Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 1
Trends in Disability at Older Ages: Summary of a Workshop Trends in Disability at Older Ages Over the past 25 years, mortality at all ages has declined dramatically in the United States (Manton and Stallard, 1984). Whether such increases in survival have been accompanied by changes in the prevalence of chronic disability remains unclear. Studies from the 1970s and early 1980s point to modest increases in the prevalence of disability at older ages (Crimmins, Saito, and Ingegneri, 1989; Riley, 1990; Verbrugge, 1989). More recent evidence suggests, however, that trends in disability may have declined during the 1980s (Manton, Corder, and Stallard, 1993a). A firm understanding of current trends in disability and their causes is critical to plan for the future of disability-related programs. Currently, more than 40 federal programs assist people with disabilities. Government programs—primarily Medicare, Medicaid, and those of the Department of Veterans Affairs—provide nearly two-thirds of the $120 billion spent annually on health care for the elderly. By identifying the population eligible for existing programs and determining current program participation rates, information on trends in disability could improve the accuracy of projected costs of such programs. Identifying trends in disability is also required to evaluate health care reform measures that seek to expand home health care benefits. At the request of the National Institute on Aging, the Committee on National Statistics held a workshop in October 1993 to review the data and methods used to determine trends in disability at older ages. The purpose of the workshop was to bring together scientists to discuss observed trends in disability, the forecasting of future trends, and the implications of findings for policy issues and future
OCR for page 2
Trends in Disability at Older Ages: Summary of a Workshop data collection. Participants included academic researchers and representatives of various federal and private-sector agencies with a policy interest in disability. The format of the workshop was informal to facilitate interaction among speakers, discussants, and observers. The specific objectives of the workshop were fourfold: (1) assess and compare findings from various studies of disability trends in the elderly population; (2) consider potential sources of bias that may explain differences in trends from different surveys; (3) identify potential explanations of observed trends in disability; and (4) examine implications of data and methods for forecasting disability trends. This report is a summary of the workshop proceedings and discussions. The first section reviews the evidence presented at the workshop regarding emerging trends in disability. The second section outlines issues raised regarding forecasting trends in disability, including implications of recent trends for further research. A summary of findings is presented in the final section. TRENDS IN DISABILITY Survey data, from both cross-sectional and prospective panel studies, provide insight into trends in disability at older ages. There is no general agreement with respect to the conceptual definition of disability, and no standard for measuring disability in old age. Furthermore, disability can be considered a multidimensional construct, encompassing both cognitive and physical limitations. Surveys of the population aged 65 and older generally focus on a person's ability to carry out activities essential to maintain independence. Such surveys usually include measures of an individual's ability to perform basic self-care activities, such as bathing, dressing, and feeding oneself. Such measures are generally referred to as activities of daily living or ADLs (Katz and Apkom, 1976). Measures that assess independence in activities requiring adaptation to the environment, such as shopping, preparing meals, and carrying out household chores, are generally called instrumental activities of daily living or IADLs (Lawton and Brody, 1969). Because ADLs and IADLs focus on the level of dependence in activities, rather than the underlying cause of that dependence, distinctions between physical and cognitive impairments cannot be drawn from such measures. Nevertheless, ADLs and IADLs are thought to be hierarchical, with ADL limitations representing more severe disability.
OCR for page 3
Trends in Disability at Older Ages: Summary of a Workshop Operational measures of functional disability vary widely across surveys (Wiener et al., 1990). The number and types of specific activities differ from survey to survey, as do definitions of dependence. Some measures target persons needing help; others focus on persons actually receiving help. Discrepancies also exist with respect to sources of assistance: some measures include only help from persons; others include help from either persons or special equipment. In addition, some definitions target persons with chronic disability (e.g., disability existing or expected to exist for a specified length of time); others have no minimum duration required to be considered disabled. The lack of uniform measures across surveys has resulted in a wide range of estimates of the size of the population with disabilities. Whether such definitional inconsistencies have led to contradictory estimates of disability trends is less clear. Disability estimates can be described with either stock or flow measures. In the disability literature, stock measures are typically referred to as prevalence, while flow measures are commonly called incidence. Disability prevalence is generally defined