Objectives. We elucidated how US late-life disability prevalence has changed over the past 3 decades.

Methods. We examined activities of daily living (ADL) and instrumental activities of daily living (IADL) disability trends by using age–period–cohort (APC) models among older adults aged 70 years or older who responded to the National Health Interview Survey between 1982 and 2009. We fitted logistic regressions for ADL and IADL disabilities and for each of the 3 APC trends with 2 models: unadjusted and fully adjusted for age, period, cohort, and sociodemographic variables.

Results. The unadjusted and adjusted period trends showed a substantial decline in IADL disability, and ADL disability remained stable across time. Unadjusted cohort trends for both outcomes also showed continual declines across successive cohorts; however, increasing cohort trends were evident in the adjusted models.

Conclusions. More recent cohorts of US older adults are becoming more disabled, net of aging and period effects. The net upward cohort trends in ADL and IADL disabilities remain unexplained. Further studies should explore cohort-specific determinants contributing to the increase of cohort-based disability among US older adults.

Knowledge of trends in late-life disability is crucial to health professionals because a high prevalence of disability leads to elevated mortality and health risks for individuals and an increased burden of health care costs for society. Studies of late-life disability trends have generally focused on 2 measures of activity limitations: (1) activities of daily living (ADL), such as bathing and toileting, and (2) instrumental activities of daily living (IADL), such as shopping and cooking. Previous studies found declining disability among older Americans during the last 2 decades of the 20th century.1,2 Disability trend experts investigated differences in disability trends across 5 data sets (Heath and Retirement Study, Medicare Current Beneficiary Survey, National Health Interview Survey [NHIS], National Long Term Care Survey, and Supplement on Aging) and concluded that ADL disability continually declined in the 1990s at a rate of 1.0% to 2.5% per year.3

Schoeni et al.1 analyzed ADL and IADL disability trends among Americans aged 70 years and older using data from the NHIS and reported a declining trend for any disability (ADL or IADL limitations) between 1982 and 2002. Although the prevalence of IADL disability declined from 14.5% in 1982 to 8.1% in 2002, the prevalence of ADL disability remained steady over this period. Schoeni et al.1 concluded that the substantial reduction in US older adult disability was primarily driven by the decline in IADL limitations over this period. This evidence corroborates the conclusion of an earlier review suggesting that most of the decline in late-life disability in the 1980s and 1990s was the result of the reduction of IADL-only disability prevalence, which decreased at a rate of 0.4% to 2.7% per year.4

More recent analyses revealed a contrary pattern indicating that late-life disability has been increasing among newer birth cohorts who are now moving into older adulthood.5,6 Seeman et al.5 used data from 2 cross-sectional waves (1988–1994 and 1999–2004) of the National Health and Nutrition Examination Survey and found that ADL disability, IADL disability, and impaired mobility increased significantly among respondents aged 60 to 69 years in the more recent survey period. For respondents aged 70 to 79 years, a significant increase in IADL disability was found, whereas respondents aged 80 years and older exhibited a modest decrease in functional limitations.5 Thus, the authors warned of potentially increasing older-age disability among younger US cohorts who are now moving into older adulthood. Contrarily, European researchers combined 2 longitudinal datasets covering 10 European countries and noted that the rates of need for help with self-care and mobility ADL decreased across birth cohorts who were interviewed in 1988 to 1991, 1993 to 1995, and 1998 to 2000.7

Despite the sobering finding of an increase in disability among older Americans, the disability prevalence comparison by Seeman et al.5 was limited to merely 2 aggregated survey periods and lacked the specificity of disability trends in more refined periods. Because the results described earlier showed potentially contradictory findings regarding late-life disability trends, we sought to better elucidate how disability prevalence has changed across time among US adults aged 70 years and older. To do so, we used age–period–cohort (APC) models to analyze late-life disability trends between 1982 and 2009.

Demographically speaking, time can be captured by 3 temporal dimensions: age, period, and cohort. Specifically, age is a proxy for biological processes that ultimately lead to disease, disability, or death. Period, or survey year, reflects changes in sociocultural, economic, technological, and environmental factors that may affect the entire population at a given time simultaneously but perhaps not equally. For example, a drought may lead to increased food prices, which largely affect those with lower incomes. Finally, cohort describes a unique set of individuals who both are born into a social system during a similar time period and experience similar social experiences over their life course.8 Each aspect of APC makes a unique contribution to the study of population health, including disability. Aging has an obvious relationship to population health, and period captures the current burden of morbidity, disability, and mortality in the entire population at a given time. Cohort, however, reflects the health of successive generations and is an important dimension for understanding how population health is changing over time. Overall, failure to isolate APC trends risks substantial bias and provides an incomplete picture of population health trends.8–10

Schoeni et al.1 presented a thorough description of cohort-unadjusted and age-adjusted period trends for ADL and IADL disability, which results in a clear snapshot of both the current and the past burden of disability among US older adults. However, their approach did not identify cohort trends. We sought to expand their study by extending their analysis of disability trends through 2009, estimating both period-based and cohort-based trends in disability, and also controlling for age effects. Doing so would allow us to confirm whether the recent increase in disability among more recent birth cohorts of US older adults, as reported by Seeman et al.,5 is found in our larger and more contemporary data set. Thus, the purposes of our study were to (1) replicate the ADL and IADL disability prevalence produced by Schoeni et al.1 and (2) estimate both unadjusted and adjusted APC trends for ADL and IADL disability among US older adults from 1982 through 2009.

Our analyses were based on responses from older adults who participated in the NHIS between 1982 and 2009. The NHIS is a repeated cross-sectional survey conducted annually to investigate the health of the civilian noninstitutionalized US population. Each survey follows a multistage area probability design allowing for representative sampling of households. The Integrated Health Interview Series (IHIS) compiled by the University of Minnesota contains a normalized set of NHIS variables that have consistent coding across each survey year to facilitate temporal analysis. We retrieved all data from the IHIS Web site11 except for the disability outcome variables (ADL and IADL disability) between 1982 and 1996, which we retrieved directly from the NHIS Web site12 and merged with our IHIS data set. Our final sample consisted of 87 612 men and 131 343 women aged 70 years and older.

Dependent Variables

The 2 outcomes (ADL and IADL disabilities) that we examined were similar to the outcomes analyzed by Schoeni et al.1; however, the inclusion criteria that we used for estimating IADL prevalence were slightly different. Before 1997, NHIS respondents aged 70 years and older (71 years and older for 1982) were asked 2 questions regarding their disability status: (1) “Because of any impairment or health problem, do you/does ______ need the help of other persons with personal care (ADL) needs, such as eating, bathing, dressing, or getting around this home?” and (2) “Because of any impairment or health problem, do you/does ______ need the help of other persons in handling routine (IADL) needs, such as everyday household chores, doing necessary business, shopping, or getting around for other purposes?” Respondents who answered “no” to question 1 were subsequently asked question 2, and those who answered “yes” to question 1 skipped question 2.

According to personnel from the Division of Health Interview Statistics (written communication, December 2010), the reason for this skip pattern was that the NHIS (1982–1996) was initially interested in the most severe limitations, such as personal care needs, and assumed that if individuals could not help themselves with personal care needs, they could not accomplish routine tasks alone. Thus, in our IADL prevalence estimates for 1982 to 1996, we assumed that respondents who needed help with ADL (those who answered “yes” to any ADL question) also needed help with IADL, and we included these individuals with those who reported only needing help with IADL. Schoeni et al.,1 however, included IADL limitations only when estimating their IADL disability prevalence.

From 1997 onward, the set of disability questions was modified slightly. The leading sentence of the 2 questions was changed to the following: “Because of a physical, mental, or emotional problem, do/does (you/anyone in the family)… .” Respondents answered the IADL question regardless of their response to the ADL question. Thus, our IADL prevalence estimate for 1997–2009 again included those who reported needing help with ADL and IADL as well as those only needing help with IADL. To estimate ADL disability prevalence for 1982 to 2009, we included those only needing help with ADL and those needing help with both ADL and IADL. We should note that reported difficulty in performing ADL or IADL is a subjective experience of coping, whereas needing help with ADL or IADL is an indicator of whether respondents require external assistance in ADL and IADL.7 Thus, trends of ADL and IADL disabilities using different operational definitions may yield incongruent results.

Independent Variables

The main variables of our APC analysis included age, period, and cohort. Age was treated as a continuous variable. We explored various functional forms for age but determined that a linear age term best represented the relationship between age and disability. Moreover, effect estimates were robust to several different age specifications, and these models are available on request. Although we included respondents aged 70 years and older in our model, the top-coded age was different for various NHIS years. The top-coded ages for 1982 to 1995, 1996, and 1997 to 2009 were 99, 90, and 85 years, respectively. To make our sample commensurable to that of Schoeni et al.,1 we allowed the top-coded age to vary by survey year; however, we restricted our modeling of the unadjusted and adjusted age trends to age 84 years and younger because we did not feel confident in estimating the predicted probability of ADL and IADL disability for ages beyond the top-coded categories. Period indicates the year (1982–2009) in which the respondent was interviewed, and we estimated these effects using single-year dummy variables. We estimated cohort by subtracting age from period. Cohorts were subsequently aggregated into 5-year bands to break the linear dependence among the APC dimensions, known in demographic research as the identification problem.13 For example, individuals born between 1883 and 1887 were grouped into the midpoint of this range, which is cohort 1885.

Because we estimated both unadjusted and adjusted models of disability, we included several control variables in a portion of our analyses. Sociodemographic control variables included region of residence (Northeast, north central or Midwest, South, and West), race (White, Black, and other race), marital status (never married, married, and other, which combines widowed, divorced, and separated), education (less than high school, some high school, high school graduate, some college, and college or more), employment status (working, unemployed, and retired), and combined family income, which was adjusted for household size and the Consumer Price Index. Finally, body mass index (BMI; defined as weight in kilograms divided by the square of height in meters) was calculated by IHIS using self-reported height and weight. Moreover, BMI information was lacking for 22% of the respondents; thus, BMI values for these people were imputed with mean. We subsequently classified respondents' BMI into 4 categories (underweight, normal weight, overweight, and obese) on the basis of the Centers for Disease Control and Prevention's BMI classification system.14 BMI was not adjusted in all models because we primarily used it to determine whether it accounted for the cohort-based trends we uncovered.

Statistical Analyses

Recent APC literature has proposed using random effects models, given the gain in efficiency obtained;8 however, we used a fixed effects approach, which yielded virtually identical results but allowed us to more easily adjust for the complex survey design and yielded effect estimates that were fully adjusted for APC. First, we replicated the unadjusted period trend of ADL and IADL disability estimated by Schoeni et al.1 to ensure the commensurability of our samples and disability definitions. We also first restricted our sample to the 1982–2002 survey years to compare the period-based trends with those of Schoeni et al., and the 2 figures were identical (Figure 1a and Figure 1b).

Second, we fitted logistic regressions for each of the 2 outcomes (ADL and IADL disabilities) and for each of the 3 unadjusted APC trends (6 regressions total) using our IADL inclusion criteria, described earlier, with additional years of data from 2003 to 2009. Predicted probabilities for each age, period, and cohort band were calculated and plotted for both outcomes. We accounted for the NHIS's complex survey design by adjusting for the appropriate design variables (primary sampling unit and stratum) and sampling weights created in the normalized IHIS data. All logistic regressions were performed in Stata/SE version 12.0 (StataCorp LP, College Station, TX). We hereinafter refer to the age-, period-, and cohort-only models as the first models, and they were unadjusted for individual-level sociodemographic covariates.

Third, we again regressed ADL and IADL disability on age, period, and cohort but added sociodemographic control variables. For both disability outcomes, we calculated and plotted the predicted probability for each age, period, and cohort by holding all other independent variables at their mean values. We hereinafter refer to these as the second models, and they were fully adjusted for all variables included in the model. In an unreported analysis, we analyzed men and women separately and found very similar APC trends; thus, we show only gender-aggregated trends in this article.

Figure 2a shows the unadjusted period trend of predicted probabilities for ADL and IADL disabilities. Confidence intervals for the trends are shown as well. The unadjusted predicted probability of ADL disability was relatively flat between 1982 and 2009. The unadjusted period trend for IADL disability, however, showed a substantial decline between 1982 and 1998, but a leveling off between 1999 and 2009. With adjustments for age, cohort, and sociodemographic variables, the period trends demonstrate very similar patterns to the unadjusted trends (Figure 2b).

Figure 3a shows that the unadjusted predicted probabilities of ADL and IADL disabilities increased substantially with age. This aging pattern was more pronounced for IADL disability than for ADL disability. The age effects remained strong after adjusting for period and cohort effects and sociodemographic variables (Figure 3b).

Figure 4a illustrates the unadjusted cohort trends for both disability outcomes. ADL and IADL disabilities decreased dramatically across cohorts; however, this rate of decline decreased in more recent cohorts. Also, the decline in IADL disability was more prominent than that in ADL disability. For example, the decrease in the predicted probability of IADL disability was approximately 93% between the 1885 cohort (0.75) and the 1940 cohort (0.05). Again, these cohort trends were unadjusted for age and period and did not include controls for sociodemographic variables. By contrast, the adjusted cohort trends were reversed from the unadjusted trends (Figure 4b). Both ADL and IADL disabilities increased among more recent cohorts, although the increase in ADL disability was much more subtle. IADL disability increased steadily between the 1890 and 1915 cohorts, followed by a brief plateau from 1915 to 1930, a slight increase from 1930 on, and then a decline in the 1940 cohort. Regardless of some cohort variation, the general trend was of increasing IADL disability across birth cohorts, net of age, period, and sociodemographic controls. The ADL disability trend showed a similar pattern, but we observed a less pronounced increase between the 1885 and 1935 cohorts, which was followed by a slightly smaller decline between the 1935 and 1940 cohorts. Overall, the ADL trend increased moderately over successive birth cohorts, net of age, period, and sociodemographic controls. Although confidence intervals for most cohort-to-cohort comparisons overlapped for both the ADL- and the IADL-adjusted cohort trends, the 1915 and 1920 cohorts appeared to have the highest levels of disability and were statistically different from the other cohorts. Moreover, although we found few significant differences when comparing adjacent cohorts, we observed a general upward trend for IADL disabilities. The upward trend was also clear but less pronounced for ADL disability.

Moreover, we found that the reversals in cohort trends from Figure 4a to Figure 4b were largely the result of controls for age and period but were not the result of adjustment for sociodemographic variables. We tested additional models controlling only for sociodemographic variables and found that the adjusted APC trends for both disability outcomes varied only slightly from the unadjusted trends. However, the reversal of the cohort trend for ADLs became apparent when age was controlled, but the reversal observed in the IADL cohort trend was the result of adjustments for both age and period (models not shown). Given this pattern of net cohort increases in disability, we tested whether the US obesity epidemic was responsible for increasing disability among more recent cohorts. When we added BMI into the adjusted cohort models (Figure 4b), the cohort trends for ADL and IADL disabilities remained unchanged because the solid (BMI-unadjusted) and dashed (BMI-adjusted) lines substantially overlapped.

The unadjusted disability period trends observed in our study confirm the findings of previous studies1,2,15,16 that older-age IADL disability has declined substantially over time in the United States, whereas ADL disability has remained quite stable over the past 3 decades (1982–2009). Late-life disability trends in Taiwan between 1989 and 2007 have shown similar results, with IADL disability declining rapidly over the period and the ADL disability trend remaining flat.17 When compared with Schoeni et al.,1 we found very minor fluctuations in unadjusted and adjusted period trends for both disability outcomes with the additional 7 survey years we added. Although the period effects may imply reduced burdens in IADL disability among noninstitutionalized older adults, our results also indicate a net increase in ADL and IADL disability in the overall cohort trends, after adjusting for period and age effects. Contrary to the period results we have described, the adjusted cohort trends suggest that older adults in more recent cohorts are experiencing more limitations in ADLs and IADLs than earlier cohorts, which best coincides with the recent findings of Seeman et al.5

To summarize, after accounting for the effects of aging and for period effects that help tap changes in sociocultural, economic, technological, and environmental factors between 1982 and 2009, successive cohorts of older adults are becoming more disabled over time. These period and cohort results, although seemingly contradictory, are both potentially important and not necessarily at odds. Indeed, the unadjusted cohort effects (Figure 4a) are perfectly consistent with the overall period effects (Figures 2a and 2b); that is, the burden of disability declined over time, with a flattening out over the past 7–8 years or so. However, once we adjusted for this temporal decline and the effects of aging, our results showed that birth cohorts are becoming increasingly more disabled (Figure 4b). Although such a pattern may not necessarily lead to any short-term changes in the overall US disability burden, the flattening out of period-based reductions and increases among more recent cohorts could result in an eventual overall uptick in the US disability burden should the cohort trends persist. We further explored the sources of net cohort-based upturns found in both disability outcomes. We tested whether the US obesity epidemic has contributed to the increasing disability cohort trends. The results indicate that BMI did not account for the cohort-based increases in ADL and IADL disabilities among older adults in our restricted age range of 70 years and older. This result is consistent with the finding that obesity is not associated with IADL or ADL disability.18,19

Several caveats to our examination of trends in US older adult disability should be considered. Older adults with disabilities in more recent cohorts may have survived longer because of advanced medical technology that prolongs the longevity of frail older adults,6 which could increase the proportion of people with disability in more recent cohorts. Older adults in newer cohorts might also be more likely to report disability to acquire disability benefits.20,21 Also, rates of institutionalization (which would exclude individuals from the NHIS sampling frame) differ across cohorts; more recent cohorts of older adults, for example, have lower rates of nursing home use.22 Moreover, the earliest and latest cohort bands in our analysis did not capture a full age distribution (e.g., the 1940 cohort only covered respondents at younger ages: 70 and 71 years), thus biasing our estimates for cohort trends. We were not able to specify an interaction (age × cohort) in our models because the collinearity of these predictors is very high. Finally, the mix of proxy reporting and self-reporting before 1996 in the NHIS interview is another limitation of our study; previous research has suggested that the 2 reporting systems may differ.23

Despite these limitations, this study provides an updated examination of disability trends among US adults aged 70 years and older. Although we used a different data set and a different methodological approach, our results validate the existence of increasing disability across successive cohorts as recently reported by Seeman et al.5 However, that study only compared disability rates among older adults (aged 60–84 years) in 2 aggregated National Health and Nutrition Examination Survey periods (1988–1994 vs 1999–2004). We investigated disability trends for a much broader spectrum of cohorts and confirmed the increasing trend of disability among recent birth cohorts moving into older adulthood. Although speculation about causes of rising ADL and IADL disability among recent cohorts has occurred, further studies are needed to investigate other social determinants that are unique to each cohort, which may result in an elevated level of cohort-based disability.


This article was supported by the National Institute on Minority Health and Health Disparities (grant R01MD004025).

We express our gratitude to Aimee Bower for her programming assistance and to the anonymous reviewers and journal editors for helpful comments.

Human Participant Protection

This research was approved by the San Diego State University institutional review board.


1. Schoeni RF, Martin LG, Andreski PM, Freedman VA. Persistent and growing socioeconomic disparities in disability among the elderly: 1982-2002. Am J Public Health. 2005;95(11):20652070. LinkGoogle Scholar
2. Field MJ, Jette AM, eds.; Committee on Disability in America. The Future of Disability in America. Washington, DC: National Academies Press; 2007. Google Scholar
3. Freedman VA, Crimmins E, Schoeni RF, et al. Resolving inconsistencies in trends in old-age disability: report from a technical working group. Demography. 2004;41(3):417441. Crossref, MedlineGoogle Scholar
4. Freedman VA, Martin LG, Schoeni RF. Recent trends in disability and functioning among older adults in the United States: a systematic review. JAMA. 2002;288(24):31373146. Crossref, MedlineGoogle Scholar
5. Seeman TE, Merkin SS, Crimmins EM, Karlamangla AS. Disability trends among older Americans: National Health and Nutrition Examination Surveys, 1988-1994 and 1999-2004. Am J Public Health. 2010;100(1):100107. LinkGoogle Scholar
6. Lakdawalla DN, Bhattacharya J, Goldman DP. Are the young becoming more disabled?Health Aff (Millwood). 2004;23(1):168176. Crossref, MedlineGoogle Scholar
7. Aijänseppä S, Notkola IL, Tijhuis M, van Staveren W, Kromhout D, Nissinen A. Physical functioning in elderly Europeans: 10 year changes in the north and south: the HALE project. J Epidemiol Community Health. 2005;59(5):413419. Crossref, MedlineGoogle Scholar
8. Yang Y, Land KC. Age-period-cohort analysis of repeated cross-section surveys: fixed or random effects?Sociol Methods Res. 2008;36(3):297326. CrossrefGoogle Scholar
9. Ryder NB. The cohort as a concept in the study of social change. In: Mason WM, Fienberg SE, eds. Cohort Analysis in Social Research: Beyond the Identification Problem. New York: Springer-Verlag; 1985:944. CrossrefGoogle Scholar
10. Mason KO, Mason WM, Winsborough HH, Poole WK. Some methodological issues in cohort analysis of archival data. Am Sociol Rev. 1973;38(2):242258. CrossrefGoogle Scholar
11. Minnesota Population Center and State Health Access Data Assistance Center. Integrated Health Interview Series: Version 3.0. Integrated Health Interview Series Web site. Available at: http://www.ihis.us. Accessed April 5, 2010. Google Scholar
12. National Health Interview Survey. Questionnaires, datasets, and related documentation1996 and prior. Centers for Disease Control and Prevention Web site. Available at: http://www.cdc.gov/nchs/nhis/quest_data_related_1996_prior.htm. Accessed May 18, 2010. Google Scholar
13. Glenn ND. Cohort Analysis. 2nd ed. Thousand Oaks, CA: Sage; 2005. CrossrefGoogle Scholar
14. Centers for Disease Control and Prevention. About BMI for adults. Centers for Disease Control and Prevention Web site. Available at: http://www.cdc.gov/healthyweight/assessing/bmi/adult_bmi/index.html#Interpreted. Accessed April 15, 2011. Google Scholar
15. Cutler DM. Declining disability among the elderly. Health Aff (Millwood). 2001;20(6):1127. CrossrefGoogle Scholar
16. Manton KG, Corder LS, Stallard E. Estimates of change in chronic disability and institutional incidence and prevalence rates in the U.S. elderly population from the 1982, 1984, and 1989 National Long Term Care Survey. J Gerontol. 1993;48(4):S153S166. Crossref, MedlineGoogle Scholar
17. Martin LG, Zimmer Z, Hurng BS. Trends in late-life disability in Taiwan, 1989-2007: the roles of education, environment, and technology. Popul Stud (Camb). 2011;65(3):289304. Crossref, MedlineGoogle Scholar
18. Martin LG, Schoeni RF, Andreski PM. Trends in health of older adults in the United States: past, present, future. Demography. 2010;47(suppl):S17S40. Crossref, MedlineGoogle Scholar
19. Alley DE, Chang VW. The changing relationship of obesity and disability, 1988-2004. JAMA. 2007;298(17):20202027. Crossref, MedlineGoogle Scholar
20. Autor DH, Duggan MG. The rise in the disability rolls and the decline in unemployment. Q J Econ. 2003;118(1):157205. CrossrefGoogle Scholar
21. Bound J, Waidmann T. Accounting for recent declines in employment rates among working-aged men and women with disabilities. J Hum Resour. 2002;37(2):231250. CrossrefGoogle Scholar
22. National Center for Health Statistics. Health, United States, 2006: With chartbook on trends in the health of America. DHHS Publication No. 2006-1232. Hyattsville, MD: National Center for Health Statistics; 2006. Available at: http://www.cdc.gov/nchs/data/hus/hus06.pdf. Accessed July 20, 2011. Google Scholar
23. Magaziner J, Zimmerman SI, Gruber-Baldini AL, Hebel JR, Fox KM. Proxy reporting in five areas of functional status. Comparison with self-reports and observations of performance. Am J Epidemiol. 1997;146(5):418428. Crossref, MedlineGoogle Scholar


No related items




Shih-Fan Lin, DrPH, Audrey N. Beck, PhD, Brian K. Finch, PhD, Robert A. Hummer, PhD, and Ryan K. Master, PhDShih-Fan Lin, Audrey N. Beck, and Brian K. Finch are with the Center for Health Equity Research and Policy, San Diego State University, San Diego, CA. Robert A. Hummer is with the Population Research Center, University of Texas at Austin. Ryan K. Master is with the Columbia Population Research Center, Columbia University, New York, NY. “Trends in US Older Adult Disability: Exploring Age, Period, and Cohort Effects”, American Journal of Public Health 102, no. 11 (November 1, 2012): pp. 2157-2163.


PMID: 22994192