of inequalities in health and disability, and whether
society is making progress toward compressing long-existing gender, racial,
and socio-economic inequalities in health. For example, according to
numerous studies, blacks are more likely to report disability than their white
counterparts (Clark, 1997; Clark and Maddox, 1992; Kelley-Moore and
Ferraro, 2001; Manton & Gu, 2001; Mendes de Leon et al., 1997). Moreover,
women are more likely to report disability than men (Leveille, Penninx, Melzer,
Izmirlian, & Guralnik, 2000; Oman, Reed, & Ferrara, 1999). Disability
differentials by gender and ethnicity have been of major concern to the
Centers for Disease Control and have led to a major initiative known as
Healthy People 2010. This document lists two health objective goals for the
first decade of the twenty-first century (U.S. Department of Health and Human
Services, 2000):
1) “Increase quality and years of healthy life” (p. 2) and
2) “Eliminate health disparities” (p.5).
3
While there is general agreement on the importance of the goals, progress
over time has been difficult to quantify. Data from various published studies
cannot be compared due to major differences in one or more of the following:
1) the definition of disability, 2) focus on different age-groups, 3) mode of data
collection, 4) level of proxy involvement, and 5) research design and
methodology (Freedman et al., 2004).
Purpose of the Study
The current study builds upon the recent availability of two very similar
longitudinal data sets coupled with a new approach for estimating healthy life
that improves the ability to measure change in ability/disability from
longitudinal surveys. This dissertation provides empirical evidence, for the
U.S. population ages 70 and above, of changes in disability onset, recovery,
and life spent with disability traced over a sixteen year time period. Further,
this dissertation compares active life expectancy trends over time
disaggregated by gender, ethnicity, and educational level.
Active life expectancy falls within a broader category of research in healthy
life expectancy. This line of research can use many definitions of health,
ranging from specific disease to various definitions of disability, as used in the
present study. Disability predicts mortality, institutionalization, depression,
and other key outcomes. It can be partially defined as the inability of any
individual to function in his/her respective role in society, and has a direct
impact on resource needs with policy implications.
4
In the following section is a description of the three major theories that
describe expected trends in active life expectancy. While all three assume
that future increases in total life expectancy are probable, the theories differ on
whether the added years will be spent in quality of life or in stages of disability.
Expansion of Morbidity
The first major theory, developed by both Gruenberg (1977) and
Kramer (1980) independently, proposed the expansion of morbidity
hypothesis, which predicts that levels of morbidity in the population will
increase over time. This theory was based on the assumption that recent
declines in mortality rates among the oldest old are due to decreased fatality
rates from chronic conditions. However, while the theory posits that the oldest
old may survive longer, it warns that these individuals will be more likely to live
extended lives with disability, dementia, and co-morbidity (Gruenberg, 1977,
1980; Kramer, 1980).
Compression of Morbidity
Alternatively, in a separate but opposing theory, Fries and Crapo
proposed that there is a natural limit to life expectancy that prevents significant
future gains in life expectancy. Thus, according to the theory, the
postponement of the onset of chronic diseases to more advanced ages will
only lead to a compression of time with morbidity (Fries, 1980; Fries & Crapo,
1981).
5
Dynamic Equilibrium
Finally, Manton (1982) introduced the dynamic equilibrium theory,
which assumed that among the oldest old, a decline in the pace of chronic
disease progression will lead to a slow increased prevalence of light to
moderate disability while mortality falls. Essentially, Manton proposes that
while life expectancy may increase, and the onset of severe disability may be
delayed, the oldest old will not escape from the problems associated with
disability.
In light of these differing theoretical arguments, researchers have
developed global health indicators that take into account both quantity and
quality of life using health expectancy to examine the empirical evidence that
might support or refute the above theories.
Description of the Dissertation Research
This study incorporates both the Longitudinal Study of Aging I and II (LSOA
I and LSOA II). The overarching goal is to determine whether the incidence
and recovery from disability and mortality differ between two cohorts of older
adults.
This dissertation focuses on the underlying processes that determine
active life expectancy, including changes in disability onset and recovery, and
changes in age-specific mortality among the disabled and non-disabled. In
this study, individuals were classified as disabled when they were unable to
perform any one or more of 10 indicators of disability based on five Activities of
6
Daily Living (ADLs): bathing, dressing, eating, getting in and out of bed, and
toileting (Katz & Akpom, 1976) as well as the Instrumental Activities of Daily
Living (IADLs): preparing meals, managing money, performing light
housework, shopping, and using the telephone (Lawton & Brody, 1969). The
focus of this study is transitions to and from disability, as well as to mortality.
Research Questions
This study will focus on the following five research questions:
1) What are the differences in age-specific rates of onset of disability
and recovery between the 1980s and 1990s?
2) What changes in rates of mortality among disabled and
non-disabled individuals occurred between the 1980s and 1990s?
3) What were the estimated changes in life expectancy, and in life
expectancy with and without disability, from the 1980s to the 1990s?
4) How do changes in active life expectancy vary by gender, race, and
education level?
5) Which groups experienced a compression of morbidity, an
expansion of morbidity, or parallel changes in life expectancy and
time spent disabled?

As indicated, data for the study come from two distinct longitudinal data
sets and measure change in active life expectancy across a decade: one
representative of the elderly population of the U.S. in the late 1980s, and the
7
other representative of the elderly population of the U.S. in the late 1990s. The
two cohorts are compared based on age-specific estimates for the risk of
decline and/or recovery from disability as well as the corresponding risk of
death for the disabled and non-disabled.
Contribution to the Literature
This dissertation represents a unique approach to examining trends in
health expectancy research, as it may be the first attempt to study two
longitudinal studies of different time periods. The second longitudinal survey
(LSOA II) was specifically designed to replicate the first study (LSOA I) for the
purpose of monitoring changes in time. To date, no studies have been
published comparing these two longitudinal studies from different decades.
Furthermore, trends in disability and active life expectancy have not
provided strong evidence either supporting or refuting the compression of
morbidity hypothesis. Stronger evidence requires extensive longitudinal data
as well as extensive analysis of covariates related to disability and mortality in
order to investigate population variability in observed trends. In
heterogeneous societies, there may be dramatically different experiences for
different subgroups of the population.
8
Organization of the Dissertation
This dissertation is organized into seven chapters that investigate both
the compression of morbidity hypothesis for the total population, as well as
whether changes in morbidity relative to mortality are occurring differentially by
gender, race, and socioeconomic status. Chapter II is a review of the relevant
literature on disability; measurement issues; trends in the prevalence and
incidence of disability in the older population; trends in old age mortality; and
recent trends in active life expectancy. Chapter III includes a detailed
description of the data, The Longitudinal Study of Aging I and II, as well as a
description of the process used to compensate for missing data. Chapter III
also describes the methodological approach using an executable algorithm
based on matrix algebra of health transition probabilities and Markov chains.
The fourth chapter begins the description of the results. Chapter IV
describes overall changes in both life expectancy and active life expectancy
between 1984 and 2000. The chapter addresses the compression of
morbidity hypothesis.
The next three chapters focus more on the issue of whether or not there
were any changes in the degree of inequality for active life expectancy among
subgroups of the population. In particular, Chapter V focuses on differences
between men and women. Chapter VI investigates inequality by race at two
points in time and compares changes in active life expectancy, and the
processes underlying it, by race and sex. Chapter VII investigates inequality
9
at two points in time by education level and compares changes in active life
expectancy, and the processes underlying it, by education and sex.
Finally, Chapter VIII contains a discussion and attempts to coalesce
conclusions based on the results in the previous chapters. This final chapter
relates these results back to the previously published literature, and
subsequently explores the implications of the results.

10
CHAPTER II: LITERATURE REVIEW
This chapter contains a review of the relevant literature in the fields of
disability, issues related to the measurement of disability, a description of
recent trends in disability, trends in mortality, and a description of active life
expectancy and recent trends. The analysis that follows this review of the
literature focuses on changes in active life expectancy in two cohorts, and thus
this literature review focuses on published studies that track changes in active
life expectancy, or its components: disability and mortality.
Disability
Disability is the inability (or limited ability) to perform expected social
roles and tasks within a given environment (Freedman, 2006; Nagi, 1991;
Putnam, 2002; Verbrugge, 1989). Disability can more simply be defined in
terms of the inability to live independently and provide self-care (Crimmins,
2003). Verbrugge and Jette (1994) clarified the ‘disability process’ by placing
disability at the end of a process of health change beginning with diseases and
impairments proceeding through impairment, functional limitation and finally to
disability. This model has greatly influenced the design of disability research,
and variations of this simple model have been expanded by Barberger-Gateau
and colleagues (2002); Femia, Zarit, and Johansson

(2001); Patrick (1997);
and Peek, Patel, and Ottenbacher (2005), among others, as well as by one of
the original authors (Jette, 1999).

11
Many researchers view disability primarily as a gap between the
demands of the environment and the capabilities of the individual (Brandt &
Pope, 1997; Hahn, 1994; Lawton & Nahemow, 1973; Nagi, 1965; Pope and
Tarlov, 1991; Putnam, 2002; Verbrugge & Jette, 1994; Murray et al., 2002).
This view emphasizes that disability is not an inherent characteristic of an
individual, and can vary by environment. Based on this concept, disability can
be measured as the inability to function independently in a given environment,
which in the context of the older population often refers to their home.
In studies of the older population, disability often refers to the inability to
perform activities needed for independent living and personal care, known
professionally as Activities of Daily Living (ADLs) and Instrumental Activities of
Daily Living (IADLs) (Crimmins, 2003; Jette, 1994; Spector & Fleishman,
1998; Weiner, Hanley, Clark, & Van Nostrand, 1990). These measures,
defined in the following paragraphs, generally encompass daily activities
necessary for self-care and independent living. A variety of other measures
have been used that emphasize physical strength and agility including those
developed by Nagi (1976) and Rosow-Breslau &

Guttman (1966), also
described in the following paragraphs.
Activities of Daily Living (ADLs). Activities of Daily Living (ADLs),
developed by Katz and Akpom (1976), assess the ability to do basic physical
activities required for self-care without help (Branch, Katz, Kniepmann, &
12
Papsidero, 1984). The ADLs are the daily basic tasks of: bathing, dressing,
eating, using a toilet, and transferring from a bed or chair.
Instrumental Activities of Daily Living (IADLs). Another common measure,
known as Instrumental Activities of Daily Living (IADLs), refers to higher level
functional

abilities needed to maintain an independent life, including preparing
meals, managing money, shopping, using a telephone, and doing light
housework (Lawton & Brody, 1969).
The ADL and IADL measures are used in both research and in determining
eligibility for government-sponsored programs including: the Medicare Part A
Skilled Nursing Facility Prospective Payment System (CMS, 2005), Medicaid
services (HHS, 2000), and Adult Day Health Care (HHS, 2006). These
indicators are also used in determining claims for long-term care insurance
and in assessing quality of care provided to the elderly (Mukamel et al., 2006).
The measures can be used as individual scales, or they can be combined in
one scale to make a psychometrically valid measure of disability (Spector &
Fleishman, 1998).
Other measures of disability. Some researchers have quantified
disability with a focus on mobility and strength. For example, Nagi (1976)
included pulling or pushing large objects; stooping,

crouching, or kneeling;
reaching or extending arms above shoulder

level; and writing or handling small
objects. Other researchers used other activities in the disability gauge.
Rosow-Breslau designed a Functional Health Index to measure the ability

to
13
perform mobility and strength tasks such as walking up and down stairs,
walking half a mile,

and doing heavy work around the house (Rosow, Breslau,
&

Guttman, 1966). Verbrugge, Merrill, and Liu (1999) rated disability using a
single self-rated question from the Center for Disease Control’s Behavioral
Risk Factor Surveillance System Survey: “During the past 30 days, for about
how many days did your poor physical or mental health keep you from doing
your usual activities, such as self-care, work, or recreation?”
Most of the literature discussed above uses inability to perform a task
as a measure of disability. Another approach in measuring disability is the
concept of dependency, defined as needing or having help to do an activity
(Gill & Kurland, 2003). Dependency approaches, however, have been
criticized because they are contingent on the presence of supportive
mechanisms, and this support may buffer and conceal disability. Furthermore,
the availability of support from others may vary in ways not directly associated
with the level of disability-need.
Performance tests are another way to measure individuals’ ability. This
has been done in several studies, including: the InCHIANTI study to measure
walking ability in a variety of conditions (Shumway-Cook et al., 2007); the
MacArthur Study of Successful Aging to test balance and ability to stand from
a chair (Seeman et al., 1996); the Chicago Health and Aging Study with
tandem standing balance tests (Mendes de Leon et al., 2005); and the 2006
collection of the Health and Retirement Survey. The majority of this data are
14
very recent, and limited analyses have been performed based on these
measures to date. On the other hand, self-reported measures have been
reported in many surveys over the last few decades, and those remain the
most common source for information on disability in the analysis of survey
data.
Measurement of disability has important implications when interpreting
trends over time. Even modest variations in the way disability is measured
may lead to significant differences in prevalence and incidence rates
(Freedman et al., 2004; Jette, 1994; Weiner et al., 1990). For example, one
may achieve varying results by asking about the duration of disability
(short-term versus three or more months) versus the extent of disability or the
effect of disability. For example, a person who is unable to prepare his/her
own food but is living with a spouse or other full-time caregiver willing to take
on the function may not be assigned as high a disability level as someone in a
similar situation sans immediate caregivers (Freedman et al., 2004; Weiner et
al., 1990).
Trends in disability
Disability Prevalence. Disability prevalence refers to the proportion of
the population that is currently affected by a limitation in their capacity for
independent functioning. Prevalence is often used to describe the health of
the population and the number of individuals who need some form of treatment
15
or assistance. By contrast, disability incidence refers to those who recently
became disabled and measures the impact of more recent events.
Data on trends in disability prevalence in the U.S. can be found going as far
back as birth cohorts between 1825 and 1844. Fogel used data from male
Union Civil War veterans and physician records to calculate a 0.6% annual
rate of decline in chronic morbidity over 75 years (Fogel, 1994).
Disability rates may have not declined monotonically throughout the
last century, however. In examining data from the National Health Interview
Survey (NHIS) for the non-institutionalized U.S. population from 1957 to 1981,
Verbrugge (1984) found an increase in the proportion of people ages 65 and
over reporting an inability to engage in major life activities due to health
problems in the 1970s. She also reported increases in the number of days in
which activities were restricted. Colvez and Blanchet (1981) as well as
Crimmins, Saito, and Ingegneri (1989) also reported increased disability
prevalence in the 1970s.
Crimmins, Saito, and Ingegneri (1997) reported decreases in the
overall population disabled in the 1980s. Crimmins, Saito, and Reynolds,
examining the older population specifically, reported fluctuations in the
prevalence of disability in the 1984-1990 Longitudinal Study of Aging, with an
increase in the prevalence between 1984 and 1986, followed by a decline
1988 through 1990 (1997a). The same study also found fluctuations in the
prevalence of the percent with personal care disability (similar to the ADL
16
measure), as well as the percent with routine need disability (similar to the
IADL measure) between 1982 and 1993 using annual data from the National
Health Interview Survey (1997a). Freedman and Soldo (1994) summarized
the findings from longitudinal studies from the U.S. and three foreign countries.
They concluded that there is evidence of modest declines in the incidence and
prevalence of mild disability over this time period. Later studies confirmed
their findings (Freedman et al., 2002).
Manton and colleagues found declines in the prevalence of chronic
disability among the population ages 65 and over population based on the
1982 to 1989 National Long Term Care Survey (NLTCS) (Manton, Corder, &
Stallard, 1993). Spillman (2004) reanalyzed the same dataset using data from
different years and found no decline in ADL disability between 1984 and 1994,
suggesting, as Crimmins had earlier (1997b), that there were no clear patterns
in disability for those age 75 and above through this time period.
Manton and colleagues (1997) and Manton and Gu (2001) used the
National Long Term Care Survey (NLTCS) to investigate trends during the
time periods of 1982-1994 and 1982-1999. Manton and Gu (2001) showed
declining prevalence of disability between 1982 and 1999, with greater annual
decline rates in the 1990s than the 1980s. Spillman (2004) showed that the
declines were primarily found in IADLs such as financial management and
shopping.
17
Further evidence came from Freedman and Martin, who found a
decline in the prevalence of disability between 1984 to 1993 at a rate of about
0.9% to 2.3% per year for each of four functions, with the fastest rate of decline
above age 85 (1998). Similarly, Freedman and Martin (1998) also found a
downward trend in functioning problems between the years 1984 and 1993 as
seen in the increasing numbers of individuals able to lift and carry a 10-pound
object, climb a flight of stairs, or walk a quarter of a mile. Slightly more recent
data, taken from the 1993 to 1996 Medicare Current Beneficiary Survey,
revealed continued evidence of a decline in disability for persons aged 65 and
older (Waidmann & Liu, 2000).
Wolf and colleagues (2007), studying the New Haven Established
Populations for Epidemiologic Studies

of the Elderly, found no evidence of
trends in the prevalence of ADL disabilities between 1982 and 1991 (ages 75
and above) or similarly between 1982 and 1994 (ages 77 and above) in the
non-institutionalized population. However, using alternative measures of
disability focused on walking up and down stairs and walking half a mile, Wolf
and colleagues were able to identify a downward trend in disability among men
ages 75 and older between 1982-1994 (2007). Upon changing the definition
to Nagi measures of disability (pushing or pulling, kneeling or crouching,
reaching up, and grasping small objects), they found a significant increase in
the prevalence of disability over this time.
18
Disability incidence. Although less is known about trends in disability
incidence, there is evidence of possible increases in the incidence of activity
limitation in the early 1970s for people in the age range of 45-74, although
there is no clear pattern for individuals age 75 and above. (Crimmins &
Ingegneri, 1993). By contrast, reductions in disability onset were observed in
the 1980s (Crimmins et al., 1997b; Manton, Corder, & Stallard, 1993) and
1990s (Cai & Lubitz, 2007; Wolf et al., 2007). The early reductions in disability
were associated with a movement towards more severe disability for those
considered disabled (Manton, Corder, & Stallard, 1993). These studies also
provide information on change in age-specific recovery rates. Crimmins and
colleagues’ study found some improvement in recovery rates from 1984-1986
to 1988-1990 (1997b); while Manton and colleagues’ study (1997) reported
that all recovery rates were lower in the 1984-1989 period than the 1982-1984
period. Wolf and colleagues (2007) also found declining recovery between
1982 and 1994.
An accumulation of multiple years of longitudinal data has allowed for
analyses that can explore how prevalence, incidence, and recovery interact to
create the disability trends observed. There are several possible reasons why
research might report declines in the prevalence of disability, such as: 1)
declines in the rate of onset, 2) rising rates of recovery, 3) falling disability
prevalence among new elders added to the population each year, and/or 4)
increased death rates among disabled people, or decreased mortality among
19
non-disabled persons (with stable death rates for disabled persons) (Crimmins
et al., 1994; Wolf et al., 2007). Furthermore, if recovery rates rise faster than
onset rates (even if onset rates are increasing), diverging trends in incidence
and prevalence may be the result.
Wolf and colleagues (2007) reasoned that declines in the onset of
disability, which should lead to declines in prevalence, have been offset by
parallel declines in recovery from disability. The combination of these two
effects is thought to lead to the flat trends observed in the New Haven data for
prevalence over this time period.
20
Mortality Rates
Mortality declined rapidly through much of the twentieth century,
although the causes of death responsible for mortality declines pre-1960 differ
from the more recent causes. Declines in the first half of the twentieth century
were driven primarily by a reduction in death rates from infectious diseases.
Since the mid-1960s, mortality decline is primarily due to a reduction in the
fatality of chronic diseases, especially heart disease (Crimmins, 1981;
Crimmins & Ingegneri, 1993). As the primary cause of mortality shifted from
infectious to chronic conditions, the associations of mortality and disability
trends may have changed as well (Crimmins & Ingegneri, 1993).
In the most recent decades, older people have become more aware of
their health needs and better able to manage their chronic conditions, and the
likelihood of surviving to very old ages has increased markedly (Vaupel, 1997;
Wilmoth et al., 2000). Further, mortality patterns, which generally follow a
Gompertz curve with death rates increasing exponentially with age (Finch and
Pike, 1996), are lower than would be predicted at the oldest ages (Thatcher et
al., 1997; Vaupel et al., 1998). As age-specific mortality declined over time,
the selectivity of survivors also declined. Whereas only 20% of people in 1930
survived to age 80, more than 50% of the population in 1997 is expected to
survive at least until age 80 (Crimmins, 2001).
As seen in Figure 2.1, death rates have declined in the older population
throughout the last century (NCHS Mortality Statistics, 2006). The decline is
21
most dramatic for those aged 85 and above, making this age group the fastest
growing population, as people survive longer and delay death to more
advanced ages. There is evidence that mortality declines in the 1980s
occurred at a slower pace than the trend in the previous 15 years (Crimmins,
Saito, & Ingegneri, 1997; Crimmins, 2004; White, 2002). Figure 2.2 shows
mortality trends in more recent years by sex. As seen in the figure, mortality
changes have been greater for males than females.
Figure 2.1. Mortality for the Older Population: 1900-2000

0
5,000
10,000
15,000
20,000
25,000
30,000
1900 1920 1940 1960 1980 2000
Rate per 100,000 po
65-74
75-84
85+

Source: NCHS Mortality Statistics:
http://www.cdc.gov/nchs/datawh/statab/unpubd/mortabs/hist290.htm
22
Figure 2.2. Mortality for Specified Age Groups: 1980-1996

Source: NCHS Lifetables 1984-1996:
http://www.cdc.gov/nchs/products/pubs/pubd/lftbls/lftbls.htm

For females, gains in life expectancy were much smaller between 1980
and 1990 than in the previous decade. Female life expectancy at age 70
increased by 1.2 years between 1970 and 1980, by 0.5 years between 1980
and 1990, and by 0.2 years between 1990 and 2000 (see table 2.1). For
males, the change was more stable, with increases of 0.7 years between 1970
and 1980 as well as between 1980 and 1990, and an increase of 1.0 year
between 1990 and 2000.
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
1980
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
Probability of Death (Qx)
Males 70-74
Females 70-74
Males 80-84
Females 80-84
23
Table 2.1. Life Expectancy at Age 70 Across Recent Decades
1970 198019902000
Males 10.6 11.3 12.0 13.0
Females 13.6 14.8 15.3 15.5
Source: NCHS Life Tables
http://www.cdc.gov/nchs/products/pubs/pubd/lftbls/lftbls.htm

Health related to mortality. Recently, studies have begun to examine
disability and health immediately preceding mortality (e.g., Hayward et al.,
1998). It appears that most deaths in men at age 70 occur among men without
disability (Hayward et al., 1998). For women at age 70, the proportion is about
50%. The proportion of deaths that occur to non-disabled individuals declines
with age, and yet, interestingly, a large portion of death in the population
occurs among non-disabled individuals. There is limited data available on
death rates by disability status that is also representative of the population, but
the growing number of longitudinal studies are opening up the possibility for
such studies.
Active Life Expectancy
Active Life Expectancy is a summary measure of population health,
integrating morbidity and mortality into a single quantifiable measure of
remaining years of life, with or without a disabling condition. Thus it addresses
both life expectancy as well as the quality of the remaining years.
Non-disabled years are deemed “active,” and disabled years (sometimes
referred to as inactive) can be further classified according to the level of
severity. Active Life Expectancy tends to be used synonymously with the term
24
Disability-Free Life Expectancy (DLFE), both of which are subsets of the
broader term healthy life expectancy. Active life expectancy is usually life
without some form of physical limitation or disability, while healthy life
expectancy can be life without a disease or impairment, such as diabetes
(Jagger et al., 2003), cognitive loss (Dubois & Herbert, 2006; Suthers et al.,
2003) or a risk factor such as obesity or hypertension.
Health expectancy was originally envisioned as a general public health
measure, not focused on any specific age group. The concept took on a more
gerontological perspective in 1983, when a study that was focused solely on
the elderly population defined disability as being unable to perform at least one
of the activities of daily living rather than relying on the earlier “bed disability”
definition (Katz et al., 1983). Wilkins and Adams (1983), studying the
Canadian population, used similar measures on morbidity and mortality data
to estimate active life expectancy for the older population. Their analysis was
based on double-decrement life tables with two absorbing states: limited
function and death.
Today, Active Life Expectancy (ALE) is used as a descriptor of
population health by government statistics in the U.S. (Wagener et al., 2001)
and Europe (Bone et al., 1995), as well as by journalists and public health
researchers who want to report information that is precise and succinct
(Verbrugge, 1997). It likely reflects health care needs and the need for social
or institutional support, and can also be used to provide a sense of the
25
magnitude of the impact of public health programs or other health
interventions (Wagener et al., 2001). The results can be compared over time
or across populations, assuming similar measurement of disability, as one way
to assess trends in both disability and mortality. The results appear to
represent the number of years an average person could expect to live in good
health, with the remaining years classified as inactive. However, in truth, a
person’s inactive or unhealthy years are not necessarily always their last years
of life; they could be scattered throughout their remaining lifetime.
Researchers have used the active life expectancy measure to study
changes in a variety of measures, including: the prevalence of disability

(Crimmins et al., 1989, 1997b; Manton, 1988); incidence rates of disability
(Crimmins et al., 1994); trends in disability over time (Cai & Lubitz, 2007);
across countries (Lopez et al., 2006; Mathers et al., 2001); disparities across
race groups (Crimmins & Saito, 2001; Hayward & Heron, 1999); disparities
across education levels (Crimmins & Saito, 2001; Melzer et al., 2001); specific
diseases and conditions (Jagger et al., 2003; Reynolds et al., 2005); causes of
death (Hayward et al., 1998); or the effect of changing threshold (e.g., level or
severity) defining disability (Lynch, Brown, & Harmsen, 2003).
Active life expectancy can be calculated by applying life table methods
to period-specific data in the synthetic cohort mode, with the goal of estimating
the average number of remaining years of life free from ADL disability at
specific ages (Land et al., 1994). Calculations have been derived based on
26
prevalence-rate methods (Crimmins, Saito, & Ingegneri, 1989; Manton &
Stallard, 1991), and incidence-rate methods (Crimmins, Hayward, & Saito,
1994; Land, Guralnik, & Blazer, 1994; Rogers, Rogers, & Belanger, 1989).
Incidence based methods, using longitudinal data, are often based on
increment-decrement life tables to calculate a multi-state life table. The
advantage of this type of model is that it can demonstrate substantial recovery
rates from functional limitation possible at any age (Crimmins, Saito, &
Hayward, 1993; Rogers et al, 1989, Rogers, Rogers & Belanger, 1992).
As with life expectancy, active life expectancy is not necessarily the true
experience of an actual population, but rather the number of years that a
hypothetical population would live in active and inactive states if it experienced
continued mortality and morbidity rates as observed in the data used for the
analysis. Further, there is no single number that is “the” active life expectancy
of a population at a given moment because the value depends on the definition
of active and the method used to calculate the estimate (Crimmins & Hayward,
1996).
Measurement of Active
While the definition of “active” can vary from study to study, it generally
relates to the ability to care for one’s own needs independently. This can be
based on scales, such as the Activities of Daily Living (e.g. can you bathe
yourself, dress yourself, eat without assistance, transferring from bed to chair)
(e.g., Jagger et al., 2003; Land, Guralnik, & Blazer, 1994; Reyes-Beaman et
27
al., 2005; Rogers, 1989); or the combination of ADL and IADL measures
(Instrumental Activities of Daily Living, including activities such as preparing
meals, shopping, managing money, using a telephone, and other related
activities) (Spector & Fleishman, 1998). Many researchers combine ADLs and
IADLs (Crimmins, Hayward, & Saito, 1996; Liu et al., 1995; Spector &
Fleishman, 1998); and some researchers add other functional limitations to
the definition (Nuseelder et al., 2005; Kaneda et al., 2005) or base it on other
ability criteria available in large surveys (Crimmins & Saito, 2001; Geronimus
et al., 2001; Hayward & Heron, 1999).
Methods used in the Measure of Active Life Expectancy
Cross-sectional methods. The first technique used to calculate healthy
life expectancy was the prevalence-based life table method, also known as
Sullivan’s Method (Sullivan, 1971). The method combines age-specific
disability prevalence with life table mortality estimates to create a snapshot of
the combined effect of disability and mortality at one point in time. It can be
interpreted as the current health composition of the population adjusted for
mortality levels (Crimmins, Saito, & Hayward, 1993). The results indicate
nothing about expected life cycle events of individuals exposed to current
mortality and morbidity conditions, as these can change over time.
This method is very popular around the world since most data sources
are cross-sectional and this method only requires data on prevalence. There
are multiple advantages of Sullivan's method: 1) it has a straightforward
28
protocol that doesn’t require special software, 2) it is based on cross-sectional
disability data that are less costly and more available than panel surveys, 3)
only moderate sample size in the disability survey is required to produce
reliable age-specific prevalence estimates, 4) an abridged life table is
sufficient for mortality since the method is not sensitive to the size of the
groups, and 5) prevalence-based methods are less influenced by survey
design and analysis strategies than longitudinal designs (Saito et al., 1991;
Deeg et al., 2003).
There are also some limitations and disadvantages of this type of
approach. For example: 1) the assumptions of the method constrain the
validity of the results to persons who are exposed to current mortality and
morbidity, 2) during times of rapidly declining mortality, health expectancy
estimates do not change much relative to the improvements in life expectancy
(e.g., Bebbington, 1991; Crimmins et al., 1997a). As an indicator of change
over time, it may lack sensitivity because mortality and disability can change
independently of each other, and also because prevalence rates, on which
disability is based, do not change as quickly as disability incidence rates.
Further, Sullivan’s estimates are not strongly affected by changes in mortality
that are not accompanied by simultaneous changes in morbidity (Crimmins et
al, 1993). Finally, Sullivan assumes that mortality rates are the same for able
and disabled individuals, and this assumption is not likely accurate.
29
Sullivan’s method has been popular around the world and has been
used in several major studies. For example, Robine and colleagues (1997)
used Sullivan’s method to calculate disability-free life expectancy (DFLE) for
30 countries, as well as to examine trends in DFLE in France between 1981
and 1991. Mathers and colleagues (2001) reported DFLE in 191 countries
using this method. Estimates of DFLE exist for many countries because of the
very minimal data requirements of this method.
The Sullivan method does not allow interpretation of changing mortality
and morbidity transitions, as this method assumes all transition rates input to
the process stay constant over some period of time (Crimmins, Saito, and
Hayward, 1993). In times of declining mortality, Sullivan’s method is likely to
show little change in active life expectancy, as declining mortality has only a
small effect on the population structure. Declines in mortality do have a
significant effect for individual life cycles (Preston, 1982), and thus Sullivan’s
method may not be appropriate for measuring the impact of declining mortality
on active life expectancy over an extended period of time. An alternative
method, using longitudinal multi-state analysis, can better account for changes
in active life expectancy that are associated with mortality decline over time.
Longitudinal methods. Analysis of ALE based on longitudinal data
from a sample of the population is of increasing importance as researchers
delve into the causal mechanisms driving change in population health.
Sometimes referred to as “Dynamic Methodology,” the analysis techniques
30
are significantly more complicated to model and interpret, but have more
potential to inform us about the interactive processes and the dynamics of
onset, recovery, and death among various subsets of the population. Health
change is based on incidence, rather than prevalence used in cross-sectional
studies, and thus is based on recent events affecting health. Many models
rely on a multi-state (increment-decrement) life table method that uses
age-specific transition rates that reflect bi-directional transitions between
active life, as well as to death from either active or inactive state.
The major disadvantage of this approach is that it requires extensive
longitudinal data, with a substantial sample size in order to model the onset of
disability and transitions to death. While different interval lengths between
observations are possible, the results could be biased if the time interval is too
long to accurately capture recent onset rates, or too short to capture the
mortality associated with disability. The model is also very sensitive to the
parameters used in the estimation process.
Several methods have been developed to analyze longitudinal data,
including a Hazard model approach (e.g., Crimmins, Hayward, & Saito, 1994);
an algorithm using Markov Chains called IMaCh (Lievre et al., 2003); a Markov
transition model developed using the SAS software (Ling et al., 2005); and a
Bayesian Monte-Carlo estimation technique (Lynch, 2007).
31
Trends in Active Life Expectancy
A study examining the health of Americans between 1964 and 1974
found overall increases in life expectancy; however, most of the increase was
in years spent disabled (McKinlay et al., 1983; Robine et al., 2003). Another
study, with a similar time frame, 1966-1976, also found that life expectancy
increased over this time by 2.2 years; however, only 0.6 years of the 2.2 years
were without disability (Colvez, 1980).
A study in Canada using morbidity and mortality data from 1950 to 1978
(Wilkins & Adams, 1983), as well as another study in England and Wales using
data from 1976 to 1985 (Bebbington, 1991), found similar results, with the bulk
of life expectancy increases attributed to additional years in disability. Further
studies in the U.S. concluded that there was a trend toward longer life
accompanied by more time with disability in the U.S. Thus, between 1970 and
1980, while life expectancy in the U.S. increased by approximately 3 years,
most of the increase was in years of disability (Crimmins et al., 1989; Crimmins
et al., 1992; Robine et al., 2003).
Crimmins examined trends across 1965, 1970, and 1980 (Crimmins et
al., 1989), as well as through 1990 (Crimmins et al., 1997a; Crimmins & Saito,
2001), and found that when examining severe disability, or bed disability, the
increased life expectancy observed between 1970 and 1980 in the U.S. was
additional years without disability (1997a). When using a less severe
definition of limitation of activity, most of the increase in life expectancy could
32
be attributed to years with activity limitation. Crimmins and Saito (2001) found
that differences between groups by education level and race were growing
larger over the 1970-1990 time period, and thus changes in active life
expectancy depend highly on the environment in which an individual lives.
The next section addresses this issue in more detail.
Continuing with the time period of the 1980s, similar observations of
increases in active life expectancy were observed in France and England. For
example, between 1981 and 1991, disability-free life expectancy increased
significantly

by 3.0 years at birth for males and 2.6 years at birth for females.
Overall life expectancy at birth increased by 2.5 years, and thus the proportion
of years expected without disability increased (Robine, Mormiche, & Sermet,
1998). In England, the 1980s was characterized by a shift towards more years
lived in less severe disability (Bone, Bebbington, Jagger et al., 1995).
Differences in Active Life Expectancy by Sex, Race, and Socioeconomic
Status
Researchers have used active life expectancy to summarize
differentials in disability and mortality for different subgroups of the population.
For example, women generally have longer life expectancy than men, and
when disability is defined in a yes or no fashion, studies show that women tend
to have both longer active and longer inactive lives than men (Barendregt,
2003). Most studies show that women live a greater proportion of their longer
lives with disability. In studies where the severity of disability is taken into
33
account, the proportion of remaining life expected in disability is equal for both
sexes (Barendregt, 2003). This could result from men spending less time with
disability, but having more severe disability.
Health expectancy indicators are comparable across groups with
different age structures and are useful for comparing socioeconomic
differences across groups, or over time. Socioeconomic (SES) differences in
active life provide a summary indicator of the effects of inequality on health or
disability. Most studies find that people who have lower socioeconomic
resources have shorter life expectancy and live a greater proportion of their
lives in a disabled state (Cambois et al., 2001; Crimmins & Cambois, 2003;
Crimmins & Saito, 2001; Melzer et al., 2000; Minicuci et al, 2005). Differences
by social class in expected active life are usually larger than differences in total
life expectancy. Reducing socioeconomic differences in active life has
become a public health goal in many countries. Trends over recent decades
suggest that differences between low and high SES groups have been
growing, led by increased morbidity differentials (Crimmins et al., 2001).
Although many different measures of socioeconomic status are used in the
literature, education level is the most common measure because it does not
generally change after adulthood, and it is not affected by poor health in later
life, as income or occupation might be.
In the United States, race and ethnic differences in active life
expectancy have become an important area of inquiry; these are, of course,
34
highly related to socioeconomic differences. Blacks in the U.S. have a shorter
active life expectancy and a longer disabled life expectancy (Crimmins &
Saito, 2001; Geronimus et al., 2001; Guralnik et al., 1993; Hayward & Heron,
1999). Race and gender also appear to interact: a number of studies conclude
that black women live a notably greater percentage of their lives with disability
than black men (Mendes de Leon et al., 2005; Geronimus et al, 2001), who
have the shortest life expectancy and largest numbers of life years lost.
Conclusions
This chapter summarized current knowledge of disability, active life
expectancy, and trends in active life expectancy and its component disability
and mortality measures. Clear statements on trends in disability or active life
expectancy are difficult to make because of the variety of measures used, and
the limited availability of nationally representative data from historical time
periods. Among the disability measures popularly used are measures of
social disability (ADL, IADL) or physical functioning (Nagi, 1976; Rosow,
Breslau, & Guttman, 1966), each of which can measure differing levels of
difficulty or impairment. Many disability measures directly measure disability
concepts (functioning within the home or inability to do a specific task), while
others take a more indirect approach (dependency or care needs).
Research in the field of disability prevalence has revealed some general
trends, including declines in disability early in the twentieth century, an
increase in disability over the 1970s, flat trends through the early 1980s, and a
35
decrease through the late 1980s and early 1990s. These estimates are based
on measures of disability prevalence, and while they are good markers of
population needs, they do not fully inform us of the processes affecting trends.
Namely, these include disability onset, recovery, and mortality among both the
disabled and the non-disabled. More recent studies have begun to look more
deeply into issues of disability onset, recovery, and mortality; however, the
longitudinal data needed for these analyses has only recently been collected.
We cannot directly infer trends in onset of disability based on estimates of
prevalence, as we have no information on duration of disability and mortality in
cross-sectional studies.
We know that mortality in general has been declining at all ages throughout
the last century, although more so for men than women in recent years. This
decline has been the driving force for increased life expectancy, and the
factors causing mortality decline contribute to changes in active life
expectancy. The actual effect of mortality decline on active life expectancy
depends largely on whether changes in mortality affect the disabled differently
from the non-disabled and how disability onset and recovery are affected.
There are many limitations to our knowledge of trends in active life
expectancy. For example, few studies have been done using representative
longitudinal data. Further, while mortality and morbidity change over time, it is
not known whether the changes are happening such that one changes more
dramatically than the other. Finally, no analysis has been done to compare
36
active life expectancy from different longitudinal studies covering different time
periods.
This study offers an opportunity to look at true longitudinal trends in the
onset and recovery of disability in two distinct cohorts of older individuals. It
will give new insight into changes in the dynamics of disability over time, and
offer a new perspective in trends in disability.
The chapters that follow build on the literature that was described in this
chapter and also include an empirical analysis of recent trends in disability.
Each chapter focuses on trends in disability among special sub-groups of the
population, examining overall trends first, and subsequently disaggregating
data by sex, and finally analyzing data by race and by level of education.
37
CHAPTER III: Data Description and Methods
Survey and data description
Data for this dissertation are from the first and second Longitudinal
Studies of Aging (LSOA I and LSOA II). The LSOA I and II were designed and
collected through a collaborative effort of the National Center for Health
Statistics (NCHS) and the National Institute on Aging (NIA) (NCHS website,
2006). The first data collection was originally known as the Supplement on
Aging, conducted in conjunction with the 1984 National Health Interview
Survey (NHIS). The Supplement on Aging asked additional questions to
persons 70 and older in order to provide baseline data about social and
environmental factors associated with the health of the elderly. The LSOA II
was also a supplement to the NHIS survey, focused on the same age group
with the same wording of questions and general layout.
These two data sets are well-suited for the analysis of trends in the
health of the elderly because they represent two distinct cohorts of older
people, and consequently, they can give us an indication of recent trends in
disability. The LSOA I was the first nationally representative longitudinal study
of older people, and it was used as the basis for the first multi-state analysis of
active life expectancy (Crimmins, Hayward, and Saito, 1994). The LSOA II is
representative of the same population approximately 10 years later, and it is
used to report more recent estimates of active life expectancy.