as the proportion of a population that is affected by disability at a point in time. In contrast, disability incidence is generally interpreted as the proportion of an initially nondisabled population that becomes disabled within a period of time. Closely related to the incidence rate is the hazard rate, which refers to an instantaneous risk of becoming disabled at a point in time. Prevalence and incidence are related mathematically, and the relationship may be complex. Whether they move in the same direction may depend, for example, on changes in the duration of disability. Below we review existing evidence with respect to trends in functional disability. We begin with a discussion of prevalence trends; this is followed by an examination of trends in disability incidence; potential threats to validity and possible causes of observed trends are then discussed. Prevalence Disability prevalence refers to the proportion of a population with disability at a specific point in time. If the content and administration of surveys are uniform over time, then trends in disability prevalence can be assessed by comparing proportions of disabled persons across years and calculating changes in either absolute or relative terms. By definition, analyses of prevalence trends implicitly
OCR for page 4
Trends in Disability at Older Ages: Summary of a Workshop control for changes in population size over time; however, this approach does not control for changes in the age composition of the population or other potential confounders. Conventional techniques employed to control for confounders include standardization and stratification. Alternatively, a regression framework may be employed to examine changes in disability prevalence over time, controlling for other potential confounders. At the workshop, findings on disability prevalence were presented from the National Long Term Care Survey (NLTCS), the National Health Interview Survey (NHIS), and the Longitudinal Study of Aging (LSOA). Trends in disability prevalence at older ages from other countries, including Australia, the United Kingdom, and Canada, were also reviewed. The National Long Term Care Survey The 1982, 1984, and 1989 NLTCS was designed to measure the prevalence and incidence of chronic disability in the U.S. elderly Medicare-enrolled population. The list sample for the initial screening was drawn from Medicare administrative files. Approximately 97 percent of the elderly population in the United States are eligible for Medicare. In 1982, a sample of 35,008 Medicare-eligible people over age 65 was drawn randomly and screened for chronic disability. A total of 6,393 chronically disabled community residents were interviewed in person to obtain more detailed information on health, functioning, and social and economic factors. In 1984, all survivors were reinterviewed in person. In addition, a sample of community residents who were not disabled in 1982 and a new sample of people who reached age 65 between 1982 and 1984 were screened for disability. The 1989 NLTCS design was similar to the 1984 survey. All those surviving to 1989 were reinterviewed. In addition, a sample of community residents who were not disabled in 1984 and those who reached age 65 between 1984 and 1989 were screened for disability. Overall response rates for the three survey years were extremely high, ranging from 97.1 percent in 1982 to 95.4 percent in 1989. In each of the three surveys, identical questions were asked about chronic disability. The NLTCS defines chronic disability as the actual or anticipated inability to perform a set of ADL or IADL activities for 90 days or longer. The ADL activities examined include bathing, dressing, eating, getting
OCR for page 5
Trends in Disability at Older Ages: Summary of a Workshop in and out of bed, getting to the bathroom or using the toilet, and getting around outside. IADL activities included meal preparation, grocery shopping, money management, using the telephone, light housework, laundry, getting around outside, and getting to places beyond walking distance. All community residents in the sample were asked about their ADL and IADL limitations in 1982, 1984, and 1989; in addition, institutionalized sample members were asked about their ADL limitations in 1984 and 1989. Estimates of chronic disability prevalence from the 1982, 1984, and 1989 NLTCS are shown in Table 1. All estimates are weighted to represent the elderly U.S. population and have been adjusted for nonresponse. Results show that the population aged 65 and older grew from 26.9 million in 1982 to over 30.8 million in 1989, an increase of 14.7 percent over the seven-year period. During the same time period, the oldest-old population (those aged 85 and older) grew twice as fast as the population aged 65 to 84. The growth of the chronically disabled population aged 65 and older, in contrast, was only 9.3 percent over the period, from about 6.4 million to nearly 7 million people. Standardizing TABLE 1 Trends in Chronic Disability Prevalence: 1982-1989 NLTCS (U.S. Population Aged 65+) 1982 1984 1989 N(000s) % N(000s) % N(000s) % Nondisabled 20,543 76.3 21,396 76.3 23,894 77.4 Disabled 6,381 23.7 6,646 23.7 6,977 22.6 IADLs Only 1,436 5.3 1,598 5.7 1,358 4.4 Any ADL 3,419 12.7 3,477 12.4 3,921 12.7 1-2 ADLs 1,750 6.5 1,823 6.5 2,007 6.5 3-4 ADLs 727 2.7 785 2.8 1,080 3.5 5-6 ADLs 942 3.5 869 3.1 834 2.7 Institution 1,535 5.7 1,542 5.5 1,698 5.5 Total 26,924 100.0 28,042 100.0 30,871 100.0 N (Unweighted) (19,142) (20,474) (22,146) Source: Adapted from Manton, Corder, and Stallard (1993a). Reprintedby permission of the Gerontological Society of America.
