Special AHEAD Issue of the Journals of Gerontology Announcement06/19/97
HRS and AHEAD are pleased to announce a special issue of the Journals of Gerontology (Series B, Vol. 52B, 1997) devoted entirely to research based on AHEAD (Assets and Health Dynamics Among the Oldest Old). Available below is information on ordering the special issue, its table of contents, and a preliminary version of the introductory text.
To order the special issue, please send $10.00 (checks payable to "GSA") to:
The Gerontological Society of America P.O. Box 79151 Baltimore, MD 21279-0151
In the near future, reprints of these papers will be available for order via the University of Michigan's Population Studies Center Publications web page.
Table of Contents
George C. Myers, F.Thomas Juster, and Richard M. Suzman
Asset and Health Dynamics Among the Oldest Old: An Overview of the AHEAD Study
Beth Soldo, Michael D. Hurd, Willard L. Rodgers, and Robert B. Wallace
A Comparative Analysis of ADL Questions in Surveys of Older People
Willard Rodgers and Baila Miller
Measures of Cognitive Functioning in the AHEAD Study
A. Regula Herzog and Robert B. Wallace
The Structure of Health Status Among Hispanic, African American, and White Older Adults
Timothy E. Stump, Daniel 0. Clark, Robert J. Johnson, and Fredric D. Wolinsky
Race, Socioeconomic Status, and Health: Accounting for Race Differences in Health
Michael Schoenbaum and Timothy Waidmann
Wealth Inequality Among Older Americans
James P. Smith
Transfer Behavior Within the Family: Results From the Asset and Health Dynamics Study
Kathleen McGarry and Robert F. Schoeni
Patterns of In-Home Care Among Elderly Black and White Americans
Theresa M. Norgard and Willard L. Rodgers
The Division of Family Labor: Care for Elderly Parents
Douglas A. Wolf, Vicki Freedman, and Beth J. Soldo
Selection of Children To Provide Care: The Effect of Earlier Parental Transfers
John C. Henretta, Martha S. Hill, Wei Li, Beth J. Soldo, and Douglas A. Wolf
Introductory TextGeorge C. Myers
Center for Demographic Studies
F. Thomas Juster
Institute for Social Research
University of Michigan
Richard M. Suzman
Office of the Demography of Aging
National Institute on Aging
The purpose of this Special Issue is to introduce the scientific community to the analytic potential of a new database: the Asset and Health Dynamics Among the Oldest Old (AHEAD) Study. AHEAD is designed as a longitudinal investigation of persons 70 years of age and over in the United States.
The contributions to this volume had their origin in an Early Results Workshop in September 1994 that was held at the Institute for Social Research, University of Michigan, with support from the National Institute on Aging (NIA). The papers were subsequently modified with updated data release files and extensively revised before they were submitted for consideration for this collection. The conventional peer review process was used to decide on the content of this Special Issue. The hope was that the reader could gain an appreciation of the comprehensive nature of the AHEAD study from a set of articles that address major methodological and substantive issues in the field of aging.
In this Introduction, we provide a brief summary of each contribution, a background sketch of the origins of the AHEAD study, and a discussion of the main features of the study to date and planned changes for the future.
CONTENTS OF THE ISSUE
In the first article, Soldo, Hurd, Rodgers, and Wallace provide a comprehensive picture of the three main theoretical perspectives that guided the AHEAD study -- aging and health transitions; dynamics of health, assets, and improverishments; and health dynamics, family resources, and care outcomes. It is quite evident from these themes that the emphasis from the outset has been on the joint dynamics of health, economic resources, family, and health care as they unfold over time in the lives of older Americans. The article covers the design features of the baseline survey, including sampling, questionnaire content, and data quality considerations. Selective descriptive results are provided to illustrate the characteristics of the sample at the beginning of the study. Finally, future plans for the longitudinal study are presented. Especially noteworthy in this regard are the merging of the AHEAD with the HRS studies and the addition of supplemental samples of aging-in cohorts and the missing cohorts between the two studies.
The two articles that follow deal with measures related to the health and functioning of older persons-- activities of daily living (ADLs) and cognitive status assessments. Rodgers and Miller compare the responses to the ADL questions from the main body of the questionnaire with responses to different types of ADL questions that were contained in so-called "experimental modules," which were asked of samples of the respondents. The use of these experimental modules in AHEAD has made it possible to try out measures of interesting new areas of investigation that might be candidates for inclusion in later waves of the study, as well as similar sets of questions that measure the same concept in different ways. The alternative versions of ADL questions were those used in the 1984 Study of Aging (SOA) and the screener for the 1982 National Long Term Care Survey (NLTCS). Extensive analyses of these comparisons and others lead the authors to suggest that there is apparently considerable, largely random, measurement error in answers to ADL questions. However, composite indexes of functioning limitations are found to have high reliability and validity. Although the conclusions from this article may be somewhat disquieting, they have served to inform changes that were introduced in the second wave of the study. In addition, they led to design experiments intended to further examine these issues.
One of the distinguishing features of the AHEAD is the inclusion of an extensive set of measures of cognitive functioning. As Herzog and Wallace note in their article, while cognitive function is acknowledged as crucial for assessing the overall functioning of persons at advanced ages, it has seldom been assessed adequately in large-scale national studies. They find the psychometric properties of the varied measures of cognitive functioning used in the study to be quite satisfactory. Moreover, the measures are correlated with other measures of health and functioning in ways that might be predicted; thus providing assurance of construct validity. Several other encouraging aspects were noted: response rates to these batteries were quite high even though individuals had the option of not responding; those who declined to answer scored poorly on other measures of mental status and health; and telephone administration did not appear to introduce any major bias.
Stump, Clark, Johnson and Wolinsky assess the replicability of a health status model, derived from other studies of older persons, with AHEAD data. They find general support for a multidimensional model that differentiates between basic, household, and advanced ADLs, along with lower body disability. However, upper body disability, which also has been included in the model in other studies, did not emerge in the AHEAD analyses. The measurement model was confirmed in separate analyses by gender, race, and ethnicity. The importance of lower body disability was particularly striking in analyses of structural models with self-rated health and depression as outcome variables.
Schoenbaum and Waidmann test a range of specific structural models of racial differences in health status in their article. This is possible because of the detailed information that is available in AHEAD on various aspects of health and socioeconomic status (SES) and the oversampling of Blacks. They find differences in health status between Blacks and Whites and that a substantial amount, but not all, of the differences can be explained by socioeconomic factors. The use of simulation in the analyses enables the authors to estimate the percentage of the racial gap that can be explained by specific SES measures for the various health outcomes.
Smith examines the distribution of wealth among households containing at least one person 70 years of age or over. The analyses confirm the advantages of using bracketing techniques, initially developed in the HRS, for obtaining further information on assets in cases in which the respondent initially either refused or was unable to provide an exact response. Using the greatly improved data on wealth available from AHEAD, Smith reports very sharp wealth disparities in the study. The variation in net worth is found to be strongly related to minority status, health, and education.
McGarry and Schoeni make use of the extensive data on intrafamily transfers of financial resources found in AHEAD to examine the characteristics of adult children receiving transfers from their elderly parents. They find that transfers are compensatory, in that parents provide large amounts of money to their less well-off children. Interestingly, they report few financial transfers from children to parents. Services provided to parents are far more common, but the services apparently are not exchanged for financial transfers.
The last three articles are concerned with caregiving provided to the older persons in the study. The AHEAD study contains extensive data on help received from informal sources, such as family members and friends, as well as formal sources. Care received is reported by the intensity of care for specific conditions, and multiple sources may be recorded. Finally, the full family roster that exists for each household member contains detailed characteristics on each person, makes it possible to derive potential networks of caregivers that can be related to actual care received. These features of the AHEAD data are brought out in the three articles.
Norgard and Rodgers examine in-home care among Blacks and Whites from both formal and informal sources. Caregiving is more likely to be from informal than formal sources. However, contrary to expectations, Blacks received less assistance and less informal care than Whites. Wolf, Freedman, and Soldo are able to use AHEAD data to examine models that specify parent-care decisions within the context of the entire family network. They report that the amount of care provided by a child is affected not only by the parent's need, but by their own circumstances and the characteristics and care behavior of their siblings. In the final article, Henretta, Hill, Li, Soldo, and Wolf extend the caregiving process to consideration of who is selected to provide care. They find strong support for the hypothesis that parental caregiving reciprocates for earlier lifecycle transfers to specific children within a family.
BACKGROUND OF THE AHEAD STUDY
The AHEAD Study began as a supplement to the Health and Retirement Study (HRS) shortly after the latter study was launched in 1992. Discussions between the NIA project officer and the principal investigator suggested the possibility of using the elderly screening sample left over from the HRS as a potentially cost-effective way to develop an ancillary study that would include sizeable samples of the oldest old (age 85 and over) population. In order to select the members of the HRS sample, in which the primary respondent was between the ages of 51 and 61, the University of Michigan had invested well over $1.5 million to screen over 56,000 households containing persons of all ages. The accuracy of the information on the older members of this screening sample was decaying steadily with time (people moved, died, became institutionalized, etc.) and the screening sample represented an enormously valuable but wasting asset. In short, the HRS team took advantage of the opportunity presented by the screening sample to develop a study on the very old, and in doing so placed their own creative stamp on the study.
While the oldest old population is has been one of the fastest growing age groups in the U.S. population, there were two pressing research questions that could not be addressed because of lacunae in the available statistical data. First, there was virtually no longitudinal data on the dynamics of economic status at older ages, especially in relation to changes in health. Although poverty rates among the elderly were highest among the oldest old, no one knew how much of the poverty resulted from adverse health events, becoming widowed, attrition of savings, or spending down to become eligible for Medicaid. Interest also was growing into the question of how low economic status experienced at late middle age might impact on health and survival in very old age.
Second, data from other surveys, such as the NLTCS and Longitudinal Study of Aging (LSOA), suggested that a considerable fraction of those who were functionally disabled in one wave appeared to have recovered at least some function in subsequent waves. However, data from these surveys were not sufficiently robust to resolve how much of this apparent improvement stemmed from improved coping resulting from adopting assistive devices (hearing aids, ramps, canes, etc.), how much came from the positive impact of medical treatment (especially surgical interventions such as lens replacement, angioplasty, and joint replacement), and how much was simply measurement error.
The initiation of the AHEAD study can best be understood if viewed in the context of several prior developments. These included: a) the NIA initiative on the Oldest Old, b) development of databases for the research community, such as the NLTCS and the LSOA, and c) more general improvement of aging-related research databases and data systems, including the HRS. Other threads important to understanding the development include the evolution of research areas, such as the economics and demography of aging, and sample survey methods applied to elderly populations.
The initiative on the Oldest Old was started at the beginning of the 1980s after several researchers (e.g., Eileen Crimmins, Ira Rosenwaike, and Kenneth Manton) reported that the 85 and over population was growing at a rate far faster than had been suggested by Bureau of the Census population projections. It soon became apparent that there was an almost total absence of suitable data on this oldest segment of the population (Suzman and Riley, 1985). Census data often were marred by inaccurate age reporting, while surveys such as the National Health Interview Survey (NHIS) did not include sufficient numbers of older persons to sustain detailed analyses.
Several surveys, initiated in the early 1980s, provided some limited opportunities for studying individuals in this age group. The NLTCS was initiated by the Office of the Assistant Secretary of Planning and Evaluation in the Department of Health and Human Services in 1982 and continued by the Health Care Financing Administration in 1984. Subsequent waves of the panel study in 1989 and 1994 were funded by the NIA through grants to Kenneth G. Manton at Duke University's Center for Demographic Studies. The NLTCS was the first national longitudinal survey of the elderly that focused on disability (as assessed by functional status) and included adequate samples of the oldest old population, both community-dwelling and institutionalized. Analyzing data from the early waves of the complex NLTCS database proved to be methodologically burdensome, and the uneven periods between waves made the analysis of transitions in functional status difficult.
The LSOA, begun in 1982 at the National Center for Health Statistics as a supplement to the NHIS, focused on the total community-dwelling elderly population rather than the disabled. In time, the LSOA, in spite of response rate problems in the transitions into institutions and a sample size that did not permit much disaggregation for the age 80 and over group, proved to be quite user-friendly. The NLTCS, the LSOA, and the National Nursing Home Survey, in addition, could be viewed as a partially integrated set of surveys that covered the full spectrum of the older population (Manton & Suzman 1992). However, except for passive monitoring of deaths the LSOA terminated with the 1990 interview wave. Aside from the NLTCS, with its focus on the disabled and five-year periodicity, the field was left with no survey that measured shorter-term transitions in functional status and adequately covered the oldest old population. Further, as the field of the demography and epidemiology of health and functioning had evolved, a set of new questions had arisen concerning transitions to improved functioning .
The more recent development of the economics of aging (Hurd, 1990) had resulted in renewed interest in the dynamics of retirement and economic status at very old ages. The issues raised included the accumulation and decumulation of wealth over the lifecycle, the strength of the bequest motive, and interactions with health. An Ad Hoc Advisory Panel to the NIA (1988) had concluded that economic survey data on the very old were severely deficient , thus posing an obstacle to research progress. Moreover, no longitudinal surveys included adequate measures of both health and economic status. The report from this meeting became an important impetus for the development of the HRS (Juster and Suzman, 1995). One project that might have provided some of the needed data on the dynamics of economic status and its impact on health and survival was the proposed reinterview of the survivors of the original Retirement History Survey. Unfortunately, this was abandoned after Title 13 concerns with the release of microdata made continued activity futile.
The way in which AHEAD was integrated into the HRS in terms of organizational dynamics and constraints on time and available resources also played a contributing role in its development, as did the operational culture that the HRS had evolved. While the HRS was one of the largest social science planning exercises ever undertaken, with a relative abundance of time and resources (Juster and Suzman, 1995), AHEAD was constrained in both respects. AHEAD was developed in a hurry. The time constraints derived from the rapidly decaying screening sample, coupled with the NIH proposal submission, review and funding cycle. AHEAD also was funded less generously than HRS. Given the existence of the NLTCS and the LSOA, it occupied a more specialized niche than the HRS. Although a Congressional Appropriations Committee had recommended that NIA fund a demographic study on the oldest old, focusing on the transitions back to improved functional status, interest in such a study was less intense and widespread than for the HRS. Nonetheless, the challenge to create AHEAD was met with enthusiasm from the research community. Moreover, many of these initial limitations were overcome when independent funding from an NIA grant was received in 1995 to support later waves of the study.
Basic Study Design
AHEAD was developed, in large part, by adapting the survey design of the HRS to make it suitable for studying older persons. The birth cohorts of 1923 and earlier (ages 70 and older in 1993) formed the sample for AHEAD, whereas HRS studied the birth cohorts of 1931-1941 (ages 51-61 at baseline in 1992). Thus, AHEAD devoted considerable attention to the characteristics and prevalence of ADL and IADL limitations, including data on the use of helpers and equipment, and lower-level physical functioning. The HRS, in contrast, focused instead on measuring higher-level functioning (e.g., running or jogging a mile). On the other hand, HRS paid a good deal of attention to the economic factors associated with retirement decisions, including job characteristics, job demands, pension characteristics, hours and hours flexibility, work history, etc. AHEAD paid much less attention to this set of variables, although it did measure some of the relevant characteristics of work for the relatively small fraction of the AHEAD sample still gainfully employed at baseline.
The basic philosophy underlying the selection of variables for AHEAD, like the HRS, was that measures were strong candidates for inclusion in the questionnaire if they represented significant variables in important models of behavior among the elderly. Thus, a variety of specialists in a number of fields--demography, sociology, economics, epidemiology, public health, medicine, cognitive psychology, and public policy--were consulted. They helped the staff shape content decisions in the major areas of concern, such as physical health, including extensive data on disease conditions; ADL/IADL limitations; cognitive functioning; family structure and transfers; income and wealth; health insurance; and work activities.
The AHEAD planning group expanded the measurement of concepts on intergenerational exchanges within the family over the lifecycle , which were measured in only embryonic ways within HRS, and pushed these issues into new territory. Based on NIA staff recommendations that were influenced by both scientific and strategic considerations, the HRS had introduced a considerable amount of cognitive assessment into the HRS interview. Given the different age range of the AHEAD sample and the association between age and the prevalence of dementia, this trend was continued and considerably strengthened within AHEAD.
There are a number of specific features of the field operations on the AHEAD survey that are worth noting. First, the survey provides incentive payments to respondents. AHEAD requires considerable time to administer; the first wave taking about 70 minutes and later waves only marginally less. Thus, respondents were provided with an incentive payment of $20.00 per case, with married couple households receiving twice that amount if both partners participated. Second, there were some mode differences in the administration of the survey that depended on age. Persons 80 years of age and older were assigned to a personal interview, although that could be converted to a telephone interview at the request of the respondent. In contrast, persons under 80 years of age were assigned to a telephone-interview mode, with the possibility of converting to a face-to-face interview at the preference of the respondent. About three quarters of the interviews used the assigned mode. Finally, a substantial number of the AHEAD interviews were completed by a proxy respondent rather than by the designated respondent. This typically occurred because the designated respondent was ill, cognitively impaired, or unable to participate in a relatively lengthy interview. The incidence of proxy responses varied with the age of the designated respondent, with almost a third of the interviews in the oldest age group being conducted with proxy respondents rather than with the designated respondent. For the proxy interviews, the survey material was generally the same as for designated respondents, except that cognitive test questions were not used on proxy interviews and the expectation/subjective perception questions were not asked.
DISTINCTIVE FEATURES OF THE STUDY
A number of the distinctive features of AHEAD have been noted in the earlier discussions of the contributions in this volume. We can summarize these briefly, as well as note some other aspects of the study that warrant attention. Most importantly, the value of AHEAD will only be fully realized as it unfolds as a longitudinal study. These matters are only lightly touched upon in the articles that are devoted primarily to analyses of baseline data.
1. Coverage. AHEAD is based on a representative sample of the continental United States, which makes it suitable for estimating national prevalence and incidence levels for many variables of interest. The oversampling of Blacks and Hispanics provides sufficiently large numbers of cases to analyze differentials within these groups, as well as comparisons with other population groups. The overall sample is large enough so that analyses of even the oldest old (n=1004) are possible using these data. Although not emphasized in any of the articles in this issue, there is considerable opportunity to examine geographic variations with AHEAD data. Moreover, the oversample of Florida residents makes it possible to conduct more intensive AHEAD investigations in the state with the largest proportion of population 65 years and over.
2. Quality of data. There has been a commitment since the inception of AHEAD to assess the validity and reliability of data from the study. In several articles included in this Special Issue, attention given to the effects of interviewing mode, nonresponse, missing data, and proxies on findings. These are issues that are particularly salient in studying persons at advanced ages.
3. Design. Although results from an experimental module are only dealt with in the Rodgers and Miller piece, further analyses can benefit from other modules that have been included in the AHEAD data collection. The use of bracketing and unfolding techniques is one of the major innovations of the HRS and AHEAD, as demonstrated in the Smith article. Another important design consideration in AHEAD is the availability of couple data for households in which both spouses have participated. None of the articles included in the Issue deals with couple data, but it is sure to receive considerable attention in secondary analyses of these data. The study also was designed to include data drawn from other record-linkage systems (most notably, the National Death Index, Medicare and Social Security files). Again, these data have not been utilized in the current articles, but should assume great importance in future research. Finally, next-of-kin data collection efforts instituted for the second wave should enhance the ability of researchers to follow panel members beyond the fixed intervals and provide outcome data of considerable interest.
4. Content. One advantage of the AHEAD database for studying the aging population lies in the availability of a rich mixture of physical and cognitive health status and functioning measures, along with extensive economic measures of income and wealth. One of the critically important issues for public policy is the degree to which aging populations consume at the same level as their current income, dip into assets as they age and thus consume more than their income, or continue to save and thus add to assets as they age. It is the interaction of economic factors with health conditions, health status, and health expenditures that substantially influences the trajectory of saving and wealth with age. Other datasets contain extensive physical health and functioning measures, but the AHEAD data are certainly more ambitious than most other studies in the area of cognitive functioning. A second feature of AHEAD is the availability of extensive data on family structures and transfers of resources, care, and support. Sufficient detail has been obtained on the complete set of potential givers and recipients of transfers, their characteristics, and recent exchanges. Finally, there was considerable attention given in the study to the expectations of respondents with regard to a variety of events future survival, macroeconomic circumstances, entry into a nursing home, etc. While the articles in this Issue have not utilized this type of data, they are likely to be used widely in secondary analyses.
Data collection for the AHEAD panel is scheduled to continue every other year. Current plans call for a new cohort to be introduced into the AHEAD sample in the year 1998, when the original AHEAD cohort will be 75 years of age or older (and the original HRS cohort will be between 57 and 67). The new AHEAD cohort will comprise the birth cohorts of 1924-1930, thus representing people who will be between 68 and 74 in 1998. The net effect of adding this new cohort to the AHEAD study will be to make the AHEAD sample fully representative of people 68 years of age and older. As the HRS sample also is planned to be augmented by a new "aging-in" cohort, HRS will be fully representative of the population age 51 through 67. Thus, the combination of both study samples will be representative of the entire United States population 51 years of age and older.
AHEAD was designed as a survey of the non-institutional population. Individuals who were institutionalized at baseline were not included in the original AHEAD sample frame. However, individuals in households who were in the community-dwelling population at baseline and who subsequently become institutionalized are continued in the AHEAD sample. It is anticipated that individual data, as well as data relating to the structure of the institution in which the AHEAD sample member is residing, will be collected for all persons who become institutionalized in succeeding waves of the survey.
Although many components of the HRS were incorporated into AHEAD, the influence has been a two-way street, and many innovations from AHEAD were incorporated into the HRS. Although natural tensions existed between the sibling surveys for the first two rounds with accommodations resulting in some disruptions in the continuity of measures, the two surveys have gradually blended into a single seamless instrument that will continue to provide valued data for the future study of America's older population.
Hurd, M. D. (1990). Research on the elderly: Economic status, retirement, and consumption and saving. Journal of Economic Literature, 28, 565-637.
Juster, F. T., & Suzman, R. (1995). An overview of the Health and Retirement Study. The Journal of Human Resources, 30 (Suppl.), S-7-S56.
Report of the Ad Hoc Advisory Panel to the Behavioral and Social Research Program, National Institute on Aging: Recommendations to the NIA Extramural Program on Priorities for Data Collection in Health and Retirement, May 1988, Administration Document.
Suzman, R., & Riley, M. W. (Eds.) (1985). The oldest old. Milbank Memorial Fund Quarterly, 63 (Special Issue).
Suzman, R. M., Willis, D. P., & Manton, K. G. (Eds.) (1992). The oldest old. New York: Oxford University Press.