38
Description of LSOA I. The LSOA I was a prospective study with a
nationally representative sample comprised of 7,527 civilian,
non-institutionalized persons 70 years of age and over at the time of their 1984
LSOA interview. The LSOA I followed this cohort of older persons through
three follow-up interviews conducted in 1986, 1988, and 1990. The baseline
interview was administered face-to-face in homes by U.S. Census Bureau
interviewers; follow-up interviews were administered by the National Opinion
Research Center at the University of Chicago using computer assisted
telephone techniques (NCHS website, 2006). The interview data for the first
LSOA was augmented by linkage to Medicare records and the National Death
Index death records.
Unlike the other follow-up interviews (1988 and 1990), the 1986 follow-up
did not attempt to find all participants. Due to budgetary limitations that year,
the researchers only re-interviewed a sample of 5,151 of the 7,527 original
respondents. This data collection strategy does not affect the analysis
presented in this dissertation, however, because an attempt to track the full
sample was made in the 1988 and 1990 surveys, and the method we use
allows unequal interval length between surveys.
The original purpose of the LSOA I survey was to facilitate research on the
older population in order to describe the continuum ranging from functionally
independent living in the community all the way through total dependence,
including institutionalization and death (NCHS website, 2006). The data are
useful for measuring change in the functional status of older people over time,
39
and represent the oldest nationally representative longitudinal data source on
older Americans.
Description of LSOA II. The Second Longitudinal Study of Aging (LSOA II)
was originally labeled the Second Supplement on Aging (SOA II) and was
conducted in conjunction with the 1994 National Health Interview Survey
(NHIS). The LSOA II was a prospective study with a nationally representative
sample comprised of 9,447 civilian non-institutionalized persons 70 years of
age and over at the time of their SOA II interviews. Data on baseline disability
status is available for 9,409 cases. The LSOA II followed this cohort of older
persons through two follow-up interviews, conducted in 1997-98 and
1999-2000. The LSOA II was designed to provide a replication of the first
LSOA in order to determine whether there had been changes in the disability
and impairment process among older persons between the 1980s and 1990s.
Like the LSOA I, the LSOA II follow-up consisted of interviews conducted over
the telephone using a computer-assisted telephone interview protocol for most
participants. For those individuals who could not be (or did not want to be)
reached by phone, a self-administered questionnaire was sent by mail.
40
Comparison of Similarities and Differences between LSOA and LSOA II.
An effort was made to design the LSOA II to be comparable in as many
ways as possible to the LSOA I. However there are a number of differences
that should be considered. Table 3.1 outlines the key differences in the
samples and the survey designs. As shown in the table, the response rate is
slightly lower in the second LSOA (87.4% as compared with 93.2% for the
earlier LSOA I). Further, while only 4% of LSOA I respondents were lost after
the baseline interview, 11% of the LSOA II respondents were similarly
interviewed only once. In addition, the response rate was lower at every
subsequent follow-up wave in the LSOA II, as compared to the similar wave in
the LSOA I (see table 3.1).
Among those known to have died during the survey time interval, the
month and year of death were reported for about 94% of the LSOA I
participants, while only about 79% of LSOA II participants reported both the
month and year of death. For participants known dead, but without a specific
date of death in the data, the midpoint between the previous survey date, and
the date of the following survey (when participant was known to be dead) was
used as the date of death.
There were some changes in the timing of the baseline interviews as well,
as the LSOA I baseline interview was conducted at the same time as the
National Health Interview Survey, while the LSOA II interviews were
conducted between 7-17 months after respondents participated in the National
Health Interview Survey. Further, the baseline interviews extended longer
41
than a 12-month period, leading to individuals aging a year between their
selecting into the sample and their interview. Thus, there is a slight
undercount of individuals at exactly the age of 70 in the LSOA II; however, this
is adjusted using the analysis weight.
Table 3.1. Description of the Similarities and Differences in the Two
LSOA Datasets
LSOA I LSOA II
Analysis Sample Size N=7,477 N=9,409
Response rate wave 1 93.2 87.4
Number of Waves of
Data Collection
4 3
Missing at every wave
after initial wave
3.9% (N=289) 11.0% (N=1039)
Missing last interval
after imputation
7.0% (N=523) 8.7% (N=822)
Response rate at wave
2
--90.8% (of alive)
4037 complete
576 dead
54 refused
354 missing
--78.8% of alive
6954 complete
975 dead
775 refused
705missing
Wave 3 --83.2% of alive
4906 complete
920 dead
191 refuse
798 missing
606 previously dead
--77.0% of alive
5493 complete
1302 dead
1260 refused
379 missing
975 previously dead
Wave 4 --79.0% of alive
4080 complete
696 dead
500 refused
584 missing
1498 previously dead

42
Table 3.1, Continued
Percent of dead with
known date of death
94.1% 79.3%
Attrition loss by final
wave
11.2% (N=6638) 16.9% (N=7825)
Interview period

Average length
between interviews

1/84-12/84
8/86-12/86
8/88-12/88
7/90-9/90

26.0 months (including
those with skipped 86
interview)
24.8 months
22.2 months
9/94-3/96
5/97-7/98
6/99-8/2000

29.3 months
27.6 months
Initial Age Age 70 in 1984 Age 70 at baseline interview
1994-1996 (one year was
anticipated for data
collection and thus the
original pool included
individuals ages 69+ in 1994
as part of the initial sample
panned for interview in
1995; because interview
period was delayed, 70 year
olds are slightly
under-represented)
Over sampling None blacks
43
Table 3.1, Continued
Sampling method Multistage complex sample
design, All SOA households
with a person aged 80 or
over. Within household, all
persons 80 and over and
their relatives 70-79. In a
second stage, all other
households with a person
70-79 were selected. From
these households, all Black
persons and their relatives
aged 70-79 were selected.
Finally, remaining
households with a person
aged 70-79 whose residents
were either white or other
non-black were randomly
sorted, and half of the
households were selected. If
more than one person 70-79
lived in the household, all
were included.
Same, except over-sample
of blacks
Proxies 13.6% - 1984
24.3% – 1986
23.8% – 1988
26.2% - 1990
17.0% - 1994
21.1% - 1998
23.6% - 2000
Institutionalized 1986 – 3.7% (n=150)
1988-4.6% (n=229)
1990 – 5.8% (n=234)
1998-2.4%(n=226)
2000-4.7%(n=311)

Follow-up Immediate at NHIS interview 7-17 months
Similarities Supplement to NHIS of 70+, in-person baseline, same
procedure for accepting proxies, general order of questions
and wording preserved.
Baseline in person using census bureau interviewers
Follow-up by telephone or mail-in survey conducted by
Univ. of Chicago National Opinion Research Center

44
Table 3.1, Continued
Length Some parts were collected
on paper to reduce interview
length.
Longer, functioning and
Disability Follow-back
Survey (DFS) questions
interspersed.
Developed as part of phase
II of 94-95 NHIS-D, DFS2
(disabled) and DFS3
(non-disabled) together
comprise LSOA baseline
Description of the 1986
subsample.
1985 reduction in number of
primary sampling units
(PSUs), selection of 2 PSUs
per non-self-representing
stratum, use of all area
frame (slight impact in
comparability of baseline
data in respective years,
corrected by weights)

Weights Use of weights allows for
control of design changes
and comparisons.
Product of four components
take into account complex
multistage probability
design: probability of
selection, household
non-response adjustment
within segment, first-stage
ratio adjustment and
post-stratification by
age-sex-race.

45
The time intervals between interviews are slightly longer in the LSOA II
study because the 6-year period was divided into 3 interviews, as compared to
the 4 interviews of the LSOA I. The unequal time intervals do not affect the
analysis, however, since the Markov chain approach used, and described later,
uses the time interval between events for each individual, regardless of
differences across individuals.
The LSOA II oversampled the black population, while there are no
oversampled groups in baseline of the LSOA I. This oversampling is
accounted for in the survey weight, and thus should not affect the
comparability of the analytical samples.
The LSOA II has a small undercount of 70-year-olds because the baseline
interview period covered 18 months, and some individuals were not reached
until their 71
st
birthday. The weight includes an adjustment for this fact,
allowing the results to remain representative of the national population.
In many other ways, the surveys are quite similar. For example, a similar
proportion of proxy respondents were used in corresponding surveys, and the
same protocol was used to choose proxy respondents. Finally, the surveys
were specifically designed to be comparable in their content, question order,
and question wording.
Weighting
Unique weights for each of the LSOA datasets were calculated by the
National Center for Health Statistics. The NHIS calculated weights for both
data sets based on four components: multistage design probability, selection
46
probability, non-response, and stratification by age-sex-race (National Center
for Health Statistics, 2002).
Additional weighting adjustments were needed to reflect the
institutionalized population. Both the LSOA I and the LSOA II samples
originally omitted the institutional population, but both followed participants
after the baseline into institutions, and thus the datasets under-represent this
portion of the population. Using estimates of the proportion of the
institutionalized population from the 1989 National Long-Term Care Survey for
the LSOA I and the 1994 National Long-Term Care Survey for the LSOA II,
then institutionalized population is weighted at later waves so that the
institutionalized are represented appropriately in transition states. The
proportions institutionalized in later follow-up interviews are shown in table 3.2.
Table 3.2. Proportion Institutionalized in Later Waves, and the Additional
Weighting Factor Used to Match National Long-term Care Survey
Estimates.
LSOA I LSOA II

Extra Weight to the
institutionalized at
wave 2
3.03 2.9
Institutionalized
sample by wave
N=150 (3.7%) wave 2
N=229 (4.6%) wave 3
N=234 (5.8%) wave 4
N=226 (2.4%) wave 2
N=311 (4.7%) wave 3
47
Variables and Measures
Variables
Calculations of active life expectancy were based on classifying
individuals as disabled or non-disabled at each time point. Able/disabled was
assessed by a dyad of questions, the first of which asked: “Because of health
or physical problem, do you have any difficulty [with a specific ADL or IADL]?”
If the subject responded “yes,” he/she was subsequently asked: “By yourself,
how much difficulty do you have [with a specific ADL or IADL]: some, a lot, or
are you unable to do it?”
Active individuals were those who were able to perform all of the ADL and
IADL tasks (i.e., they did not report being unable to perform any one of the 10
tasks). If an individual responded to fewer than 5 of the 10 ADL or IADL
functioning questions, he/she was classified as missing (coded as “–1” in the
data). If an individual was known to be institutionalized at the time of interview,
he/she was classified as disabled.
Independent Variables
Age, measured in years, was calculated by the difference between
recorded date of birth in the survey and the date of each interview. The
sample age varied between 70 to 104 years. The demographic variables of
gender, education, and race were included as dichotomous independent
variables in the analyses. Race was dichotomized as black or non-black. The
non-black population is substantially white, and is thus referred to as white.
Other races are not represented in significant numbers. Education was
48
dichotomized as having twelve or more years of education, or less than twelve
years of education.
Table 3.3 provides the basic frequencies of the variables in the
analyses in each data set. Note that in the LSOA II, collected ten years later,
the sample was composed of proportionately fewer females (60.3% versus
62.0%), more educated (59.0% versus 43.8% with a high school degree or
more), and more racially diverse (10.6% versus 7.4% black).
Comparing the first wave of the surveys to the final wave, also shown in
Table 3.3, over the course of the 6-year survey time frame, the samples at the
end were about 5 years older on average (less than 6 years due to higher
mortality at the oldest ages), included a slightly higher proportion of females at
the final wave, had a slightly better educated pool of survivors at the final
wave, and the race composition did not change.

49
Table 3.3. Basic Sample Frequencies at Baseline and Final Wave
Initial Wave
1984 LSOA I N 1994 LSOA
II
N
Mean Age (in years) 77.2 (SD 5.6) 7477 76.3 (SD 5.7) 9382
Percent Female 62.0 60.3
% 12 Years Education
or more
43.8 59.0
% Black 7.4 7477 10.6 9382

Final Wave
1984
LSOA I
N 1994
LSOA II
N
Mean Age (in years) 82.2 (4.8) 4139 81.4 (5.3) 7006
Percent Female 65.4 4139 62.5% 7006
% 12 Years
Education or more
46.7 4139 61.0% 6889
% Black 7.4 4139 10.2 7006
% Dead after 6 years 32.4 2427 24.9 2340
50
Tables 3.4 and 3.5 show frequencies of the functioning and mortality
data across the waves of data collection. Table 3.4 shows the data in percent
format for both the data as originally collected, as well as the values used in
the analysis. The analysis data included imputation for missing data in order
to accommodate sample attrition without compromising the representative
nature of the baseline sample. The imputation procedure and the number of
cases imputed will be described in detail in a subsequent section. Table 3.5
shows the number of cases that reported the able, disabled, missing or dead
states at each wave.
As shown in table 3.4, the proportion of the sample that is disabled
increases over time, starting at about 12% of the sample at baseline, and
increasing to about 25% of survivors in LSOA I after 6 years, and 21% of
survivors in LSOA II. Between 9 and 19 percent of the sample reported
missing disability data in each wave, although this was reduced by the
imputation process. As shown in table 3.5, the number missing at the final
waves is very different before and after imputation. This issue will now be
described in further detail.
51
Table 3.4. Distribution of disability status variables in the LSOA I and
LSOA II
LSOA I
1984 1986 1988 1990
Active* (1) 87.6% [87.5%] 77.0% [78.7%] 77.8% [79.6%] 75.1% [78.4%]
Disabled (2) 12.4% [12.5%] 23.0% [21.3%] 22.2% [20.4%] 24.9% [21.6%]
Dead** (3) 12.9% [12.9%] 16.5% [16.5%] 16.4% [12.8%]
Missing*** (4) 9% (of attempted
interviewees,
36.6% overall)
(N=2740)
13.1% [14.7%] 0.0% [19.5%]
Previously Dead N=629 N=1619
N=7,496 N=4,667 N=6,867 N=5,785

LSOA II
1994 1997 2000
Active* (1) 86.8% [86.8%] 82.3% [82.6%] 79.0% [79.7%]
Disabled (2) 13.2% [13.2%] 17.7% [17.4%] 21.0% [20.3%]
Dead** (3) 12.0% [12.0%] 16.1% [15.4%]
Missing*** (4) 13.7% [15.7%] 0.8% [18.8%]
Previously Dead N=975
N=9409 N=9,409 N=8,434
* The percent active refers to the number of respondents who were active
relative to all respondents with known active or disabled status.
** The percent dead refers to the number of respondents who were reported
dead at the time of interview relative to all respondents who were not
previously known to be dead.
*** The percent missing refers to the number of respondents who have missing
physical functioning data relative to all respondents who were known alive.

Percent used in analysis shown first, percent in the raw data as collected in
brackets.
52
Table 3.5. Frequency of disability status variables in the LSOA I and
LSOA II
LSOA I
1984 1986 1988 1990
Active* (1) 6565 [6541] 3178 [3178] 3873 [3873] 3149 [3143]
Disabled (2) 931 [931] 949 [860] 1102 [993] 1045 [866]
Dead** (3) 629 [629] 990 [990] 825 [740]
Missing*** (4) [24] 2740 [2829] 902 [1011] 0 [1128]
Previously Known
Dead (subtracted
from missing)
629 1619
Total Sample N=7,496 N=7,496 N=7,496 N=7,496
Removing
Previously Known
Dead
N=6867 N=5877

LSOA II
1994 1997 2000
Active* (1) 8167 [8167] 5875 [5741] 5535 [4378]
Disabled (2) 1242 [1242] 1267 [1213] 1471 [1115]
Dead** (3) 975 [975] 1358 [1302]
Missing*** (4) 1292 [1480] 76 [1583]
Previously Known Dead
(subtracted from missing)
975
Total Sample N=9,409 N=9,409 N=9,409
Removing Previously
Known Dead
N=8,434
* The percent active refers to the number of respondents who were active
relative to all respondents with known active or disabled status.
** The percent dead refers to the number of respondents who were reported
dead at the time of interview relative to all respondents who were not
previously known to be dead.
*** The percent missing refers to the number of respondents who have missing
physical functioning data relative to all respondents who were known alive.

Number used in analysis shown first, percent in the raw data as collected in
brackets.
53
Missing Data Imputation Methods and Effects of Missing Data
Missing Data
Issues of missing data affecting the analysis can be broken into two
groups: data missing as a result of non-response to specific questions and
data missing because of sample attrition. In the LSOA I dataset, there were
312 cases (4.2%) missing because of non-response to questions within the
survey, and 546 cases (7.3%) missing because of sample attrition (in other
words, those cases were lost after the initial wave). In the LSOA II data set,
there were 80 cases (0.9%) missing because of non-response to questions,
and 1,503 cases (16.0%) because of sample attrition. The high attrition rate in
the LSOA II cannot be ignored, as this affects a sizable proportion of the
sample. Table 3.6 illustrates the missing data issues in both the LSOA I and
the LSOA II. There are four pairs of descriptions in table 3.6, and each pair
shows the proportion of the sample with missing data before as well as after
the imputation process.
54
Table 3.6. Distribution of Missing Data in the LSOA I and LSOA II

LSOA I LSOA II
Complete Data
(all functioning, or
interviewed until death)
before imputation
88.6% (N=6,638) 83.2% (N=7,826)
Complete Data
(all functioning, or
interviewed until death)
after imputation

100% (N=7,496) 99.7% (N=9,378)
Not interviewed after initial
wave before imputation
3.9% (N=289) 8.1% (N=762)
Not interviewed after initial
wave after imputation
0% (N=0) 0.3% (N=31)

Missing data for 2 intervals
before imputation
6.3% (N=474) 11.0% (N=1,039)
Missing data for last 2
intervals after imputation
0% (N=0) 0.3% (N=31)

Missing data for last
interval before imputation
15.0% (N=1,128) 16.8% (N=1,583)
Missing data for last
interval after imputation
0% (N=0) 0.8% (N=76)
TOTAL N=7,496

N=9,409

55
As part of the data analysis, a categorization process was used to
identify the underlying reasons for non-response. These categories included
the following: vital status (dead or alive), interview type (community dwelling,
institutionalized, proxy decedent, or none), and missing data. These data
appear in Table 3.7, which displays a layered cross tabulation by interview
type and vital status at the second wave by the same characteristics at the
third wave. The actual number of respondents is reported in this table. The
areas highlighted in gray represent cases that were reported dead, and thus
no functioning data is needed at these waves. The 28 respondents in the first
column were not included in the analysis, as these participants refused to
participate in the longitudinal component of the LSOA survey.
The 16.0% attrition loss in the LSOA II is illustrated in the second and
fourth columns of Table 3.7, headed by the titles “Alive” (the first alive) and
“Unknown,” which include individuals not interviewed at the third wave, and for
whom vital status was either alive or unknown at the final wave.
56
Table 3.7. Cross Tabulation of Second and Third Interview with Vital
Status at Each Time Point
Third wave Interview

Not interviewed wave 3
Comm
unity
Institution Proxy
dead
Second Wave Interview
Col 1 Col 2 Col 3 Col 4 Col 5 Col 6 Col 7
Int Type Vital
Status
Not
interv
Alive Dead Unknown Alive Alive Dead
Alive . 435 51 55 171 11 48
Dead 2 12 138 24 5 63
Not
interviewed
Wave 2
Unknown 28 73 59 150 68 10 36
Community Alive . 711 150 98 4987 199 667
Institution Alive . 18 18 5 9 89 85
Proxy Dead 934

Attrition is less complicated in the LSOA I, as far fewer cases were lost
to follow-up, and vital status was known on all individuals. A similar table for
LSOA I would not be informative, since there are additional data points on
which to base the imputation (four as opposed to three), and a single variable
identifying both vital and interview status is available for each wave. No such
summary variable exists in the LSOA II, and only a complex cross-tabulation
can fully inform us which individuals are actually lost to follow-up and known to
be alive because of contact with a proxy. Most proxy respondents were
relatives or a respondent’s spouse.
Method of imputation. Missing data for cases known alive, but not
responding to questions (missing from the sample), were assigned values
based on the proportion of the sample with the same previously known
health-transition patterns, as observed across time, as those with known
health status in both the initial and final interviews. In the 1984 data,
information on the previous three waves was used to match cases with
57
unknown final health status (fourth wave) to those with the same entries for the
first three waves. If information on only two prior waves was available, the
previous two waves were used. If only the first wave was known, then the
proportion was based on all cases with the same initial health status and a
known live health status at the final interval.
The first step in creating the imputations was to look at the series of
disability states that were observed over time. A four-digit code was
constructed consisting of the individual’s disability status at each of the four
interviews for the LSOA I. This frequency of this code groups individuals
based on transitions between disability states over time. Similarly, a
three-digit code was constructed for the LSOA II (with each digit representing
the individual’s disability state at each of the 3 interviews). The coding of this
variable was similar to the disability variable (1=able, 2=disabled, 3=dead)
except for missing, which is coded as 4, rather than the –1 used by the
analysis program. The frequency of the disability codes representing states at
each wave for both LSOA I and II is shown in Tables 3.8 and 3.9. The tables
show the frequency before and after imputations. The imputations were
created using a process explained in the following paragraphs.
Expanding on the logic of Table 3.7, Table 3.8 shows specific disability
detail at each wave before and after imputation, using the three and four digit
codes previously described. In the LSOA II data, those initially active in the
first 2 waves, but known alive with missing data in the final wave (active code
114 in Table 3.8), had a 0.91 likelihood of being randomly assigned to an
58
active state, and 0.09 likelihood of being assigned a disabled state. These
proportions were calculated based on the sample with three interviews and the
same states at waves 1 and 2 (see dark bordered box in Table 3.8).
59
Table 3.8. Detailed View of Health Transitions Before and After
Imputation (left side original code and frequency, right side final code
and frequency)

Not interviewed wave 3

Alive Dead Unknown

N=435
Before After
144-410 141-373
142-37
244-25 241-8
242-17

N=51
Before After
144-42 143-42
244-9 243-9
444-1

N=55
Before After
144-49 114-35
124-8
134-6
244-6 214-4
444-3 224-1
234-1

N=12
Before After
144-9 141-8
244-3 142-1
241-2
242-1

N=138
Before After
144-91 134-91
244-47 234-47

N=24
Before After
144-16 134-16
244-8 234-8

N=73
Before After
144-67 141-59
142-8
244-6 241-2
242-4

N=59
Before After
144-41 143-41
244-18 243-18

N=150
Before After
144120141-91
142-11
143-18
244-30 241-14
444-1 242-4
243-12

Third wave Interview

Not interviewed wave 3
Comm
unity
Instit
ution
Proxy
dead
Second Wave Interview
Int Type Vital
Status
Not
interv
Alive Dea
d
Unkn
own
Alive Alive Dead
Alive . 435 51 55 171 11 48
Dead 2 12 138 24 5 63

Unknown 28 73 59 150 68 10 36
Community Alive . 711 150 98 4987 199 667
Institution Alive . 18 18 5 9 89 85
Proxy Dead 934

Not interviewed
Alive Dead unknown
60
Table 3.8, Continued

Not interviewed wave 3

Alive Dead Unknown

N= 711
Before After
114-546 111-491
112-55
124-53 121-18
144-46 122-35
141-41
142-5
214-22 211-12
212-10
224-36 221-5
244-8 222-31
241-5
242-3
414-4
444-4

N= 150
Before After
113-90 113-90
123-28 123-28
143-6 143-6
213-6 213-5
223-20 223-20
243-1
413-1

N= 98
Before After
114-69 111-45
112-2
113-22
124-12 121-3
144-4 122-6
123-3
141-2
142-4
143-1
214-3 211-2
224-10 212-1
221-1
222-5
223-4

N= 18
Before After
124-5 121-3
144-6 141-2
142-6 224-6
222-6
244-1 242-1
N= 18
Before After
123-8 123-8
143-3 143-3
223-3 223-3
243-4 243-4

N=5
Before After
124-2 122-2
144-1 142-1
224-1 222-1
244-1 242-1

Third wave Interview

Not interviewed wave 3
Commu
nity
Institu
tion
Proxy
dead
Second Wave Interview
Int Type Vital
Status
Not
interv
Alive Dead Unkn
own
Alive Alive Dead
Alive . 435 51 55 171 11 48
Dead 2 12 138 24 5 63

Unknown 28 3 59 150 68 10 36
Community Alive . 711 150 98 4987 199 667
Institution Alive . 18 18 5 9 89 85
Proxy Dead 934

Community
Institution
Alive Alive
61
Table 3.8, Continued

Community institutionalized Proxy dead

Alive Alive Dead

N= 171
Before After
141-133 141-137
142-22 142-25
144-7 241-3
241-3 242-6
242-6

N=11
Before After
142-6 142-9
144-3 242-2
242-2

N=48
Before After
143-40 143-40
243-8 243-8

N=5
Before After
141-3 141-4
142-2 144-1

N=63
Before After
133-42 133-42
233-21 233-21
433-1

N=68
Before After
141-49 141-53
142-8 142-8
144-4 241-3
241-3 242-4
242-3
244-1

N=10
Before After
142-5 142-7
144-2
242-2 242-3
244-1

N= 36
Before After
143-29 143-29
243-7 243-7

Third wave Interview

Not interviewed wave 3
Commu
nity
Inst
ituti
on
Proxy
dead
Second Wave Interview
Int Type Vital
Status
Not
interv
Alive Dead Unkn
own
Alive Aliv
e
Dea
d
Alive . 435 51 55 171 11 48
Dead 2 12 138 24 5 63

Unknown 28 73 59 150 68 10 36
Community Alive . 711 150 98 4987 199 667
Institution Alive . 18 18 5 9 89 85
Proxy Dead 934

Not interviewed
Alive Dead unknown
62
Table 3.8, Continued

Community institutionalized Proxy dead

Alive Alive Dead

N= 4987
Before After
111-3869 111-3913
112-335 112-340
114-49 121-118
121-116 122-205
122-203 141-67
124-4 142-5
141-58 211-70
142-4 212-75
144-10 221-38
211-70 222-153
212-74 241-2
214-1 242-1
221-37
222-152
224-2
241-2
242-1
411-4
422-3
444-2
N= 199
Before After
111-2 112-98
112-59 122-46
114-37 142-3
122-37 212-15
124-9 222-37
142-2
144-1
212-12
214-3
222-32
224-5
411-4
422-3
444-2

N=667
Before After
113-386 113-385
123-134 123-134
143-10 143-12
213-30 213-35
223-99 223-99
243-7 243-2
413-1
423-4
443-1

N=9
Before After
121-1 121-2
122-1
142-3 141-2
144-2 142-3
242-1 241-1
244-1 242-1
N= 89
Before After
121-2
122-31 122-33
142-7
144-14 142-21
221-3 222-28
222-25
242-4
244-3 242-7

N= 85
Before After
123-41 123-41
143-14 143-14
223-22 223-22
243-8 243-8
423-1
443-1

Third wave Interview

Not interviewed wave 3
Comm
unity
Insti
tutio
n
Prox
y
dea
d
Second Wave Interview
Int Type Vital
Status
Not
interv
Alive Dead Unknown Alive Aliv
e
Dea
d
Alive . 435 51 55 171 11 48
Dead 2 12 138 24 5 63

Unknown 28 73 59 150 68 10 36
Community Alive . 711 150 98 4987 199 667
Institution Alive . 18 18 5 9 89 85
Proxy Dead 934

Community
Institution
Alive
63
The active status codes for these individuals was based on the
summed number of individuals with active codes 111 (able during times 1, 2,
and 3; N=3,869) and 112 (able during times 1 and 2 but disabled at time 3;
N=394). These two codes were the only possible combinations for individuals
known alive at time three but with an unknown ability/disability status. Thus,
among those individuals reporting ability status in all three interviews, 3,869
reported that they were able to perform all of the ADL and IADL tasks while
394 reported being disabled at time three. Thus, of the total pool of 4,263
cases there was a 91% (3,869/4,263) likelihood that an individual not disabled
at times 1 and 2 and alive at time 3 would not be disabled at time 3 (active
code 111). Similarly, there was a 9 percent probability that an individual not
disabled in times 1 or 2 would become disabled at time 3, or active code 112.
A similar procedure was used for other cases. Among those who were active
in the first wave but were disabled in the second wave (active status
code=124), the respondents had a 0.32 likelihood of being in an active final
state and a 0.68 likelihood of being assigned a disabled final state.
A similar technique was used for cases known to be disabled in the first
interval. If a case began as disabled, then transitioned to active in the second
time period, and was known alive in the third time period (active status
code=214), the person had a 0.49 likelihood of being in an active state in the
third interval and 0.51 chance of being assigned a disabled state. If a case
was known disabled in two intervals and alive in the third interval (active status
64
code=224), a 0.20 likelihood was assigned for a final active status and a 0.80
likelihood of being assigned to a disabled state.
For 1,139 cases, only health status at the first interval was known, while
vital status at the final interval was known to be alive. These cases were
assigned a transition likelihood based on all cases beginning with the same
initial health status. For those initially active (active status code=144), there
was a 0.94 likelihood of being assigned a final state of active and 0.06
likelihood to be assigned disabled in the final interval. For those initially
disabled, and missing data for the other two time periods (active status
code=244), there was a 0.67 likelihood of being assigned a final status of
active and 0.33 likelihood of being assigned a final status of disabled based on
the transition probabilities of all cases with initial status disabled and known
final functioning status.
For 205 of the cases, vital status was unknown at the third interval, and
thus it was possible that many were alive while some may have died. These
cases were assigned to active, disabled, or dead states based on proportions
of all known data in the entire dataset with the same initial health status.
Allowing 24 percent of these cases to be assigned as dead (based on 2,340
deaths and 9,330 known alive), 76 percent were assumed alive divided into 86
percent able and 14 percent disabled.
Table 3.9 provides the final frequency of the disability status code in
both the original data, as well as the data after imputation, in both LSOA I and
II. Fewer cases have missing data (coded as 4) in the last time period after
65
imputation, although missing data still exists in non-terminal interviews, as well
as for cases for whom no vital status information is known. In general, the
percent of transitions did not change much, except those transitions ending in
a missing status. The most dramatic change was observed among those who
were initially able but were subsequently lost after the baseline survey in the
LSOA II (disability code 144). Originally, ten percent of the sample was coded
this way. For most of these individuals, they were alive (based on the Table
3.8 cross-tabulation) and thus they were assigned either able or disabled end
disability codes (codes 141 or 142). Thus, we observe large changes for these
codes. The changes are more subtle for LSOA I, where only 128 imputations
were done, since far fewer cases were lost to follow-up.