OCR for page 6
Trends in Disability at Older Ages: Summary of a Workshop results for changes in the age distribution between 1982 and 1989 yields even smaller increases in the chronically disabled population. Source: Adapted from Manton, Corder, and Stallard (1993a). Reprintedby permission of the Gerontological Society of America. Despite the absolute increases in the number of older persons with chronic disability, the proportion declined from 23.7 percent in 1982 to 22.6 percent in 1989—a decline of 1.1 percentage points. There is some disagreement as to whether this change is occurring at all levels of chronic disability or mostly among people with chronic IADL disability. Manton et al. (1993a) argue that declines have occurred in both chronic IADL and chronic ADL disability. Indeed, from 1982 to 1989 declines were observed in the proportion of the older population with IADL limitations (0.9 percentage points), 5-6 ADL limitations (0.8 percentage points) and living in institutions (0.2 percentage points). Standardizing these results for changes in the age distribution over the period yields even larger declines in the proportion of the older population with chronic disability; results show declines in the proportion of the older population with IADL disability (1.0 percentage points), 1-2 ADL limitations (0.2 percentage points), 5-6 ADL limitations (0.9 percentage points), and living in institutions (0.7 percentage points). The authors also point out that these declines are consistent with disability reductions found in other studies (e.g., Bebbington, 1988; Svanborg, 1988) and with reductions in specific causes of disability, such as osteoporosis (Naessen et al., 1989) and circulatory diseases (Manton, 1990). Looking at the same figures, others conclude that declines in the proportion of the older population with chronic IADL disability have occurred along with shifting across levels of ADL disability. The percentage of the older population with IADL disability declined from 5.3 percent in 1982 to 4.4 percent in 1989, a decline of 0.9 percentage points. During the same time period, the percentage with any ADL disability remained relatively constant over the decade, at about 12.7 percent. In the nonstandardized figures, increases in the percentage of people with 3-4 ADL limitations (0.8 percentage points) completely offset declines in the percentage with 5-6 ADL limitations. Even in the age standardized estimates, the declines are not observed across all levels of ADL limitations: the percentage of older people with 3-4 ADL limitations increases by 0.7 percentage points.
OCR for page 7
Trends in Disability at Older Ages: Summary of a Workshop The National Health Interview Survey The National Health Interview Survey is an annual household survey based on nationally representative samples of the U.S. civilian, noninstitutionalized population. Data are collected via personal interviews with household respondents. Annual samples are comprised of approximately 40,000 households, representing over 100,000 people. Noninterview rates are relatively low, ranging from 3.3 percent in 1983 to 5.1 percent in 1988 and 1989. Weighted data represent the U.S. noninstitutionalized population. Each year since 1983, the NHIS core interview schedule has included identical questions about activity limitations. The survey identifies disabled persons with a series of three questions. For each household member, respondents are first asked whether because of any impairment or health problem that household member needs the help of other people with personal care activities (“such as eating, bathing, dressing or getting around this home”). The second question addresses the need for help from other persons with what are considered routine needs (“such as everyday household chores, doing necessary business, shopping, or getting around for other purposes”). A third question asks whether that household member is limited in any way in any activities because of an impairment or health problem. In the NHIS, the definition of disability does not include a minimum duration restriction. It is important to note that the questions in the NHIS are not, strictly speaking, ADL and IADL measures because they do not address limitations separately for each activity. Nevertheless, the types of activities referred to as personal care activities and routine care activities are generally considered to be ADLs and IADLs, respectively. Thus, for purposes of this report, in order to facilitate comparisons across surveys, we use the terms ADL and IADL disability interchangeably with personal care and routine care limitations, respectively, when discussing the NHIS. Plots of age- and sex-specific disability estimates over time were presented at the workshop using data from the NHIS. These plots suggest that trends in personal and routine care limitations from 1983 to 1992 have been substantially linear. Summary measures of these linear trends are presented in Table 2. Coefficients (ß) represent, for each age and sex group, the average changes in the proportion with disabilities. P-values are for t-tests indicating whether ß coefficients are statistically different from 0.0.
OCR for page 8
Trends in Disability at Older Ages: Summary of a Workshop TABLE 2 Trends in Disability Prevalence: 1983-1992 NHIS (U.S. Population Aged 60+) Males Females ß (p-value) ß (p-value) Personal Care Limitations Age Group 60-64 −0.05 (0.17) 0.03 (0.24) 65-69 −0.05 (0.47) 0.02 (0.65) 70-74 0.04 (0.66) 0.05 (0.35) 75-79 0.07 (0.52) −0.08 (0.43) 80-84 0.14 (0.32) 0.19 (0.18) 85+ −0.005 (0.99) 0.14 (0.50) Routine Care Limitations Age Group 60-64 −0.11 (0.01) −0.12 (0.09) 65-69 −0.07 (0.06) −0.12 (0.05) 70-74 −0.05 (0.50) −0.31 (0.01) 75-79 −0.03 (0.85) −0.10 (0.45) 80-84 −0.43 (0.01) −0.27 (0.04) 85+ −0.59 (0.001) −0.57 (0.09) Note: p-values are for t-tests on ß coefficients. Source: Feldman et al. (1993). Results indicate a statistically significant decline in the proportion of the population with routine care limitations for most age groups; the relative number of those with personal care limitations, however, has not changed appreciably. Declines in routine care limitations are observed for both men and women in nearly all age groups with the greatest declines at the oldest ages. The Longitudinal Study of Aging The LSOA began in 1984 as a supplement to the core NHIS. The supplement was administered to 7,541 noninstitutionalized people aged 70 and older in 1984. Subjects were reinterviewed in 1986, 1988, and 1990. In 1988 and 1990, all of the original sample members (including those in institutions) were eligible for reinterview; in 1986 only a subset of the original respondents were reinterviewed (N = 5,151). Baseline interviews were conducted in person; follow-up interviews were administered by
OCR for page 9
Trends in Disability at Older Ages: Summary of a Workshop telephone. Attrition rates increased over the survey period from about 8 percent at the first follow-up to over 15 percent in 1990. In all four years, the LSOA asked a series of questions to assess whether respondents were unable to perform (or had difficulty performing) activities alone and without using special equipment. Activities of daily living assessed include: bathing or showering, dressing, eating, walking, getting in and out bed or chairs, and using the toilet (including getting to the toilet). IADLs include: preparing meals, shopping for personal items, money management, using the telephone, and light housework. Weighted data are representative of the noninstitutionalized elderly population. Results from the LSOA are presented in Table 3.1 The analysis is limited to the noninstitutionalized population aged 76 and over, in order to ensure comparability across the four years. Estimates have been adjusted using a combination of imputation and weighting procedures to account for sample attrition. Overall, the LSOA results indicate no consistent change in the prevalence of ADL or IADL disability over the six-year period. ADL disability prevalence increased modestly for men from 6.1 percent in 1984 to 7.5 percent in 1990 but showed no trend for women. In contrast, the proportion of TABLE 3 Trends in Disability Prevalence: 1984-1990 LSOA (U.S. Population Aged 76+) 1984 1986 1988 1990 ADL Disabled Males 6.1 6.5 6.8 7.5 Females 8.1 10.6 8.4 10.2 Total 7.3 9.1 7.8 9.2 IADL Disabled Males 9.2 12.5 10.1 10.7 Females 12.7 12.6 11.5 11.6 Total 11.4 12.6 11.0 11.3 (N) (5,135) (4,293) (3,643) (2,942) Source: Crimmins and Saito (1993). 1 These results reflect more refined estimates submitted by Crimmins following the workshop presentation. Conclusions drawn from the revised estimates were essentially unchanged from the original presentation.
OCR for page 10
Trends in Disability at Older Ages: Summary of a Workshop women with IADL disability declined modestly from 1984 to 1990, from 12.7 percent to 11.6 percent but showed no trend for men. Logistic regression models yielded no consistent trend: controlling for age and sex, sample members were more likely to be disabled in 1986 and 1988 than in 1984, but there was no significant difference in the prevalence of disability in 1990 and 1984. International Data In order to explore whether declines in disability are occurring outside the United States, trends from other countries were also reviewed at the workshop. Despite widespread collection of disability statistics, few countries outside the United States have collected the data necessary for describing trends in disability over time. Countries with repeated cross-sectional data include Germany, Japan, Australia, the Netherlands, Taiwan, the United Kingdom, and Canada. In Germany, samples have been drawn from censuses and registries. In Japan, surveys of disabled people have been fielded approximately every five years since 1950. In most other cases, repeated cross-sectional surveys have been administered in the form of household surveys. The primary source for international data on disability trends presented at the workshop is the United Nations Disability Statistics Data Base. At the workshop, data from Australia, the United Kingdom, and Canada were presented. In Australia, two National Household Health Surveys have been undertaken in the past decade. In both 1981 and 1988, surveys were fielded using independent nationally representative samples, covering both the community and the institutionalized populations of all ages. In both years, the severity of impairments and disability was assessed using identical instruments. A person is considered disabled if he or she has any of the following specific impairments or disabilities for at least six months: loss of sight or hearing, speech difficulties, blackouts, slowness at learning, incomplete use of limbs, long-term treatment for emotional problems, restricted activities, disfigurement, or the need for supervision. Handicapped persons are defined as disabled persons aged 5 years and older who are also limited to some degree in their ability to perform certain tasks. Such tasks include: self-care (i.e., bathing, dressing, eating), verbal communication (understanding and being understood by others), mobility (i.e., using transportation, moving around inside and outside the home), schooling, and employment. Severely handicapped refers to being unable to perform one of these activities without help or supervi
OCR for page 11
Trends in Disability at Older Ages: Summary of a Workshop sion; moderately handicapped refers to having difficulty with one or more tasks; and mildly handicapped is defined as having no difficulty but using an aid in carrying out such tasks. A more detailed description of this survey is found in Mathers (1991). In 1971, a community-based survey of persons aged 65 and older was undertaken in Manitoba, Canada. Approximately 3,850 people, sampled from the master registry of the provincial health insurance system, were interviewed regarding health and disability. In 1983 survivors were reinterviewed along with a sample of elderly respondents, for a total sample size of 4,145. Measures of disability included the ability to carry out activities of daily living, self-perceived health, mental performance, interviewer rating, and high-risk diagnosis. See Roos, Havens, and Black (1993) for a more detailed description of this survey. In Melton Mowbray, U.K., cross-sectional surveys of the population aged 75 and older were undertaken in 1981 and 1988. Because the Physician Registry was used as a sampling frame, the surveys cover both the community and the institutionalized populations. Approximately 1,200 and 1,580 people aged 75 and older were interviewed in 1981 and 1988, respectively. Instruments to assess disability were identical in the two years. Disability was assessed with a series of ADL questions; in addition, the instrument included questions about incontinence and cognitive functioning. Details on this study are found in Jagger, Clark, and Clark (1991). Table 4 shows crude prevalence rates of disability for the United States; Australia; Melton Mowbray, U.K.; and Manitoba, Canada. Estimates from the NLTCS represent the percentage of people with IADL or ADL disability or in an institution. Estimates from the Australia Health Surveys identify. people in both the community and in institutions who require personal help or supervision or are unable to carry out self-care activities, verbal communication, mobility, schooling, or employment. Disability is defined as needing help in performing any of six activities of daily living for both Melton Mowbray, U.K., and Manitoba, Canada. The Melton Mowbray survey includes the institutionalized elderly; the Manitoba survey is strictly community-based and refers to an earlier time period (1971-1983). These results show no consistent trends in disability in the four countries. In Australia and Manitoba, for example, the proportion of older persons with disability has increased. In contrast, in Melton
OCR for page 18
Trends in Disability at Older Ages: Summary of a Workshop ranging from 2.9 percent in 1984 to 4.6 percent in 1989. In the LSOA, however, attrition is sizable (ranging from 8 percent in 1986 to 15 percent in 1990). Large increases in attrition observed over the study period will exaggerate declines in disability prevalence if people lost to follow-up are more likely than those retained to be disabled; similarly, declines in incidence will be overstated if people lost to follow-up are more likely than those retained to become disabled. Adjustments for attrition often involve recalculating sample weights under the assumption that those lost to follow-up are essentially the same as those followed. Trend lines will be biased depending on the accuracy of such assumptions. Fifth, the use of longitudinal data can also influence prevalence trend estimates if only people with the most severe disability are selected into the sample; the trend will appear to improve due to the statistical phenomenon of regression to the mean. The LSOA is not vulnerable to this statistical artifact because it began in 1984 as a random sample of the noninstitutionalized elderly population. The NLTCS, however, screens for chronic disability. Two factors mitigate the effects of regression toward the mean in the NLTCS: (1) the aged-in and screened-in subsamples help minimize correlation across years and (2) because of detailed questions on disability, NLTCS respondents are distributed over a wide spectrum of disability states, including those who recover completely from disability. Finally, assumptions underlying estimates pose potential threats to the validity of both prevalence and incidence trends. Comparisons of crude prevalence measures do not take into account, for example, shifts in the age and sex composition of the population. Because the elderly population has been itself aging, crude prevalence measures will underestimate the decline in disability during the 1980s. Age-standardized estimates, such as those estimated with the NLTCS, show expected trends in the absence of changes in the age structure of the population. Standardized estimates also involve assumptions, which vary with the method of standardization, but robustness can be demonstrated by considering a variety of approaches. In sum, although there are a number of potential threats to the validity of the results presented at the workshop, it is unlikely that the sources reviewed here are of sufficient magnitude to account for the observed declines in disability. Given the uniform content and administration of the NHIS and the NLTCS, their relatively low losses to follow-up, and small declines in the proxy interviews, the observed declines in disability are unlikely to be spurious effects of survey design or statistical meth
OCR for page 19
Trends in Disability at Older Ages: Summary of a Workshop odology. However, large losses to follow-up and a shift toward proxy interviewers may effect the prevalence and incidence trends from the LSOA. Such sources of bias could have opposite effects on trend estimates, with large losses to follow-up resulting in an overstatement of the decline in disability and shifts to proxy respondents dampening declines. Possible Causes of Observed Trends The existing data point to modest declines in the prevalence of IADL limitations and in the incidence of both IADL and ADL disability. In addition, although there is some debate in the interpretation of the evidence with respect to trends in the prevalence of chronic ADL disability, there is general agreement that the prevalence of ADL disability is not increasing. These trends are particularly striking in light of the fact that disability appeared to have increased during the 1970s. Currently there is insufficient evidence to conclude that there was a definite break in trend during the 1980s. The apparent shift in trend could be a consequence of a change in survey content. In the case of the NHIS, for example, before 1982 questions were asked about activity limitations; after 1982 questions were changed to reflect ADL and IADL concepts. Alternatively, the shift in trend may reflect changes in the way individuals interpret and respond to survey instruments. Nevertheless, if reported disability is truly declining, as workshop findings suggest, understanding the dynamics underlying this phenomena is an important goal, particularly if the rate of decline intensifies in the future. Workshop participants considered several explanations for observed declines in disability, including an improving education profile of succeeding cohorts, improvements in the delivery of health care services, and changes in interpretation of the concept of disability. Education has been linked to lower morbidity rates and greater self-sufficiency across the life-cycle (Preston, 1992). Although the exact mechanism remains unknown, education most likely operates on health and functioning through a combination of health promotion and disease prevention activities. Projected distributions of educational attainment among the older population indicate a rapid upgrading of schooling among the oldest cohorts during the 1980s and beyond (Preston, 1992). Hence, as better-educated cohorts enter old age, declines in disability at older ages may be witnessed
OCR for page 20
Trends in Disability at Older Ages: Summary of a Workshop either because the age of disease onset increases or the rate of disease severity is slowed due to earlier interventions. Findings from the LSOA suggest that, for persons aged 70 and older, the chances of needing help with personal care has actually increased slightly over time after controlling for improvements in education. It remains unclear whether education has a direct impact on the health of elderly cohorts or whether education operates on older people 's perceptions and reports about their health. Dramatic changes in infrastructure, particularly the availability of medical devices, also may have had a large effect on disability reports over the decade. A recent study by Manton, Corder, and Stallard (1993b) suggests that there has been an increase in the use of assistive devices—such as ramps and wheelchairs—and a decrease in the use of personal assistance from 1982 to 1989. If assistance devices alter older people's perceptions about their health or functional independence, increased use of technologies could improve self-reports of health, leading to apparent declines in disability. Alternatively, it is important to keep in mind that questions typically used to index disability are self-reports of subjective needs. As a result, it is possible that what has changed are individuals' interpretation of behaviors associated with disability (e.g., reports of dependence on mechanical aides or personal helpers) rather than the actual clinical markers of disability. Downward trends in disability may therefore reflect the extent to which older individuals are less inclined to report dependence behaviors or the extent to which they can more easily adjust to daily behaviors to compensate for losses in function. Similarly, downward trends in chronic disability, which in part rely on self-reports of expected length of disability, may reflect changes in respondents' perceptions about the future course of their disability. FORECASTING DISABILITY Effective planning for medical, rehabilitative, and social services requires reliable estimates of the numbers and characteristics of current and future disabled people. Projections of the elderly disabled population are used by a number of federal agencies with disability benefit program responsibilities, including the Social Security Administration, the Health Care Financing Administration, and the Department of Veterans Affairs. At the workshop, participants discussed a number of forecasting
OCR for page 21
Trends in Disability at Older Ages: Summary of a Workshop issues, including the importance of a clear conceptual framework and the advantages and disadvantages of various statistical models. Conceptual Frameworks There is no general agreement with respect to the conceptual definition of old-age disability. Disparities stem in part from the complex nature of disability, which involves not only physical and cognitive capacity and performance, but also social circumstances and perceptions about ability. Furthermore, disability is inherently a dynamic process in which function can decay or improve over time, even among those with the same underlying condition. Given the multidimensional and dynamic nature of disability, clear conceptual definitions are necessary to provide guidance in constructing operational definitions for forecasting models. There was a general consensus at the workshop that conceptual definitions should flow from a broader theoretical model of health in old age. Yet the concepts of etiology, pathology, and manifestation of disease embedded in the classical medical model are not necessarily appropriate for the study of chronic disabling disease (for further discussion of this point see Stoto and Durch, 1993). Two alternative models were proposed at the workshop as possible frameworks for the study of disability. One model was developed at a World Health Organization Scientific Group meeting on the Epidemiology of Aging (World Health Organization, 1984). The model describes changes in population health as a series of survival curves, representing the proportion of a cohort surviving to a given age without morbidity, disability, or mortality (Manton and Soldo, 1992). Areas between the curves indicate the number of person-years spent in a given health state: disease-free, diseased but unimpaired, and disabled. Because this model incorporates both health changes at the individual level and patterns of disability at the population level, it is especially useful for assessing aggregate needs for health-related services. Another model, proposed by the World Health Organization for comparing disability measures across countries, is the International Classification of Impairments, Disabilities and Handicaps (ICIDH) (World Health Organization, 1980). Although not a theoretical model per se, the ICIDH is a useful classification system for guiding the construction of disability measures and relating disability to a
OCR for page 22
Trends in Disability at Older Ages: Summary of a Workshop range of policy measures. The ICIDH defines impairment as disturbances at the level of the organ; disability is defined as restrictions in function at the level of the person; and handicap refers to the difficulties encountered by people who are impaired and/or disabled in relation to their environments and culture. From a policy perspective, measures of impairments are most useful for questions focusing on prevention; disability measures help answer questions about rehabilitation programs; and handicap measures are relevant to discussions of how to equalize opportunities for persons with impairments and disabilities (United Nations, 1990). Statistical Models The health of elderly populations can be projected using a number of different forecasting strategies, including actuarial, economic, demographic, and epidemiologic models (Manton, Singer, and Suzman, 1993). In general, federal government programs have employed actuarial forecasting methods, which traditionally have been used to anticipate the future fiscal risk of programs based on well-defined past experiences. At the same time, the research community has developed more sophisticated models of health and functioning that describe the risk of morbidity and death as a function of individuals' health states. The Social Security Administration, for example, uses the actuarial method to project applicants and beneficiaries of the disability insurance (DI) program. The DI program provides monthly cash payments for disabled workers under age 65 and their dependents. Disability is defined as the inability to engage in any “substantial gainful activity” due to a medically determinable physical or mental impairment. Approximately 5 million persons currently receive benefits from the DI program. The methodology for forecasting the DI program population involves the identification of new beneficiaries by applying age- and sex-specific disability rates to population projections. New beneficiaries are then “aged” forward through time using standard mortality schedules. Beneficiaries are removed from the rolls once they no longer meet the Social Security Administration's definition of disability. Common reasons for termination are death, recovery, and reaching age 65 (conversion). Sensitivity analyses reflect “optimistic” and “pessimistic” population assumptions and are not true confidence intervals.
OCR for page 23
Trends in Disability at Older Ages: Summary of a Workshop In the recent past, projections have been consistently lower than actual beneficiary levels, in part because actual disability rates have been much higher than projected. As a result, the number of DI beneficiaries has exceeded projected numbers in recent years. In addition to anticipated changes in demographic conditions (most notably the aging of the baby boom generation), other prominent reasons accounting for the large increases in DI beneficiaries include unanticipated changes in economic conditions, programmatic policies and public outreach programs. Surprising to some, AIDS-related awards have had very little impact on the dramatic increases. It can be reasoned, however, that awards resulting from HIV impairments have a minimal effect due to an abbreviated lifespan and consequently a shortened period of time on the rolls. An alternate approach to modeling disability has been undertaken by Manton and colleagues who address the dynamic nature of health and functioning, particularly at advanced ages. The centerpiece of this approach is the Grade of Membership (GOM) model (Woodbury and Manton, 1977, 1983). Rather than compartmentalizing individuals into discrete categories, the GOM model quantifies the degree of membership of each individual in a number of predetermined health states. In contrast to classical modeling techniques, GOM allows for the incorporation of many dimensions of functioning while remaining robust to any single measurement. Disability forecasts that model health states with GOM techniques can build in correlations between morbidity, disability, and mortality and allow for shifting relationships among these processes over time. In addition, the sensitivity of projections to the modification of behavioral risk factors can be incorporated into models. Generally, projections from models with such interdependencies will not agree with simpler Markov models, which assume, given the present state of health, no effect of past history on the future course of health and no individual heterogeneity. The GOM framework can be used to address not only the traditional economic factors affecting disability but also social factors, such as behavioral risk factors for disease and developing infrastructure (e.g., medical devices). Because this approach to forecasting incorporates more information than government projection methodologies, it is likely to provide a more realistic prediction model of health and disability. Although GOM offers a relatively large degree of flexibility, it still requires specifica
OCR for page 24
Trends in Disability at Older Ages: Summary of a Workshop tion of structural parameters that involves assumptions about the unobserved pure types. Empirical checking of such assumptions is key to ensure the robustness of GOM estimates. Workshop participants also discussed the importance of enumerating and defending model assumptions. Potential threats to external and internal validity are, for example, embedded in the construction of sampling weights, adjustments for sample design and attrition, and the use of proxy responses. In order to defend findings as robust, sensitivity analyses should be conducted. In addition, the fit of forecasting models should be assessed statistically, and statistically based confidence intervals should be placed around projections. Future Research In general, participants agreed that trends in disability should be monitored continuously by researchers. Data on disability are useful for policy makers to evaluate the impact of varying eligibility criteria and participation rates on the projected size and cost of disability programs. Survey data are particularly important because, unlike administrative data, they can be used to evaluate proposed rule changes. Several new sources of data to address trends in disability at older ages were identified during the workshop. Recently the University of Michigan completed the first wave of the Health and Retirement Survey (HRS). The HRS, which targets the cohort of 1931-1941 (adults aged 51 to 61 in 1992), includes a core set of questions on functional, health, and disease states as well as a disability module designed to assess occupational disability and employer accommodations made to disabilities. The University of Michigan also is undertaking the first wave of the survey of Asset and Health Dynamics (AHEAD), which will offer insight into trends in disability and resource consumption among persons aged 70 and older in the coming decade. Researchers at Duke University are preparing to collect another round of the NLTCS, scheduled to be in the field in June 1994. In addition, several federal initiatives will address disability at older ages. The National Center for Health Statistics is planning a supplement to the NHIS in 1994-1995 that focuses exclusively on disability, and the Social Security Administration is planning a medical examination study for a sample of DI program participants.
OCR for page 25
Trends in Disability at Older Ages: Summary of a Workshop Simplifying linkages between survey data and administrative data sources (i.e., Medicare, National Death Index) would facilitate analyses by both government analysts and academic researchers. Such linkages are now carried out on an ad hoc basis. Both the NLTCS and the LSOA are linked to Medicare and the National Death Index, for example, but the NHIS is not linked to these administrative data bases. The Committee on National Statistics and the Social Science Research Council have recently reviewed issues of data confidentiality and linkages in the report of their Panel on Confidentiality and Data Access (Duncan, Jabine, and de Wolf, 1993). Research on disability at older ages will continue to guide public policy makers. Surveys using questions tailored to existing programs can guide incremental program rule changes. In addition, longer term surveys that use conceptual frameworks to guide disability questions may be useful in the future to help shape new programs. The need for information of the latter kind is illustrated by the recent shift in policy embodied in the Americans With Disabilities Act (ADA). Unlike previous disability-related legislation, this act focuses on the equalization of opportunity for people with limitations. Because federal surveys are generally tied to existing programs, little information on disability accommodations is currently available. The new HRS provides coverage of disability accommodation in the workplace, but only for relatively old workers. Future research also should attempt to bridge research on disability in old age and earlier in life. Although the risk of disability is quite low prior to old age, the nonelderly disabled population is of particular interest because their limitations often last for extensive periods of time. Currently an appropriate conceptual framework and corresponding measures are not available to study disability across the lifespan. SUMMARY Over the past decade, the population aged 65 and older grew from about 26.9 million to more than 30.8 million. During the same time, the population of those aged 85 and older grew even faster. Despite this aging of the population, functional disability has not increased either in terms of prevalence (i.e., the proportion disabled) or incidence (i.e., the proportion becoming disabled).
OCR for page 26
Trends in Disability at Older Ages: Summary of a Workshop Instead, modest declines in the proportion of older persons with limitations in instrumental activities of daily living (e.g., shopping, preparing meals, carrying out household chores) have occurred. Analysis of the National Long Term Care Survey shows that a smaller percentage of older persons had chronic difficulty performing IADL activities in 1989 than in 1982. Findings from the National Health Interview Survey point to declines in the proportion of the older population with IADL disability for older men and women of all ages, with declines greatest at the oldest ages. Whether or not declines in disability associated with activities of daily living (e.g. bathing, dressing, feeding) have occurred remains open to debate. Analysis of the National Health Interview Survey and the Longitudinal Study of Aging shows no discernible trend in the proportion of the older population with any ADL limitations. Furthermore, data from selected other countries do not support declines outside the United States. Findings from the National Long Term Care Survey, standardized to control for age shifts in the population, suggest that chronic ADL disability may have declined, but not consistently across all levels of disability. Evidence presented at the workshop points to modest declines in the onset of old-age disability over the last decade. According to analysis of the National Long Term Care Survey, declines in chronic ADL disability have slowed for all levels of disability. Results from the Longitudinal Study of Aging also point to declines in the onset of disability through the 1980s for all age groups, with the greatest declines at the oldest ages. Potential sources of bias, such as changes in the content and administration of surveys, shifts to proxy respondents, and increases in attrition, are unlikely to account for these declines in disability. Given their relatively uniform content and administration, the relatively low losses to follow-up and small declines in the proxy interviews, bias is unlikely to account for the declines in disability reported from the NHIS and the NLTCS. In contrast, large losses to follow-up and a shift toward proxy interviewers may introduce bias into the prevalence and incidence estimates from the LSOA. Such sources of bias are likely to have opposing effects on trend estimates, with large losses to follow-up resulting in an overstatement of the decline in disability and shifts to proxy respondents dampening declines. A number of factors may be driving observed disability trends. Improvements in education of more recent cohorts of elderly people may explain, in part, declines in disability. Although the exact
OCR for page 27
Trends in Disability at Older Ages: Summary of a Workshop mechanism remains unclear, education is most likely to operate on health and functioning through a combination of health promotion and disease prevention activities. Dramatic changes in infrastructure during the 1980s, particularly the availability of medical devices, also may have had a large effect on disability reports over the decade. Increased use of technologies may improve self-perception and thus self-reports of health, leading to apparent declines in disability. Alternatively, changes in perception and interpretation of disability may be driving observed trends. Effective planning for medical, rehabilitation and social services requires reliable estimates of the numbers and characteristics of current and future disabled people. Projections of the elderly disabled population are used by a number of federal agencies with disability benefit program responsibilities and by policy makers interested in modifying programs and proposing new services for this population. Useful projections of this population will, of necessity, rely on the specification of both a clear conceptual model and a flexible statistical model.
OCR for page 28
Trends in Disability at Older Ages: Summary of a Workshop This page in the original is blank.
Representative terms from entire chapter: