Recent Trends in Disability and Functioning Among Older Adults in the United States: Title and subTitle BreakA Systematic Review FREE

Disability and underlying physical, cognitive, and sensory limitations are not inevitable consequences of aging. Yet 20% of older US adults have chronic disabilities,¹ 7% to 8% have severe cognitive impairments,² roughly one third have mobility limitations,³ 20% have vision problems,⁴ and 33% have hearing impairments.⁴ Women, minorities, and persons of low socioeconomic status are especially vulnerable.^{5 - 6}

The cost of medical care for a disabled older person averages 3 times that for a nondisabled senior.⁷ Moreover, long-term care expenditures for older US residents with disabilities (including those receiving nursing home or community-based care) reached $123 billion in 2000, with more than 65% paid by government.⁸ Families also provide substantial uncompensated informal care.⁹ As the number of older persons burgeons, the proportion needing assistance with daily tasks may also increase. Indeed, a fundamental question in geriatrics is whether recent mortality decline has been accompanied by a compression or expansion of periods of morbidity.^{10 - 12}

In recent years, more than a dozen published studies have detailed changes in self-reports of disability and underlying functioning problems of older US adults. Although the evidence is sometimes conflicting, several well-publicized studies have suggested that rates of disability, severe cognitive impairment, and functional limitations have declined substantially.^{1 - 3 ,13 - 14} Whether such improvements extend to all types of such difficulties and all groups remains unclear. If improvements are pervasive and continue, the effect on US health and economic well-being could be far reaching, with potentially more older persons able to work longer and relatively fewer needing medical and long-term care.^{15 - 18}

To our knowledge, no systematic review of the quality, quantity, and consistency of this literature has been conducted to date. Herein, we synthesized and critiqued the evidence addressing 2 specific questions: What are the most recent trends in the prevalence of late-life disability and functioning? Are all socioeconomic and demographic groups benefiting equally, or are disparities widening or narrowing?

To identify pertinent studies, we focused on 2 key dimensions of late-life health: (1) physical disability, defined as the inability to carry out independently specific roles or activities within a given environment; and (2) functional, cognitive, and sensory limitations, defined as the underlying difficulty a person has with specific physical, memory-related, and vision- and hearing-related tasks. For example, needing help with bathing or medication management may be considered a disability, whereas difficulty bending, remembering, or seeing may be considered limitations.

Although in practice not completely distinct, each of these broad categories reflects a critical step in the "disablement process," a framework developed by the Institute of Medicine and others.^{19 - 20} The framework describes the process of progressive changes through which disease and injury lead to limitations (in either organ function or in the function of the individual) and ultimately lead to disability (defined as the inability to carry out a specific role in a given environment). Although the questions vary from survey to survey, disability is most often measured by self-reports of either needing help or having difficulty with activities of daily living (ADL)²¹ and instrumental activities of daily living (IADL).²² Physical (also called functional) limitations are often measured by self-reported difficulty with specific body tasks proposed by Nagi²³ (such as reaching, bending, stooping); cognitive limitations have been measured with cognition tests or questions about memory; and sensory limitations are often measured by self-reports of vision or hearing difficulties. Although often internally consistent²⁴ these self-reported measures may not always agree with performance-based measures.²⁵ Moreover, these measures capture different dimensions of the disablement process and therefore will not necessarily move in concert over time at the population level or in the same direction for all subgroups.²⁶

To identify studies focused on the late 1980s and 1990s, we searched MEDLINE and AGELINE for relevant articles published in English from January 1990 through May 2002. We used 3 categories of keywords: trends; aging or older adults; and disability, cognitive impairment, functioning, vision, or hearing. We contacted authors to identify additional articles, including those in press (however, conference presentations and reports in progress were not included). We also reviewed reference lists in all relevant published articles, including reviews^{27 - 29} and commentaries.^{15 ,26} More than 800 titles and/or abstracts were reviewed in all. We identified 27 articles and reports that related to the topic of old age disability or functioning trends for further review.

This set of studies was narrowed to those pertaining to prevalence estimates for the older US population (generally ages ≥65 or 70 years). Moreover, because of our focus on understanding recent changes, all studies chosen had to include at least 1 data point in the mid-1990s. We excluded studies focusing on the early 1980s or earlier,^{13 ,30 - 33} on century-long trends,^{34 - 35} or exclusively on trends in the institutionalized population^{36 - 38} or health care utilization.¹⁷ The remaining 16 reports^{1 - 4 ,14 ,16 ,39 - 48} were based on 8 unique surveys.

To assess trends, we evaluated these surveys on 10 criteria, each of which represents a potential threat to the validity of comparisons over time (Table 1). These criteria were developed based on a 1994 workshop report on disability trends prepared by the National Research Council,²⁸ as well as on a subsequent report to the government.²⁹

We did not explicitly evaluate the reliability of outcome measures, including intrarater reliability, for 3 reasons. First, survey question wording varies widely and no widely acceptable standard for measuring self-reported disability exists. Second, these reliabilities are rarely reported in published studies. Third, the national surveys often rely on previously validated and widely accepted self-assessment scales^{21 - 24} that generally demonstrate good internal consistency.^{24 - 25}

Several of the studies^{2 ,16 ,42 ,47} included sensitivity analyses that demonstrated robustness of findings to particular survey features that otherwise would have contributed to a designation as fair. For example, several studies explicitly addressed the exclusion of the institutionalized population by combining their results with data from national nursing home studies,^{2 ,16 ,42 ,47} and other studies have demonstrated robustness to missing data and loss to follow-up.^{2 ,42} In these instances we considered the relevant survey feature to be good rather than fair.

We assigned a summary rating (good, fair, or poor) to each survey based on the following rules: surveys with 2 or more poor features (out of 10) at the time of the analysis were designated to be poor; a rating of good was reserved for surveys with at least 5 good and no poor features; the remaining surveys were designated to be fair. To assess trends in disparities, we also considered whether the study included statistical tests for differences in trends over time, but we did not explicitly rate this factor.

As summarized in Table 2, the studies considered herein were based on 8 surveys, of which 2 were rated as good,^{1 ,14 ,41 ,47} 4 were rated as fair,^{2 - 4 ,40 ,42 - 44 ,46 ,48} 1 was rated as poor,³⁹ and 1 was given a mixed rating (fair or poor, depending on the outcome).⁴⁵

The 2 surveys classified as good offered different strengths and weaknesses. Studies based on the National Health Interview Survey,^{41 ,47} for example, shared the advantage of annual surveys of independent cross-sections of the population, but excluded the institutional population and contained only global (nondetailed) assessments of disability. In contrast, the National Long Term Care Survey,^{1 ,14} one of the best designed surveys for analyzing national disability trends, shared the following strengths: coverage of the full 1990 decade; inclusion of the institutional population; identical field procedures; detailed disability questions; low loss-to-follow-up rates. The only relatively weak feature of this survey for assessing trends (which may fluctuate from year to year) was that it was administered only once every 5 years.

Five data sets were rated as fair for assessing trends because at least 1 criterion was considered poor. For example, at the time it was analyzed, the Asset and Health Dynamics of the Oldest Old Study allowed comparisons of cognition in only 2 years, 1993 and 1998. The Medicare Current Beneficiary Survey was limited to a 5-year span of data (1992-1996) at the time it was analyzed.⁴⁸ The National Mortality Followback Studies changed its survey mode from in-person to telephone; moreover, although the cognition-related questions were stable, the disability questions were changed extensively and thus the survey is considered fair for assessing cognition but poor for assessing trends in ADL limitations prior to death. The Supplements on Aging to the 1984 and 1994 National Health Interview Surveys (SOA I and II) (analyzed in 4 studies^{4 ,40 ,44 ,46} ) used different timing in their field procedures (ie, the SOA I was administered at the same time as the core National Health Interview Survey, and the SOA II was administered 7-17 months after the core). Finally, more than one third of reports in the Survey of Income and Program Participation (SIPP), analyzed in 2 studies,^{3 ,43} were provided by proxy respondents.

Only the Framingham Heart Study³⁹ was consistently assigned a poor rating. The study had 2 flaws for assessing national trends: it drew on a select sample and changed how it administered the survey (from in-person to over the telephone) and made extensive question changes.

Table 3a provides highlights of major findings from each survey rated as good or fair along with our calculations, where data allowed, of the average annual percent change (calculated as a percentage of the base year). In Table 4, we summarize the findings across all surveys by survey rating for each major outcome category. For outcomes with estimates from at least 3 surveys rated either good or fair, we also summarize the high and low estimates of the average annual percent change.

Among the 3 surveys providing trend estimates for the prevalence of "any disability" (ie, defined as having ADL or IADL disability, or in some cases ADL or IADL or being institutionalized),^{1 ,14 ,41 ,47 - 48} all 3 showed statistically significant declines, with a high of −1.55% per year and a low of −0.92% per year.

Of the 6 surveys providing trend estimates for ADL disability, 4 were rated as good or fair but offered conflicting evidence resulting in a wide range of estimates of the average annual percent change, ranging from −1.38% per year to 1.53% per year.

Four surveys provided trend estimates for IADL disability and all were rated as good or fair. Although 3 of the 4 surveys assessed trends in only IADL disability (ie, IADL but not ADL disability)^{1 ,14 ,41 ,47 - 48} and the fourth assessed trends in any IADL disability (irrespective of ADL disability),^{40 ,46} all 4 surveys showed significant declines in the prevalence of IADL disability ranging from −2.74% per year to −0.40% per year.

Of the 4 surveys analyzed for functional limitation trends, 3 were rated as fair and 1 as poor. Of those surveys rated as fair, 2 showed declines.^{3 ,40 ,43 - 44 ,46} The only increase in functional limitations relied on the Medicare Current Beneficiary Survey⁴⁸ ; however, that analysis focused on a group with only functional limitations (but no ADL or IADL disability) and is thus difficult to interpret.

The 2 surveys, both receiving a rating of fair, that studied trends in cognitive limitations^{2 ,42 ,45} showed significant declines in severe cognitive impairment, 1 among noninstitutionalized persons aged 70 years and older and the other in the last year of life.

Finally, the 2 surveys^{3 - 4 ,40 ,43} that assessed sensory limitations had mixed results. Analysis of the Survey of Income and Program Participation^{3 ,43} showed large declines from 1984 to 1993 in the percentage of US adults aged 50 years and older and aged 65 years and older with difficulty seeing. Evidence from the Supplements on Aging to the National Health Interview Survey showed that rates of being blind or deaf or having hearing impairment remained constant between 1984 and 1995.^{4 ,40}

It is a common finding that the oldest old, women, blacks, and those with the least education have the greatest disability and limitation,⁶ but few studies explicitly have focused on trends in disparities for major demographic and socioeconomic groups. Of the 8 surveys reviewed herein, 6 were analyzed for trends stratified by age, race, sex, or educational attainment or some combination thereof (Table 5a) and all but 1 survey⁴⁵ were rated as fair or good. However, only 3 of these analyses included statistical tests for disparities.^{43 ,45 ,47} In Table 6, we categorized for each survey through inspection trends in disparities for each major outcome as narrowing, widening, or not changing and indicated in footnotes where statistical tests were conducted. In one case¹ that rendered no clear pattern to the stratified trends by age, race, and education, this information was omitted from the summary table.

No narrowing or widening of differences is apparent in outcomes across age groups. Of the surveys rated as fair or good whose data were used in studies that included statistical tests, one showed no significant change for age disparities in any disability⁴⁷ and the other no significant change for age disparities in cognitive dysfunction prior to death.⁴⁵

The results for trends in sex disparities give the impression of narrowing, if any change at all. Narrowing sex differences in the mean number of IADLs and the percentage with 3 or more IADL disabilities,^{40 ,46} any functional limitations,^{3 ,40} and vision impairments³ were apparent (although not tested for) in some cases but were not evident in others focusing on the presence of any disability,⁴⁷ ADL disability,^{40 - 41 ,46} any IADL disability,⁴¹ or severe cognitive impairment.² The only study to test for sex disparities⁴⁷ did not find statistical evidence of a difference for men and women in trends in the prevalence of any disability.

Three of the surveys have been used to assess trends in disparities by race. Data from the Asset and Health Dynamics of the Oldest Old Study² showed larger declines in severe cognitive impairment between 1993 and 1998 for nonwhites than whites—suggesting the disparity between the races may be narrowing. The 1984 and 1993 Survey of Income and Program Participation³ showed larger declines among blacks than among whites or persons of other races with respect to 3 functional limitations and vision limitations. However, in the only study that included tests for trends in disparities by race, Schoeni and colleagues⁴⁷ reported no statistically significant differences in disability declines between nonwhites and whites between 1982 and 1996, using data from the National Health Interview Survey.

Reports of trends in disparities by educational level have been inconsistent. Declines in any disability were significantly larger for those with more than a high school education compared with those with just a high school education or less.⁴⁷ Declines in severe cognitive impairment appeared to be largest among those with less than a high school education,² but there was no significant change in educational disparities in functional limitations and vision limitations over time.⁴³

We found that for older US adults the prevalence of any disability declined significantly during the 1990s and that estimates of the average annual decline ranged from −1.55% to −0.92% per year. However, these improvements did not hold across all specific measures of disability. To the contrary, late-life disability declines have been concentrated among IADL limitations, such as household chores, shopping, going outside, and medication management, and among limitations in basic physical tasks, such as lifting, climbing stairs, and walking. Estimates of the average annual rate of decline of the former ranged from −2.74 to −0.40. The prevalence of severe cognitive impairment also may have declined, but this preliminary finding needs to be verified with additional data sources and a longer study period. Limitations in hearing appear to have been constant over the last decade. The evidence is mixed for self-reported vision.

Perhaps most importantly, conflicting evidence exists about ADL disability, the most severe type of disability generally associated with long-term care needs. Of the surveys of fair or better quality for evaluating trends, only the National Long Term Care Survey found declines; the remaining surveys showed increases or no change. It remains unclear why the national surveys evaluated herein provided inconsistent evidence with respect to old age ADL trends. Potential methodological explanations include differences across surveys in how questions are worded and in defining the specific ADL activities, whether the institutional population is included in the sampling frame, and whether the design is cross-sectional or panel based (the latter of which is subject to loss to follow-up). Analytic decisions about missing data, nonresponse weights, and the age-standardization of results may also contribute. Resolving these inconsistencies is an important next step for understanding whether declines in severe disability have been occurring.

Considerable gaps in our understanding of trends in disparities across major demographic groups remain. Although none of the results we reviewed suggested that the gaps were widening between old and young, men and women, or whites and nonwhites, whether the gaps have narrowed or have remained stable for these groups over the last decade remains unclear. For educational disparities, the only study with statistical tests for disparities and based on survey data with a rating of good found the disability gap was widening. Notably, only 3 of the studies that analyzed disparities included statistical tests of disparities. Future work would do well to focus on rigorously examining trends in important disparities. A thorough understanding of trends in disparities is critical not only for identifying groups that might benefit from various health-related interventions but also for projecting the future course of population-level health trends.

Not all studies of late-life disability and functioning trends are equally valid. Only 2 out of the 8 surveys reviewed herein received a rating of good. The majority had at least 1 poor feature, which varied from survey to survey. Many studies failed to report critical pieces of information relevant for assessing the validity of findings or did not attempt to investigate sensitivity to missing data. And, with 2 exceptions,^{2 ,42} the studies reviewed herein relied on ratios of disability and functioning. Supplementation of self- and proxy-reported items with performance-based measures, which may evaluate capacity independent of changes in attitudes, environments, and adaptations, could help minimize issues related to item consistency and interpretation in future studies of disability trends.⁵⁰

Pinpointing explanations—including the role of medical care—for the population-level shifts in late-life health continues to be a high priority for future research. Thus far the search for potential explanations for improvements has been extremely limited in scope,¹⁶ with investigations focusing largely on shifts in the demographic and socioeconomic status of older adults^{2 - 3 ,43 ,47 - 48} and on changes in the late-life chronic disease profile and treatment of disease.⁴⁴ To date, the empirical evidence has not provided overwhelming support for any causal hypotheses, although the increase in educational attainment of the older population has been consistently identified as correlated with these shifts^{2 - 3 ,43 ,47 - 48} and improvements in functioning have occurred despite increased reports of chronic conditions.⁴⁴ Our synthesis suggests that the search should be broadened to examine sets of factors related to the performance of IADLs and the physical and cognitive abilities that underlie them. For example, future research could focus on the role of the physical environment (meaning assistive technology and modifications to the home environment that may make these tasks easier to carry out); efforts aimed at preventing, reversing, and generally slowing the progression of physically and cognitively disabling conditions (including innovations in the growing field of rehabilitative medicine); and events that occur earlier in life that may be linked to both education and late-life functioning.

Finally, despite studies suggesting continued declines in disability will offset the effects of population aging on the size of the older disabled population,^{18 ,48} the implications of our findings for the future demand for medical care suggest that caution is in order. Without better insight into the causes of these improvements, it remains unclear whether medical expenditures have fueled health improvements or whether health improvements will help save medical costs in the future. The lack of consensus on trends in severe personal care disability—clearly the most expensive form of disability—further contributes to the uncertainty as to whether the improvements in old age health witnessed to date will yield cost savings to public programs paying for home- and community-based and institutional care. Certainly the relatively wide range of estimates across studies and dearth of consensus on trends in disparity suggests that predictions of the future size and composition of the older disabled population based on findings from a single study may be misleading. Predictions aside, the framework presented herein could well serve as a guide for assessing and maximizing the validity of future studies of old-age disability and functioning trends.