66
Table 3.9. Distribution of Health Statuses in Each Survey
LSOA I Number of cases Percent
Before After Before After
1111 1653 1828 22.1 24.9
1112 224 223 3.0 3.0
1113 181 178 2.4 2.4
1114 180 0 2.4 0
1121 57 57 0.8 0.8
1122 155 180 2.1 2.5
1123 89 89 1.2 1.2
1124 31 0 0.4 0
1134 269 260 3.6 3.5
1141 78 163 1.0 2.2
1142 36 35 0.5 0.5
1143 48 48 0.6 0.7
1144 87 0 1.2 0
1211 38 45 0.5 0.6
1212 30 30 0.4 0.4
1213 11 11 0.2 0.2
1214 7 0 0.1 0
1221 17 17 0.2 0.2
1222 112 129 1.5 1.8
1223 90 83 1.2 1.1
1224 21 0 0.3 0
1234 167 152 2.2 2.1
1241 3 3 0.0 0.0
1242 9 25 0.1 0.3
1243 17 16 0.2 0.2
1244 16 0 0.2 0
1344 418 405 5.6 5.5
1411 1100 1252 14.7 17.1
1412 111 111 1.5 1.5
1413 111 111 1.5 1.5
1414 152 0 2.0 0
1421 41 41 0.6 .6
1422 111 136 1.5 1.9
1423 50 49 0.7 .7
1424 26 0 0.4 0
1434 307 306 4.1 4.2
1441 114 375 1.5 5.1
1442 48 48 0.6 0.7
1443 86 86 1.2 1.2
67
Table 3.9, Continued
LSOA I Number of cases Percent
Before After Before After
1444 264 0 3.5 0.0
2111 8 7 0.1 0.1
2112 10 16 0.1 0.2
2113 3 3 0.0 0.0
2114 6 0 0.0 0.0
2121 5 8 0.0 0.1
2122 19 18 0.3 0.3
2123 11 11 0.2 0.2
2124 3 0 0.0 0.0
2134 10 8 0.1 0.1
2141 1 5 0.0 0.1
2142 5 5 0.0 0.1
2143 5 5 0.1 0.1
2144 4 0 0.1 0.0
2211 7 7 0.1 0.1
2212 13 15 0.2 0.2
2213 4 2 0.1 0.0
2214 2 0 0.0 0.0
2221 4 4 0.1 0.1
2222 104 117 1.4 1.6
2223 69 59 0.9 0.8
2224 19 0 0.3 0.0
2234 157 135 2.1 1.8
2241 2 2 0.0 0.0
2242 8 15 0.1 0.2
2243 15 15 0.2 0.2
2244 7 0 0.1 0.0
2344 211 176 2.8 2.4
2411 12 14 0.2 0.2
2412 5 7 0.1 0.1
2413 1 1 0.0 0.0
2414 4 0 0.1 0.0
2421 5 5 0.1 0.1
2422 40 43 0.5 0.6
2423 19 18 0.3 0.3
2424 4 0 0.1 0.0
2434 80 78 1.1 1.1
68
Table 3.9, Continued

2441 4 4 0.1 0.1
2442 5 30 0.1 0.4
2443 15 15 0.2 0.2
2444 25 0 0.3 0
Total 7496 7496 100% 100%

LSOA II Number of cases Percent
Before After Before After
111 3871 4449 41.1 47.3
112 394 495 4.2 5.3
113 475 497 5.1 5.3
114 701 35 7.5 0.4
121 117 144 1.2 1.5
122 272 329 2.9 3.5
123 211 214 2.2 2.3
124 87 8 0.9 0.1
134 608 721 6.5 7.7
141 248 837 2.6 8.9
142 57 151 0.6 1.6
143 146 248 1.6 2.6
144 969 28 10.3 0.3
211 70 84 0.7 0.9
212 86 101 0.9 1.1
213 40 40 0.4 0.4
214 29 4 0.3 0.0
221 37 44 0.4 0.5
222 209 261 2.2 2.8
223 144 148 1.5 1.6
224 63 1 0.7 0.0
234 326 382 3.5 4.1
241 9 40 0.1 0.4
242 19 55 0.2 0.6
243 51 90 0.5 1.0
244 170 3 1.8 0.0
Total 9409 9409 100% 100%

69
Transition between disability states. In order to compare the two datasets
directly, a listing of the total number of data points by dataset (LSOA I and
LSOA II) is shown in Table 3.10. The transition types between two time points
considered in the present analysis include:
Active Disabled
Active Dead
Disabled Active
Disabled Dead
Active Active
Disabled Disabled
Table 3.10. Number of Health Transitions for Adults Aged 70 and Older
Sample States 1984 LSOA

1994 LSOA II
Remained Healthy 9,838 62.5% 10,695 66.9%
Became Disabled 1,618 10.3% 1,454 9.1%
Died From Healthy
State
1,454 9.2% 1,477 9.2%
Recovered from
Disability
362 2.3% 692 4.3%
Remained Disabled
1,478 9.4% 840 5.3%
Died From Disabled
State
990 6.3% 822 5.1%
16,598100% 15,980 100%
70
These transitions are illustrated in Figure 3.1, which illustrates the
transitions listed above in a multi-state transition model. This figure illustrates
bi-directional flow between disability states for individuals who were alive, and
the absorbing mortality state.
Table 3.9 shows the number of these transitions pooled across all waves of
data. As shown in the table, the majority of the sample remained healthy
across each time point (62.5% in LSOA I and 66.9% in LSOA II). The key
differences between the surveys are seen in the proportion of the sample
experiencing recovery, which was 2.3% in LSOA I and 4.3% in LSOA II, and in
the proportion remaining disabled, which was 9.4% in LSOA I and 5.3% in
LSOA II.
Figure 3.1. Multi-state Table Representation of Transition between Live
States, and to Death

No functional
impairment (1)
Impaired (2)

Dead (3)
71
Methodology
Analysis of the LSOA data was based on a multi-state life table that was
calculated through the IMaCH (Version 0.98h) algorithm designed by Brouard
and Lievre (2001). The program calculates weighted estimates of life
expectancy and implied disability prevalence based on the recent rates of
onset and recovery from disability and age-specific rates of death by disability
state (Lievre et al., 2003). One advantage of this program is that variable
interval lengths within individuals can be used. This will adjust for the
differences in interval length between the two surveys and allow cases with
some missing data to contribute to the results. Other methods, such as the
hazard equation approach, exclude any interval with missing data, and
assume only one transition between points of observation.
Laditka and Wolf (1998) described how the parameters of an
embedded Markov Chain could be used to model disability change over time
in longitudinal data with a maximum likelihood estimation technique. The
advantage of this Markov Chain approach is that it can capture change that
occurs between survey interviews. Markov Chains have been used in similar
studies before modeling disease onset and progression (Jackson & Sharples,
2002).
The IMaCh method has been used in a variety of health transition
studies using longitudinal data to examine health expectancy in relationship to
gender (Minicuci et al., 2005; Kaneda, Zimmer, & Tang, 2005; Jagger, Goyder,
Clarke, Brouard, & Arthur, 2003), education (Minicuci et al., 2005; Kaneda,
72
Zimmer, & Tang, 2005), and diabetes (Jagger et al., 2003). The IMaCh
approach allows the computation of age-specific transition schedules for the
onset of disability, recovery from disability, mortality among the disabled, and
mortality among the non-disabled. The calculations are done through an
estimation of a Markov Chain, a statistical technique that allows controllable
levels of random variation over time to simulate health changes in observed
data (Laditka & Wolf, 1998). The random nature of the Markov model is used
to estimate variation in disability status over time, including between survey
intervals.
The Markov chain in the IMaCh program uses conditional probabilities
to estimate the likelihood for being in a specific state at a given time. In
mathematical notation the probabilities could be described as hP
ijx
where the
initial state is represented by i, age is x, j is the final state, and h is the time
period between observations to which the probability refers. This notation
indicates that the result is conditional to the observed state i at age x for a
given period of allowed transitions, 'h'. This transition is modeled in the
following multinomial logistic equation:
Ln [(p)/(1-p)] = Bx (Age) + Cx (Covariate1) + Dx (Covariate2) + Error
This equation is solved using a maximum likelihood estimation and the
multiplication of a series of matrices. The hPx matrix (where h is the set time
period and P is an estimated probability), is the product of the nh*stepm
matrices, where n is the number of matrices. The contribution of each
individual to the likelihood of a transition is denoted as hP
ijx
.
73
The three states are able (coded 1), disabled (coded 2), or dead (coded 3).
The transition probabilities were estimated based on a series of 3 x 3 matrices:
= = Ρ ) (
ij
x h x h
p

11
x h
p
12
x h
p
13
x h
p

21
x h
p
22
x h
p
23
x h
p
0 0 1

The transition probabilities are then used as the inputs to a multi-state life
table using the maximum likelihood approach. The multiple states are
illustrated in Figure 3.1. The model can have two or more states of health, and
an absorbing state (in the case of this study, death). The multi-state method
calculates active life expectancy from a set of transitions between active and
inactive states of life and to death. Since this method is based on incidence
rates from longitudinal data, results represent only the most recent
occurrences in the population. The method provides standard errors for the
model parameters which are then used to derive standard errors for the life
expectancies derived from the transition probabilities (Liévre et al., 2003).
The calculation of active life expectancy used in this dissertation relied on
having at least two data points, showing a transition between health states,
and thus the imputation scheme described earlier was important in order to
accurately reflect the life expectancy of the actual population. Individuals who
did not report any transitions (e.g., those lost to follow-up) would not be
included in the analysis. This would bias the results, if those individuals were
not included in some way. The imputations were made only if an individual
74
had a disability status at the baseline interview, and was known to be alive (but
not interviewed) at the time of a subsequent wave.
The multi-state calculation of active life expectancy is an estimate of the
average length of active life expected in a ‘synthetic cohort.’ A synthetic cohort
is a hypothetical group of people who experience the morbidity and mortality
rates described in the life table at all ages (Crimmins, Saito, & Hayward, 1993).
The results of multi-state life tables demonstrate the effect that current
disability and mortality rates would have if applied to the life cycle of the
current cohort. The results of the calculation are estimates of total active and
disabled life expectancy.
Comparison of estimated life expectancy to vital statistics. The U.S.
Department of Health and Human Services, Centers for Disease Control,
National Center for Health Statistics (NCHS) collects mortality data each year
in order to create life tables for the U.S. population. These life tables include
life expectancy at each age for the entire population, as well as by gender and
race. Since the NCHS life tables are based on the NCHS final mortality report
and population data from the Census Bureau, as well as data from the
Medicare program, they represent the largest source of data on mortality in the
U.S., and are the “gold standard” by which any calculation of life expectancy
can be compared.
As shown in Tables 3.11 through 3.17, the estimates of total life
expectancy derived from the sample data match closely the values reported by
the vital statistics. Comparing the life expectancy estimates from IMaCh to
75
vital statistics (Table 3.11), it is clear that the estimates at age 70 are very
close to vital statistics. The difference is a little larger at age 80, where the
1997 estimate is six-tenths of a year higher than vital statistics, but the 1987
estimate matches vital statistics.
Table 3.11. Total Life Expectancy Reported by NCHS for the United
States: Vital Statistics Comparison with Total Life Expectancy Estimates
at Two Time Points
Age 1987 1997
LSOA I Vital Statistics LSOA II Vital Statistics
70 13.7 13.6 14.4 14.3
80 8.2 8.2 9.1 8.5
90 4.8 n/a 5.8 4.5
Source: NCHS Life tables: NCHS 1987; 1997

Comparing the sex-specific estimates of life expectancy against vital
statistics (Table 3.12), the results for males are within one tenth of a year of
vital statistics values. The estimates for females are slightly more disparate,
as the sample values for life expectancy appear to overestimate life
expectancy by about four-tenths of year at age 70 in both years, and by nearly
a year (10.0 as compared with 9.1) at age 80 in 1997.
76
Table 3.12. Vital Statistics Comparison with Total Life Expectancy
Estimates By Sex
Male LSOA Data Male Vital
Statistics
Female LSOA
Data
Female Vital
Statistics
1987 1997 1987 1997 1987 1997 1987 1997
70 11.60 12.62 11.8 12.7 15.51 15.82 15.1 15.5
80 6.83 7.61 6.9 7.5 9.37 10.00 8.8 9.1
90 4.02 4.53 4.0 5.44 6.24 4.7

Comparing the estimates for each race (Table 3.13), it is clear that the
estimate for blacks is about one year above vital statistics (13.6 as compared
with 12.6) in 1987 and just two tenths of a year above vital statistics (13.3 as
compared with 13.1) in 1997. The sample size is relatively small for blacks in
1987 (total N=555), and thus the estimates of age-specific mortality rates over
six years are less precise than in the later LSOA data, which oversampled the
black population. The estimates for whites are far closer to vital statistics
values in both years, just one-tenth of a year higher in 1987 (13.8 as compared
with 13.7), and three-tenths of a year higher in 1997 (14.6 as compared with
14.3).
Table 3.13. Vital Statistics Comparison with Total Life Expectancy
Estimates By Race
Black LSOA
Data
Black Vital
Statistics
White LSOA
Data
White Vital
Statistics
1987 1997 1987 1997 19871997 1987 1997
70 13.6 13.3 12.6 13.1 13.8 14.6 13.7 14.3
75 10.7 10.6 10.1 10.7 10.9 11.6 10.7 11.2
80 8.3 8.4 7.9 8.3 8.4 9.2 8.1 8.5
85 6.5 6.7 6.4 6.4 6.5 7.2 6.0 6.2
90 5.2 5.4 5.0 5.0 5.8 4.5

77
Turning to race and sex comparisons (Tables 3.14 and 3.15), the estimates
for black males in 1987 is about eight-tenths of a year above the vital statistics
(11.7 as compared with 10.9), and about three-tenths a year below in 1997
(11.2 as compared with 11.5). The estimates for white males are within
two-tenths of a year of vital statistics.
The estimates for black females are 1.4 years higher (15.3 as compared
with 13.9) than vital statistics in 1987, and five-tenths of a year lower (13.8 as
compared with 14.3) in 1997 (Table 3.15). The estimates at ages 75 and
above conform to within half a year of vital statistics. For white females
differences in 1987 are within four-tenths of a year of vital statistics, and
smaller at ages 70 and 75 in 1997.
Table 3.14. Vital Statistics Comparison with Total Life Expectancy
Estimates: Males by Race
Black Male
LSOA Data
Black Male
Vital Statistics
White Male LSOA
Data
White Male
Vital Statistics
1987 1997 1987 1997 1987 1997 1987 1997
70 11.66 11.16 10.9 11.5 11.56 12.52* 11.8 12.7
75 8.64 8.21 8.7 9.3 8.87 9.67 9.1 9.9
80 6.68 6.54 6.8 7.3 6.70 7.36 6.9 7.4
85 4.87 4.88 5.6 5.7 5.19 5.56 5.2 5.4
Table 3.15. Vital Statistics Comparison with Total Life Expectancy
Estimates: Females by Race
Black Female
LSOA Data
Black Female
Vital Statistics
White Female
LSOA Data
White Female
Vital Statistics
1987 1997 1987 1997 1987 1997 1987 1997
70 15.26 13.81 13.9 14.3 15.48 15.55 15.1 15.5
75 11.60 11.39 11.1 11.5 12.10 12.38 11.8 12.1
80 9.08 8.77 8.6 8.9 9.23 9.82 8.8 9.1
85 7.19 7.26 6.8 6.7 6.99 7.57 6.4 6.6

78
The vital statistics reports do not show mortality rates by education level.
However, nationally representative data of education-specific mortality does
exist at the Centers for Disease Control, and a report is available to
researchers that can form a baseline with which our survey estimates can be
compared (Molla, 2002). Some of the results were published and are publicly
available (Molla et al., 2004).
The comparison of the life expectancy results with the NCHS data is less
precise than the previous life tables. The comparison is imperfect, since the
NCHS data is based only on 1998 mortality data, and the education groups are
defined slightly differently. The NCHS data divide the population into 3
education groups (0-8 years, 9-12 years, and 13 or more years), while the
estimates from the LSOA are based on having less than 12 years, or 12 or
more years. Nonetheless, the NCHS data provide the “gold standard”
estimates, as they are based on national mortality data.
Briefly, the LSOA estimates of total life expectancy for males appear to
slightly underestimate the NCHS results, although by less than a year in most
cases, except for those in the higher education group, which appears to
underestimate total life expectancy by about 1.5 years (Tables 3.16 and 3.17).
There is the difference that the NCHS estimate excludes those with only 12
years of formal education, while the LSOA includes this group; however, the
underestimate appears to be consistent for males.
For females, the LSOA estimates very closely match the NCHS data,
usually within one-tenth of a year, but by no more than half a year. With the
79
exception of age 80, our female estimates also slightly underestimate total life
expectancy.
Comparisons of active life expectancy (Tables 3.16 and 3.17) show that
the LSOA reports active life expectancy that is about 1 year longer than the
NCHS data for males, and about 3 years longer for females. The definition of
active is different, as NCHS is based on “any” activity limitation prevalence in
the National Health Interview Survey 1997-1999, while the LSOA definition is
based on the onset of the inability to perform 10 specific tasks of independent
functioning.
Table 3.16. Comparison of Total Life Expectancy by Education level with
NCHS Education Life Tables
Males <12 yrs
LSOA Data
Males 9-12 yrs.
Molla
Males >=12 yrs
LSOA Data
Males 13+ yrs.
Molla
1987 1997 1987 1997 1998
70 11.4 11.8 12.7 12.9 14.5
75 8.7 8.9 9.8 10.0 11.4
80
6.5 7.0
1998
12.2
9.4
7.5 7.5 7.6 8.5

Females <12 yrs
LSOA Data
Females 9-12
yrs. Molla
Females >=12
yrs LSOA Data
Females 13+
yrs. Molla
1987 1997 1998 1987 1997 1998
70 14.6 14.7 14.6 15.7 15.9 16.4
75 11.4 11.8 11.4 12.6 12.7 12.7
80 9.0 9.4 8.5 9.8 10.0 9.1
(Molla, 2002)
80
Table 3.17. Comparison of Active Life Expectancy by Education level
with NCHS Education Life Tables
Males <12 yrs
LSOA Data
Males 9-12 yrs.
Molla
Males >=12 yrs
LSOA Data
Males 13+ yrs.
Molla
1987 1997 1998 1987 1997 1998
70 9.5 9.6 7.2 10.8 11.2 9.3
75 6.6 6.6 5.1 7.8 8.4 6.6
80 4.4 4.7 3.3 5.5 5.9 4.4

Females <12 yrs
LSOA Data
Females 9-12
yrs. Molla
Females >=12
yrs LSOA Data
Females 13+
yrs. Molla
1987 1997 1998 1987 1997 1998
70 10.6 10.0 7.5 11.6 12.3 8.8
75 7.3 7.2 5.0 8.6 9.1 5.9
80 4.9 5.0 2.9 5.8 6.5 3.4
(Molla, 2002)
Additional Results. The multi-state life table results include additional
measures that are helpful in understanding disability trends as well. The
IMaCh program outputs a series of files that describe a number of measures:
the transition rates between live states, and from live to dead states; the active
and disabled life expectancies for each age; the variance for active and
disabled life expectancy; and the implied prevalence of disability in the life
table population and status-based life tables. Each of these output files
describe one of the underlying disability processes or the combined effect of
these processes and can be used to compare the processes in cohorts as well
as patterns in different years. The transition rates can be further
disaggregated by gender and education so that comparisons across the two
decades could be made for each group. In addition, the death rate, divided
into death amongst the disabled and the non-disabled, is reported and can be
compared across groups.
81
Implied Prevalence. The multi-state life table provides an estimate of
disability prevalence in the life table population based on the current level of
disability and mortality transitions. These values determine the distribution of
disability, or other health status, in the life table population, assuming that all
transition rates input to the process stay constant over some period of time.
Due to the stationarity assumption, the estimate is not necessarily the same as
the observed prevalence of disability. The actual future prevalence may be
different, as the implied prevalence does not take into account the
heterogeneity that exists between birth cohorts or the changes that will occur
in the future. However, the implied prevalence can be an especially useful tool
for comparing the implication of current rates. The implied prevalence may be
a more sensitive indicator of the effects of change in the disability process than
observed prevalence because it is based on recent events, and thus is less
affected by disability incurred at younger ages.
Status-based life tables. Most measures of life expectancy and active
life expectancy are derived from population-based life tables which represent
the expected average life for the entire population. However, individuals arrive
at old age either disabled or not, and their health status (e.g., already disabled
or non-disabled) at a particular age can have profound implications for their
remaining life expectancy and life free of disability. Status-based life tables
divide the population into groups based on their disability status at the
beginning of the observation period. Status-based life tables can show
82
whether changes in total and active life expectancy were larger for the
previously non-disabled or the previously disabled.
Application of the Active Life Expectancy in the LSOA I and II
Chapters IV, V, VI, and VII will provide the results of the analyses using the
methodology described in this chapter. The two data sources will be
referenced by the midpoint year of their collection: 1987 in the case of the
LSOA I and 1997 in the case of the LSOA II. Chapter four looks at overall
differences between the LSOA I cohort and the LSOA II cohort. Chapter five
focuses on differences between males and females in active life expectancy
across the time of the two surveys. Chapter six estimates active life
expectancy and disability transitions among individuals by race across the two
surveys. Chapter seven investigates how changes may be different over time
for two different levels of education.
83
Chapter IV: Change in Disability-Free Life Expectancy for Americans 70
Years Old and Over between 1984-2000

This chapter provides an assessment of changes in disabled and non-
disabled life expectancy from experiences covering the period from 1984
through 2000 for the population 70 years of age and older in the United
States. It is the first assessment of changes in healthy life expectancy based
on repeated longitudinal surveys of the older population. This chapter
examines change in the length of disabled and non-disabled life expectancy,
as well as change in the age-specific rates of onset and recovery from
disability, and change in death rates for the disabled and non-disabled
segments of the population. The implications of these changes for the future
disability status of the older U.S. population are investigated using estimates
of the implied prevalence of disability. This paper extends current knowledge
about disability by indicating the source of observed changes in prevalence
of disability and clarifies how the links between mortality and disability affect
the average life course of older persons.
Background
Studies in recent years have provided evidence of a decline in the
prevalence of disability among the older population over more than two
decades (Manton, Gu, & Lamb 2006; Freedman Crimmins, Schoeni, et al.,
2004). First, less severe disability declined beginning in the 1980s; and then
84
more severe disability declined beginning in the 1990s (Crimmins 2004;
Freedman, et al., 2004).
While we have less information on changes in incidence and recovery
from disability, three studies report reductions in the onset of disability or
movement to more severe disability over the 1980s to the early 1990s
(Crimmins, Saito, & Reynolds, 1997; Manton, Corder, & Stallard, 1993; Wolf,
Mendes de Leon, & Glass, 2007). Two of these studies are based on
national data sets, including the Longitudinal Study of Aging (Crimmins et al.,
1997) and the National Long Term Care Survey (Manton et al., 1993); the
third is based on the New Haven Populations for Epidemiologic Studies of
the Elderly (Wolf et al., 2007). These studies provide conflicting information
on change in age-specific recovery rates, with the Crimmins et al. study
reporting some improvement in disability recovery rates from 1984-1986 to
1988-1990; the Manton et al. study reporting that all recovery rates were
lower in the 1984-1989 period than the 1982-1984 period; and Wolf et al.
finding decreases in recovery among older persons in New Haven from
1982-1994.
Most published studies of active life expectancy trends are based on
estimates from cross-sectional approaches. These studies have found
increases in disability-free life expectancy in the United States since the
1980s (Crimmins & Saito, 2001; Crimmins, Saito, & Ingeneri, 1997), as well
as in France (Robine, Mormiche, & Sermet, 1998), and longer life with less
85
severe disability in England (Bone, Bebbington, Jagger et al., 1995). Only
one study has been published using longitudinal data (Cai & Lubitz, 2007)
and it found increases in active life expectancy during the 1980s, but only for
males.
Hypotheses:
The aforementioned literature suggests the following research
hypotheses to be tested: 1) incidence of disability will be lower in the 1997
cohort than in the 1987 cohort, 2) recovery rates will be higher in the 1997
cohort, 3) death rates will be higher in the 1987 cohort, 4) the 1997 cohort
will have higher estimates of active life expectancy, 5) disabled life
expectancy will be shorter in the 1997 cohort, and 6) active life expectancy
will increase faster than disabled life expectancy over the time period. This
last hypothesis is the most important, as faster increase in active life
expectancy, relative to disabled life expectancy, is evidence of a
compression of morbidity.
This chapter examines trends in active life expectancy across two
cohorts of older people. The trends include transition rates for disability
onset, recovery, mortality, and estimates of active and disabled life
expectancy. In the first section, transition rates to and from disability are
presented. These rates, in conjunction with transitions to mortality, form the
basis of the second section, which contains estimates of the total, active, and
disabled life expectancies. The total life expectancy estimates are then
86
compared to national vital statistics. The third section is a summary of the
implied prevalence of disability, based on recent onset rates. The fourth and
final section is a description of status-based life tables showing how total,
active, and disabled life expectancies vary depending on disability status at
age 70.
Transitions between disability states
The transition probabilities were derived from a maximum likelihood
estimation of 4 equations, each depicting the age-specific effects of the
likelihood of the following four transition types: disability onset, recovery from
disability, death among the previously able, or death among the previously
unable. As shown in Table 4.1, the coefficients for age are positive for every
transition type except recovery, for which increased age results in a reduced
transition probability. The coefficient for onset in 1987 is slightly higher than
the coefficient in 1997. We see no difference in the age-effect for recovery in
both years, nor in death among the previously able. The age coefficient for
death among the previously unable is slightly higher in 1987, reflecting
perhaps a slight relative decline in the 1997 cohort.
87
Table 4.1. Disability Transition Equation Coefficients by Transition Type

Onset Recovery
1987 LSOA 1997 LSOA 1987 LSOA 1997 LSOA
Constant -12.780 -11.368 3.194 3.400
Age 0.137 0.120 -0.057 -0.057

Death Among Able Death Among Unable
1987 LSOA 1997 LSOA 1987 LSOA 1997 LSOA
Constant -9.486 -9.401 -3.153 -2.481
Age 0.093 0.091 0.032 0.023

The aforementioned equations form the basis of the graphs that will
be used to examine changes over time in the age-specific onset of disability
(Figure 4.1). First, the age pattern indicates, not surprisingly, that at both
dates, older persons are more likely than younger persons to become
disabled. The probability of disability onset is relatively similar for the 1987
and 1997 observation cohorts from age 70 to age 83. After age 83, however,
probability of disability onset is significantly lower in 1997 than it was in 1987.
This represents a delay in the onset of disability in the later decade,
suggesting that the more recent cohort is less likely to become disabled at
the oldest ages when compared to people at the same age a decade earlier.
88
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95
1 987
1 997
Figure 4.1. Probability of Becoming Disabled: LSOA I and LSOA II

89
The rates of recovering from disability – i.e. of changing to non-
disabled from disabled – in the two samples are shown in Figure 4.2. Again,
not surprisingly, the age pattern is negative and opposite to the onset
pattern, older persons are less likely to recover. Above age 72, we observe
that there is a significantly greater likelihood of recovery in the 1997 cohort
than in the 1987 cohort. Thus, it appears that the probabilities of both
disability onset and recovery improved over the 10-year period.
Figure 4.2. Probability of Recovery from Disability: LSOA I and LSOA II

0.00
0.10
0.20
0.30
0.40
0.50
70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95
1987
1997
90
Figure 4.3 shows that the probability of dying from the disabled state
is considerably higher than that of dying from the non-disabled state at both
dates. However, in contrast to the changes in disability incidence and
recovery, the probability of death from either state does not differ significantly
between the two surveys (Figure 4.3). Although not significantly different, the
death rates appear somewhat lower among the disabled in the more recent
cohort.
Figure 4.3. Death Rates for the Non-Disabled and Disabled

0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95
1987
1997
1987
1997
Disabled
Non-disabled
91
Life Expectancy
Total life expectancy, disability-free life expectancy, and disabled life
expectancy implied by the transitions are shown in Table 4.2. In 1997, at
age 70, life expectancy was 14.43; non-disabled life expectancy was 11.76;
and disabled life expectancy was 2.67. Both total life expectancy and
disability-free life expectancy were shorter in the earlier cohort at age 70.
Note that while the mortality rates did not differ significantly for the two
cohorts, the change in total life expectancy at age 70 was significantly
different (a function of interacting disability and mortality processes in the life
table model), increasing by three-fourths of a year (from 13.66 to 14.43).
Disability-free life expectancy increased by almost a full year (from 10.87 to
11.76). However, there is no significant change over the 10-year period in
life expectancy with disability (from 2.79 to 2.67).
One other multistate analysis of life expectancy without ADL/IADL
impairment based on LSOA I produced a similar estimate of ADL/IADL
disabled life expectancy at age 70 (2.88 versus 2.79) but quite different
estimates of total life expectancy (12.16 versus 13.66) (Crimmins et al.,
1994). The use of the IMaCH method produces the longer estimates of total
life expectancy because it is based on a logistic model and the more
traditional multi-state life table uses central death rates. The earlier method
of estimation could not accommodate different interval lengths, so many data
points used in this analysis were not used in that analysis.

92
Table 4.2. Total, Disability-free, and Disabled Life Expectancy (with confidence intervals): LSOA I and LSOA II

Age Total Life Expectancy Disability-free Life Expectancy Disabled Life
Expectancy
1987 1997 1987 1997 1987 1997
70 13.66
(13.35-13.97)
14.43
(14.09-14.77)
10.87
(10.59-11.14)
11.76
(11.46-12.05)
2.79
(2.64-2.95)
2.67
(2.52-2.82)
80 8.18
(7.92- 8.45)
9.12
(8.80-9.43)
5.15
(4.96-5.34)
6.34
(6.10-6.57)
3.04
(2.85-3.22)
2.78
(2.60-2.96)
90 4.82
(4.51-5.13)
5.76
(5.38-6.13)
1.69
(1.50-1.89)
2.92
(2.67-3.18)
3.13
(2.84-3.42)
2.83
(2.55-3.11)

93
Total life expectancy and disability-free life expectancy are also
significantly longer at ages 80 and 90 for the 1997 cohort than for the 1987
cohort. Increases in the length of life without disability are even greater at
theses ages, exceeding a year. The proportion of expected life free of
disability also increased between the two dates, from 80 to 82 percent at age
70, from 62 to 70 percent at age 80, and from 35 to 51 percent at age 90.
Implied Prevalence
The prevalence of ADL/IADL disability in the stationary population implied by
continuing the rates of disability onset, recovery, and death is shown in
Figure 4.4. This population health structure is independent of initial health
state prevalence and age structure and provides an indication of the long-
term impact of continuation of current processes. The implication of these
rates is that further reduction in disability is predicted, especially at the older
ages. Another way of understanding the long-term impact of continuing the
rates is to examine the shift in the age at which a given level of prevalence
would occur. For example, if the rates for the 1987 cohort were continued, a
disability prevalence of 40% would occur at roughly age 86. However, if the
rates for the 1997 cohort were continued, a prevalence of 40% would not be
achieved until age 92 – a change of six years. Although the age change
depends on the level of prevalence chosen, the key point is that the shift in
the incidence rates implies that there would be substantively meaningful
declines in disability at advanced ages.
94
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95
1987
1997
Figure 4.4. Implied Prevalence of Disability

Status-based Life Tables
To this point, the focus has been on population-based life tables, which
represent the expected average length of life by disability state for the entire
population. This section investigates differences in active life expectancy
observed for individuals of different prior disability status.
As shown in Table 4.3, for those free of disability at age 70, life
expectancy increased by 0.6 year (13.9 to 14.5), and almost all of this
increase (0.5 year) was in disability-free life expectancy. These levels of life
expectancy are fairly similar to those for the entire population. On the other
hand, people who are disabled at age 70 have lower life expectancy and
more disabled and less disability-free life expectancy that those without
95
disability. At age 70, the disabled gained a year in life expectancy between
the two studies (from 9.4 to 10.4) and they gained even more (1.2 years) in
non-disabled life expectancy. While not a significant change, this group also
has experienced a decrease in disabled life expectancy. These results point
to greater improvement over this period in length of life free of disability
among those who arrive in old age with disability.

96
Table 4.3: Status Based Health Expectancies at Age 70 by Year and Initial Health Status

Initial Disability State at age 70
Disability-free Disabled
1987 1997 1987 1997
Total Life Expectancy 13.9(13.57-14.19) 14.5(14.15-14.81) 9.4(8.69-10.15) 10.4(9.75-11.11)
Disability-free Life
Expectancy 11.2(10.93-11.43) 11.7(11.39-11.91) 4.8(4.13-5.41) 6.0(5.41-6.59)
Disabled Life
Expectancy 2.7 (2.56 - 2.85) 2.8 (2.67 - 2.99) 4.7(4.31-4.99) 4.4(4.15-4.71)

97
SUMMARY AND CONCLUSIONS
In conclusion, the first hypothesis, which stated that the incidence of
disability will be lower in the 1997 cohort than in the 1987 cohort, was
supported by data analysis at the older ages, since the onset of disability was
significantly lower in the 1997 cohort after age 83. The second hypothesis,
that recovery rates will be higher in the 1997 cohort, was also supported, as
recovery was lower after age 73. The third hypothesis, that death rates will
be higher in the 1987 cohort, was not supported, as no significant differences
were observed. The fourth hypothesis, stating that the 1997 cohort will have
higher estimates of active life expectancy, was supported by the data. The
gain in active life expectancy was about 1 year at age 70, about 1.2 years at
age 80, and 1.3 years at age 90. The fifth hypothesis, that disabled life
expectancy will be shorter in the 1997 cohort, was not supported, as there
was no change in disabled life expectancy over this time. The sixth
hypothesis, predicting that active life expectancy will increase faster than
disabled life expectancy over the time period, was supported. At all ages,
the number of years lived in an active state increased, while the number of
years lived as disabled remained unchanged. It thus appears that the gains
in life expectancy observed over this time we contained fully in active years.
The decline in the rate of onset, and increase in the rate of recovery,
together combined to create a predicted decline in the future prevalence of
disability among those ages 82 and above, assuming that the current rates of
98
onset and recovery continued. Furthermore, while life expectancy overall
increased in the population, it was not because of decreases in mortality
among either the disabled or the non-disabled at a given age; but because of
a shift of the population into the lower mortality non-disabled group. This
indicates not only the importance of disability processes in determining life
expectancy free of disability but also the importance of preventing and
delaying disability, as well as promoting recovery from disability, in
determining the prevalence of disability in the population.
While both processes appear to have changed, it is not possible to
do more than speculate about the reasons for change. The decrease in
onset of disability could be linked to the fact that diseases have become
less disabling over time (Cutler, 2001; Crimmins et al., 1999). Recent
cohorts may arrive having reached a given age with physical damage that
would result in less likelihood of ADL/IADL disability (Soldo et al., 2006) and
this could result from the fact that they have had a lifetime advantage
resulting from more education and less physically demanding jobs, although
a recent analysis showed little effect of improving education on reduction in
ADL disability (Freedman et al., 2004b). It is also possible that changes in
housing environments result in less environmental challenge in newer
cohorts of older persons. Technological and market changes may offer
better opportunities for self-maintenance in more recent years, thus
reducing the disability associated with functional loss. Increasing recovery
99
from disability may also reflect change in housing conditions, if people have
become more likely to move in order to ameliorate a disability. Direct
attempts to remediate disability may also have improved. For instance, it is
also possible that changes in the ability to treat stroke have resulted in
better return to functioning for people who have an acute cerebrovascular
event, or that greater use of therapy has enabled people to better regain the
ability to provide self-care.
Our findings repeat those of earlier analysis of LSOA I in that a trend
toward increased recovery is found (Crimmins et al., 1997). One of the
reasons that trends based on the LSOA may differ from those based on
analysis of the NLTCS is that disability in the NLTCS is limited to that which
has lasted, or is expected to last, 90 days. It is possible that eliminating
disability lasting less than three months modifies the results of trend
analysis, as recovery is more likely from short term disability (Crimmins and
Saito, 1993).
We cannot compare the years of disability-free life estimated here to
those developed using different definitions of disability-free life expectancy
and cross-sectional methods; there is no reason to expect the numbers to be
identical with different definitions of disability and different methodological
assumptions. However, the conclusion that most of the recent increase in
life expectancy at the older ages has been concentrated at the older years is
the same using both approaches. The richness of the multi-state method is
100
obvious, however, because one is able to clarify the processes underlying
this conclusion: a decrease in the rate or disability onset and an increase in
the rate of recovery.
101
Chapter V: Changes in Life Expectancy and Active Life Expectancy by Sex
between 1984 and 2000
This chapter examines changes in active life expectancy for males
and females across two cohorts of older people. The changes include
transition rates for disability onset, recovery, and mortality among the
disabled and non-disabled for each sex, as well as estimates of active and
disabled life expectancy for each sex.
Gender is an important factor to consider in investigations of disability and
active life expectancy because of the pattern of differences in mortality and
morbidity between men and women. Women clearly report

a higher
prevalence of disability (Crimmins & Saito, 2001; Fried et al.,

1994; Merrill,
Seeman, Kasl, and Berkman, 1997; Verbrugge, 1985;Verbrugge, 1989;
Wingard, 1984), as well as lower mortality rates than men at all ages (Fried
et al.,

1994; Merrill, Seeman, Kasl, & Berkman, 1997; Verbrugge, 1985,

Verbrugge, 1989; Wingard, 1984). In studies of active life expectancy, one
consistently finds that women have both a longer life expectancy and a
longer expected life with disability (Crimmins & Saito, 2001; Laditka &
Laditka, 2002; Waidman & Liu, 2000).
Gender Differences in Disability Incidence
Evidence of gender differences in disability onset is mixed. Some
studies find no clear gender difference in incidence (Crimmins, Hayward, &
Saito, 1996; Oman, Reed, & Ferrara, 1999; Wolf et al., 2007). Others have
102
found higher incidence of disability among females (Leveille et al., 2000).
Age-specific disability onset rates were estimated to be about 3.5% lower for
males in one study, resulting in about a 5-year lag for men in incidence rates
so that women’s onset is similar to men 5 years older (Leveille et al., 2000).
Some research has shown that the gender differences decrease after health
and demographic characteristics are controlled (Guralnik & Kaplan, 1989;
Lawrence & Jette, 1996).
Trends and Incidence
Wolf et al. (2007), studying the time coefficient of transition equations,
found similar time coefficients for disability incidence among males and
females, suggesting no difference in the rate of change during the 1980s
between males and females. While the probability of incidence may be
similar, studies of physical decline in both sexes have shown steeper
declines for females (Beckett, 1996; Crimmins, Saito, & Reynolds, 1997;
Taylor, 2005), and females report longer duration of disability (Branch et al.,
1991).
Gender Differences in Recovery
Studies appear to show limited gender differences in age-specific
rates of recovery. Leveille et al. (2000) found higher recovery rates among
males aged 90 and above. Another study, by Wolf et al. (2007), found no
age-specific differences between males and females in recovery rates
through the 1980s. Looking at general recovery rate differences that are not
103
age-specific, Crimmins et al. (1996) found that females are 88% less likely to
experience a recovery from severe disability to no functioning problems than
males. If the threshold is set to recovery from some function problems, a
less strict definition, women are only 17% less likely than males to recover to
a state of having no problems.
Trends in Recovery
Crimmins and colleagues (1997a) found general improvement in recovery
rates from 1984-1986 to 1988-1990, while the Manton and colleagues study
(1997) reported that all recovery rates were lower in the 1984-1989 period
than in the 1982-1984 period. Wolf et al. (2007), using a non-representative
sample, also found declines in recovery for both men and women during the
1980s.
Mortality by gender
Mortality has declined throughout recent decades for older males, who
show consistent declines in absolute terms dying at each age, while mortality
rates for older females appear to be leveling off (NCHS mortality data, 1998).
Crimmins et al. (1996) found non-disabled females to be 50% less likely to
die relative to non-disabled males between 1984 and 1990, and disabled
females 24% less likely to die relative to disabled males.
Gender Differences in Active Life Expectancy
Some studies have found women have longer active lives (Crimmins et
al., 1997a; Manton & Land, 2000; Reynolds et al., 2005) while others find
104
that men have longer active lives (Cai & Lubitz, 2007), and others have found
no particular difference (Crimmins et al., 1996). There does seem to be
consensus that females live more years as disabled than males do (Cai &
Lubitz, 2007; Crimmins et al., 1996; Crimmins et al., 1997a; Manton & Land,
2000; Reynolds et al., 2005).
Trends in Active Life Expectancy
Crimmins et al. (1997a) found an increase between 1970 and 1980 in
disabled life expectancy, and also found there to be a generally stable
proportion of disabled life between 1980 and 1990. Cai and Lubitz found a
continued pattern of stable disabled life expectancies through the 1990s (Cai
& Lubitz, 2007). General life expectancy gains have been greater for older
men than for women. For both men and women, the increase in life
expectancy during the 1990s appears to be accounted for by an increase in
ALE.
The results in this chapter can be divided into 4 sections. In the first
section, the transition rates to and from disability and mortality for both men
and women are presented. Differences between men and women are also
presented. These rates form the basis of the second section, which reports
estimates of the total, active, and disabled life expectancies. The third
section is a summary of the implied prevalence of disability, based on recent
onset rates. The fourth section is a description of status-based life tables
105
showing how total, active, and disabled life expectancies vary depending on
initial disability status.
Hypotheses
The previous literature suggests that the current analysis of gender
differences in active life expectancy, as well as gender differences in trends
in active life expectancy, can test the following hypotheses: 1) females will
have higher disability incidence than males during the study period, 2)
declines in incidence will be greater for females, 3) females will have lower
recovery rates than males, 4) the rate of recovery increase will be higher for
males, 5) death rates will be higher for males in both the disabled and able
states, 6) the trend in death rates will lead to faster decline for males, 7)
females will have higher estimates of active life expectancy, 8) males will
experience larger gains in active life expectancy, and 9) females will have
higher estimates of disabled life expectancy. While little literature exists to
suggest trends in improvement (or decline) over time, the following
hypotheses are made: 10) males will show greater improvement during the
study period, and 11) males will experience a faster increase in active life
expectancy.
Changes in Transition Rates
Table 5.1 provides the distribution of the almost 16,000 intervals of
observation by sex and status at the beginning and end of the survey
intervals. Most intervals at both times for both men and women began and
106
ended disability-free. As shown in Table 5.1, for both male and female
respondents to the LSOA I, approximately 63% of the time intervals began
and ended disability-free. In the LSOA II, the percentage was higher for men
(68.7%) than for women (65.0%), and both genders reported a higher
percentage of respondents disability-free at two consecutive time points in
the later survey. Among males, the percent disabled dropped at two
consecutive interviews, (8.9% to 7.9%), while there was essentially no
change among females (10.9% to 10.2%). One of the largest differences
between the sexes appears to be in the proportion that experiences the
onset of disability between two consecutive time points. Among males, there
was a drop from 13.3% to 11.8%, while among females, there was an
increase from 6.5% to 7.3%. Thus females had lower onset, but the trend for
women was toward rising onset rates, while the opposite was true for males.
Table 5.1. Distribution of Disability-state Transitions for Males and
Females Between Observation Intervals

Males
States at Beginning and End LSOA I LSOA II
NPercentN Percent
Disability-free at both
interviews 356563.14420 68.7
Disabled at both interviews 503 8.9 510 7.9
Became Disabled 750 13.3 757 11.8
Recovered from Disability 107 1.9 135 2.1
Died From Disability- free State 382 6.8 288 4.5
Died From Disabled State 340 6.0 322 5.0
Total 5647 6432

107
Table 5.1, Continued
Females
States at Beginning and End LSOA I LSOA II
NPercentN Percent
Disability-free at both
interviews 609063.26536 65.0
Disabled at both interviews 1050 10.9 1029 10.2
Became Disabled 628 6.5 736 7.3
Recovered from Disability 244 2.5 319 3.2
Died From Disability- free State 1033 10.7 911 9.1
Died From Disabled State 586 6.1 518 5.2
Total 963110049

Recovery rates were higher for female, with the difference increasing
from 2.5% to 3.2%, as compared to an increase of 1.9% to 2.1% for males.
There was a higher proportion of transitions to recovery for females in the
LSOA II (3.2%) as compared to a decade earlier (2.5%), although there was
no apparent change for males, who appeared to have lower recovery rates
overall, at about 2% of the transitions in both surveys.
Turning to transitions to death, the proportion dying among the disabled
appear to be similar for males and females, about 6% in the LSOA I and 5%
in the LSOA II. Among the disability-free, however, we see higher
proportions among females, 10.7% to 9.1%, compared to 6.8% to 4.5% for
males.
The previous description focused on general time and sex trends,
without controlling for age. This next section delves into more detail about
trends in disability onset, recovery, and mortality in the two cohorts by
examining age-specific rates.
108
The transition probabilities were derived from a maximum likelihood
estimation of four equations, each depicting the age- and gender-specific
effects of the likelihood of the given transition (onset, recovery, death among
the able, or death among the disabled). As shown in Table 5.2, the
coefficients for age are positive for every transition type except for recovery.
The coefficient for females (as compared to males) is positive for onset in
1987, but negative in 1997, showing no stable relationship between onset
and gender over time. A similar story is observed for recovery, where the
coefficient for females is negative in 1987 and positive in 1997. There is a
consistent gender association with mortality, however, as females have lower
mortality in both years, from either the able or disabled state.
Table 5.2. Disability Transition Equation Coefficients By Transition
Type
Onset Recovery
1987 LSOA 1997 LSOA 1987 LSOA 1997 LSOA
Constant -12.798 -11.072 3.243 3.337
Age 0.136 0.119 -0.057 -0.057
Female 0.137 -0.232 -0.078 0.052

Death Among Able Death Among Unable
1987 LSOA 1997 LSOA 1987 LSOA 1997 LSOA
Constant -9.463 -9.360 -3.134 -2.594
Age 0.098 0.092 0.036 0.025
Female -0.773 -0.150 -0.461 -0.042

The next section is a graphical representation based on the equations
just described. Figure 5.1 shows differences in the incidence of disability
between males and females in 1987 and 1997. First, the age pattern
109
indicates, not surprisingly, that at both dates, older persons were more likely
than younger persons to become disabled. Next, it is clear in the first figure
that females experienced higher rates of onset than males at ages 80 and
above in the 1987 data. The gender difference in onset is not apparent in the
1997 data, however.
Figure 5.1. Probability of Becoming Disabled for the 1987 and 1997
Cohorts: Males and Females
1987
0
0.1
0.2
0.3
0.4
0.5
0.6
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
Male
Female

110
Figure 5.1, Continued
1997
0
0.1
0.2
0.3
0.4
0.5
0.6
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
Male
Female

Figure 5.2 shows the same data, although arranged to compare time
trends (separate by gender) for each sex. There appears to be a slight
increase in onset of disability for males in the later 1997 cohort between ages
74 and 85, but the change is very slight (see Figure 5.2). For females, the
rate of onset dropped at ages 90 and above.
111
Figure 5.2. Probability of Becoming Disabled for Males and Females:
LSOA I and LSOA II

Males
0
0.1
0.2
0.3
0.4
0.5
0.6
70
72
74
76
78
80
82
84
86
88
90
92
94
1987
1997

Females
0
0.1
0.2
0.3
0.4
0.5
0.6
70
72
74
76
78
80
82
84
86
88
90
92
94
1987
1997

112

The rates of recovery from disability, changing to non-disabled from
disabled, in the two samples are shown in Figure 5.3. Again, not
surprisingly, the age pattern was negative and opposite to the previous one;
older persons were less likely to recover. First, looking at sex differences in
Figure 5.3, we see that recovery rates were indistinguishable for males and
females in both 1987 and 1997 (Figure 5.3). Both sexes show exactly the
same shape, level, and trend in age-specific recovery rates.
Figure 5.3: Probability of Recovering from Disability for the 1987 and
1997 Cohorts: Males and Females

1987
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
Male
Female
113
Figure 5.3, Continued
1997
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
Male
Female

Next, looking at the time trend in recovery rates (Figure 5.4), there
appear to be no significant changes over time for either sex, although there is
some hint of a higher level of recovery in the 1997 survey at the younger
ages for both males and females (see Figure 5.4).
114
Figure 5.4: Probability of Recovery from Disability for Males and
Females: LSOA I and LSOA II

Males
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
70
72
74
76
78
80
82
84
86
88
90
92
94
1987
1997

Females
0.0
0.1
0.1
0.2
0.2
0.3
0.3
0.4
0.4
70
72
74
76
78
80
82
84
86
88
90
92
94
1987
1997

115
Transitions to death
Figures 5.5 through 5.8 indicate the probability of dying from either a
disabled state or a non-disabled state. Note that the scales of the graphs
differ slightly, particularly on Figures 5.7 and 5.8 as compared to 5.5 and 5.6,
as the death rates for those previously active are lower (less than .07 at age
70) than for those previously disabled (death rates about 0.2 and above at
age 70). The advantage of prior active functioning will be discussed in more
detail in the status-based results section.
Looking at the differences between males and females, there is clear
evidence of higher death rates among males in 1987 than among females
(Figure 5.5 and 5.7) among both the previously non-disabled and previously
disabled. In the 1997 data, however, we see no sex difference in mortality
among either the active (Figure 5.6) nor the disabled (Figure 5.8).
116
Figure 5.5: Probability of Death Separated by Survey Year for the 1987
and 1997 Cohorts: Males and Females: Probability of Death among
Non-Disabled in the 1987 Survey

1987
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
Male
Female

Figure 5.6. Probability of Death among Non-Disabled in the 1997 Survey

1997
0
0.05
0.1
0.15
0.2
0.25
0.3
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
Male
Female

117
Figure 5.7. Probability of Death among Disabled in the 1987 Survey

1987
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
Male
Female

Figure 5.8. Probability of Death among Disabled in the 1997 Survey

1997
0
0.1
0.2
0.3
0.4
0.5
0.6
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
Male
Female
118

Among males, there appears to be a drop in the death rate at age 77
and above among those previously active (Figures 5.9 and 5.11). Among
males previously disabled, there appears to be a recent drop at ages 82 and
above. Among females, the patterns appear to be very similar for males and
females. Death from the non-disabled state decreased for men and
increased for women over this time (see Figure 5.10). These diverging
trends were also reflected in death from the disabled state, which declined
for men but not for women over this time.
Figure 5.9. Probability of Death Separated by Sex for Males and
Females from the Non-Disabled and Disabled States: A. Probability of
Death among Non-Disabled Males in Each Survey
0
0.1
0.2
0.3
0.4
70
72
74
76
78
80
82
84
86
88
90
92
94
1987
1997

119
Figure 5.10. Probability of Death among Non-Disabled Females in Each
Survey

0.0
0.1
0.2
0.3
70
72
74
76
78
80
82
84
86
88
90
92
94
1987
1997

Figure 5.11. Probability of Death among Disabled Males in Each Survey
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
70
72
74
76
78
80
82
84
86
88
90
92
94
1987
1997

120
Figure 5.12. Probability of Death among Disabled Females in Each
Survey
0
0.1
0.2
0.3
0.4
0.5
70
72
74
76
78
80
82
84
86
88
90
92
94
1987
1997
121

Life Expectancy
Increases in total life expectancy between the 1987 and the 1997 cohorts
were significant only for men. For example, at age 70 the increase in total
life expectancy was 1.02 years for males and an insignificant 0.31 years for
females. Furthermore, although increases in active life expectancy occurred
at most ages for both men and women, increases at age 70 were significant
only for men. Of similar interest is the proportion of life that is expected in an
active (non-disabled) state. There were no significant changes in disabled
life expectancy for either sex. As Table 5.3 shows, males at each age have
a higher proportion of expected active life than females. The change over
time for the sexes appears to be similar.

122
Table 5.3. Total, Active, and Disabled life expectancy (with confidence intervals) for Males and Females: LSOA
I and LSOA II

Males
Age Total Life Expectancy Active Life Expectancy Disabled Life Expectancy Proportion Active
1987 1997 1987 1997 1987 1997 1987 1997
70 11.60
(11.18-12.01)
12.62*
(12.18-13.07)
9.78
(9.40-10.15)
10.94 *
(10.53-11.35)
1.82
(1.64-2.00)
1.69
(1.52-1.85)
.84 .87
80 6.83
(6.51-7.15)
7.61*
(7.26-7.95)
4.73
(4.46-5.00)
5.85 *
(5.55-6.16)
2.10
(1.88-2.31)
1.75
(1.57-1.93)
.69 .77
90 4.02
(3.69-4.34)
4.53
(4.21-4.85)
1.68
(1.44-1.92)
2.72 *
(2.45-2.99)
2.34
(2.02-2.66)
1.81
(1.58-2.04)
.42 .60

Females
Age Total Life Expectancy Active Life Expectancy Disabled Life Expectancy Proportion Active
1987 1997 1987 1997 1987 1997 1987 1997
70 15.51
(15.07-15.95)
15.82
(15.36-16.28)
11.84
(11.48-
12.20)
12.40
(12.01-12.79)
3.67
(3.42-3.92)
3.42
(3.20-3.65)
.76 .78
80 9.37
(8.98-9.75)
10.00
(9.61-10.40)
5.58
(5.31-5.84)
6.60*
(6.31-6.89)
3.79
(3.50-4.08)
3..40
(3.16-3.65)
.60 .66
90 5.44
(4.98-5.89)
6.24
(5.81-6.66)
1.80
(1.56-2.05)
2.99*
(2.70-3.27)
3.63
(2.89-3.87)
3.25
(2.92-3.57)
.33 .48

123
Implied Prevalence
The prevalence of disability in the stationary population implied by
continuing the rates of disability onset, recovery, and death is shown in
Figure 5.13. The implication of these rates is that reduction in disability is
predicted at the older ages, with a dramatic decline apparent for males
beginning at ages 78 and above. For example, at age 85, males in the 1987
cohort had an implied prevalence of 27%, while in the 1997 cohort that was
reduced to 20%. By age 90, the difference changes from 43% in 1987 to
30% in 1997. Females also show a decline in implied prevalence in 1997
beginning at age 83. While the decline is not as apparent as for males, the
trend appears to parallel the declines observed for males at the later ages.
For example, if the 1987 cohort rates were continued, a 40% prevalence rate
would occur at roughly age 88 for females, and if the rates for the 1997
cohort were continued, a 40% prevalence would not be achieved until age 90
for females.
124
Figure 5.13. Implied Prevalence of Disability by Sex

0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
70
73
76
79
82
85
88
91
94
1987 Male
1997 Male
1987 Female
1997 Female
125
Status-based Life Tables
To this point, the focus has been on population-based life tables which
represent the expected average life for the entire population. In the following
paragraphs, life tables representing prior health status will be presented.
As shown in Table 5.4, for males classified as active at age 70, life
expectancy increased by 1.1 years (11.77 to 12.84 years) with essentially all
of the increase in disability-free life expectancy. For females classified as
active at age 70, life expectancy increased by 0.4 years (15.63 to 16.09
years) and again, this increase corresponds to an increase in active life
expectancy (although the difference across the decade is not significant).
For those who reach age 70 with a disability, the data indicate a likelihood
for fewer remaining years of life. Further, those remaining years are more
likely to be spent in a disabled state. At age 70, there was no significant
change in life expectancy for disabled males, although they appeared to
experience gains in active life expectancy. However, those gains were not
statistically significant. For disabled females, the apparent increase in total
life expectancy is not significant; however, the increase in active life
expectancy is significant. These results point to greater improvement of
active life among those who arrive in old age with or without a disability. In
addition, disabled females appear to have made small gains in active life
expectancy.

126
Table 5.4. Status Based Health Expectancies at age 70 by year and Initial Health Status

Males

Active Disabled

1987 1997 1987 1997
Total Life Expectancy 11.77 (11.36-12.18) 12.84 (12.40-13.28)* 7.69 (6.91-8.47) 8.30 (7.52-9.07)
Active Life Expectancy 10.03 (9.68-10.38) 11.23 (10.84-11.62)* 3.96 (3.25-4.67) 5.10 (4.41-5.80)
Disabled Life Expectancy 1.74 (1.58-1.90) 1.61 (1.46-1.76) 3.73 (3.36-4.10) 3.20 (2.91-3.49)

Females
Active Disabled

1987 1997 1987 1997
Total Life Expectancy 15.63 (15.21-16.04) 16.09 (15.64-16.55) 10.82 (9.94-11.70) 12.09 (11.28-12.89)
Active Life Expectancy 12.13 (11.81-12.46) 12.80 (12.44-13.16) 5.44 (4.66-6.21) 6.92 (6.24-7.60)*
Disabled Life Expectancy 3.49 (3.26-3.72) 3.29 (3.08-3.51) 5.38 (4.96-5.81) 5.17 (4.80-5.54)

127

Conclusions
In conclusion, the first hypothesis, that females will have high levels of
disability incidence, was supported in the 1987 cohort at ages 81 and above.
Interestingly, there were no gender differences in disability incidence in 1997.
The second hypothesis, that the change in incidence will be larger declines
for females was supported, as females declined, albeit only at ages 90 and
above. Males, in fact, had higher incidence of disability in 1997. The third
hypothesis, that females will have lower recovery rates than males, was not
supported, as no gender differences in recovery rates were found. The
fourth hypothesis, that the rate of recovery increase is higher for males, was
not supported, as no significant differences in recovery rates were found, and
point estimates suggest that females experience higher age-specific
recovery. The fifth hypothesis, that death rates are higher for males in both
the non-disabled and disabled states, was supported in 1987 at every age,
but not in 1997, where no gender differences could be found. The sixth
hypothesis, that the trend in death rates will lead to faster decline for males,
was supported in both years. Interestingly, female death rates increased
among those in the active state in 1997, at the same time that their male
counterparts were experiencing lower death rates than a decade earlier. The
death rates for disabled females were stagnant, at the same time that their
male counterparts were declining. This divergence in trends for the two

128
sexes is likely driving the observed improvements in total and active life
expectancy for males, while we observe largely stagnant trends for females
of the same age. The seventh hypothesis, that females will have higher
average active life expectancy than males, was supported, as females in
both 1987 and 1997 reported more years of remaining active life than males.
The eighth hypothesis, that males will experience larger relative gains
in active life expectancy, was supported, as the difference in active life
expectancy at age 70 was about 2 years active in 1987, and 1.5 years active
in 1997. This is also observed in the slight difference in the proportion of life
remaining active, where a growth of 3% was observed for males, and only
2% for females. The change over time at ages above 70 was not significant.
The ninth hypothesis, that females will have higher estimates of disabled
life expectancy than males, was supported, as females appear to survive
more years with disability (about 3.5 years on average) compared to males,
who can expect about 2 years lived with disability.
The analysis provides clear evidence that in the decade from 1987 to
1997, life expectancy for older males has increased, while it has been
stagnant for females. Males have experienced across the board increases in
active life expectancy, while for females, increases in active life expectancy
are only significant at ages 80 and above. Females have longer active and
disabled life expectancy at each age, but also have a smaller proportion of
remaining life expected active.

129
Males experienced a slight increase in the onset of disability, while
females at the oldest ages experienced a slight decline in the onset of
disability. In both cohorts, there are no significant differences between males
and females in the recovery from disability.
Mortality rates were higher at every age for both disabled and non-
disabled males than for comparable females in the 1987 cohort, but the
gender difference in age-specific mortality disappeared in the 1997 cohort.
This change is driven by declines in male mortality, as female mortality
appears to be the same in both the 1987 and 1997 cohorts.
An examination of status-based results show increases in active life
expectancy at age 70 between 1987 and 1997 for active males, but no
change for disabled males. For females, an increase in active life
expectancy is observed only for disabled individuals.
Implications
The increased survival of males, combined with the increased length
of active life expectancy, can have an important impact on the daily lives of
both males and females. The data suggest that older females may spend
fewer years in widowhood, and have more years in which they do not need to
care for a disabled husband or male cohabitant. Lower disability for males
could reduce some of the stress observed among female caregivers, leading
to greater fulfillment in late life relationships.

130
As increases in male life expectancy continue, there will be more older
males in the older population in the future, which might mean either that
more males will live alone or in institutions, or that perhaps the increased
survival among males is driven by married males, who have lower mortality
rates than unmarried males (Avlund, Damsgaard, and Holstein, 1998).
The results suggest that the increase in life expectancy and active life
expectancy observed in the older population is driven by changes among
males. As more men are surviving to older ages, their generally smaller
proportion of life spent disabled is lowering the observation of average
changes over time. The results did not show any positive changes for
females over this time, suggesting that there may not be any systematic
improvement in active life expectancy, at least not one that affects both
genders equally.
The reason why little change is observed for females may be
complicated by a potential ceiling effect, where females may be closer to
some sort of natural limit in low mortality, while males may have more room
for improvement. James Fries proposed that the population was
approaching some sort of limit in life expectancy, and that limit is an
important part of the compression of morbidity hypothesis. While limits to
population life expectancy have not been proven in any study, and maximum
population life expectancy has continued to increase linearly for the past 160
years (Oeppen & Vaupel, 2002), biologists would argue that, barring some

131
sort of dramatic change in medicine, there is a natural limit to life expectancy.
It is possible, and in fact likely, that life expectancy will eventually stagnate
over time.

132
Chapter VI: Changes in Active Life Expectancy by Race between 1984-2000

This chapter examines trends in active life expectancy for blacks and
whites across two cohorts of older people in order to examine whether
changes in active life expectancy are occurring differently for each race. The
trends include transition rates for disability onset, recovery, and mortality for
each race, as well as estimates of active and disabled life expectancy for
each race.
Background
A variety of studies have found older blacks to have a higher prevalence
of disability than whites (Bound, Schoenbaum, & Waidmann, 1995; Clark,
1997; Hayward & Heron, 1999; Manton & Gu, 2001), resulting in lower active
life expectancy (Crimmins & Saito, 2001). Prevalence, of course, takes into
account all historical cases of disability affecting a population, and may not
accurately indicate recent health changes. Researchers attempting to
understand recent changes in health in the U.S. population have found
different patterns depending on the measures and population studied.
According to Crimmins & Saito (2001), differentials in disability onset have
widened between 1970 and 1990. In contrast, Margellos, Silva, and
Whitman (2004) found a decline in disparities in several health measures,
including stroke mortality. Furthermore, with advancing age, racial
differences in disability may decline (Johnson, 2000; Mendes de Leon et al.,
2005).

133
Another issue is whether race differences in disability are larger for men
or women, as women generally report higher levels of disability (Leveille et
al., 2000; Crimmins et al., 1996), as do blacks (Clark, 1997; Hayward &
Heron, 1999; Manton & Gu, 2001). Differences in chronic conditions like
diabetes are larger between races among women (Robbins et al., 2000) than
among men. Measures of physical functioning in the National Health and
Nutrition Examination Survey also show larger differences between black
women and white women than are found between black men and white men
(Ostchega et al., 2000).
Active Life.
While total life expectancy at older ages is similar for whites and blacks, a
study representative of the population 70 years old and over between 1984
and 1990 found large differences by race in active life expectancy such that
blacks have lower active life expectancy than whites of the same education
level (Crimmins et al., 1996). Earlier studies, based on a sample living in
North Carolina, found higher life expectancy and active life expectancy
among blacks than whites at ages 75 and higher (Guralnik et al., 1993; Land
et al., 1994).
Hypotheses
The previous literature suggests that the current analysis of race
differences in active life expectancy, and of race differences in trends in
active life expectancy, can test the following hypotheses: 1) blacks will have
higher disability incidence than whites during the study period, 2) declines in

134
incidence will be larger for whites, 3) blacks will have lower recovery rates
than whites, 4) the rate of recovery increase will be higher for whites, 5)
death rates will be higher for blacks in both the disabled and able states, 6)
the trend in death rates will lead to faster decline for blacks, 7) whites will
have higher estimates of active life expectancy, 8) whites will experience
larger gains in active life expectancy, and 9) blacks will have higher
estimates of disabled life expectancy.
The first section of this chapter discusses transition rates to and from
disability for each race. These rates, in conjunction with transitions to
mortality, form the basis of the second section, which contains estimates of
the total, active, and disabled life expectancies. The third section is a
summary of the implied prevalence of disability, based on recent onset rates.
The final section is a description of status-based life tables showing how
total, active, and disabled life expectancies vary depending on initial disability
status.
Transitions between disability states
The transition probabilities were derived from a maximum likelihood
estimation of four equations, each depicting the age-specific and gender-
specific effects of the likelihood of the given transition (onset, recovery, death
among the able, or death among the disabled). As shown in Tables 6.1 and
6.2, the coefficients for age are positive for every transition type except
recovery. The coefficient for blacks (as compared to whites) is positive for
onset in 1987, but negative for onset in 1997, showing an apparent change in

135
the relationship between onset and race over time. When gender is added to
the equation (see Table 6.2), we see that blacks had a higher probability of
onset in both years, but also showed a higher probability of recovery.
Without accounting for gender in the equation (Table 6.1), a less clear picture
is observed with lower recovery in 1987 and higher recovery in 1997.
Turning to the equations predicting the probability of mortality,
excluding the potential impact of gender (Table 6.1), we observe higher
mortality among previously able blacks as compared to whites, and higher
mortality among the disabled in 1997 for blacks (with a small inverse
coefficient in 1987 for blacks). After controlling for sex, the size of the race
coefficients predicting mortality among the previously able drops, although it
remains slightly higher for blacks (Table 6.2). Among the previously
disabled, the probability of mortality was higher for blacks than whites in
1987, and was essentially no different from whites in 1997.
Table 6.1. Disability Transition Equation Coefficients for Race and Age
Onset Recovery
1987 LSOA 1997 LSOA 1987 LSOA 1997 LSOA
Constant -12.572 -10.767 2.484 0.605
Age 0.139 0.114 -0.051 -0.016
Black -0.455 -.336 0.303 0.027

Death Able Death Unable
1987 LSOA 1997 LSOA 1987 LSOA 1997 LSOA
Constant -9.071 -9.209 -3.147 -3.037
Age 0.088 0.091 0.029 0.032
Black -0.113 -0.176 0.255 -0.020

136
Table 6.2. Disability Transition Equation Coefficients for Race, Sex, and
Age
Onset Recovery
1987 LSOA 1997 LSOA 1987 LSOA 1997 LSOA
Constant -12.995 -11.152 3.308 3.299
Age 0.138 0.118 -0.057 -0.054
Black 0.136 0.265 -0.081 -0.257
Female 0.447 -.225 -0.286 0.051

Death Able Death Unable
1987 LSOA 1997 LSOA 1987 LSOA 1997 LSOA
Constant -9.410 -9.290 -3.100 -2.690
Age 0.097 0.095 0.036 0.033
Black -0.774 -0.511 -0.461 -0.759
Female 0.016 -0.168 -0.254 -0.053

137
Tables 6.3 through 6.8 show the distribution of the transitions between
two consecutive time points in the data by race as well as by race and
gender. First, racial differences will be discussed, then change over time
within each race group, and within each race-sex group. To begin, we
observe that about 60% of the transitions for blacks (Table 6.3) were in the
“remained healthy” at two time points category, while for whites (Table 6.4),
about two-thirds of the transitions fit this category (referring to 1997). We
observed a slightly higher proportion of blacks in the “became disabled”
category, and a higher number who remained disabled (13.0% and 8.9%
respectively). For whites, the majority of death occurred among those
previously healthy, while for blacks it was more common to die from the
previously disabled state.
Turning to changes over time, we see an increase in the proportion of
transitions that remain healthy among whites (Table 6.4) over this time period
(63.1 to 67.0%), while for blacks (Table 6.3), there is no change in this
category. Both whites and blacks reported a decline in the proportion of
transitions to disability in the later dataset, as well as a slight increase in the
number of transitions to recovery from disability, with a larger relative
increase for blacks, who have the highest proportion of recovery transitions.
Both races experienced a decline in the proportion that remained disabled at
two time points, but blacks were still disproportionately represented in this
category, even in the later dataset. The proportion of transitions to death

138
from the non-disabled state and the disabled state increased for blacks over
the decade, while they decreased for whites.
Further dividing the sample by both race and sex, we can see that
black males (Table 6.5) are the only group to experience a decline in the
proportion of transitions that remain healthy, as all other groups improved
over this time. Black females (Table 6.6) have the lowest representation in
this category (58.7% in 1997), dramatically lower than white males (69.5% in
1997 in Table 6.7) who had the highest representation. All race-sex groups
experienced a decline in the proportion that became disabled, with higher
proportions for females than males in both races.

139
Table 6.3. Number of Health Transitions by Race for Adults Aged 70 and
Older: Blacks
1987 LSOA 1997 LSOA II
N%N %
Remained Healthy 754 59.5 1026 59.5
Became Disabled 144 11.4 181 10.5
Died From Healthy State 87 6.9 156 9.0
Recovered from Disability 36 2.8 60 3.5
Remained Disabled 166 13.1 184 10.7
Died From Disabled State 80 6.3 117 6.8
Total 1267100%1724 100%

Table 6.4. Number of Health Transitions by Race for Adults Aged 70 and
Older: Whites
1987 LSOA 1997 LSOA II
N%N %
Remained Healthy 8979 63.1 9930 67.3
Became Disabled 1450 10.2 1358 9.2
Died From Healthy State 1342 9.4 1337 9.1
Recovered from Disability 320 2.2 394 2.7
Remained Disabled 1264 8.9 1015 6.9
Died From Disabled State 877 6.2 723 4.9
Total 14232100%14757 100%

Table 6.5. Number of Health Transitions by Race for Adults Aged 70 and
Older: Blacks Males
1987 LSOA 1997 LSOA II
N%N %
Remained Healthy 276 62.6 367 61.1
Became Disabled 46 10.4 57 9.5
Died From Healthy State 32 7.3 80 13.3
Recovered from Disability 8 1.8 19 3.2
Remained Disabled 53 12.0 45 7.5
Died From Disabled State 26 5.9 33 5.5
Total 441100%601 100%

140
Table 6.6. Number of Health Transitions by Race for Adults Aged 70 and
Older: Blacks Females
1987 LSOA 1997 LSOA II
N%N %
Remained Healthy 478 57.9 659 58.7
Became Disabled 98 11.9 124 11.0
Died From Healthy State 55 6.7 76 6.8
Recovered from Disability 28 3.4 41 3.7
Remained Disabled 113 13.7 139 12.4
Died From Disabled State 54 6.5 84 7.5
Total 826100%1123 100%

Table 6.7. Number of Health Transitions by Race for Adults Aged 70 and
Older: White Males
1987 LSOA 1997 LSOA II
N%N %
Remained Healthy 3317 62.8 4053 69.5
Became Disabled 468 8.9 453 7.8
Died From Healthy State 739 14.0 677 11.6
Recovered from Disability 99 1.9 116 2.0
Remained Disabled 334 6.3 243 4.2
Died From Disabled State 322 6.1 289 5.0
Total 5279100%5831 100%

Table 6.8. Number of Health Transitions by Race for Adults Aged 70 and
Older: White Females
1987 LSOA 1997 LSOA II
N%N %
Remained Healthy 5662 63.2 5877 66.8
Became Disabled 982 11.0 905 10.1
Died From Healthy State 603 6.7 660 6.4
Recovered from Disability 221 2.5 278 3.2
Remained Disabled 930 10.4 772 9.0
Died From Disabled State 555 6.2 434 4.5
Total 8953100%8926 100%

141

Recovery increased for black males, although the number of recovery
transitions was extremely small (8 in LSOA I and 19 in LSOA II). White
females also showed an increase in the percent of recovery transitions (2.5%
to 3.2%), while other groups remained approximately equal over time.
All race-sex groups showed declines in the proportion of transitions
that fit the remained disabled category. While all groups declined, there is a
clear relationship with gender as females in both race groups reported rates
that were higher than their male counterparts.
Transitions to death show divergent patterns for the race groups.
While white males and females demonstrated a decline in the proportion of
mortality transitions from either disabled or non-disabled states, black
females showed an increase in death from the disabled state (with
essentially no change in death from healthy state), and black males showed
an increase in death from the healthy state (with essentially no change in
death from the disabled state).
The transition rates between disabled, non-disabled, and death are
the basis of the active life expectancy estimates. These probabilities were
used to calculate the active life expectancies in the multistate life table.
These transitions were plotted by gender for each time point in Figures 6.1
through 6.24. On the left axis is the two-year probability of a given transition,
plotted against age.

142
Differences by race. Figures 6.1 through 6.16 show race differences
in the age-specific onset of disability, recovery from disability, and the
probability of death among the disabled and non-disabled. These figures
show differences between the races, as well as illustrating different patterns
in the trends over the decade for each race.
Differences in onset by race. Blacks in both 1987 and 1997 (Figure
6.1) had higher rates of disability onset than whites (significant for ages 76
and above in 1987). The race gap was still apparent in the 1997 cohort,
although the differences were no longer statistically significant. The race
difference was relatively substantial, as for example, the probability of
disability incidence was 20% at age 82 for blacks in 1987, while that same
probability was not reached until 4 years later, age 86, among whites in the
same year. Also, note that the scale on Figure 6.1 is much higher (maximum
0.7) than the scale on Figure 6.2 (maximum 0.5), as onset rates for blacks
were exceptionally high in 1987 for those in their late 80s and above.
Change in onset over time. Both blacks (Figure 6.3) and whites
(Figure 6.4) appear to have experienced a general decline in disability onset
over this time, although primarily at the oldest ages. The change is
significant at ages 90 and above for blacks, and ages 87 and above for
whites. Although the trend suggests a decline for both groups, the rate of
decline is higher for blacks at the oldest ages. The difference in the rate of
decline was not enough to completely off-set the appearance of race-
differences in onset at the later time, however.

143
Figure 6.1. Probability of Transition to Disability in 24 Months by Race:
Race Differences in Disability Onset the 1987 cohort

Figure 6.2. Probability of Transition to Disability in 24 Months by Race:
Race Differences in Disability Onset the 1997 cohort

0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
70
72
74
76
78
80
82
84
86
88
90
92
94
Black 1987
White 1987
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
70
72
74
76
78
80
82
84
86
88
90
92
94
Black 1997
White 1997

144
Figure 6.3. Probability of Transition to Disability in 24 Months by Race:
Change in the Probability of Disability Onset for Blacks

Figure 6.4. Probability of Transition to Disability in 24 Months by Race:
Change in the Probability of Disability Onset for Whites

Differences in recovery by race. Figures 6.5 and 6.6 show the race
differences in the probability of recovery. As shown in Figure 6.5, while no
difference is significant, it appears that whites may have had a slightly higher
rate of recovery in 1987 at the younger ages. The 1997 data shows that
there are no apparent differences by race in the probability of recovery
(Figure 6.6).
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
70
72
74
76
78
80
82
84
86
88
90
92
94
Black 1987
Black 1997
0
0.1
0.2
0.3
0.4
0.5
0.6
70
72
74
76
78
80
82
84
86
88
90
92
94
White 1987
White 1997

145
Change in recovery over time. Observations of recovery probability
show a clear increase over time in recovery rates for blacks at ages 77 and
above (Figure 6.7), with a large decline across the age range of interest. For
example, at age 84, the probability of recovery was 0.10 in 1987, and in 1997
it rose to 0.21, about double that from a decade earlier. A similar story is
shown for whites (Figure 6.8), with larger declines at the older ages. The
rate of change appears to be roughly similar for both race groups.

146
Figure 6.5. Probability of Recovery from Disability in 24 Months by
Race: Race Differences in Disability Recovery for the 1987 cohort

Figure 6.6. Probability of Recovery from Disability in 24 Months by
Race: Race Differences in Disability Recovery for the 1997 cohort

0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
70
72
74
76
78
80
82
84
86
88
90
92
94
Black 1987
White 1987
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
70
72
74
76
78
80
82
84
86
88
90
92
94
Black 1997
White 1997

147

Figure 6.7. Probability of Recovery from Disability in 24 Months
by Race: Change in the Probability of Disability Recovery for
Blacks

Figure 6.8. Probability of Recovery from Disability in 24 Months by
Race: Change in the Probability of Disability Recovery for Whites

0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
70
72
74
76
78
80
82
84
86
88
90
92
94
Black 1987
Black 1997
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
70
72
74
76
78
80
82
84
86
88
90
92
94
White 1987
White 1997

148
Race differences in the probability of death. Figures 6.9 through 6.12
show the probability of death among those previously non-disabled, while
Figures 6.13 through 6.16 show the probability of death among those
previously disabled. Interestingly, there are few differences between whites
and blacks in mortality at these ages. There is a hint that mortality may be
higher among able whites than blacks at ages 90 and above, although the
difference is not statistically significant due to small sample size. Race
differences in death are generally less apparent in the 1997 data (Figure 6.10
and 6.14), than in the 1987 data (Figure 6.9 and 6.13).
Changes in the probability of death over time. Again, no change over
time in mortality was statistically significant; however, the data suggest a
slight increase in mortality among non-disabled blacks ages 90 and above
(Figure 6.11), and perhaps for whites as well (Figure 6.12). The data are so
limited at this age that it is not possible to make this point conclusively,
however. Changes among those previously disabled are so small that it
seems as if no change occurred over the decade among either blacks or
whites previously disabled (Figures 6.15 and 6.16).

149
Figure 6.9. Probability of Transition to Death from the Non-disabled
State in 24 Months by Race: Differences in the Probability of Death the
1987 cohort

Figure 6.10. Probability of Transition to Death from the Non-disabled
State in 24 Months by Race: Differences in the Probability of Death the
1997 cohort

0
0.05
0.1
0.15
0.2
0.25
70
72
74
76
78
80
82
84
86
88
90
92
94
Black 1987
White 1987
0
0.05
0.1
0.15
0.2
0.25
0.3
70
72
74
76
78
80
82
84
86
88
90
92
94
Black 1997
White 1997

150
Figure 6.11. Probability of Transition to Death from the Non-disabled
State in 24 Months by Race: Change in the Probability of the Probability
of Death for Blacks

Figure 6.12. Probability of Transition to Death from the Non-disabled
State in 24 Months by Race: Change in the Probability of Death for
Whites

0
0.05
0.1
0.15
0.2
0.25
0.3
70
72
74
76
78
80
82
84
86
88
90
92
94
Black 1987
Black 1997
0
0.05
0.1
0.15
0.2
0.25
70
72
74
76
78
80
82
84
86
88
90
92
94
White 1987
White 1997

151
Figure 6.13. Probability of Transition to Death from the Disabled State
in 24 Months by Race: Differences in the Probability of Death the 1997
Cohort

Figure 6.14. Probability of Transition to Death from the Disabled State
in 24 Months by Race: Differences in the Probability of Death the 1997
Cohort

0
0.1
0.2
0.3
0.4
0.5
0.6
70
72
74
76
78
80
82
84
86
88
90
92
94
Black 1987
White 1987
0
0.1
0.2
0.3
0.4
0.5
0.6
70
72
74
76
78
80
82
84
86
88
90
92
94
Black 1997
White 1997

152
Figure 6.15. Probability of Transition to Death from the Disabled State
in 24 Months by Race: Change in the Probability of Death for Blacks

Figure 6.16. Probability of Transition to Death from the Disabled State
in 24 Months by Race: Change in the Probability of Death for Whites

0
0.1
0.2
0.3
0.4
0.5
0.6
70
72
74
76
78
80
82
84
86
88
90
92
94
Black 1987
Black 1997
0
0.1
0.2
0.3
0.4
0.5
0.6
70
72
74
76
78
80
82
84
86
88
90
92
94
White 1987
White 1997

153
Disability onset for females by race. As shown in Figure 6.17, black
females above the age of 79 were the only females to demonstrate a
reduction in the likelihood of the onset of disability. They also were the only
group to show increased recovery from disability over 10 years. Black and
white women did not differ in disability onset in the second survey. The
likelihood that a white woman would either became disabled or recovered
from disability was unchanged over this 10-year period.
Figure 6.17. Probability of Transition to Disability in 24 Months for Females
by Race
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85
1987 white female 1987 black female
1997 white female 1997 black female

154
Disability onset for males by race. A view of the onset of disability for men
revealed higher rates of onset for black men starting at age 78, as compared
to white men in both surveys (see Figure 6.18). Neither group showed any
significant improvement over this decade. There was one positive change
for black males, as they experienced higher recovery rates in the second
survey. This improvement placed them in a position indistinguishable from
white men (see Figure 6.18).
0.00
0.05
0.10
0.15
0.20
0.25
0.30
70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85
1987 white male 1987 black male
1997 white male 1997 black male
Figure 6.18. Probability of Transition to Disability in 24
Months for Males by Race

155

Recovery for females by race. There was an increase in the
probability of recovery for black females in the 1997 cohort (Figure 6.19). No
other groups experienced a changed in recovery rates, and all groups
appeared to have approximately the same probability of recovery, with the
exception of black females in the 1987 cohort.
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85
1987 white female 1987 black female
1997 white female 1997 black female
Figure 6.19. Probability of Recovery from Disability in 24
Months for Females by Race

156

Recovery for males by race. There was an increase in the probability
of recovery for black males (Figure 6.20) in the 1997 cohort between ages 70
and 76. No other differences are apparent between these groups.
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85
1987 white male 1987 black male
1997 white male 1997 black male
Figure 6.20. Probability of Recovery from Disability in 24
Months for Males by Race

157
Transitions to mortality. Transitions to death are estimated from both the non-
disabled and the disabled states. The probability of dying (qx values in a
lifetable) from either of these states is shown in Figures 6.21 through 6.24, as
differentiated by the disability status immediately preceding the survey in
which they were known to be dead. The transition rates are higher for the
disabled (Figures 6.22 and 6.24) than the non-disabled (Figures 6.21 and
6.23). The mortality transition rates are higher for males (Figures 6.23 and
6.24) relative to females (Figures 6.21 and 6.22).
Race differences in mortality are not obvious once divided by previous
disability state. There are no significant improvements over time in mortality
among the non-disabled, and only slight improvement for white females (see
Figure 6.22) and black males (see Figure 6.24) among the disabled group.

158
0.00
0.02
0.04
0.06
0.08
0.10
0.12
0.14
70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85
87 white female 87 black female
97 white female 97 black female
Figure 6.21. Probability of Death from the Non-Disabled State:
Females by Race

159
0.00
0.10
0.20
0.30
0.40
70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85
1987 white female 1987 black
1997 white female 1997 black
Figure 6.22. Probability of Death from the Disabled State: Females by
Race

160
Figure 6.23. Probability of Death from the Non-Disabled State: Males by
Race
0.00
0.04
0.08
0.12
0.16
0.20
0.24
70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85
1987 white male 1987 black male
1997 white male 1997 black male

161
Figure 6.24. Probability of Death from the Disabled State: Males by
Race
0
0.1
0.2
0.3
0.4
0.5
70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85
1987 white male 1987 black male
1997 white male 1997 black male

162
Active Life Expectancy
The above transition rates give rise to a series of multistate life tables
divided by race and subsequently subdivided by sex. These tables are
summarized in five-year age groups in Tables 6.9 through 6.15. Overall,
there were observed gains between 1987 and 1997 estimates in life
expectancy for men, and overall gains in life expectancy for whites. Life
expectancy changes for women overall and for blacks were not significant.
The results indicate no significant changes at age 70 in total life
expectancy or in active or inactive life expectancy in the black population
between 1987 and 1997 (Table 6.9). The point estimate of total life
expectancy for blacks was actually 0.2 years lower at the later time, with a
0.7 increase in active life expectancy, and a decrease of 1.0 year in disabled
life expectancy (see Table 6.9). In contrast, among the white population,
there was a significant increase in total life expectancy at age 70 of 0.7 years
with the increase coming from the significant increase (0.8) in active life
expectancy plus a small non-significant change in disabled life (Table 6.10).
The result is a growing gap at the older ages between blacks and whites in
total life expectancy. In addition, the proportion of active life for whites at all
older ages remains higher than that for blacks (Table 6.11). While both
blacks and whites appear to have experienced a significant increase
between 1987 and 1997 in the proportion of remaining life in the active state,
the changes are generally significant only for the white sample, due to the
smaller sample size, and thus larger standard error, in the black sample.

163
Table 6.9. Life Expectancy by Decade by Race: Total, Active, and Disabled: Black Population

1987 1997
Age Total LE Active LE Disabled LE Total LE Active LE Disabled LE
70 13.55 9.68 3.87 13.33 10.44 2.90
12.45-14.65 8.85-10.51 3.18-4.56 12.24-14.42 9.52-11.35 2.38-3.42
75 10.68 6.65 4.03 10.60 7.65 2.95
9.71-11.64 5.95-7.35 3.32-4.73 9.67-11.53 6.88-8.42 2.43-3.47
80 8.33 4.18 4.15 8.37 5.38 2.99
7.46-9.19 3.60-4.75 3.41-4.89 7.57-9.17 4.74-6.01 2.47-3.51
85 6.52 2.35 4.17 6.65 3.64 3.01
5.69-7.35 1.89-2.81 3.38-4.96 5.94-7.35 3.11-4.17 2.48-3.53
90 5.23 1.19 4.04 5.38 2.40* 2.98
4.17-5.87 0.70-1.35 3.15-4.84 4.72-6.03 1.96-2.84 2.44-3.51

* denotes 1997 value significantly different from 1987, p<.05

164

Table 6.10. Life Expectancy by Decade by Race: Total, Active, and Disabled: White Population

1987 1997
Age Total LE Active LE Disabled LE Total LE Active LE Disabled LE
70 13.82 11.06 2.76 14.55* 11.89* 2.65
13.48-14.16 10.77-11.35 2.59-2.92 14.19-14.90 11.59-12.20 2.50-2.81
75 10.87 8.00 2.90 11.60* 8.90* 2.71
10.57-11.16 7.74-8.19 2.72-3.08 11.29-11.91 8.65-9.14 2.54-2.87
80 8.41 5.35 3.06 9.15* 6.39* 2.76
8.10-8.72 5.14-5.56 2.84-3.27 8.83-9.47 6.15-6.63 2.57-2.95
85 6.47 3.28 3.18 7.21 4.41* 2.80
6.12-6.81 3.06-3.50 2.91-3.46 6.87-7.56 4.16-4.66 2.58-3.03
90 5.03 1.81 3.22 5.75 2.94* 2.82
4.64-5.43 1.59-2.03 2.86-3.58 5.37-6.13 2.67-3.20 2.53-3.10

* denotes 1997 value significantly different than 1987, p<.05

Table 6.11. Proportion of Remaining Expected Life in Active State by Race, Sex and Decade
Age
1987 Black 1997 Black 1987 White 1997 White
70
0.71 0.78 0.80 0.82*
75
0.62 0.72 0.74 0.77*
80
0.50 0.64 0.64 0.70*
85
0.36 0.55 0.51 0.61*
90
0.23 0.45* 0.36 0.51*

165

When the data are divided by both sex and race simultaneously (Table
6.12), the only group that experiences a significant increase in total life
expectancy at age 70 are white males. White men experienced an
improvement of 1.1 year in life expectancy (see Table 6.12), including a 0.7
year increase in active life expectancy (Table 6.13).
In contrast, black males did not show significant gains in life expectancy,
nor in active life expectancy. In fact, the results suggested that static
conditions existed in total and active life expectancy between 1987 and 1997
for both black males and females. This led to a growing divide as white
males gained both total and active life over this same time period.
Furthermore, it appears that black females had higher disability onset than
whites after age 80, and lower recovery from disability at age 70 than whites,
contrary to earlier findings by Clark and colleagues (1993) and Land (1994).
Furthermore, the onset of disability increased for black females, while it had
not changed for whites over the ten-year period. Recovery improved for
black males with no change in disability onset over this period.

166
Table 6.12. Total Life expectancy by Sex and Race: Total, Active, and
Disabled
Black Male White Male Black Female White Female
1987 1997 1987 1997 1987 1997 1987 1997
70 11.66 11.16 11.56 12.52* 15.26 13.81 15.48 15.55
75 8.64 8.21 8.87 9.67 11.60 11.39 12.10 12.38
80 6.68 6.54 6.70 7.36 9.08 8.77 9.23 9.82
85 4.87 4.88 5.19 5.56 7.19 7.26 6.99 7.57

Table 6.13. Active Life expectancy by Sex and Race: Total, Active, and
Disabled
Black Male White Male Black Female White Female
1987 1997 1987 1997 1987 1997 1987 1997
70 9.24 9.09 9.92 10.60 10.62 9.41 12.02 11.49
75 5.85 5.96 6.98 7.69 6.76 7.27 8.58 8.31
80 3.84 4.39 4.68 5.32 4.36 4.60 5.66 5.80
85 1.79 2.66 3.10 3.45 2.68 3.36 3.43 3.55

Table 6.14. Disabled Life expectancy by Sex and Race: Total, Active,
and Disabled
Black Male White Male Black Female White Female
1987 1997 1987 1997 1987 1997 1987 1997
70 2.43 2.07 1.73 1.92 4.64 4.40 3.46 4.06
75 2.78 2.25 1.89 1.98 4.84 4.12 3.52 4.07
80 2.85 2.15 2.02 2.04 4.72 4.17 3.58 4.01
85 3.08 2.22 2.09 2.11 4.51 3.90 3.56 4.02

Table 6.15. Proportion of Remaining Expected Life in Active State by
Race, Sex and Decade
Black Male White Male Black Female White Female
1987 1997 1987 1997 1987 1997 1987 1997
70 0.79 0.81 0.86 0.75 0.70 0.77 0.78 0.74
75 0.68 0.73 0.79 0.75 0.58 0.67 0.71 0.67
80 0.57 0.67 0.70 0.63 0.48 0.61 0.61 0.59
85 0.37 0.55 0.60 0.60 0.37 0.48 0.49 0.47

167
Status-based Results
The previous estimates of active life expectancy were based on averages
of the entire population regardless of their health. However, it is well known
that life expectancy is closely linked to previous health history, and thus it is
of interest to investigate status-based life tables, where active and disabled
life expectancy are calculated separately for individuals who were initially
active and those initially disabled. As seen in Tables 6.16 and 6.17, the
differences in total life expectancy by active or disabled status are large:
approximately 4 years longer total life expectancy for blacks (about 13.7 for
active, about 9.7 for disabled), and 5 years for whites (LE about 14.4 for
active, 9.4 for disabled) who are initially healthy. Due to the large confidence
intervals, race differences in total LE between active whites and active
blacks, or between disabled whites and disabled blacks, are not significant.
The data present a different picture regarding active life expectancy.
Initially, active whites did experience a significant increase of about 1 year in
active life expectancy between 1987 and 1998, as did disabled whites, who
initially experienced about a 1.5 year increase in active life expectancy
(Table 6.17). The confidence intervals for the black sample are larger due to
their smaller sample size, and thus there are no significant changes over
time observed for blacks, although the point estimates suggest a decline in
disabled life expectancy among both the initially active and initially disabled
black groups.

168
Further dividing this by both race and sex (Tables 6.18 through 6.21),
we can see both differences between the groups as well as different trends
over time. Starting with differences between the race and sex groups, we
can see a longer total and active life expectancy for active white males
(12.93 total, 11.33 active) than their active black male counterparts (11.76
total, 10.02 active) in the 1997 cohort. There was no significant difference
between black and white males in the 1987 cohort. Similar differences are
apparent for active females in both the 1987 and 1997 cohorts. In both
cases, white females (Table 6.19) have longer active life expectancy (12.34
years in 1987 and 12.92 years in 1997) than comparable black females
(10.87 years in 1987 and 11.42 years in 1997 in Table 6.18). Among the
disabled, however, there are no differences between the races for each sex.
Broadly speaking, life expectancy for disabled males of either race is about 8
years (Tables 6.20 and 6.21), with about 4 years expected active, while for
females (Tables 6.18 and 6.19), life expectancy among the disabled is about
11 years, with between 4.5 to 7 years active. Thus, women can expect
somewhere between 5 and 7 years disabled, while men of the same age
expect only 3 or 4 years.
Turning to trends over time for the race-sex groups, it is apparent that
there is no change over this time period for black males, while active white
males showed significant increases in both total (from 11.79 years to 12.93
years) and active life expectancy (from 10.12 years to 11.13 years). There
were no changes for disabled males over this time period.

169
Among females, the opposite was true: there was no change over
time for those active, but there was a decrease in disabled life expectancy for
disabled black females (from 6.68 years to 5.25 years). Among disabled
white females, there was a nearly significant increase in active life
expectancy as well.

170
Table 6.16. Status Based Health Expectancies at age 70 by Year and Initial Health Status among Blacks

Active Disabled

1987 1997 1987 1997
Total Life Expectancy 13.88 (12.79-14.96) 13.63 (12.54-14.72) 9.60 (8.08-11.11) 9.78 (8.40-11.16)
Active Life Expectancy 10.16 (9.40-10.92) 10.86 (9.99-11.73) 3.89 (2.80-4.98) 5.39 (4.29-6.49)
Disabled Life Expectancy 3.72 (3.06-4.38) 2.77 (2.28-3.26) 5.71 (4.76-6.66) 4.38 (3.73-5.04)

Table 6.17. Status Based Health Expectancies at age 70 by Year and Initial Health Status among Whites

Active Disabled

1987 1997 1987 1997
Total Life Expectancy 14.03 (13.70-14.37) 14.79 (14.44-15.13) 9.31 (8.51-10.10) 10.47 (9.77-11.17)
Active Life Expectancy 11.37 (11.11-11.63) 12.23 (11.95-12.52) 4.73 (4.04-5.42) 6.18 (5.58-6.78)
Disabled Life Expectancy 2.67 (2.51-2.83) 2.55 (2.40-2.70) 4.58 (4.21-4.95) 4.29 (4.00-4.58)

171
Table 6.18. Status Based Health Expectancies at age 70 by Year and Initial Health Status among Black Males

Active Disabled

1987 1997 1987 1997
Total Life Expectancy 11.70 (10.66-12.74) 11.76 (10.70-12.81) 8.00 (6.69-9.30) 7.69 (6.45-8.94)
Active Life Expectancy 9.26 (8.47-10.05) 10.02 (9.13-10.91) 3.46 (2.43-4.49) 4.46 (3.41-5.52)
Disabled Life Expectancy 2.44 (1.95-2.93) 1.74 (1.39-2.09) 4.53 (3.77-5.30) 3.23 (2.73-3.72)

Table 6.19. Status Based Health Expectancies at age 70 by Year and Initial Health Status among White Males
Active Disabled

1987 1997 1987 1997
Total Life Expectancy 11.79 (11.37-12.21) 12.93 (12.48-13.37) 7.69 (6.89-8.49) 8.35 (7.56-9.14)
Active Life Expectancy 10.12 (9.76-10.48) 11.33 (10.93-11.72) 4.04 (3.30-4.77) 5.16 (4.46-5.87)
Disabled Life Expectancy 1.67 (1.51-1.83) 1.60 (1.45-1.75) 3.65 (3.28-4.02) 3.19 (2.90-3.48)

172
Table 6.20. Status Based Health Expectancies at age 70 by Year and Initial Health Status among Black Females
Active Disabled

1987 1997 1987 1997
Total Life Expectancy 15.54 (14.42-16.67) 14.93 (13.78-16.08) 11.16 (9.48-12.85) 11.32 (9.77-12.87)
Active Life Expectancy 10.87 (10.08-11.67) 11.42 (10.52-12.32) 4.49 (3.25-5.72) 6.07 (4.85-7.29)
Disabled Life Expectancy 4.67 (3.91-5.44) 3.51 (2.91-4.11) 6.68 (5.59-7.77) 5.25 (4.44-6.06)

Table 6.21. Status Based Health Expectancies at age 70 by Year and Initial Health Status among White Females
Active Disabled

1987 1997 1987 1997
Total Life Expectancy 15.77 (15.34-16.20) 16.20 (15.73-16.67) 10.86 (9.94-11.78) 12.16 (11.33-12.99)
Active Life Expectancy 12.34 (12.00-12.67) 12.92 (12.55-13.29) 5.51 (4.69-6.32) 7.01 (6.31-7.71)
Disabled Life Expectancy 3.43 (3.20-3.66) 3.28 (3.05-3.50) 5.36 (4.91-5.80) 5.15 (4.77-5.53)

173
Implied Prevalence
The implied prevalence is a measure of the prevalence in a stationary
population implied by continuing the rates of disability onset, recovery, and
death. This measure provides an indication of the long-term impact of
current disability and mortality processes. The implication of the disability
transition rates is that differences between blacks and whites are declining
(Figure 6.25). To demonstrate this, Figure 6.25 shows that the highest
implied prevalence rates are found among blacks in 1987, with the difference
being significant at ages 83 and above. The line representing blacks in 1997
is much lower, leading to a lack of differences between the races in the 1997
data. These results suggest that if these transitions rates continue, we might
observe a decline in the difference in disability between blacks and whites.
Figures 6.26 through 6.29 further divide the implied prevalence by
race and sex. First, an examination of differences in the 1987 cohort (Figure
6.27) shows that race was more closely linked to implied disability than
gender. In fact, black males and black females had statistically the same
implied prevalence of disability, as did white males and white females. The
difference between black and white males was significant at ages 73 and
above, and so was the difference between black and white females at ages
71 and above. Essentially, the same pattern was seen in the 1997 cohort.
Turning to an examination of the rate of change over time, we can see
that both black males and black females experienced a decline in the implied

174
prevalence of disability of approximately equal and parallel levels (Figure
6.29). Changes for whites were similar, although white males experienced a
proportionately larger decline than did white females.

Figure 6.25. Implied Prevalence of Disability by Race and Year
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
70
72
74
76
78
80
82
84
86
88
90
92
94
1987 Black
1997 Black
1987 White
1997 White

175
Figure 6.26. Implied Prevalence of Disability by Race and Sex for the
1987 Cohort

Figure 6.27. Implied Prevalence of Disability by Race and Sex for the
1997 Cohort

0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
70
73
76
79
82
85
88
91
94
Black Male
Black Female
White Male
White Female
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
70
73
76
79
82
85
88
91
94
Black Male
Black Female
White Male
White Female

176

Figure 6.28. Implied Prevalence of Disability by Race and Sex for Blacks

Figure 6.29. Implied Prevalence of Disability by Race and Sex for
Whites

0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
70
73
76
79
82
85
88
91
94
87 Black Male
87 Black Female
97 Black Male
97 Black Female
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
70
73
76
79
82
85
88
91
94
87 White Male
87 White Female
97 White Male
97 White Female

177

Conclusion
In conclusion, the first hypothesis, stating that blacks would have higher
disability incidence than whites, was supported at ages 75 and above in the
1987 data, but not in the 1997 cohort, where no differences in onset by race
were found. The second hypothesis, stating that declines in incidence would
be larger for whites was not supported, as time trends revealed essentially
no difference in the rate of change for blacks and whites. The third
hypothesis, that blacks would have lower recovery rates than whites, also
was not supported. Although it appeared that whites indeed had higher point
estimates of the probability of recovery in 1987, the difference was not
significant, and it was clearly not significant in 1997. The fourth hypothesis,
that the increase in the rate of recovery would be higher for whites, was not
supported, as both blacks and whites experienced a similar increase in the
probability of recovery over the decade. The fifth hypothesis, that death
rates would be higher for blacks in both the disabled and able states, was not
supported, as there were no apparent differences by race in the probability of
mortality at a given age. The sixth hypothesis, which suggested that the
trend in death rates would lead to faster decline for blacks, was not
supported, as there were no differences in the rate of change for either race.
The seventh hypothesis, stating that whites would have higher estimates of
active life expectancy, was supported in both data sets. Although the

178
difference was not significant at every age, the data revealed that particularly
above age 80, whites generally reported more years active, and a slightly
greater proportion of active life than their black counterparts. The eighth
hypothesis, that whites would experience larger gains in active life
expectancy, was not supported, as changes in active life expectancy were
very similar for both races. Finally, the ninth hypothesis, that blacks would
have higher estimates of disabled life expectancy, was not supported, as
both races reported an approximately equal number of years disabled.
The study revealed limited changes in the onset of disability over the 10-
year period between 1987 and 1997. Black females demonstrated a slight
decrease (improvement) in their onset of disability as well as increased
recovery from disability such that it appears that for this group, the racial gap
was closed. There were higher rates of disability onset for black men as
compared to white men in both years, with neither racial group of males
showing any significant improvement over this time. The one positive effect
for black males was their higher recovery rate in the second survey. This
improvement placed their recovery in a position indistinguishable from that of
white men.
Although the elderly population in general is living less time with disability
in more recent years, not all groups have benefited equally. Men enjoyed
the largest increase in life expectancy-- particularly white men. There also
appears to be a 1.1 year increase in life expectancy for white men at age 70,

179
while at the same time no change for black men. Women of both races did
not show any improvement; and in fact, there appears to be a decline in
active life expectancy for white women, although not quite significant. The
news that the gender gap in life expectancy is narrowing could be a sign that
men are making modest gains in their health behaviors. The news that
virtually no gains have been made among blacks suggests that there is room
for future improvement in this area. In the 1997 data, the gap between
blacks and whites appeared to have grown larger, suggesting diverging
trends by race.
Implications
Implications
As life expectancy continues to increase, there will be large gains in
the size of the older population, as well as an aging of the older population as
a whole. With greater numbers of people living to the oldest ages, the U.S.
will likely need to support rapidly expanding numbers of disabled elderly for
decades to come. Although the results of the present study suggest that
recent increases in life expectancy have not been additional years with
disability, it should be noted that the definition accounts for only severe
disability. While severe disability is likely a good approximation for need of a
supportive environment, less severe disability, not captured in this analysis,
can also lead to the need for a more assistive environment.

180
The results suggest that the observed trends are not so much caused
by improving health for all people in the older population, but rather by the
changing composition of the population. As more men are surviving to older
ages, their generally smaller proportion of life spent disabled is lowering the
observation of average changes over time. In fact, the results did not show
any positive changes for females over this time, suggesting that there may
not be any systematic improvement in the proportion of the population
reporting disability due to technological improvements between the 1980s
and 1990s.
Even in the face of declining number of years spent disabled, the
fiscal impact of the disabled elderly population will likely grow due to the
parallel increase in life expectancy observed at all ages. Although increases
in total life and active life were observed for males, and not females, males
still lag behind their female age-counterparts in both total and active life
expectancy. It is not likely that male life expectancy will catch up to female
life expectancy any time soon; nor is it likely that female disabled life
expectancy will decline to male levels. There is clearly room for continued
research in this area.

181
Chapter VII: Changes in Life Expectancy and Active Life Expectancy
by Education Between 1984 and 2000

This chapter examines trends in active life expectancy for two education
levels, those with less than 12 years compared to those with 12 years or
more, in order to examine whether changes in active life expectancy are
occurring differently for individuals with different levels of education. The
trends include transition rates for disability onset, recovery, and mortality for
each education group, as well as estimates of active and disabled life
expectancy for each education group. The focus is on differences in
disability onset, recovery, and mortality, as well as on differences in active
life expectancy by education level. There are clear differences between the
1987 and 1997 cohorts, as mentioned in the third chapter: while 43% of the
1987 cohort had 12 or more years of education, that proportion rose to 59%
in the 1997 cohort. This 37% rise in the proportion of the sample with 12 or
more years of education may be an important factor to consider when
examining change in disability and active life expectancy observed over time.
Background
Socioeconomic status, particularly education, has been closely linked
to almost all health outcomes throughout the world including disability and life
expectancy (Crimmins & Cambois, 2003; Guralnik et al., 1993; Martinez-
Sanchez et al., 2001; Melzer et al., 2001; Minicuci et al., 2005; Nusselder et
al., 2005; Valkonen et al., 1997). The general consensus in studies appears
182
to be that more highly educated groups have longer lives, and also have less
disability, leading to longer lives without disability (Crimmins & Saito, 2001;
Guralnik et al., 1993; Martinez-Sanchez et al., 2001; Melzer et al., 2001;
Minicuci et al., 2005; Nusselder et al., 2005; Valkonen et al., 1997). Trends
between 1970 and 1990 (Crimmins & Saito, 2001) and the mid-1980s
through the late 1990s (Schoeni et al., 2005) suggest that there might be a
growing divergence in disability prevalence between the lower and higher
educated older people over recent years.
There are several pathways by which education may work to protect
individuals from functional limitations and major diseases. Higher education
levels are likely to increase one’s ability to understand the risks to health, and
to alter one’s life to reduce those risks. Knowledge and use of preventive
care health services and good health habits can delay age-associated
diseases. Education is linked to less smoking (Barbeau, Krieger, &
Soobade, 2004; Escobedo & Peddicord, 1996) and higher cessation rates
among those who did smoke (Wray, Herzog, Willis et al., 1998). Education is
also linked to increased physical activity, improved diet, and weight control
(Laditka & Laditka, 2002). In addition, these behaviors are linked to lower
levels of some chronic conditions including diabetes, arthritis, and
osteoporosis (Wister, 1996).
Education is often used as a proxy for social status in studies because
education remains consistent throughout later adulthood, unlike income and
183
occupation which can vary over time and do not always reflect the lifetime
circumstances of an individual. Furthermore, education is not likely to be
affected by later-life living arrangements, marital status, child-rearing issues,
or health as much as other measures of socioeconomic status, such as
income or occupation (Crimmins & Cambois, 2003).
Studies of both males and females have found an association
between education and health. For example, women with less education
have substantially more behavioral and biological risks associated with
coronary artery disease, like increased likelihood of smoking, low exercise
levels, and lower high-density lipoprotein levels than women with more
education (Matthews, Kelsey, Meilahn, Kuller, & Wing, 1989). The effects of
education may be even more pronounced in men, as a U.K. study found that
education has a larger impact on active life expectancy for men as compared
to women (Matthews et al., 2006).
Crimmins and Saito (2001) found education to be closely linked to
healthy life expectancy (the measure used also could have been called
active life expectancy). Among males with 13 or more years of education in
1990, healthy life expectancy was 10.8 years higher at age 30 than for males
with 8 or fewer years of education. Among females, the equivalent difference
was 9.5 years. Further, Crimmins and Saito found a decline in healthy life
expectancy among white females with 8 or fewer years of education between
1980 and 1990, with no change in healthy life expectancy for similar white
184
males. Among those with 13 or more years of education, both white males
and females showed an increase in healthy life expectancy, although white
males showed a larger relative increase than did white females.
Hypotheses
The previous literature suggests that the current analysis of education
differences in active life expectancy, and education differences in trends in
active life expectancy, can test the following hypotheses: 1) Individuals with
less than 12 years of education would have higher disability incidence than
those with 12 or more years during the study period, 2) declines in incidence
would be larger for those with higher levels of education, 3) those with less
than 12 years of education would have lower recovery rates than those with
12 or more years of education, 4) the rate of recovery increase would be
higher for those with 12 or more years of education, 5) death rates would be
higher for those with less than 12 years of education, especially for males
with less than 12 years of formal schooling, in both the disabled and able
states, 6) the trend in death rates would lead to faster decline for those with
12 or more years of education, 7) those with 12 or more years of education,
particularly females, would have higher estimates of active life expectancy, 8)
those with 12 or more years of formal education, particularly males, would
experience larger gains in active life expectancy, and 9) those with 12 or
fewer years of education, particularly females, would have higher estimates
of disabled life expectancy.
185
The first section presents transition rates to and from disability and to
mortality for each education group. These rates form the basis of the second
section, which contains estimates of the total, active, and disabled life
expectancies. The third section is a summary of the implied prevalence of
disability, based on recent onset rates. The fourth section is a description of
status-based life tables showing how total, active, and disabled life
expectancies vary depending on initial disability status.
Transitions between disability states.
Table 7.1 shows the frequency of the transitions between two
consecutive time points for each education group, as well as for males and
females of each education level. As shown in the table, the more highly
educated group has a larger percentage of individuals disability-free at two
consecutive interviews. For example, approximately 73% of the total
transitions were in the category that had remained disability-free in the 1997
data, as compared to 58% for those with less than 12 years of education.
Although less dramatic, this same relationship is found in the 1987 cohort.
The more highly educated group was slightly less like to become disabled,
but was also less likely to recover. The group with less education was more
likely to be disabled at two consecutive interviews, and was more likely to die
from either the disability-free or the disabled state.
An examination of the differences by both sex and education level
revealed a pattern of higher stability in a non-disabled state for the higher
186
educated group (about 70% for either sex in either year). Sex differences
are observed in the proportion experiencing the onset of disability; however,
within each sex, the lower educated group experiences a higher proportion of
transitions to disability. Remaining disabled at two time points is also more
closely linked to gender, but within each sex, education affects the proportion
of the sample remaining disabled.
Changes over time were slight for either education group, as well as
each sex-education group. Among the more highly educated, there was
generally an increase in remaining disability-free, while females with less
education showed declines in remaining non-disabled, and increases in
becoming disabled (Table 7.5). Males with less education, on the other
hand, appeared to show increases in remaining non-disabled and recovery,
as well as declines in disability onset and death among the non-disabled
(Table 7.3).

187
Table 7.1. Number of Health Transitions By Education: Transitions
among Those with Less Than 12 Years of Education
1987 LSOA 1997 LSOA II
States Beginning and End N % N %
Remained Healthy 4992 57.3 3782 58.0
Became Disabled 1034 11.9 780 12.0
Died From Healthy State 857 9.8 637 9.8
Recovered from Disability 229 2.6 248 3.8
Remained Disabled 976 11.2 643 9.9
Died From Disabled State 625 7.2 427 6.6
Total 8713 100% 6517 100%

Table 7.2. Number of Health Transitions By Education: Transitions
among Those with 12 or More Years of Education
1987 LSOA 1997 LSOA II
States Beginning and End N % N %
Remained Healthy 4741 69.9 7044 72.9
Became Disabled 560 8.3 721 7.5
Died From Healthy State 572 8.4 822 8.5
Recovered from Disability 127 1.9 196 2.0
Remained Disabled 454 6.7 500 5.2
Died From Disabled State 332 4.9 375 3.9
Total 6786 100% 9658 100%

Table 7.3. Number of Health Transitions By Education: Low Educated
Males
1987 LSOA 1997 LSOA II
States Beginning and End N % N %
Remained Healthy 1939 57.4 1525 61.7
Became Disabled 350 10.4 244 9.9
Died From Healthy State 488 14.4 334 13.5
Recovered from Disability 69 2.0 65 2.6
Remained Disabled 299 8.8 147 5.9
Died From Disabled State 234 6.9 157 6.4
Total 3379 100% 2472 100%

188
Table 7.4. Number of Health Transitions By Education: High Education
Males
1987 LSOA 1997 LSOA II
States Beginning and End N % N %
Remained Healthy 1654 70.7 2847 73.9
Became Disabled 164 7.0 251 6.5
Died From Healthy State 283 12.1 409 10.6
Recovered from Disability 38 1.6 67 1.7
Remained Disabled 88 3.8 127 3.3
Died From Disabled State 114 4.9 153 4.0
Total 2341 100% 3854 100%

Table 7.5. Number of Health Transitions By Education: Low Education
Females
1987 LSOA 1997 LSOA II
States Beginning and End N % N %
Remained Healthy 3053 57.2 2257 55.8
Became Disabled 684 12.8 536 13.3
Died From Healthy State 369 6.9 303 7.5
Recovered from Disability 160 3.0 183 4.5
Remained Disabled 677 12.7 496 12.3
Died From Disabled State 391 7.3 270 6.7
Total 5334 100% 4045 100%

Table 7.6. Number of Health Transitions By Education: High Educated
Females
1987 LSOA 1997 LSOA II
States Beginning and End N % N %
Remained Healthy 3087 69.4 4197 72.3
Became Disabled 396 8.9 470 8.1
Died From Healthy State 289 6.5 413 7.1
Recovered from Disability 89 2.0 129 2.2
Remained Disabled 366 8.2 373 6.4
Died From Disabled State 218 4.9 222 3.8
Total 4445 100% 5804 100%

189
Age-Specific Transition Rates by Education Level
The transition probabilities were derived from a maximum likelihood
estimation of 8 equations. Of these, four equations depicted the age and
education-specific effects of the likelihood of the given transition (onset,
recovery, death among the able, or death among the unable), and four
equations depicted the age, sex, and education effects of the same possible
transitions. As shown in Table 7.7, the coefficients for age are positive for
every transition type except recovery. The coefficient for the higher educated
(as compared to the lower educated) is negative for onset in both years,
without controlling for sex, but after controlling for sex (Table 7.8), the
reverse is true, as females have lower onset in 1987, with a lower onset rate
for females in 1997. Controlling for sex dramatically influences the effect of
education predicting the onset of disability. Recovery is less consistent by
race, as the coefficient was slightly higher for blacks relative to whites in
1987, and slightly lower for blacks relative to whites in 1997. The probability
of death for more highly educated individuals was lower among the able, and
slightly higher among the disabled.
If we introduce sex into the equations, we can see that the size of the
education coefficients change slightly, where the coefficient for females is
negative in 1987 and positive in 1997. There is a consistent gender
association with mortality; however, since females have lower mortality in
both years, from either the able or unable state. With gender controlled,
education was consistently associated with lower mortality (Table 7.8).
190
Table 7.7. Disability Transition Equation Coefficients for Education and
Age
Onset Recovery
1987 LSOA 1997 LSOA 1987 LSOA 1997 LSOA
Constant -12.323 -10.513 3.042 3.363
Age 0.133 0.113 -0.056 -0.057
Education 12+ -0.434 -.0565 0.176 -0.054

Death Able Death Unable
1987 LSOA 1997 LSOA 1987 LSOA 1997 LSOA
Constant -9.154 -9.130 -3.228 -2.779
Age 0.090 0.089 0.033 0.026
Education 12+ -0.275 -0.210 0.108 0.151

Table 7.8. Disability Transition Equation Coefficients for Education,
Sex, and Age
Onset Recovery
1987 LSOA 1997 LSOA 1987 LSOA 1997 LSOA
Constant -12.382 -10.388 3.148 3.183
Age 0.133 0.109 -0.056 -0.052
Education 12+ 0.161 0.277 -0.091 -0.264
Female -0.443 -.0557 0.185 -0.093

Death Able Death Unable
1987 LSOA 1997 LSOA 1987 LSOA 1997 LSOA
Constant -9.187 -9.360 -3.239 -2.731
Age 0.096 0.095 0.037 0.032
Education 12+ -0.760 -0.517 -0.473 -0.758
Female -0.234 -0.178 0.141 0.106

Incidence of disability. Information on recent changes in disability may
be best illustrated by changes in disability incidence. An examination of the
data without stratification for gender reveals that there are no significant
changes in incidence among either those with less than 12 years of
education nor those with 12 or more years of education (Figures 7.1 and
191
7.2). However, after disaggregating the sample by gender, and looking first
at changes for males, the higher educated group reported a decline in
incidence, between ages 76 and 90, with a clear gradient of higher levels of
onset for less educated males in both years (Figure 7.3). The changes for
the education groups over time appeared to be parallel, leading to no change
in this relationship over time. For females, the same pattern was observed,
with statistical significance at ages 78-93 (Figure 7.4).
With regard to changes among those with fewer than 12 years of
education, the effect of gender is clear, with higher probability of incidence
for females at ages 85 and above in 1987, and for females at all ages in
1997. Among those with 12 or more years of education, the sex difference is
still apparent, although not significant.
192
Figure 7.1. Probability of Disability Onset Among those with <12 years
of Education
0
0.1
0.2
0.3
0.4
0.5
0.6
70 75 80 85 90 95
1987
1997

Figure 7.2. Probability of Disability onset among those with >=12 years
of Education
0
0.1
0.2
0.3
0.4
0.5
70 75 80 85 90 95
1987
1997

193
Figure 7.3. Probability of Disability Onset Among Males by Education
Level
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
70
72
74
76
78
80
82
84
86
88
90
92
94
<12 Males 1987
<12 Males 1997
>=12 Males 1987
>=12 Males 1997
194
Figure 7.4. Disability Onset Among Females by Education Level
0.00
0.10
0.20
0.30
0.40
0.50
0.60
70
72
74
76
78
80
82
84
86
88
90
92
94
<12 Females 1987
<12 Females 1997
>=12 Females 1987
>=12 Females 1997
195
Figure 7.5. Disability Onset Among those with <12 Years of Education

0.00
0.10
0.20
0.30
0.40
0.50
0.60
70
72
74
76
78
80
82
84
86
88
90
92
94
<12 Males 1987
<12 Males 1997
<12 Females 1987
<12 Females 1997
196
Figure 7.6. Disability Onset Among those with >=12 Years of Education

0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
0.45
0.50
70
72
74
76
78
80
82
84
86
88
90
92
94
>=12 Males 1987
>=12 Males 1997
>=12 Females 1987
>=12 Females 1997
197
Recovery
Although there are clear differences by level of education in the onset
of disability, there are no clear differences in recovery by education
classification when the rates are modeled in the IMaCh algorithm (Figures
7.7 through 7.12). This lack of difference between groups, as well as lack of
significant change over time, suggests that, while recovery from disability is
observed, it is not likely to be the mechanism by which education affects
active life expectancy.
Transitions to Death
Figure 7.13 shows the rate of transition to death from a previously
non-disabled state. As the overlapping confidence intervals in Figures 7.13
and 7.14 illustrate, there were no significant differences in the mortality rates
from the non-disabled state, and there was no change over time in the
mortality rates for these education groups.
The same was observed for transitions to death from a previously
disabled state (Figures 7.15 and 7.16). As seen in the earlier chapters,
transitions to mortality were more frequent among those previously disabled;
however, education does not seem to significantly affect this transition.
However, dividing the sample by gender causes some differences to
become more apparent. For example, among previously active males, those
with less than 12 years of education in 1987 had the highest rates of
transition to death, although the difference is not statistically significant
198
(Figure 7.17). For females, the relationship is a bit more confused, as those
with less than 12 year of education in 1997 had lower mortality at ages 86
and above than their more highly educated counterparts in 1987 (Figure
7.20). Figure 7.21 shows that among the less educated, there was no
change for males or females in the probability of death (previously active),
with clear gender differences in mortality. The same picture is observed in
Figure 7.22 among the more highly educated. Figures 7.23 and 7.24
demonstrate that changes in death rates are not apparent among the
previously disabled, as these rates seem to be clearly defined for males and
females.
Figure 7.7. Probability of Recovery from Disability Among those with
<12 Years of Education
0
0.1
0.2
0.3
0.4
70 75 80 85 90 95
1987
1997

199
Figure 7.8. Probability of Recovery from Disability Among Those with
>=12 Years of Education

0
0.1
0.2
0.3
0.4
70 75 80 85 90 95
1987
1997

200
Figure 7.9. Probability of Recovery from Disability Among Males by
Education Level
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
70
72
74
76
78
80
82
84
86
88
90
92
94
<12 Males 1987
<12 Males 1997
>=12 Males 1987
>=12 Males 1997
201
Figure 7.10. Probability of Recovery from Disability Among Females by
Education Level

0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
70
72
74
76
78
80
82
84
86
88
90
92
94
<12 Females
1987
<12 Females
1997
>=12 Females
1987
>=12 Females
1997
202
Figure 7.11. Probability of Recovery from Disability Among those with
<12 Years of Education

0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
0.40
70
72
74
76
78
80
82
84
86
88
90
92
94
<12 Males 1987
<12 Males 1997
<12 Females
1987
<12 Females
1997
203
Figure 7.12. Probability of Recovery from Disability Among those with
>=12 Years of Education
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
70
72
74
76
78
80
82
84
86
88
90
92
94
>=12 Males 1987
>=12 Males 1997
>=12 Females
1987
>=12 Females
1997
204
Figure 7.13. Probability of Mortality Among Lower-Educated (<12yrs)
Previously Active

0
0.05
0.1
0.15
0.2
0.25
70
73
76
79
82
85
88
91
94
1987
1997

205
Figure 7.14. Probability of Mortality Among Lower-Educated (<12yrs)
Previously Disabled
0
0.1
0.2
0.3
0.4
0.5
70
73
76
79
82
85
88
91
94
1987
1997

206
Figure 7.15. Probability of Mortality Among Higher-Educated (12+yrs)
Previously Active
0
0.05
0.1
0.15
0.2
0.25
70
74
78
82
86
90
94
1987
1997

207
Figure 7.16. Probability of Mortality Among Higher-Educated (12+yrs)
Previously Disabled

0
0.1
0.2
0.3
0.4
0.5
0.6
70
73
76
79
82
85
88
91
94
1987
1997
208
Figure 7.17. Probability of Mortality Among Males Previously Active

0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
70
72
74
76
78
80
82
84
86
88
90
92
94
<12 Males 1987
<12 Males 1997
>=12 Males 1987
>=12 Males 1997
209
Figure 7.18. Probability of Mortality Among Females Previously Active
-0.05
0.00
0.05
0.10
0.15
0.20
0.25
70
72
74
76
78
80
82
84
86
88
90
92
94
<12 Females 1987
<12 Females 1997
>=12 Females
1987
>=12 Females
1997
210
Figure 7.19. Probability of Mortality Among Males Previously Disabled

0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
70
72
74
76
78
80
82
84
86
88
90
92
94
<12 Males 1987
<12 Males 1997
>=12 Males 1987
>=12 Males 1997
211
Figure 7.20. Probability of Mortality Among Males Previously Active
0.00
0.10
0.20
0.30
0.40
0.50
0.60
70
72
74
76
78
80
82
84
86
88
90
92
94
<12 Females 1987
<12 Females 1997
>=12 Females 1987
>=12 Females 1997
212

Figure 7.21. Probability of Mortality Among those with <12 years of
education previously active

0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
70
72
74
76
78
80
82
84
86
88
90
92
94
<12 Males 1987
<12 Males 1997
<12 Females
1987
<12 Females
1997
213
Figure 7.22. Probability of Mortality Among those with >=12 years of
education previously active

-0.05
0.00
0.05
0.10
0.15
0.20
0.25
0.30
0.35
70
72
74
76
78
80
82
84
86
88
90
92
94
>=12 Males 1987
>=12 Males 1997
>=12 Females 1987
>=12 Females 1997
214

Figure 7.23. Probability of Mortality Among those with <12 years of
education previously disabled

0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
70
72
74
76
78
80
82
84
86
88
90
92
94
<12 Males 1987
<12 Males 1997
<12 Females
1987
<12 Females
1997
215
Figure 7.24. Probability of Mortality Among those with >=2 years of
education previously disabled

0.00
0.10
0.20
0.30
0.40
0.50
0.60
0.70
70
72
74
76
78
80
82
84
86
88
90
92
94
>=12 Males 1987
>=12 Males 1997
>=12 Females 1987
>=12 Females 1997
216
Active Life Expectancy
The results show significant differences in life expectancy and active
life expectancy by education group. Without exception, those in the higher
education group (12 or more years) have longer total and active life
expectancies than those in the lower education group (less than 12 years).
The difference is about 1.5 years in 1987 at age 70, and about 1.1 years at
age 70 in 1997 (see Table 7.9). The differences in life expectancy by
education start to fade after age 80, but differences in active life expectancy
remain. In fact, individuals with 12 or more years of education can expect
approximately 2 more years without disability at age 70 than individuals with
less than 12 years. Furthermore, as shown in tables 7.11 and 7.12, there are
clear differences in the proportion of life spent active across the education
groups. For example, a 70 year old with less than 12 years of education in
1987 would expect to spend approximately 78% of their remaining life in an
active state, while other 70-year-old individuals with 12 or more years of
education would expect to spend about 83% of their remaining life without
disabling conditions.
Overall, there appears to be a decrease in the proportion of active life
among the low-educated in every age-group between 1987 and 1997. The
proportion of active life at age 70 is 78% in 1987 and then decreases to 73%
by 10 years later.
217
Table 7.9. Life Expectancy, Active Life Expectancy, Disabled Life
Expectancy, Among Those with <12 years of Education
1987 1997
Age Total LE Active LE Disabled LE Total LE Active LE Disabled LE
70 13.07 10.17 2.90 13.37 9.76 3.61

12.68-13.47 9.86-10.48 2.71-3.09 12.89-13.85 9.40-10.12 3.35-3.87
75 10.10 7.01 3.10 10.56 6.86 3.70

9.79-10.41 6.78-7.23 2.90-3.29 10.17-10.95 6.60-7.12 3.45-3.95
80 7.84 4.64 3.19 8.45 4.77 3.69

7.54-8.13 4.46-4.82 2.99-3.40 8.09-8.81 4.56-4.97 3.42-3.95
85 6.15 2.93 3.22 6.69 2.97 3.72

5.85-6.45 2.78-3.08 2.98-3.46 6.32-7.00 2.80-3.14 3.42-4.02
90

4.47-5.13 1.47-1.70 2.92-3.50 5.02-5.83 1.63-1.91 3.31-4.01

Table 7.10. Life Expectancy, Active Life Expectancy, Disabled Life
Expectancy, Among Those with 12 or More Years of Education
1987 1997
Age Total LE Active LE Disabled LE Total LE Active LE Disabled LE
70 14.66 12.13 2.54 14.54 11.77 2.77

14.18-15.15 11.73-12.52 2.31-2.76 14.12-14.97 11.43-12.11 2.56-2.98
75 11.54 8.91 2.63 11.60 8.79 2.81

11.11-11.97 8.58-9.24 2.39-2.87 11.22-11.99 8.51-9.08 2.59-3.03
80 8.84 6.08 2.76 9.12 6.26 2.86

8.44-9.24 5.79-6.36 2.51-3.02 8.74-9.50 6.00-6.52 2.62-3.11
85 6.72 3.87 2.85 7.04 4.08 2.96

6.34-7.11 3.63-4.10 2.57-3.13 6.65-7.42 3.85-4.31 2.68-3.24
90

4.60-5.35 1.88-2.21 2.62-3.25 5.04-5.84 2.23-2.59 2.70-3.36
Table 7.11. Proportion of Remaining Life Expected Active Among Those
with <12 years of Education
Age 1987 1997
70 0.78 0.73
75 0.69 0.65
80 0.59 0.56
85 0.48 0.44
Table 7.12. Proportion of Remaining Life Expected Active Among Those
with 12 or more Years of Education
Age 1987 1997
70 0.83 0.81
75 0.77 0.76
80 0.69 0.69
85 0.58 0.58

218

Meanwhile, although there was essentially no change in the proportion
of expected life spent in the active state across time, there were clear
differences between the proportions of remaining life between the education
groups in both surveys. The differences between the education groups are
such that the more highly educated group has a similar remaining proportion
of active life as those 5 years older in the group with less education.
For males with a lower education level (Table 7.13), there was no
statistically significant change in the percent of remaining life that is active
across the time period. For both data, the percent of remaining active life
ranged from approximately 80% at age 70 to approximately 45% at age 90.
For females with a lower education level (Table 7.14), the difference
across time was again not statistically significant. This group had a smaller
proportion of remaining active life than their male counterparts, ranging from
70% at age 70 to 30% at age 90.
For more highly educated males (Table 7.15), there was an increase
in the proportion of expected life active. At age 70, males in the 1987 and
1997 cohorts could expect 85% and 87%, respectively, of their remaining
years to be active. This is slightly longer than their counterparts with less
education. A similar pattern existed throughout the age range, in that the
1997 cohort had a greater proportion of their expected remaining years
219
active, and the more highly educated were better-off than those with less
education in both cohorts.
For females with a higher level of education (Table 7.16), there was
also an increase in the percent of remaining life that is spent active. For the
1987 cohort, at age 70, approximately 74% of remaining years were
expected active, while the 1997 cohort had a slightly greater 77% expected
active. The pattern continued very clearly throughout the age range, and
large differences were seen between women of the two education groups.
Table 7.13. Life Expectancy Among Males with <12 Years of Education
1987 1997
Age Total LE Active LE Disabled LE Total LE Active LE Disabled LE
70
11.44 9.47 1.97 11.83 9.63 2.21

10.44-12.45 8.64-10.31 1.60-2.34 11.29-12.37 9.17-10.08 1.98-2.43
75
8.66 6.58 2.08 8.90 6.57 2.34

7.87-9.46 5.95-7.21 1.72-2.45 8.47-9.34 6.23-6.90 2.11-2.56
80
6.54 4.39 2.15 6.95 4.66 2.29

5.91-7.17 3.95-4.84 1.79-2.50 6.58-7.32 4.40-4.92 2.07-2.51
85
5.07 2.92 2.15 5.31 3.00 2.31

4.56-5.58 2.61-3.24 1.79-2.50 4.97-5.64 2.80-3.20 2.07-2.54
90
3.99 1.85 2.14 4.12 1.82 2.30

3.55-4.43 1.65-2.06 1.79-2.48 3.80-4.44 1.67-1.96 2.05-2.56
* denotes 1994 value significantly different from 1984, P<.05
220

Table 7.14. Life Expectancy Among Females with <12 years of
Education
1987 1997
Age Total LE Active LE Disabled LE Total LE Active LE Disabled LE
70
14.61 10.60 4.01 14.69 10.01 4.68

13.50-15.72 9.76-11.43 3.33-4.69 14.10-15.28 9.58-10.44 4.31-5.04
75
11.44 7.34 4.10 11.76 7.18 4.59

12.23-13.09 6.68-7.99 3.43-4.77 11.26-12.26 6.86-7.50 4.23-4.94
80
8.99 4.93 4.06 9.39 4.95 4.44

10.47-12.40 4.44-5.41 3.40-4.72 8.93-9.85 4.70-5.20 4.09-4.80
85
7.08 3.10 3.98 7.39 3.06 4.33

8.14-9.84 2.76-3.43 3.32-4.63 6.93-7.84 2.87-3.26 3.94-4.71
90
5.53 1.66 3.87 5.97 1.87 4.10

6.30-7.85 1.44-1.87 3.20-4.54 5.48-6.46 1.71-2.03 3.67-4.53

Table 7.15. Life Expectancy Among Males with 12 or More Years of
Education
1987 1997
Age Total LE Active LE Disabled LE Total LE Active LE Disabled LE
70
12.66 10.79 1.87 12.86 11.16 1.70

12.23-13.09 10.42-11.17 1.70-2.03 12.36-13.35 10.73-11.60 1.52-1.87
75
9.77 7.82 1.94 10.07 8.37 1.70

9.41-10.12 7.53-8.11 1.77-2.11 9.64-10.49 8.01-8.73 1.52-1.88
80
7.47 5.46 2.00 7.63 5.87 1.76

7.14-7.79 5.22-5.71 1.82-2.18 7.25-8.02 5.57-6.18 1.57-1.95
85
5.73 3.69 2.04 5.73 3.91 1.83

5.42-6.03 3.48-3.89 1.84-2.24 5.38-6.08 3.65-4.16 1.62-2.03
90
4.18 2.00 2.18 4.37 2.50 1.87

3.88-4.47 1.85-2.14 1.95-2.41 4.04-4.69 2.31-2.70 1.64-2.09

221
Table 7.16. Life Expectancy Among Females with 12 or More Years of
Education
1987 1997
Age Total LE Active LE Disabled LE Total LE Active LE Disabled LE
70
15.66 11.64 4.01 15.90 12.28 3.62

15.20-16.11 11.30-11.98 3.75-4.28 15.36-16.44 11.86-12.69 3.31-3.93
75
12.55 8.57 3.99 12.71 9.11 3.60

12.16-12.95 8.30-8.83 3.72-4.25 12.22-13.20 8.76-9.46 3.28-3.91
80
9.82 5.80 4.01 10.06 6.52 3.54

9.43-10.20 5.58-6.03 3.73-4.30 9.59-10.53 6.22-6.82 3.20-3.87
85
7.67 3.69 3.98 7.75 4.21 3.54

7.27-8.07 3.50-3.89 3.65-4.30 7.29-8.21 3.95-4.46 3.18-3.91
90
5.94 2.03 3.91 5.92 2.39 3.53

5.51-6.37 1.88-2.18 3.53-4.29 5.44-6.40 2.20-2.58 3.12-3.94

Table 7.17. Percent of Life Active Among Males with <12 Years of
Education
1987 1997
70 0.83 0.81
75 0.76 0.74
80 0.67 0.67
85 0.58 0.56
90 0.46 0.44

Table 7.18. Percent of Life Active Among Females with <12 Years of
Education
1987 1997
70 0.73 0.68
75 0.64 0.61
80 0.55 0.53
85 0.44 0.41
90 0.30 0.31

Table 7.19. Percent of Life Active Among Males with 12 or More Years
of Education
1987 1997
70 0.85 0.87
75 0.80 0.83
80 0.73 0.77
85 0.64 0.68
90 0.48 0.57

222
Table 7.20. Percent of Life Active Among Females with 12 or More
Years of Education

1987 1997
70 0.74 0.77
75 0.68 0.72
80 0.59 0.65
85 0.48 0.54
90 0.34 0.40

Differences in the implied prevalence of disability were very clear by
education group, especially at ages 85 and older (see Figure 7.25). In both
surveys, the lower education group had higher prevalence of disability than
the higher educated group, but there were no changes over the time period
for either education group.
Figure 7.25. Implied Prevalence of Disability by Education and Survey
Year
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
70
72
74
76
78
80
82
84
86
88
90
92
94
1987 <12
1987 >=12
1997 <12
1997 >=12

223
Further disaggregation of the results by sex revealed greater
consistency among males than females. Among males, there was clear
separation of the education groups at a younger age, with less distinction at
the older ages (Figures 7.26 and 7.27). For females, the distinction is less
evident in the 80s, but the differences were more pronounced in the 90s.
Figure 7.26. Implied Prevalence by Sex and Education: Males
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
70
72
74
76
78
80
82
84
86
88
90
92
94
1987 males <12
1987 males >=12
1997 males <12
1997 Males >=12

Figure 7.27. Implied Prevalence by Sex and Education: Females
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
70
72
74
76
78
80
82
84
86
88
90
92
94
1987 females <12
1987 females >=12
1997 females <12
1997 females >=12
224
Status-based Life Expectancies
The status-based results demonstrate the different effect of initiating
the survey in an already able state compared to initiating the first survey in a
disabled state. Most notable are the large differences between Tables 7.21
and 7.22, as well as between 7.23 and 7.24. Tables 7.21 and 7.23 refer to
the average expected life expectancy in each state among those who were
not disabled at the time of the first survey. For 70-year-old individuals with
less than 12 years of education who were initially healthy (Table 7.21),
approximately three-quarters of their remaining life is expected without
disability (about ten and a half years), with about 3 remaining years expected
disabled. As shown in the table, this does not appear to change across the
decade. Similarly, 70-year-old individuals who were initially able and had 12
or more years of education could expect nearly 2 additional years of active
life expectancy, but a similar amount of years disabled. This longer active
life has the effect of increasing the proportion of remaining life expected in
the active state, and thus leads to a higher proportion of active life among
those initially able, as well as higher relative proportion of active life for those
with 12 or more years of education, relative to those in the same initial ability
state (Tables 7.25 and 7.26).
225
Table 7.21. Status Based Active and Disabled Life Expectancies:
Expected Able and Disabled Years by Decade among Those Initially
Able and <12 Years of Education

1987 1997
Age able disabled able disabled
70 10.41 2.82 10.22 3.46
75 7.68 2.87 7.62 3.45
80 5.52 2.91 5.56 3.42
85 3.91 2.92 4.04 3.37
90 2.80 2.90 2.96 3.29
95 2.07 2.87 2.23 3.20

Table 7.22. Status Based Active and Disabled Life Expectancies:
Expected Able and Disabled Years by Decade among Those Initially
Disabled and <12 years of Education
1987 1997
Age able disabled able disabled
70 4.35 4.81 5.05 5.13
75 2.54 4.56 3.11 4.93
80 1.43 4.24 1.85 4.66
85 0.79 3.88 1.09 4.34
90 0.45 3.51 0.65 4.01
95 0.26 3.17 0.39 3.68

Table 7.23. Status Based Active and Disabled Life Expectancies:
Expected Able and Disabled Years by Decade among Those Initially
Able and >=12 Years of Education
1987 1997
Age able disabled able disabled
70 12.29 2.49 12.09 2.69
75 9.25 2.54 9.20 2.70
80 6.75 2.58 6.83 2.70
85 4.82 2.61 5.00 2.68
90 3.42 2.62 3.65 2.66
95 2.47 2.61 2.70 2.62

226

Table 7.24. Status Based Active and Disabled Life Expectancies:
Expected Able and Disabled Years by Decade among Those Initially
Disabled and >=12 Years of Education
1987 1997
Age able disabled able disabled
70 5.43 4.38 5.44 4.49
75 3.22 4.19 3.33 4.33
80 1.82 3.92 1.95 4.11
85 1.00 3.60 1.12 3.85
90 0.55 3.28 0.64 3.57
95 0.31 2.96 0.37 3.30

Table 7.25. Expected Proportion of Remaining Years Able by Decade
among Those with <12 Years of Education

Initially Able Initially Disabled
1987 1997 1987 1997
70 0.79 0.75 0.48 0.50
75 0.73 0.69 0.36 0.39
80 0.65 0.62 0.25 0.28
85 0.57 0.55 0.17 0.20
90 0.49 0.47 0.11 0.14
95 0.42 0.41 0.08 0.10

Table 7.26. Expected Proportion of Remaining Years Able by Decade
among Those with >=12 Years of Education
Initially Able Initially Disabled
1987 1997 1987 1997
70 0.83 0.82 0.55 0.55
75 0.78 0.77 0.44 0.43
80 0.72 0.72 0.32 0.32
85 0.65 0.65 0.22 0.23
90 0.57 0.58 0.14 0.15
95 0.49 0.51 0.09 0.10

227
Conclusion
In conclusion, the first hypothesis, stating that individuals with less than
12 years of education would have higher disability incidence than those with
12 or more years during the study period, was supported, as both males and
females with less than 12 years of education showed higher incidence of
disability in both the 1987 and 1997 cohorts. The second hypothesis, stating
that declines in incidence would be larger for those with higher levels of
education, was not supported, as changes in incidence appeared to happen
in parallel patterns for both education groups. The third hypothesis, which
predicted that those with less than 12 years of education would have lower
recovery rates than those with 12 or more years, was not supported, since
there were no clear differences in recovery rates for any education level, nor
for any sex-education group. The fourth hypothesis, stating that the rate of
recovery increase would be higher for those with 12 or more years of
education, was not supported, as no changes in recovery were observed for
any group. The fifth hypothesis, that death rates would be higher for those
with less than 12 years of education, was not supported, as no clear
differences in mortality across education groups was observed. The sixth
hypothesis, stating that the trend in death rates would lead to faster decline
for those with 12 or more years of education, was not supported, as there
were no clear differences in the rate of decline in mortality among any
groups. The seventh hypothesis, stating that those with 12 or more years of
228
education would have higher estimates of active life expectancy, was
supported, and these differences were meaningful in both 1987 and 1997.
When further divided by sex, education differences are still apparent, with a
difference of about 2 years longer active life among those with the higher
level of education. The eighth hypothesis, stating that those with 12 or more
years of formal education would experience larger gains in active life
expectancy, was not supported, as significant gains in active life expectancy
were not observed for ether education level, or for any sex-education group.
The ninth hypothesis, stating that those with 12 or fewer years of education
would have higher estimates of disabled life expectancy, was supported in
the 1997 data, but not in the 1987 data. In fact, disabled life expectancy in
the lower educated group in the 1997 data was higher than even estimates in
1987, a difference that was significant at every age, suggesting that a lower
level of education was more closely associated with functioning in the more
recent data.
Discussion
Differences in education are associated with persistent disparities in
both disability and life expectancy. Those with 12 or more years of education
can expect about 10% longer life without disability, as well as a life that is
15% longer on average at age 70. The results from the present study show
that these disparities have not changed in the recent decade. Another study
covering a longer range of time has shown that the disparities across
229
education groups may in fact have increased between 1970 and 1990
(Crimmins & Saito, 2001).
The methodology used in this study limited the analysis to binary
covariates, and thus it was not possible to compare gradients of education
level. However, other studies have shown that education does in fact follow
a gradient, with additional years beyond 12 affording continued improvement
in both total as well as active life expectancy (Crimmins & Saito, 2001; Molla
et al., 2004).
The results suggest that increases in active life expectancy observed
over the last decade have not been the result of improvements for any one
education group. Taken in conjunction with the change in education between
the two LSOA cohorts, the logical conclusion is that compositional change is
driving the increase in observed active life expectancy.

230
Chapter VIII: Summary and Conclusions
A small change in the health needs of the population could have an
enormous effect on the need for expensive supportive care, especially if it
affects the ability to live independently. As the population of older people
grows in both numbers and proportion, it will be of increasing interest to know
what factors are influencing trends in disability and health in the older
population. Measures of active life expectancy provide data to monitor
health trends, examine equity between subgroups of populations, and
provide a basis for planning future health care needs, as well as for linking
interventions to potential outcomes (Bone et al., 1995; 1998; Crimmins,
2003).
At the time that Fries first proposed the compression of morbidity
hypothesis (1981), there was little conclusive evidence to support or refute
the concept. Since then, the concept of active life expectancy has developed
as a way to summarize both health and mortality simultaneously. It can be
useful for understanding the ways in which health status and length of life
change in real populations, and whether there has been a compression or an
expansion of morbidity (Crimmins, 2003).
Summary of Findings for the Overall Population
Each of the chapters in this dissertation has focused on trends in
active life expectancy from a different perspective. The first chapter of
results, Chapter 4, reflected national trends in the dynamics of disability and
231
active life expectancy for the population at two time points. The second
section of results, Chapter 5, reflected trends for males and females in the
dynamics of disability and active life expectancy, as well as differences
between males and females. The third section of results, Chapter 6,
examined trends in the dynamics of disability and active life expectancy for
the white and black populations, as well as differences between whites and
blacks. Finally, the fourth section of results, Chapter 7, explored the
dynamics of disability and active life expectancy for those with different
education levels, as well as differences between the education levels in each
component of disability and active life expectancy.
Briefly, Chapter 4 found overall declines in the incidence of disability in
the 1997 cohort, compared to the 1987 cohort at ages 83 and above. It also
found an increase in the probability of recovery from disability among those
ages 73 and above in the 1997 cohort, as compared to the 1987 cohort.
Although no significant differences in mortality were observed between the
two cohorts, there was a gain in active life expectancy of about 1.0 years at
age 70, about 1.2 years at age 80, and about 1.3 years at age 90. There
was no change in disabled life expectancy over this time, however.
The fourth chapter generally supported the compression of morbidity
hypothesis, as the number of years lived active showed an increase, while
the number of years lived disabled did not. It thus appears that the gains in
life expectancy observed over this time were contained fully in active years.
232
If the observed rates of onset and recovery continued into the future, there
would be a decline in the future prevalence of disability among those ages 82
and above.
Finally, the fourth chapter suggested that the increase in life
expectancy between 1987 and 1997 was not because of decreases in
mortality among either the disabled or the non-disabled at a given age;
rather, it was because of a shift of the population into the lower mortality non-
disabled group.
Results for Males and for Females
The results were further tested among males and females separately.
Differences were apparent across the sexes. In the 1987 cohort, females
had higher disability onset at ages 81 and above, but there were no gender
differences in onset in the 1997 cohort. This change appears to be explained
by an increase in onset for males, and not by a decline among females,
making this shift in gender difference less good news than it might seem at
first glance. No gender differences in recovery rates were found, and neither
males nor females experienced any change in the probability of recovery
over this time period.
Death rates were higher for males in both the non-disabled and
disabled states, as compared to females in 1987 at every age, but not in
1997, where no gender differences could be found. This change appears to
be driven by declines for males. The probability of death for females
233
increased among those in the active state in 1997, at the same time that their
male counterparts were experiencing lower death rates than a decade
earlier. Similarly, the death rates for disabled females were stagnant, at the
same time that their male counterparts had declining death rates. This
divergence in trends for the two sexes likely led to observed increases in
total and active life expectancy for males, while largely stagnant trends in
active and total life expectancy were observed for females of the same age.
Females in both 1987 and 1997 reported more years of remaining
active life than males. Females, however, showed stagnant trends in active
life expectancy, while males showed increases in active life expectancy over
the decade, leading to shrinking gender differences in active life expectancy.
Finally, females had higher estimates of disabled life expectancy than males,
with estimates of about 3.5 years on average of disabled years, compared to
males who could expect to live approximately 2 years in disability.
Results for Blacks and Whites
The results for blacks and whites pointed to some of the inequalities in
different subpopulations in the U.S. Some inequalities improved over the
decade; however, race continues to impact life expectancy and active life
expectancy. Beginning with the dynamics of disability, the 1987 data showed
blacks to have a higher probability of disability onset at ages 75 and above,
but there were no race differences in onset in the 1997 cohort. Race did not
play a role in determining the probability of recovery in either cohort, although
234
there was a statistically insignificant hint that white recovery might have been
slightly higher relative to blacks in 1987. There were no apparent differences
by race in the probability of mortality at a given age, and no differences in the
rate of change for either race.
Turning to active life expectancy, whites did have higher estimates of
active life expectancy in both data sets, although the difference was not
significant above age 80. Whites also had a slightly greater proportion of
active life than their black counterparts. Changes in active life expectancy
were very similar for both races, and no clear race effect for change was
observed. Finally, both blacks and whites reported approximately equal
number of years disabled, and thus, while blacks live a greater proportion of
their lives disabled, the number of years disabled is approximately the same.
When further disaggregated by races and sex, it was apparent that black
females experienced a slight decrease (improvement) in their onset of
disability as well as increased recovery from disability such that it appears
that for this group, the racial gap was closed. There were higher rates of
disability onset for black men as compared to white men in both years, with
neither racial group of males showing any significant improvement over this
time. The one positive effect for black males was their higher recovery rate
in the second survey. This improvement placed their recovery in a position
indistinguishable from that of white men.
235
White males enjoyed the largest increase in life expectancy, with 1.1
years gained at age 70, while at the same time no change was observed for
black males. Females of both races did not show any increase in life
expectancy; in fact, there was a decline in active life expectancy for white
females, although not quite significant. The news that the gender gap in life
expectancy is narrowing could be a sign that men are making modest gains
in their health behaviors. The news that virtually no gains have been made
among blacks suggests that there is room for future improvement in this
area. In the 1997 data, the gap between blacks and whites appeared to
have grown larger, suggesting trends that diverge by race.
For Those of Differing Levels of Education
The results for two education levels pointed to inequalities linked to
socioeconomic status in the U.S. Examining the dynamics of disability for
individuals of 2 education levels found that those with less than 12 years of
education, both male and female, have a higher probability of experiencing
the incidence of disability in both the 1987 and 1997 cohorts. Further, the
difference between the groups did not change. In contrast, there were no
clear differences in recovery rates for any education level, nor for any sex-
education group. Differences in the probability of death for population
subgroups were not found between either education group, and no clear
differences in the rate of decline in mortality were observed between the two
groups.
236
Those with 12 or more years of education had higher estimates of active
life expectancy in both 1987 and 1997. When further divided by sex and
education, differences were still apparent for each group, with a difference
among those with the higher level of education of about 1 year longer active
life for males and 2 years longer active life for females. When examining
increases in active life expectancy over time, significant gains in active life
expectancy were not observed for either education level, nor were gains
observed for any sex-education group. Finally, those with 12 or fewer years
of education had higher estimates of disabled life expectancy in the 1997
data, but not in the 1987 data. In fact, disabled life expectancy in the lower
educated group in the 1997 data was higher than even estimates in 1987, a
difference that was significant at every age. This suggests that a lower level
of education was more closely associated with functioning in the more recent
data.
Substantive Value of Findings
The results suggest that the observed trends are not so much caused
by improving health for all people in the older population, but rather the
changing composition of the population. The older population is gradually
becoming more highly educated as younger cohorts age-in to the older ages.
The older population is becoming increasingly male, as mortality rates for
males decline faster than females. The older population is becoming
increasingly diverse, as minority populations, such as blacks, make up a
237
larger proportion of the population each year. The fact that improvements
were found between 1987 and 1997 in active life expectancy for males, those
with 12 or more years of education, and declining disability onset among
blacks, all suggest increases in active life expectancy for the overall
population.
Few consistent changes were observed over the time period within
any specific category. For example, although black females narrowed the
gap on disability onset and death among the disabled, active life expectancy
was unchanged over the time period for black females. In the example of
education, no changes in onset, recovery, death, or active life expectancy
were observed among any of the education groups, with disparities in
disability and active life expectancy persisting in the later cohort. The only
subgroup for which there was an increase in active life expectancy was for
males, who also showed declines in disability onset and mortality among
both the able and disabled.
This means that efforts to improve the health and disability status of
the elderly have thus far had little impact on disability-based active life
expectancy in the population ages 70 and above. It may be that new
interventions are designed to help those who are already disabled, and thus
might in fact increase disabled years. Programs that expand health services
through collaborations between community-based providers and managed
care, for example, offer services that expand health and social opportunities
238
for seniors, but often require that individuals meet certain disability criteria in
order to qualify (Wilber, Allen, Shannon et al., 2003).
The actual impact of living a disabled life can vary, of course,
depending on the environmental demands on the body. For example, for
married individuals, easy access to caregivers might mitigate the importance
of a disability, and lead to a happier life with disability. Marriage or
cohabitation is linked to quality of life, and thus can impact one’s general life
outlook (Musick, 2004). As male life expectancy increases, so to do the
chances that older people will live in a marital or cohabitation relationship,
thus affecting the actual impact of disabled life years on quality of life.
Methodological Contributions
The IMaCh program is a relatively new approach in the calculation of
active life expectancy, but it is uniquely well-suited for the comparison of
trends over time, as it can account for uneven survey time intervals, and
generate confidence intervals to make determinations of statistical
significance.
Many studies of time trends compare different cross-sectional studies,
or use only one longitudinal data set. The advantage of this investigation is
that it used two longitudinal studies and one method of analysis, leading to
results that are easily comparable, and the statistical significance of change
of time can be conclusively determined.
239
This study uses a measure of disability that is highly comparable to
past studies (Crimmins et al., 1994, 1996; Crimmins & Saito, 2001; Cai &
Lubitz, 2007; Wolf et al., 2007). The results are largely similar to Crimmins
and Saito (2001), who found substantial differences in active life expectancy
(although not total life expectancy) for blacks as compared to whites, and
widening differentials between blacks and whites, as well as differentials
between those less educated as compared to the more highly educated
between 1970-1990. Cai and Lubitz (2007) focused only on gender
differences and trends between 1992 and 2002, and found increases in
active life expectancy only for males. This dissertation builds on these
studies by offering further support for their findings of a decline in years lived
in severe disability among males in the early 1990s and compared to the
1980s, as well as additional analyses on the dynamics of disability processes
for each group.
Implications
The results of this study suggest a relatively bright future for the
disability status of the older population, as the results show an overall trend
towards a smaller proportion of life spent disabled, with life expectancy
increases characterized by additional years without disability. This apparent
compression of morbidity could lead to decreased years spent in institutions,
and more years of living independently.
240
There remains much potential for future disability decline, as it should
be theoretically possible to improving the lives of health-disadvantaged
groups, including minorities and those with lower education levels. Further
attention of policies can be directed towards addressing the reasons why
disadvantaged groups report higher levels of disability, and lower active life
expectancy.
While sex differences in health measures are always likely to exist to
some extent, this dissertation demonstrated that males have begun to close
the gap in some respects, as they showed larger increases in total and active
life expectancy over this time period. It was unclear why males experienced
more dramatic change than females over this time period, but it might be
evidence that some gender differences in health are not determined
biologically, and in fact can be controlled.
Limitations
This dissertation involved a complex analysis of two large datasets,
and was likely affected by several major limitations. The LSOA data,
although designed to comparable, present some challenges to the data
analyst. Ideally, data on multiple waves should be available for most
participants. Sample attrition, which was particularly high in the second
LSOA, limits the number of transitions that can be used in the analysis.
Although several methods were used to test the effect of the missing data, it
is not known why certain cases were lost.
241
The size of the minority population is substantially smaller in the first
LSOA than in the second. This greatly shrinks the available data to stratify
by all of the variables we would have liked, such as race, sex, education, and
age. The second LSOA contained an over-sample of the black and the
population above age 85, while the first LSOA did not. Although these
differences were controlled for by sample weights, the limited sample N in
the first LSOA limits the statistical strength of our comparisons across time.
The IMaCh algorithm only allows for dichotomous covariates, and thus
the analysis of education is based on a semi-arbitrary cut-point of 12 years.
While there is a logical reason for using this value, since a high-school
degree is the doorway to advanced study or managerial level work, it is
possible that different results might be seen with other cut-points.
The sample was limited to the population aged 70 and above. Many
disability and mortality processes begin at mid-life, and thus the age
restriction on the data may limit our understanding of the age-pattern of some
of the dynamics of disability.
The time period of analysis is limited, and patterns may not be as
obvious during a short time-frame, such as the decade of 1987 to 1997, as
they might be over an extended period of time. Disability trends tend to be
cyclical, and the natural increase and decrease in markers of health and
disability may depend on the vantage point of the baseline year chosen, and
the final year defining the data. In order to more fully understand the
242
dynamics of disability, data covering a long period of time, including several
cohorts, would be ideal.
Longitudinal analyses can be affected by the specification of the
model used in the analysis. Each method of analysis has its advantages and
disadvantages, and the bias of a given model may not be apparent unless it
is compared to other methods. The IMaCh program assumes that the
underlying functional form of the association between age (or time) and the
risk of an event follows a Gompertz distribution. While this assumption is
reasonable for death in old age and the onset of certain types of disabilities,
it may not be appropriate for all measures of health.
Policy Implications
The results suggest that the near future is likely to show declines in
disability onset for the elderly, and likely declines in the prevalence of
disability as younger cohorts reach the older age groups. While these trends
represent positive news for policy-makers, by suggesting that the demand for
supportive services may increase more slowly than the growth of the older
population, there are still places in which policy interventions could lead to
further declines. Needing particular attention are sub-populations not
experiencing improvement over time, such as blacks, females, and those
with less than 12 years of education.
The elderly population is constantly changing in its composition of
individuals by gender, race, and education. The trend over the last decade
243
shows a growing proportion of men surviving to the oldest ages, and this may
be a sign of declining differences in mortality and health by gender. The
older population is also gradually becoming composed of a higher proportion
of individuals with a high school education and above. Since higher levels of
education are associated with improved health and decreased mortality (and
therefore longer life expectancy), this change in the composition of the
population alone should lead to a decline in the prevalence of some health
conditions, and an increase in life expectancy.
The elderly population is also becoming more diverse, with growing
proportions of blacks and other minorities. The black population has not
enjoyed increases in active life expectancy in the recent decade; however,
there were some signs that incidence has declined and recovery has
increased among blacks.
Although the results of the present study suggest that recent
increases in life expectancy have not been associated with additional years
with disability, it should be noted that the definition only accounts for severe
disability. While severe disability is likely to be a good approximation for
needing a supportive environment, it is also true that less severe disability,
which is not captured in this analysis, can also lead to the need for a more
assistive environment.
244
Recommendations for further studies
The definition of disability is open to contentious debate, as disability
is a complex concept. Not all disabilities prevent individuals from living in
their own home, as the level of severity is important to consider, as well as
the effect of the disability on quality of life. A variety of approaches have
been developed to operationalize variations in the level of disability, ranging
from preclinical mobility disability (Fried et al., 2000) to subclinical functional
limitation (Wolinsky et al., 2007), or even looking at biological markers that
predict later-disability onset (Crimmins, 2007). Others look at more severe
activity limitations that use only functioning inability or limitations in mobility
(Wolf et al., 2007).
This dissertation attempted to show some of the demographic
characteristics that might be related to differences in active life expectancy.
The analysis was not exhaustive of all such measures, most notably the
combination of race and education and tests of interactive effects.

245
BIBLIOGRAPHY

Albert, S.M. , Bear-Lehman, J., Burkhardt, A., Merete-Roa, B.,
Noboa-Lemonier, R., and Teresi, J. (2006). Variation in Sources of
Clinician-Rated and Self-Rated Instrumental Activities of Daily
Living Disability. J. Gerontol. A Biol. Sci. Med. Sci., 61(8):826 - 831.
Anderson, R.T., James, M.K., Miller, M.E., Worley, A.S., and Longino,
C.E. (1998). The timing of change: Patterns in transitions in
functional status among elderly persons. J Gerontol B-Psychol,
53:S17-27.
Avlund, K, Damsgaarda, M.T. and Holstein, B.E. (1998). Social
relations and mortality. An eleven year follow-up study of
70-year-old men and women in Denmark. Social Science &
Medicine, 47(5):635-643.
Barbeau, E.M., Krieger, N. and Soobader, M.J. (2004). Working class
matters: socioeconomic disadvantage, race/ethnicity, gender, and
smoking in NHIS 2000. Am. J. Public Health 94 (2):269–278.
Barberger-Gateau, P., Fabrigoule, C., Amieva, H., Helmer, C. and
Dartigues, J.F. (2002). The Disablement Process: A Conceptual
Framework for Dementia-Associated Disability. Dement Geriatr
Cogn Disord, 13:60–66.
Bandeen-Roche, K., Xue, Q.L., Ferrucci, L., Walston, J., Guralnik, J.M.,
Chaves, P., Zeger, S.L., and Fried, L.P. (2006). Phenotype of frailty:
characterization in the women's health and aging studies. Journals
of Gerontology:-Biological Sciences & Medical Sciences,
6:262-266.
Barendregt, J. (2003). Disability-adjusted Life Years (DALYs) and
Disability-adjusted Life Expectancy (DALE). In Determining Health
Expectancies edited by J.-M. Robine, C. Jagger, C.D. Mathers, E.M.
Chichester, England: John Wiley & Sons:247-262.
Barendregt, J. J., Bonneux, L., & van der Ma, P. J. (1998). Health
expectancy: from a population health indicator to a tool for policy
making. Journal of Aging and Health, 10:242-258.
Bebbington, A.C. (1991). The Expectation of Life without Disability in
England and Wales: 1976-88. Population Trends, 66:26-29.
246

Bebbington, A.C. (1993). Regional and Social Variations in
Disability-Free Life Expectancy in Great Britain. Calculation of
health expectancies: harmonization, consensus achieved and
future perspectives. In J.-M. Robine, C. D. Mathers, M. R. Bone and
I. Romieu, Colloque INSERM: John Libbey Eurotext. 226: 175-191.
Bebbington, A. & Bajekal, M. (2003). Sub-national Variations in Health
Expectancy. In Robine JM, Jagger C, Mathers CD, Crimmins EM,
Suzman RM, (Eds). Determining Health Expectancies. Chichester,
England: John Wiley & Sons:127-147.
Berkman, L.B., Singer, B., and Manton, K. (1989). Black/White
Differences in Health Status and Mortality Among the Elderly.
Demography, 26: 661-678.
Blackwell, D.L., Hayward, M.D., Crimmins, E.M. (2001). Does
Childhood Health Affect Chronic Morbidity in. Later Life? Social
Science and Medicine, 52:1269-1284.
Bone, M.R., Bebbington, A.C., Jagger, C., Morgan, K. and Nicolaas, G.
(1995). Health Expectancy and Its Uses. London, HMSO.
Bossuyt, N., Gadeyne, S., Deboosere, P., and Van Oyen, H. (2004).
Socio-economic inequalities in health expectancy in Belgium. Public
Health,118:3-10.
Branch, L.G., Guralnik, J.M., Foley, D.J., Kohout, F.J., Wetle, T.T.,
Ostfeld, A., and Katz, S. (1991). Active life expectancy for 10,000
Caucasian men and women in three communities. Journal of
Gerontology: Medical Sciences, 46:M145-M150.
Branch, L.G., Katz, S., Kniepmann, K. and Papsidero, J.A. (1984). A
prospective study of functional status among community elders.
American Journal of Public Health, 74(3):266-268.
Brandt, E.N., Pope, A.M. (1997). (Eds). Enabling America: Assessing
the Role of Rehabilitation Science and Engineering. Washington,
DC: National Academy Press.
Brody, J.A. (1984). Facts, projections and gaps concerning data on
aging. Public Health Reports, 99, 5:468-475.
Brody, J.A. (1982). Life expectancy and the health of older persons.
American Geriatric Society, 20:681-683.
247

Bronnum-Hansen, H., Andersen, O., Kjoller, M., and Rasmussen, N.K.
Social gradient in life expectancy and health expectancy in
Denmark. Soz Praventivmed,:49:36-41.
Brouard, N. & Lievre, A. (2001) Computing health expectancies using
IMaCh: A maximum likelihood computer program using interpolation
of Markov Chains (Version V91h, May 2001) [Computer software].
Institute National d'Etudes Deemgraphiques (INED, Paris) in the
Mortality, Health and Epidemiology Research Unit.
Cai, L. & Lubitz, J. (2007). Was There Compression of Disability for
Older Americans From 1992 to 2003? Demography, 44(3):479-495.
Cambois, K., Robine, J.M., and Hayward, M. (2001). Social inequalities
in disability-free life expectancy in the French male population.
Demography, 38:513-524.
Cambois, E., Robine, J.M., and Hayward, M.D. (2000). Social
Inequalities in Disability Free Life Expectancy in the French Male
Population, 1980-1991. Demography. 38(4):513-24.
Centers for Disease Control (1986). Premature mortality in the United
States: Public health issues in the use of years potential life lost.
MMWR, 35 (Suppl. 25): 1S-11S.
Chi, D.Q., and Labat, J-C. Projection de population totale por la France:
1975-2020. Institut national de la statistique et des etudes
economiques (Paris 1979) 10 pp., (Les collections de I’INSEE:
demographie et employ; 63-D).
Chiang, C.L. (1965). An index of health: Mathematical models. NCHS,
Vital and Health Statistics, Series 2, No. 5, (Washington, 1965).
Chiang, C.L. (1968) Introduction to stochastic processes in biostatistics.
J Wiley (New York 1968)
Chirikos, T.N., and Nestel, G. (1985). Longitudinal analysis of functional
disabilities in older men. Journal of Gerontology, 40(4):426-433.
Clark, D.O., Maddox, G.L., and Steinhauser, K. (1993). Race, aging,
and functional health. Journal of Aging and Health, 5, 536-553.
248

CMS, (2005). “Medicare Learning Network Matters: Medicare Part A
Skilled Nursing Facility (SNF) Prospective Payment System (PPS)
Pricer Update and Health Insurance Prospective Payment
System(HIPPS) Coding Update Effective January 1, 2006.” Center
for Medicare and Medicaid Services. Available online at:
http://www.cms.hhs.gov/MLNMattersArticles/downloads/MM3962.p
df
CMS, (2006). “Home Health Agency Risk-Adjusted and Descriptive
Outcome Reports: Description and Definitions.” Center for Medicare
and Medicaid Services. Available online at:
http://www.cms.hhs.gov/HomeHealthQualityInits/downloads/HHQI
OBQIMeasureCalculation.pdf
Colvez, A. (1980). Evolution de l’etat de sante au cours de la derniere
decennie: peut-on continuer a parler d’amelioration. Ministere des
Affaires Sociales, Service des etudes epidemiologiques (Quebec
1980) 10 pp.
Colvez, A. (1979). Quatre documents sur l’introduction de nouveaux
indicateurs de sante pour la planification. Ministere des Affaires
Sociales (Quebec 1979).
Colvez A. and Blanchet, M. (1981). Disability Trends in the United
States Population 1966-1976: Analysis of Reported causes.
American Journal of Public Health, 71:464-471.
Costa, D. (2000) Working Paper Series W7605 (Natl. Bureau Econ.
Res., Cambridge, MA), pp. 1-40.
Crimmins, E.M. (1981). The changing pattern of American mortality
decline, 1940-77, and its implications for the future. Population &
Development Review, 7(2):229-254.
Crimmins, E.M. (2003). Disability, demography of. In R. Dement & G.
McNicoll (Eds.) Encyclopedia of Population. New York:McMillan
Reference: 241-242.
Crimmins, E.M. (2004). Trends in the Health of the Elderly. Annual
Review of Public Health, 25:79-98
Crimmins, E.M., Cambois, E. (2003). Social inequalities in health
expectancies. In: Robine JM, Jagger C, Mathers CD, Crimmins EM,
Suzman RM (Eds.) Determining Health Expectancies. Chichester,
England: John Wiley & Sons:111-125.
249

Crimmins, E. and Hayward, M. (1995). What Can We Learn About
Competence at the Older Ages From Active Life Expectancy.
Societal Mechanisms for Maintaining Competence in Old Age. S.
Willis, K. W. Schaie and M. Hayward.New York, Springer Publishing
Company.
Crimmins, E. M., Hayward, M.D., and Saito, Y. (1994). Changing
Mortality and Morbidity Rates and the Health Status and Life
Expectancy of the Older Population. Demography, 31:159-175.
Crimmins, E., and Hayward, M.D. (1992). The relationship between
changing mortality and morbidity rates and life cycle health status
and the prevalence of health status in the older population. Annual
Meeting of the Population Association of America, Denver,
Colorado.
Crimmins, E. M., Hayward, M.D., and Saito, Y. (1996). Differentials in
Active Life Expectancy in the Older Population of the United States.
Journal of Gerontology 51B(3):S111-S120.
Crimmins, E.M., Kim, J.K., Hagedorn, A. (2002). Life with and without
disease: women experience more of both. J Women
Aging;14(1-2):47-59.
Crimmins, E.M. and Ingegneri, D. (1993). Trends in Health Among the
American Population. Pp. 225–42 in Demography and Retirement:
The Twenty First Century, edited by A.M. Rappaport and S.J.
Schieber. Westport, CT: Praeger.
Crimmins, E.M. & Saito, Y. (1993). Getting better and getting worse:
Transitions in functional status among older Americans. Journal of
Aging and Health, 5:3-36.
Crimmins, E.M. and Saito, Y. (2001). Trends in Disability- free Life
Expectancy in the United States, 1970-1990: Gender, Racial, and
Educational Differences. Social Science and Medicine, 52:1629-41.
Crimmins, E.M., Saito, Y., and Hayward, M.D. (1993). Sullivan and
Multistate Methods of Estimating Active Life Expectancy: Two
Methods, Two Answers. In J.M. Robine, C. D. Mathers, M.R. Bone,
and I. Romieu (Eds.) Calculation of Health Expectancies:
Harmonization, Consensus Achieved and Future Perspectives.
Paris, John Libbey, Ltd., pp. 155-160.
Crimmins, E.M., Saito, Y., Ingegneri, D. (1989). Changes in Life
Expectancy and Disability-Free Life Expectancy in the United
States. Population and Development Review, 15:235-67.
250
Crimmins, E.M., Y. Saito, and D. Ingegneri. (1997a). Trends in
Disability-free Life Expectancy in the U.S.: 1970-1990. Population
and Development Review, 23(3):555-72.
Crimmins, E.M., Saito, Y. and Reynolds, S. (1997b). Further Evidence
on Recent Trends in Two Sources: The LSOA and the NHIS.
Journal of Gerontology: Social Sciences, 52B(2):S59-S71.
Crimmins, E.M., Suzman, R.M., Laditka, S.B., and Laditka, J.N. (2006).
Effects of Diabetes on Healthy Life Expectancy: Shorter Lives with
More Disability for Both Women and Men. Pp. 71-90 in Longer Life
and Healthy Aging edited by Zeng Yi, E.M. Crimmins, Y. Carriere,
and J.-M. Robine. Dordrecht, The Netherlands: Springer.
Cutler, D.M. (2001).The reduction in disability among the elderly. Proc.
Natl. Acad. Sci. USA 98:6546–47
Deeg, D.J.H., Verbrugge, L.M., & Jagger, C. (2003). Disability
Measurement. In: Robine JM, Jagger C, Mathers CD, Crimmins
EM, Suzman RM (Eds.) Determining Health Expectancies.
Chichester, England: John Wiley & Sons:203-220.
Diehr, P., Derleth, A., Cai, L., and Newman, A.B. (2007). The effect of
different public health interventions on longevity, morbidity, and
years of healthy life. BMC Public Health, 7:52.
Doblhammer, G., Kytir, J. (1998). Social inequalities in disability-free
and healthy life expectancy in Austria. Wiat Klin Wochensehr.
1998:110:393-396.
Dubois, M. F, and Hebert, R. (2006). Cognitive-impairment-free life
expectancy for Canadian seniors. Dementia and Geriatric Cognitive
Disorders, 22(4):327-333.
Elias, M. F. and Elias, P.K. (1997). Commentary: Active life expectancy:
Concept or model for research on competency? In S. L. Willis, K. W.
Schaie and M. Hayward (eds). Societal Mechanisms for Maintaining
Competence in Old Age. New York, Springer Publishing Company:
35-49.
Elo, I.T. (2001). New African American life tables from 1935-1940 to
1985-1990. Demography, 38, 97-114.
Elo, I.T., & Preston, S.H. (1997). Racial and ethnic differences in
mortality at older ages. In L.G. Martin and B.J. Soldo (Eds.), Racial
and ethnic differences in the health of older Americans (pp. 10-42).
Washington, DC: National Academy Press.
251

Escobedo, L.G. and Peddicord, J.P. (1996). Smoking prevalence in US
birth cohorts: the influence of gender and education, Am. J. Public
Health, 86 (2): 231–236.
Feinleib, M. (1980). Recommendations for future research. In US
Department of Health and Human Services, Second conference on
the epidemiology of aging, Hyattsville, MD. Public Health Service,
National Institutes of Health 1980, pp. 359-362.
Femia, E.E., Zarit, S.H., and Johansson, B. (2001). The disablement
process in very late life: A study of the oldest-old in Sweden. Journal
of Gerontology: Psychological Sciences 56B:P12–P23.
Finch, C.E. and Pike, M.C. (1996). Maximum life span predictions from
the Gompertz mortality model. Journals of Gerontology Series A:
Biological Sciences and Medical Sciences, 51(3) B183-B194.
Fogel, R.W. (1994) Economic Growth, Population Theory, and
Physiology: The Bearing of Long-Term Processes on the Making of
Economic Policy. Am. Econ. Rev. 84:369-395.
Fogel, R. W. and Costa, D.L. (1997). A theory of technophysio
evolution, with some implications for forecasting population, health
care costs and pension costs. Demography, 34:49-66.
Franco, O.H., de Laet, C., Peeters, A., Jonker, J., Mackenbach, J.,
Nusselder, W. (2005). Effects of physical activity on life expectancy
with cardiovascular disease. Arch Intern Med, 14;165(20):2355-60.
Freedman, V.A., and Soldo, B.J. (1994). Forecasting disability:
Workshop summary. Convened by the Committee on National
Statistics of the Commission on Behavioral and Social Sciences and
Education, National Research Council. October 4-5 1993.
Washington DC: National Academy PressSoldo.
Freedman, V.A., and Martin, L.G. (1998). Understanding trends in
functional limitations among older Americans. American Journal of
Public Health 88(10):1457-1462.
Freedman, V.A., Martin, L.G., Schoeni, R.F. (2002). Recent trends in
disability and functioning among older adults in the United States: a
systematic review. JAMA, 25;288(24):3137-46.
252

Freedman, V.A., Crimmins, E.M., Schoeni, R.F., Spillman, B.C., Aykan,
H., Kramarow, E., Land, K., Lubitz, J., Manton, K.G., Martin, L.G.,
Shinberg, D. and Waidmann, T.A. (2004). Resolving
Inconsistencies in Trends in Old-Age Disability: Report From a
Technical Working Group. Demography 41(3):417-41.
Freedman, V.A. (2006). Late-Life Disability Trends: An Overview of
Current Evidence. In Marilyn J. Field, Allan M. Jette, and Linda
Martin (Eds.) Workshop on Disability in America: A New Look –
Summary and Background Papers. Washington, DC : National
Academies Press. Also available online at:
http://www.nap.edu/catalog/11579.html
Fried, L.P., Bandeen-Roche, K., Chaves, P.H.M., and Johnson, B.A.
(2000). Preclinical Mobility Disability Predicts Incident Mobility
Disability in Older Women. The Journals of Gerontology, Series A,
M43(10)
Fries, J. F. (1980). Aging, natural death, and the compression of
morbidity. The New England Journal of Medicine, 303:130-135.
Fuchs, V.R. (1984). Though much is taken: reflections on aging, health
and medical care. Milbank Memorial Fund Quarterly/Health and
Society, 62(2):143-165.
Geronimus, A.T., Bound, J., Waidmann, T.A., Colen, C.G., and Steffick,
D. (2001). Inequity in life expectancy, functional status, and active
life expectancy across selected black and white populations in the
United States. Demography, 38:227-251.
Gibson, R.C. (1994). The Age-by-Race Gap in Health and Mortality in
the Older Population: A social science research agenda. The
Gerontologist, 34:454-462.
Gill, T.M. and Gahbauer, E.A. (2005). Overestimation of Chronic
Disability Among Elderly Persons. Archives of Internal Medicine,
165(22):2625-2630.
Gill, T.M. and Kurland, B. (2003). The Burden and Patterns of Disability
in Activities of Daily Living Among Community-living Older Persons.
The Journals of Gerontology Series A: Biological Sciences and
Medical Sciences, 58:M70-M7.
Gori, G.B., Richter, B.J., Yu, W.K. (1984). Economics and extended
longevity: a case study. Preventive medicine, 13:396-410.
253

Gruenberg, E.M. (1977). The failures of success. Milbank Memorial
Fund Quarterly– Health and Society, 55(1):3–24.
Gruenewald TL. Karlamangla AS. Greendale GA. Singer BH. Seeman
TE. (2007). Feelings of Usefulness to Others, Disability, and
Mortality in Older Adults: the MacArthur Study of Successful Aging.
Journals of Gerontology Series B-Psychological Sciences & Social
Sciences, 62(1):P28-37.
Grundy, E and G. Holt. (2001). The socioeconomic status of older
adults: How should we measure it in studies of health inequalities?
J. Epidemiol. Community Health, 55(12): 895 - 904.
Grundy, E. & Glaser, K. (2000). Socio-demographic differences in the
onset and progression of disability in early old age: A longitudinal
study. Age Ageing, 29:149-57.
Guralnik JM, Land KC, Blazer D, Fillenbaum GG, Branch LG. (1993).
Educational status and active life expectancy among older blacks
and whites. N Engl J Med., 329:110-116.
Hayward, M., Crimmins, E., and Saito, Y. (1998). Cause of Death and
Active Life Expectancy in the Older Population of the United States.
Journal of Aging and Health, 10(2):192-213
Hayward, M. and M. Heron (1999). Racial Inequality in Active Life
Among Adult Americans. Demography, 36(1):77-91.
Hayward, M. D. and D. Lichter, T. (1998). A life cycle model of labor
force inequality. Sociological Methods & Research 26(4): 487-510.
Hayward, M. D. and M. A. Hardy (1985). Early retirement processes
among older men. Occupational differences. Research On Aging
7(4): 491-515.
Hayward, M. D. and W. R. Grady (1990). Work and Retirement Among
a Cohort of Older Men in the United States, 1966-1983.
Demography 27(3): 337-56.
Hayward, M. D., D. K. McLaughlin, et al. (1996). Inequality in Men's
Mortality: The SES Gradient and Geographic Context.
Hayward, M.D., E.M. Crimmins, and Y. Saito. 1998. Cause of Death
and Active Life Expectancy in the Older Population of the United
States. Journal of Aging and Health 10:192-213.
254

Hayward, M. D., E. M. Crimmins, et al. (1994). The relationship between
retirement life cycle changes and older men's labor force
participation rates. Journal of Gerontology, 49(5): S219-30.
Hayward, M. D., E. M. Crimmins, et al. (1994). Cognitive Functioning
Changes Among the Chronically Impaired Elderly in the United
States. Advances in Health Expectancies: Proceedings of the 7th
Meeting of the International Network on Health Expectancy
(REVES), Canberra, Australian Institute of Health and Welfare.
Hayward, M. D., S. Friedman, et al. (1996). Race Inequities in Men's
Retirement. Journal of Gerontology 51B(1): S1-S10.
Hayward, M. D., S. Friedman, et al. (1998). Career trajectories and
older men's retirement. Journal of Gerontology: Social Sciences
53B(2): S91-S103.
Hayward, M. D., W. R. Grady, et al. (1988). Recent Changes in Mortality
and Labor Force Behavior Among Older Americans: Consequences
for Nonworking Life Expectancy. Journal of Gerontology: Social
Sciences 43(6): S194-199.
Hayward, M.D. and W.R. Grady. (1990). Work and Retirement Among a
Cohort of Older Men in the United States, 1966-1983. Demography
27:337-356.
Hayward, M. D., W. R. Grady, et al. (1988). Changes in the retirement
process among older men in the United States: 1972-1980.
Demography, 25(3):371-86.
Hayward, M. D., W. R. Grady, et al. (1989). Occupational influences on
retirement, disability, and death. Demography, 26(3): 393-409.
Hayward, M., W. R. Grady, et al. (1988). The retirement process among
older women in the United States. Changes in the 1970s. Research
on Aging, 10(3): 358-82..
Hayward, M.D., E.M. Crimmins, T.P. Miles and Y. Yang (2000). The
Significance of Socioeconomic Status in Explaining the Race Gap in
Chronic Health Conditions. American Sociological Review
65:910-930.
Hayward-M, D. and A. Hardy-M (1985). Early retirement processes
among older men. Occupational differences. Research On Aging
7(4): 491-515.
255

Hemingway H, Nicholson A, Stafford M, Roberts R, Marmot M. The
impact of socioeconomic status on health functioning as assessed
by the SF-36 questionnaire: the Whitehall II Study. Am J Public
Health. 1997;87:1484-1490.[Abstract]
HHS, 2000. Cost-Effectiveness of Home and Community-Based
Long-Term Care Services. U.S. Department of Health and Human
Services. Available online at
http://aspe.hhs.gov/daltcp/reports/costeff.htm
Fries, J.F. and L. Crapo. (1981). Vitality and Aging: Implications of the
Rectangular Curve, New York: W.H. Freeman.
Jackson, C.H., and Sharples, L.D. (2002). Hidden Markov Models for
the onset and progression of bronchiolitis obliterans syndrome in
lung transplant recipients. Statistics in Medicine, 21: 113-128.
Jagger, C., Goyder, E., Clarke, M., Brouard, N., Arthur, A. (2003).
Active life expectancy in people with and without diabetes. Journal
of Public Health Medicine, 25(1):42-46.
Jagger, C., Hauet, E. & Brouard, N. (2001). Health Expectancy
Calculation by the Sullivan Method: A Practical Guide. Reves
International Network on Health expectancy. Working Paper.
Available online at: http://www.reves.net/pdf/RP408.pdf
Jette, A.M. (1999). Disentangling the process of disablement. Soc Sci
Med., 48(4):471-2.
Jonker, J.T., De Laet. C., Franco, O.H., Peeters, A., Mackenbach, J.,
Nusselder, W.J. (2006). Physical activity and life expectancy with
and without diabetes: life table analysis of the Framingham Heart
Study. Diabetes Care, 29(1):38-43.
Kandrack, M, Grant, KR, Segall, A. (1991). Gender differences in health
related behavior: some unanswered questions. Soc Sci Med., 32:
579-590.
Kaneda, T., Zimmer, Z. and Tang, Z. (2005). Socioeconomic Status
Differentials in Life and Active Life Expectancy Among Older Adults
in Beijing. Disability and Rehabilitation. 27(5): 241-251.
Kaprio, J., Sama, S., Fogelholm, M., Koskenvuo, M.. (1996). Total and
occupationally active life expectancies in relation to social class and
marital status in men classified as healthy at 20 in Finland. J
Epidemiol Community Health, 50:653-660.
256

Karup, J. (1893). Die Finanziage der Gothaischen
Staatsdiener-Wittwen-Societat, Desden.
Katz, S., Akpom, C.A. (1976). A measure of primary sociobiological
functions. Int J Health Serv. 6(3):4 3-508.
Katz, S., Branch, L.G., Branson, M.H., Papsidero, J.A., Beck, J.C.,
Greer, D.S. (1983). Active life expectancy. The New England
Journal of Medicine, 309, 20: 1218-1224.
Koster, A., B. W. J. H. Penninx, H. Bosma, G. I. J. M. Kempen, T. B.
Harris, A. B. Newman, R. N. Rooks, S. M. Rubin, E. M. Simonsick, J.
T. M. van Eijk, and S. B. Kritchevsky. (2005). Is There a Biomedical
Explanation for Socioeconomic Differences in Incident Mobility
Limitation? J. Gerontol. A Biol. Sci. Med. Sci., 60(8):1022-1027.
Kovar MG, Poe G. The National Health Interview Survey design,
1973-84, and procedures, 1975-83. National Center for Health
Statistics. Vital Health Stat 1(18). 1985.
Kramer, M. (1980) The rising pandemic of mental disorders and
associated chronic diseases and disabilities. Acta Psychiatrica
Scandinavica, 62(285):282–297.
Kramer, M. (1980). The rising pandemic of mental disorders and
associated chronic diseases and disability. Acta Psychiatr Scand
Suppl.;285:382-397.
Laditka, S.B., and J.N. Laditka. (2006). Effects of Diabetes on Healthy
Life Expectancy: Shorter Lives with More Disability for Both Women
and Men. Pp. 71-90 in Longer Life and Healthy Aging edited by
Zeng Yi, E.M. Crimmins, Y. Carriere, and J.-M. Robine. Dordrecht,
The Netherlands: Springer.
Laditka. S.B. and Laditka, J.N. (2002). Recent perspectives on active
life expectancy for older women. Journal of Women & Aging.
14(1-2):163-84
Laditka, S. B. (1998). Modeling lifetime nursing home use under
assumptions of better health. Journal of Gerontology: Social
Sciences 53B(4): S177-S187.
Laditka, S. B. and D. A. Wolf (1995). Microsimulation of Active Life
Expectancy Methods and Policy Applications. The 8th Meeting of
REVES,Chicago, Center for Policy Research, Syracuse University.
Laditka, S.B. and D. Wolf. (1998). New Methods for Analyzing Active
257
Life Expectancy. Journal of Aging and Health 10:214-41.
Laditka, S.B., M.D. Hayward. (2003). The Evolution of Demographic
Methods to Calculate Health Expectancies. In Determining Health
Expectancies edited by J.-M. Robine, C. Jagger, C.D. Mathers, E.M.
Chichester, England: John Wiley & Sons:221-234.
Lamb, V. L. (1999). Active life expectancy of the elderly in selected
Asian countries. NIHON University Population Research Institute,
Tokyo, Japan.
Land, K., J. Guralnik, and D. Blazer. (1994). Estimating
Increment-Decrement Life Tables with Multiple Covariates from
Panel Data: The Case of Active Life Expectancy. Demography
31:289-319.
Lawton MP, Brody EM. (1969). Assessment of older people:
self-maintaining and instrumental activities of daily living.
Gerontologist. 9(3):179-86.
Lawton M. P., Nahemow L., (1973). Ecology and the aging process.
Eisdorfer C., Lawton M. P., , ed.Psychology of Adult Development
and Aging 619-674. American Psychological Association,
Washington, DC.
Le Bras H, Tapinos G. (1979). Perspectives a long terme de la
population francaise et leurs implications economiques. Population
numero special:1403-1407.
Leveille, S.G., Penninx, B.W.J.H., Melzer, D., Izmirlian, G., & Guralnik,
J.M. (2000). Sex differences in the prevalence of mobility disability
in old age: The dynamics of incidence, recovery, and mortality.
Journal of Gerontology: Social Sciences, 55B, S41-S50.
Li, MF. (2003). Active Life Expectancy in Taiwan. Dissertation,
University of South Carolina.
Lichter, D. T., R. N. Anderson (1995). Marriage markets and marital
choice. Journal of Family Issues 16(4): 412-431.
Liévre, A., N. Brouard, and C. Heathcote. (2003). The Estimation of
Health Expectancies from Cross-Longitudinal Surveys.
Mathematical Population Studies 10:211-48. 18
258

Ling, S.M., Qian Li Xue, Eleanor M. Simonsick, Jing Tian, Karen
Bandeen-roche, Linda P. Fried, Joan M. Bathon. (2005).Transitions
to mobility difficulty associated with lower extremity osteoarthritis in
high functioning older women: Longitudinal data from the Women's
Health and Aging Study II. Arthritis Care & Research, 55(2): 256 -
263.
Liu X, Liang J, Muramatsu N, Sugisawa H. Transitions in functional
status and active life expectancy among older people in Japan.
Journal of Gerontology: Social Sciences 1995;50B(6):S383-S394.
Lopez, A.D., Mathers, C.D., Ezzati, M., Jamison, D.T., & Murray, C.J.L.
(2006) Global Burden of Disease and Risk Factors New York:
Oxford University Press.
Lubitz J, Cai L, Kramarow E, Lentzner H. 2003. Health, life expectancy,
and health care spending among the elderly. N Engl J Med.,
11;349(11):1048-55.
Lych, S. (2007). Bayesian Monte-Carlo program for estimating active
life expectancy. Contact Scott Lynch for the documentation and
code: slynch@princeton.edu
Lynch, S.M., J.S. Brown, and K.G. Harmsen. 2003. The Effect of
Altering ADL Thresholds on Active Life Expectancy Estimates for
Older Persons. Journals of Gerontology Series B-Psychological
Sciences & Social Sciences 58:S171-S178.
McKinlay, J.B., McKinlay, S.M., Jennings, S., and Grant, K. (1983).
Mortality, morbidity, and the inverse care law. In Greer, A.L. and
Greer, S. (eds.). Cities and Sickness. Beverly Hills: Sage.
Maddox GL, Clark, DO, Steinhauser K. (1994). Dynamics of functional
impairment in late adulthood. Soc Sci Med, 38: 925-36. et al., 1994
Manton, K.G. (1982). Changing concepts of morbidity and mortality in
the elderly population. Milbank Memorial Fund Quarterly/Health and
Society, 60(2):183-244.
Manton, K.G. (1988). A longitudinal Study of Functional Change and
Mortality in the United States. Journal of Gerontology: Social
Sciences, 43:S153-S161
Manton, K.G., Corder, L. and Stallard, E. (1997). Chronic disability
trends in elderly United States populations: 1982-1994 Proc. Natl.
Acad. Sci. USA Vol. 94, pp. 2593-2598.
259
Manton, K.G., Corder, L. and Stallard, E. (1993). Estimates of Change
in Chronic Disability and Institutional Incidence and Prevalence
Rates in the U.S. Elderly Population From the 1982, 1984, and 1989
National Long Term Care Survey. Journal of Gerontology: Social
Sciences 48:S153-S166.
Manton, K.G. and Gu, X. (2001). Changes in the Prevalence of Chronic
Disability in the United States Black and Nonblack Population Above
Age 65 From 1982 to 1999. Proceedings of the National Academy of
Sciences 98:6354–59.
Manton, K.G., X. Gu, X. and Lamb, V.L. (2006). Change in chronic
disability from 1982 to 2004/2005 as measured by long-term
changes in function and health in the U.S. elderly population.
Proceedings of the National Academy of Sciences of the United
States of America. 103(48):18374-9.
Manton, K.G., and Land, K. (2000). Active Life Expectancy Estimates
for the US Elderly Population: A Multidimensional Continuous
Mixture Model of Functional Change Applied to Completed Cohorts,
1982-1996. Demography 37:253-265.
Manton, K.G., and Soldo, B.J. Dynamics of health changes in the oldest
old: new perspectives and evidence. Milbank Memorial Fund
Quarterly/Health and Society 1985, 63, 2: 206-285.
Manton, K.G., and Stallard, E. (1982). Temporal trends in US multiple
causes of death mortality data: 1968 to 1977. Demography, 19(4):
527-547.
Manton, K.G., and Stallard E. (1991). Cross-sectional estimates of
active life expectancy for the U.S. elderly and oldest-old
populations. J Gerontol., 46(3):S170-182.
Manton, K.G., and Stallard, E. (1997). Health and disability differences
among racial and ethnic groups. In L.G. Martin & B.J. Soldo (Eds.),
Racial and ethnic differences in the health of older Americans (pp.
43-105). Washington, DC: National Academy Press.
Martikainen, P., Stansfeld, S., Hemingway, H., Marmot, M. (1999).
Determinants of socioeconomic differences in change in physical
and mental functioning. Soc Sci Med.;49:499-507.
Martinez-Sanchez, E., Gutierrez-Fisac, J.L., Gispert, R., and Rcgidor,
E. (2001). Educational differences in health expectancy in Madrid
and Barcelona. Health Policy, 55:227-231.
260

Mathers, C. D., Robine, J. M., & Wikins, R. (1994). Health expectancy
indicators: Recommendations for terminology. In C.D.Mathers, J.
McCallum, & J. M. Robine (Eds.), Advances in health expectancies
(pp. 18-33). Canberra: Australia: Australian Institute of Health and
Welfare.
Mathers, C.D., Sadana, R., Salomon, J.A., Murray, C.J., and Lopez AD.
(2001). Healthy life expectancy in 191 countries, 1999. Lancet,
26;357(9269):1685-91.
Matthews, R.J., Jagger, C., and Hancock, R.M. (2006). Does
socioeconomic advantage lead to a longer, healthier old age? Social
Science & Medicine, 62:2489-2499.
Melzer, D., G. Izmirlian, S.G. Leveille, and J.M. Guralnik. (2001).
Educational differences in the prevalence of mobility disability in old
age: The dynamics of incidence, mortality, and recovery. Journal of
Gerontology: Social Sciences, 56B (5):S294-S301.
Melzer, D., McWilliams, B., Brayne, C., Johnson, T., and Bond, J..
(2000). Socioeconomic status and the expectation of disability in old
age: estimates for England. J Epidemiol Community
Health,54:286-292.
Mendes de Leon, C.F., Barnes, L.L., Bienias, J.L., Skarupski, K.A., and
Evans, D.A. (2005). Racial Disparities in Disability: Recent Evidence
From Self-Reported and Performance-Based Disability Measures in
a Population-Based Study of Older Adults. The Journals of
Gerontology Series B: Psychological Sciences and Social Sciences
60:S263-S271.
Minicuci, N., Noale, M.; and the ISLA group (2005). Influence of level of
education on disability free life expectancy by sex: the ILSA study.
Exp Gerontol., 40(12):997-1003.
Molla, M (2002). Life tables by education level for the United States.
NCHS. Unpublished. Available upon request, ztg5@cdc.gov.
Molla, M., Madans, J.H., and Wagener, D.K. (2004). Differentials in
Adult Mortality and Activity Limitation by Years of Education in the
United States at the End of the 1990s. Population and Development
Review, 30(4):625-646.
Moore, D. E. and M. D. Hayward (1990). Occupational Careers and
Mortality of Elderly Men. Demography, 27(1): 31-53.
261

Mukamel, D.B., Temkin-Greener, H., Delavan, R., Peterson, D.R.,
Gross, D., Kunitz, S. and Williams, T.F. (2006). Team performance
and risk-adjusted health outcomes in the Program of All-Inclusive
Care for the Elderly (PACE). Gerontologist, 46(2):227-37.
Murray, C.J.L. (2005). Eight Americas: Investigating Mortality
Disparities across Races, Counties, and Race-Counties in the
United States. PLoS Med 3(9):e260.
DOI:10.1371/journal.pmed.0030360
Murray, C.J. and Lopez, A.D. (1997). Regional patterns of disability-free
life expectancy and disability-adjusted life expectancy: global
Burden of Disease Study. Lancet. 349(9062):1347-52.
Murray, C.J.L., Salomon, J.A., Mathers, C.D., and Lopez, A.D. (Eds.)
(2002). Summary Measures of Population Health: Concepts, Ethics,
Measurement and Applications. Geneva: World Health
Organization.
Murtagh, K.N., and Hubert, H.B. (2004). Gender Differences in Physical
Disability Among an Elderly Cohort. American Journal of Public
Health, 94 (8): 1406-1411.
Musick, K. (2004). Does Marriage Make People Happier? Marriage,
Cohabitation, and Trajectories in Well-Being. Pp. 103-4 in
Sourcebook on Family Theory and Research, edited by Vern L.
Bengtson, Alan C. Acock, Katherine R. Allen, Peggye
Dilworth-Anderson, and David M. Klein. London: Sage Publications.
Nagi, S.Z. (1965). Practitioners as participants in research.
Rehabilitation Record 6(4).
Nagi, SZ, 1976. An epidemiology of disability among adults in the
United States. Milbank Mem Fund Q Health Soc. 54:439-467
Nagi, SZ. Disability concepts revisited: implications for prevention.
Institute of Medicine. Disability in America: Toward a National
Agenda for Prevention. Washington, DC: National Academy Press;
1991:309–327.
National Center for Health Statistics. The Second Longitudinal Study of
Aging, 1994-2000, Wave 3 Survivor File, Version SF 2.1. Public-use
data file and documentation.
http://www.cdc.gov/nchs/about/otheract/aging/w3sf.htm. October
2002.
262

NCHS (1990). National Vital Statistics Reports. Unites States Life
Tables, 1987. Vol II, Section 6. DHHS Publication No. (PHS)
90-1104. U.S. Department of Health and Human Services, Public
Health Service, Centers for Disease Control, National Center for
Health Statistics. Available online at:
http://www.cdc.gov/nchs/data/lifetables/life87_2acc.pdf
NCHS (1999). National Vital Statistics Reports. Unites States Life
Tables, 1997. Volume47, No. 28. Available online at:
http://www.cdc.gov/nchs/data/nvsr/nvsr47/nvs47_28.pdf.
NCHS Data warehouse (2002). 2004 Data Warehouse on trends in
health and aging: Data from the National Health Interview Survey.
http://www.cdc.gov/nchs/agingact.htm
NCHS (2002). The 1994-2000 Second Longitudinal Study of Aging:
Wave 3 Survivor File, version SF 2.1 October 2002. National Center
for Health Statistics, Centers for Disease Control. Available online
at http://www.cdc.gov/nchs/about/otheract/aging/w3sf.htm
NCHS (2006). Mortality data. Available online at:
http://www.cdc.gov/nchs/products/pubs/pubd/lftbls/lftbls.htm
Viewed Dec. 2006.
Newacheck, P.W., Stein, R.E.K., Bauman, L., Hung, Y-Y. (2003).
Disparities in the prevalence of disability between black and white
children. Archives of Pediatrics & Adolescent Medicine, 157.(3):244
Nusselder, W.J., Looman, C.W., Mackenbach, J.P., Huisman, M., van
Oyen, H., Deboosere, P., Gadeyne,S., Kunst, A.E. (2005). The
contribution of specific diseases to educational disparities in
disability-free life expectancy. Am J Public Health, 95(11):2035-41.
Oeppen, J. and Vaupel, J.W. (2002). Enhanced: Broken Limits to Life
Expectancy. Science 10 (296) 5570:1029-1031.
Oman, D., Reed, D., Ferrara, A. (1999). Do Elderly Women Have More
Physical Disability than Men Do? American Journal of
Epidemiology, 150(8):834-842.
Papoulis, A. (1984). Brownian Movement and Markov Processes. In
Probability, Random Variables, and Stochastic Processes, 2nd ed.
New York:McGraw-Hill, pp.515-553,
263

Peek, M.K., Kushang, V., Patel and Ottenbacher, K.J. (2005).
Expanding the Disablement Process Model Among Older Mexican
Americans. The Journals of Gerontology Series A: Biological
Sciences and Medical Sciences, 60:334-339.
Perenboom, R.J., Boshuizen, H.C., Breteler, M.M., Ott, A., Van de
Water, H.P. (1996). Dementia-free life expectancy (DemFLE) in The
Netherlands. Social Science & Medicine, 43(12):1703-1707.
Pienta, A. M., Hayward, M. D., Jenkins, K. R. (2000). Health
Consequences of Marriage for the Retirement Years. Journal of
Family Issues 21(5): 559-586.
Pierce, J.P., Fiore, M.C, Novotny, T.E., Hatziandreu, E.J. and R.M.
Davis. (1989). Trends in cigarette smoking in the United States.
Projections to the year 2000, JAMA, 261 (1):61–65.
Pope, A, and Tarlov, A. (1991). Disability in America: Toward a National
Agenda for Prevention. Institute of Medicine (U.S.) Committee on a
National Agenda for the Prevention of Disabilities. Washington, DC:
National Academy Press..
Preston, S.H., Elo, I., Rosenwaike, I., & Hill, M. (1996). African
American mortality at older ages: Results from a matching study.
Demography, 33, 193-209.
Putnam, M. (2002). Linking Aging Theory and Disability Models:
Increasing the Potential to Explore Aging With Physical Impairment.
The Gerontologist 42:799-806
Reyes-Beaman, S., Jagger, C., Garcia-Pena, C., Munoz, O., Beaman,
P.E., and Stafford, B. National Group of Research on Ageing (2005).
Active life expectancy of older people in Mexico. Disabil Rehabil,
27:213-219
Reynolds, S.L., Saito, Y., Crimmins, E.M. (2005). The impact of obesity
on active life expectancy in older American men and women.
Gerontologist, 45(4):438-44.
Reynolds, S.L., Saito, Y., and Crimmins, E.M. (2005). The Impact of
Obesity on Active Life Expectancy in Older American Men and
Women. The Gerontologist 45(4):438-444.
Rice, D.P., and Felman, J.J. (1983). Living longer in the United States:
Demographic changes and health needs of the elderly. Milbank
Memorial Fund Quarterly/Health and Society 1983, 61, 3: 362-396
264
Riley, M.W., and Bond, K. (1983). Beyond Ageism: Postponing the
onset of disability. In Riley MW, Hess BB, and Bond K. (Eds.) Aging
in Society: Selected Reviews of Recent Research, Lawrence
Erlbaum Associates: Hillschale NJ: pp 243-252.
Robine, J-M, Blanchet, M., Dowd, J.E. (Eds.) (1992). Health
Expectancy: First workshop of the International Healthy Life
Expectancy Network (REVES) Quebec, September 1989. Studies
on medical and population subjects No. 54. London: HMSO
Robine, J-M., Brouard, N., Colvez, A. (1987). Les indicateurs
d’esperance de vie sans incapacite (EVSI): des indicateurs globaux
de l’etat de sante des populations. Rev. Epidem. Et Sante Publique
1987, 35: 206-224.
Robine, J-M., Colvez, A., Bucquet, D., Hatton, F., Morel, B., Lelaidier, S.
(1986). L’esperance de vie sans incapacite en France en 1982.
Population, 6: 1025-1042.
Robine, J.-M., Mathers, C., & Brouard, N. (1996). Trends and
differentials in disability-free life expectancy: Concepts, methods,
and findings. In G.Caselli & A. D. Lopez (Eds.), Health and mortality
among elderly population (pp. 182-201). New York: Oxford
University Press Inc.
Robine, J-M, Ritchie, K. (1991). Healthy life expectancy: evaluation of
global indicator of change in population health. BMJ,
302(6774):457-60.
Robine, J-M., Romieu, I., & Cambois, E. (1997). Health expectancies
and current research. Reviews in Clinical Gerontology, 7, 73-81.
Robine, J-M. Jagger, C., Mathers, C.D., Crimmins, E.M., Suzman,
R.M., and Peron, Y. (2003). Introduction In Jean-Marie Robine,
Carol Jagger, Colin D. Mathers, Eileen M. Crimmins, and Richard
Suzman (Eds.) Determining Health Expectancies. Wiley: West
Sussex, England.
Robine, J-M., Mormiche, P. and Sermet, C. (1998). Examination of the
Causes and Mechanisms of the Increase in Disability-free Life
Expectancy. Journal of Aging and Health 10(2):171-91.
Rogers, A. (1989). The elderly mobility transition: growth,
concentration, and tempo. Research on Aging, 11:3-32.
265

Rogers, R. G., Rogers, A., and Belanger, A. (1992). Disability-free life
among the elderly in the United States: Sociodemographic
correlates of functional health. Journal of Aging and Health, 4,
19-42.
Rosow, I., Breslau, N. (1966). A Guttman health scale for the aged. J
Gerontol., 21:556-559.
Saito, Y., Crimmins, E.M., and Hayward, M.D. (1999). Health
expectancy: An overview. NIHON University Population Research
Institute, Tokyo, Japan.
Saito, Y., Crimmins, E.M., and Hayward, M. (1991). Stability of
Estimates of Active Life Expectancy Using Two Methods of Life
Table Construction. Cashiers Quebecoes de Demographie 20(2):
291-327.
Saito, Y., Crimmins, E.M., and Hayward, M. (1995). Approaches to
Measuring Active Life Expectancy: Prevalence vs. Incidence,
Population vs. Status-based Life Tables. REVES 8.
Saito, Y., Qiao, X., & Jitapunkul, J. (2003). Health expectancy in Asian
Countries. In: Kobine JM, Jagger C, Mathers CD, Crimmins EM,
Suzman RM, (Eds.). Determining Health Expectancies. Chichester,
England: John Wiley & Sons:289-318.
Sanders, B.S. (1964) Measuring community health levels. American
Journal of Public Health, 54(7):1063–1070.
Schoeni, R.F., Martin, L.G., Andreski, P.M., and Freedman, V.A.
(2005). Persistent and growing socioeconomic disparities in
disability among the elderly: 1982-2002. American Journal of Public
Health, 95 (11): 2065-70.
Seeman, T.E., Bruce, M.L., and McAvay, G.J. (1996). Social network
characteristics and onset of ADL disability: MacArthur studies of
successful aging. Journals of Gerontology Series B-Psychological
Sciences & Social Sciences. 51(4):S191-200.
Seeman, T.E., Charpentier, P.A., Berkman, L.F. (1994). Predicting
changes in physical performance in a high-functioning elderly
cohort: MacArthur studies of successful aging. J Gerontol A Biol Sci
Med Sci., 49:M97-M108.
Shea, D. G., Miles, T. and Hayward, M. (1996). The Health-Wealth
Connection: Racial Differences. The Gerontologist 36(3): 342-349.
266
Shumway-Cook, A., Guralnik, J.M., Phillips, C.L., Coppin, A.K., Ciol,
M.A., Bandinelli, S., and Ferrucci L. (2007). Age-associated
declines in complex walking task performance: the Walking
InCHIANTI toolkit. Journal of the American Geriatrics Society,
55(1):58-65.
Sihvonen, A.P., Kunst, A.E., Lahelma, H., Valkonen, T., and
Mackenbach, J.P. (1998). Socioeconomic inequalities in health
expectancy in Finland and Norway in the late 1980s. Soc Sci Med.,
47:303-315.
Soldo, B.J., and Manton, K. (1985). Health status and service needs of
the oldest old: current patterns and future trends. Milbank Memorial
Fund Quarterly/Health and Society, 63, 2: 286-317.
Soldo, B.J., Mitchell, O.S., Tfaily, R., and McCabe, J.F. (2006).
Cross-Cohort Differences in Health on the Verge of Retirement.
PRC WP 2006- 17 Pension Research Council Working Paper
Pension Research Council, The Wharton School, University of
Pennsylvania Available online at:
http://www.pop.upenn.edu/rc/parc/aging_center/2006/PARCwps06
-13.pdf
Spector, W.D. and Fleishman, J.A. (1998). Combining Activities of Daily
Living with Instrumental Activities of Daily Living to Measure
Functional Disability. Journals of Gerontology Series B:
Psychological Sciences and Social Sciences 53:S46-S57.
Spillman, B.C. (2004). Changes in Elderly Disability Rates and the
Implications for Health Care Utilization and Cost. The Milbank
Quarterly, 82(1):157-194
Struijs, J.N., van Genugten, M.L., Evers, S.M., Ament, A.J., Baan, C.A.,
and van den Bos, G.A. (2005). Modeling the future burden of stroke
in The Netherlands: impact of aging, smoking, and hypertension.
Stroke, 36(8):1648-55.
Sullivan, D.F. (1966). Conceptual problems in developing an index of
health. NCHS, Vital and Health Statistics, Series 2, No. 17,
(Washington 1966).
Sullivan ,D.F. (1971). A single index of mortality and morbidity. HSMHA
Health Rep., 86(4):347-54.
Suthers, K., Kim, J.K., and Crimmins, E.M. (2003) Life Expectancy With
Cognitive Impairment in the Older Population of the United States.
The Journals of Gerontology Series B: Psychological Sciences and
Social Sciences, 58:S179-S186.
267
Tabbarah, M., Crimmins, E.M., and Seeman, T.E. (2002). The
relationship between cognitive and physical performance:
MacArthur Studies of Successful Aging. Journals of Gerontology
Series A-Biological Sciences & Medical Sciences, 57(4):M228-35.
Taylor, M.G. (2005). Disaggregating disability trajectories: Exploring
differences in the disability experience of older adults in the United
States. Ph.D. Dissertation, Duke University. Available through
ProQuest AAT 3181026.
Teachman, J.D. and Hayward, M.D. (1993). Interpreting Hazard Rate
Models. Sociological Methods & Research, 21(3): 340-371.
Thomas, L. (1977). Biomedical science and human health: the
long-range prospect. Proceeding of the American Academy of Arts
and Sciences, 106, 3.
Tuma, N.B., and Hannan, M.T., (1984). Social Dynamics: Models and
Methods. New York Academic Press, Inc.
Ubbink, J. B., R. Delport, et al. (1996). Plasma homocysteine
concentrations in a population with a low coronary heart disease
prevalence. J Nutr, 126: 1254S-1257S.
U.S. Department of Health and Human Services. Healthy People 2010.
2nd ed. With Understanding and Improving Health and Objectives
for Improving Health. 2 vols. Washington, DC: U.S. Government
Printing Office, November 2000.
http://www.healthypeople.gov/About/whatis.htm).
Valkonen, T., Sihvonen, A.P., Lahelma, E. (1997). Health expectancy
by level of education in Finland. Soc Sci Med., 44:801-8.
Vaupel, J.W., Carey JR, Christensen, K, et al. (1998). Biodemographic
trajectories of longevity. Science 280:855-860.
Verbrugge, LM. (1984). Longer life but worsening health? Trends in
health and mortality of middle-aged and older persons. Milbank
Memorial Fund Quarterly/ Health and Society, 62, 3: 475-519.
Verbrugge, L.M. (1989). The Twain Meet: Empirical Explanation of Sex
Differences in Health and Mortality. Journal of Health and Social
Behavior, 30: 282-304.
Verbrugge, L.M. and A.M. Jette. (1994). The disablement process.
Social Science & Medicine. 38(1):1-14.
268

Verbrugge, L.M., Merrill, S.S., Liu, X. (1999). Measuring disability with
parsimony Disabil Rehabil 21,295-306
Verbrugge, L.M., Rennert, C., Madans, J.H. (1997). The great efficacy
of personal and equipment assistance in reducing disability. Am J
Public Health, 87:384–392
Wagener, D.K., Molla, M.T., Crimmins, E.M., Pamuk, E., Madans, J.H.
(2001). Summary measures of population health: Addressing the
first goal of healthy people 2010, improving health expectancy.
Healthy People: Statistical Notes, No. 22, September, 2001.
Waidmann, T.A., and K.G. Manton (1998). International Evidence on
Disability Trends among the Elderly (June 18, 1998). Available
online at http://aspe.os.dhhs.gov/daltcp/reports.htm#Waidmann1
Waidmann T, Liu K. (2000). Disability trends among the elderly and
implications for the future. J Gerontol B Psychol Sci Soc Sci.,
55:S298-S307.
White, K.M. (2002). Longevity Advances in High-income Countries,
1955–96. Population and Development Review, 28:59–76.
Wiener, J.M., Hanley, R.J., Clark, R., Van Nostrand, J.F. (1990).
Measuring the activities of daily living: Comparisons across national
surveys. Journal of Gerontology: Social Services, 45:S229-S237.
Wilber, K.H., Allen, D., Shannon, G.R. & Alongi, S. (2003). Partnering
managed care and community-based services for frail elders: The
Care Advocate Program. Journal of the American Geriatric Society,
51, 807-812.
Wilkins R, Adams OB. (1983). Health expectancy in Canada, late
1970s: demographic, regional, and social dimensions. Am J Public
Health, 73:1073-1080.
Wilkins, R. (1982). La distribution de l’esperance de vie parmi les
differents etats de sante: composantes, methods de calcul et
resultants pour le Quebec, 1978 Communication presentee au
50eme congres de l’ACFAS, section demographie, 14 mai 1982,
Institut de recherches politiques (Montreal 1982).
Wilmoth JR, 2000. Demography of longevity: past, present, and future
trends. Exp Gerontol, 35:1111-1129.
269

Wilson, R.W., Drury, T.F. (1984). Interpreting trends in illness and
disability: health statistics and health status. Ann. Rev. Public
Health, 5: 83-106.
Wilson, R.W. (1981). Do health indicators indicate health? American
Journal of Public Health, 71(5):461-463.
Winkleby, M.A., Jatulis, D.E., Frank, E. and S.P. Fortmann. (1992).
Socioeconomic status and health: how education, income, and
occupation contribute to risk factors for cardiovascular disease. Am.
J. Public Health, 82: 816–820.
Wolinsky, F.D. Miller, DK, Andresen, EM, Malmstrom, TK et al. (2007).
Effect of Subclinical Status in Functional Limitation and Disability on
Adverse Health Outcomes 3 Years Later. The Journals of
Gerontology 162A(1). 101—106.
Wolf, D.A., Mendes de Leon, C., Glass, T. (2007). Trends in rates of
onset of and recovery from disability at older ages: 1982-1994.
Journal of Gerontology: Social Sciences: S3-S10.
Wolfbien, S (1949). The length of working life. Population Studies,
(3):286-294.
World Health Organization (2000). WHO issues new healthy life
expectancy ranking Japan numbers one in new “Healthy Life"
system. World Health Organization.
World Health Organization. The uses of epidemiology in the study of the
elderly: report of a WHO scientific group on the epidemiology of
aging. World Health Organization (Geneva 1984) 84 pp (Technical
report series; 706).
Wray, L.A. & Blaum, CS. (2001). Explaining the role of sex on disability:
a population-based study. Gerontologist, 41: 499-510.
Wray, L.A., Herzog, A.R., Willis, R.J., Wallace, R.B. (1998). The Impact
of Education and Heart Attack on Smoking Cessation among
Middle-Aged Adults. Journal of Health and Social Behavior:
271-294.
270

Wu, S.C., S.Y. Leu, and C.Y. Li. 1999. Incidence of and Predictors for
Chronic Disability in Activities of Daily Living Among Older People in
Taiwan. Journal of the American Geriatric Society 47:1082-86.
Yount, K.M., Agree, E.M. (2005). Differences in disability among older
women and men in Egypt and Tunisia. Demography, 42 (1):
169-187.

Abstract (if available)

Abstract The purpose of this dissertation was to examine trends in active life expectancy between the mid 1980s and the late 1990s, and to examine changes in disability onset, recovery, and mortality in different subgroups of the population. Two nationally representative longitudinal datasets of the population ages 70 and older were used to examine an incidence-based measure of disability in a multi-state life table program known as IMaCh.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Childlessness and psychological well-being across life course as manifested in significant life events

PDF

The effect of body weight on mortality: different countries and age groups

PDF

Translating two physical activity programs for older adults into home- and community-based settings: "Active Start" and "Healthy Moves for Aging Well"

PDF

Using mixed methods to identify the characteristics of older fraud victims

PDF

Aging in a high infection society

PDF

Health differences between the elderly in Japan and the United States

PDF

Is stress exposure enough? Race/ethnic differences in the exposure and appraisal of chronic stressors among older adults

PDF

Social determinants of physiological health and mortality in China

PDF

The role of the modern aging network: measuring innovations of Area Agencies on Aging

PDF

International sex and age differences in physical function and disability

PDF

Examination of the long-term psychosocial and functioning effects of a healthy living, life management behavior intervention for older adults

PDF

Predicting cognitive decline and dementia in elderly twins from indicators of early life oral health

PDF

The measurement, life course patterns, and outcomes of intergenerational ambivalence among parent-adult child dyads

PDF

Air pollution neurotoxicity throughout the lifespan: studies on the mechanism of toxicity and interactions with effects of sex and genetic background

PDF

Internet communication use, psychological functioning and social connectedness at older ages

PDF

Racial disparities in advance care planning and directives completion and end-of-life outcomes in the United States

PDF

The economic security of an aging minority population: a profile of Latino baby boomers to inform future retirees

PDF

Exploring end-of-life care experiences and disparities among Hispanic populations in the United States

PDF

Statistical algorithms for examining gene and environmental influences on human aging

PDF

Thinking generatively versus acting generatively: exploring the associations of generative self-concept and generative activity with cognitive function among older adults

Asset Metadata

Creator Hagedorn, Aaron Timothy (author)

Core Title Longitudinal change in active life expectancy: the longitudinal studies of aging 1984-2000

School Leonard Davis School of Gerontology

Degree Doctor of Philosophy

Degree Program Gerontology

Publication Date 04/22/2008

Defense Date 02/04/2008

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag active life expectancy,Disability,longitudinal,OAI-PMH Harvest,trends

Language English

Advisor Crimmins, Eileen M. (committee chair), Musick, Kelly (committee member), Wilber, Kathleen H. (committee member)

Creator Email ahagedor@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m1178

Unique identifier UC1173727

Identifier etd-Hagedorn-20080422 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-64016 (legacy record id),usctheses-m1178 (legacy record id)

Legacy Identifier etd-Hagedorn-20080422.pdf

Dmrecord 64016

Document Type Dissertation

Rights Hagedorn, Aaron Timothy

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu