Change in chronic disability from 1982 to 2004/2005 as measured by long-term changes in function and health in the U.S. elderly population

Abstract

Changes in the health and functioning of the Medicare-enrolled population aged 65+ are tracked by using the 1982–2004/2005 National Long-Term Care Surveys. We found a significant rate of decline in the prevalence of chronic disability that accelerated from 1982 to 2004. These declines are significant for both persons with less severe chronic disability, which might be compensated by modifying the built environment and providing assistive devices, and for persons with more serious disability, which may be affected by reductions in the incidence and severity of disease through biomedical interventions. Declines in chronic disability continued over the 22-year period at a rate fast enough (i.e., 1.52% per annum) to contribute significantly to the long-term fiscal stability of the Medicare (and Medicaid) programs. Changes in the rate and substance of disability declines seem consistent with the intentions of policy interventions in Medicare and Medicaid.

Declines in the prevalence of chronic disability (i.e., an impairment in a specific function existing for 90 or more days), defined in terms of the difficulty in performing both activities of daily living (ADLs, which include basic personal care activities such as eating, grooming, and bathing; ref. 1) and instrumental activities of daily living (IADLs, which include the capability of performing more complex activities such as meal preparation, shopping, and managing money; ref. 2) in the U.S. elderly population can have major effects on the long-run (75-year) fiscal stability of the Medicare, Medicaid, Veterans Administration, and Social Security programs and in the rate of increase of total U.S. health care expenditures. Singer and Manton (3) showed that if chronic disability continued to decline at 1.5% per annum, there could be significant improvements in the long-run fiscal stability of Social Security and Medicare. Moderation of the rate of growth of Medicare and other public and private health expenditures, combined with increases in disposable income and a propensity to increase consumption of health care as disposable income grows (4), may define new economic conditions in the U.S. where health improvements and increasingly productive health care investments (e.g., refs. 5–7) could become a major stimulus for economic growth and for research and technical innovation (K.G.M., G. Lowrimore, D. Tolley, and A. Ullian, unpublished work).

Using data from the National Long-Term Care Survey (NLTCS), Manton and Gu (8) found a per annum decline in chronic disability prevalence of 1.7% from 1982 to 1999. Combining NLTCS data with results of early to mid-20th-century disability trends by Fogel (4) and Costa (9), Manton et al. (10) calculated active and total-life expectancy estimates from 1935 to 1999, and using projected disability rate changes, they made active-life expectancy forecasts to 2080. The forecasts showed significant long-term improvements in the ratio of active to total life expectancy at age 65, with larger rates of improvement in the ratio at ages 85+. The rapid decline in chronic disability at ages 85+ is significant in that it is at such advanced ages where future long-term improvements in health, functioning, and survival must occur to maintain the current rate of improvement in the fiscal status of Medicare and Medicaid to 2050, the year by which the last of the post-World War II baby boom cohorts has reached at least age 85.

Linkage of Medicare cost and service use records with NLTCS survey data allowed the examination of the relation of 1982–1999 changes in Medicare costs to chronic disability declines (11). The analysis of changes from 1982 to 1999 showed that Medicare per capita costs, adjusted for inflation, declined significantly in the growing nondisabled U.S. elderly population aged 65–84, whereas Medicare per capita costs increased in the severely disabled and institutional components of the U.S. elderly population, especially at ages 85+.

Those analyses showed that, over time, a larger proportion of Medicare expenditures were being targeted to the most severely disabled individuals in the U.S. elderly population. The severely disabled population was, over the same time period, declining in prevalence. This improved resource targeting may have been affected in part by the 1997 Balanced Budget Act (BBA; ref. 12), which mandated use of prospective payment systems (PPS) for postacute care services. PPS reduced the proportion of hospital discharges to skilled nursing facilities (SNFs), although it had little effect on SNF length of stay. The payment system had financial incentives to stimulate the growth of rehabilitation and long-term care (LTC) hospitals for the most severely impaired (13). Its provisions were adjusted by the Balanced Budget Refinement Act (BBRA) of 1999.

The combined effects of declines in Medicare per capita costs in the nondisabled elderly population aged 65–84 and the increase in the size of the nondisabled population reduced Medicare expenditures in 1999 by $26 billion; i.e., Medicare expenditures were $26 billion less than they would have been had improvements in functional status and declines in Medicare expenditures related to them not occurred (11). Projection of these declines to 2009 suggests that a savings of $73 billion, or 17% of all Medicare expenditures expected for the elderly, could result if declines in chronic disability prevalence continued at their 1982–1999 rate (21).

Federal actuarial projections suggest that combined Medicare and Medicaid costs could grow from 6% to 7% of gross domestic product (GDP) in 2005 to 24% of GDP by 2080, outstripping the impact of Social Security on the overall U.S. economy because Social Security costs were projected to remain fairly consistent at 6–7% of GDP to 2080. Using an economic growth model for federal research investment developed initially by Romer (14) and modified for the health care sector by Hall and Jones (22), Manton et al. (K.G.M., G. Lowrimore, D. Tolley, and A. Ullian, unpublished work) showed that health status improvements and associated human capital growth might reduce Medicare and Medicaid expenditures to a combined cost of 12% of GDP in 2080. That level of health care spending is consistent with levels of total GDP that could be expected to be dedicated to health spending (21% of GDP) in a society with a rapidly growing level of disposable income and a short-term elasticity of spending on health care of 1.6% (4).

For federal health care expenditures to decline at the projected rate because of a per annum decline of 1.5% in chronic disability, future disability declines will have to occur at increasingly older ages. An evaluation of the relative age trajectories of Medicare and Medicaid costs in the final two years of life shows that per capita, per annum Medicare costs appear to decline with age, whereas Medicaid and LTC costs tend to increase. The reason for this phenomenon is that Medicare covers primarily short postacute-care institutional stays, whereas Medicaid can cover LTC costs for the indigent. A cross-over of these costs currently occurs at ≈90 years of age (K.G.M., G. Lowrimore, D. Tolley, and A. Ullian, unpublished work). The importance of the cross-over is that Medicare costs tend to be more responsive to biomedical research and therapeutic innovations than LTC services, which are likely more subject to “Baumol's disease” (i.e., resistance to technological innovation) because of the more personal and labor-intensive, and often less biomedically skilled, nature of most services provided in LTC settings. Thus, to have the desired effects on Medicare and Medicaid spending, investments in biomedical research will likely have to be effective in improving health and rehabilitating function at increasingly advanced ages.

Investment in biomedical research and the production of innovative and more effective therapeutic and preventive medical care are possible explanations for the continued decline in elderly disability rates. However, it is far too soon to evaluate fully the population health effects of the recent (1998–2003) doubling of the National Institutes of Health budget. Other possible contributing factors include increasing levels of education, better nutrition, greater emphasis on exercise at later ages, and behavioral risk factor changes such as smoking cessation and moderation of alcohol consumption. Changes in Medicare provisions and reimbursement (e.g., BBA, 1997; BBRA, 1999) may also have more efficiently directed more and better care to the more severely disabled elderly (e.g., more emphasis on rehabilitation services) through Medicare-reimbursed home health and SNF care. Also potentially important may be increased use of home- and community-based LTC services as currently being evaluated in the Medicaid program. Identifying how each of these factors may have contributed to the 22-year disability declines, and possible future changes, is a challenging but extremely important area of social, public health, and biomedical research. It may be crucial in dealing effectively with the medical and LTC needs of baby boomers over the next 45 years.

By using the most recent NLTCS, the current work examines how age-standardized chronic disability prevalence rates changed from 1982 to 2005 specific to the level of disability. We present four specific types of results. First are age-standardized disability prevalence changes 1982–2004/2005. Second are disability changes 1982–2004/2005 in three broad age groups. Third are changes 1982–1994 and 1994–2004/2005 in the prevalence of light-to-moderate levels of disability (i.e., only limitations in IADLs and up to four ADLs) and of severe levels of disability (five or more ADLs or institutional residence). This result is to determine whether the size of moderately and severely disabled groups changed at the same, or different, rates over the two time periods. The prevailing opinion in the literature (e.g., ref. 15) appears to have been that light or moderate levels of chronic disabilities declined at a faster rate than higher-level chronic disability, suggesting that declines may have been primarily the result of social and housing innovations and expanded use of assistive devices. The final analysis shows how Medicare costs, stratified by level of disability and age group, changed over time, 1982–2004/2005.

Compared with previous work, a number of specific changes in sample weighting and disability status determination were made to improve longitudinal estimates of disability declines calculated from the five NLTCS 1982–1999. These measurement and sample weight calculation changes are discussed in Materials and Methods.

Results

Age-Standardized Disability Trends: 1999–2004/2005.

The overall declines in chronic disability prevalence rates observed in 1982–1999 (8) continued, and they accelerated to 2004/2005 (see Table 1). The prevalence of nondisabled persons older than 65 years increased from 73.5% in 1982 (standardized to the 2004 age distribution) to 81.0% in 2004/2005. This decline in the chronic disability prevalence rate is from 26.5% in 1982 to 19.0% in 2004/2005, or a 1.52% per annum relative decline. The 1982–2004/2005 decline of 7.5% is highly significant (t = 18.8, P < 0.0001), as is the 2.2% decline observed 1999–2004/2005 (t = 5.8, P < 0.0001). Indeed, each intersurvey disability decline is statistically significant with the exception of the short 1982–1984 period (t = 0.7, P = 0.48). The 1982–2004/2005 rate of decline is slightly higher than the 1.5% argued to be needed on a long-term basis to preserve the fiscal stability of the Social Security and Medicare programs (3). The per annum rate of decline was 1.8% during the 1994–1999 period, and it increased further to 2.2% for the 1999–2004/2005 survey interval. Thus, the decline in the acceleration of the rate of disability suggests that the average rate of change may improve further in the future.

Table 1.

Percentage of disability group estimates, NLTCS 1982–2004/2005

	1982	1984	1989	1994	1999	2004/2005
Nondisabled	73.5	73.8	75.2	76.8	78.8	81.0
IADL only	5.7	6.0	4.5	4.4	3.3	2.4
One or two ADL	6.8	6.9	6.6	6.1	6.3	5.6
Three or four ADL	2.9	3.0	3.7	3.4	3.7	3.8
Five or six ADL	3.5	3.3	3.1	2.9	3.0	3.2
Institution	7.5	7.0	6.9	6.3	4.9	4.0
Per annum declines, %		0.6	1.1	1.3	1.8	2.2

Reported declines before 2004/2005 (cf. 8) are reduced in magnitude in Table 1 by roughly 0.2% because of age standardization to the 2004 population age distribution. The distribution of ages in the older-than-65 population shifted upward in 2004 to more advanced ages (e.g., 85+), ages with typically higher base prevalence rates in 1982 than in 2004/2005. Thus, age standardization is necessary to control for changes over time in the age distribution of Medicare enrollees aged 65+.

The range of estimates is consistent with the 1.7% decline observed in 1982–1999, although recent Census Bureau sample weight adjustments cause more of the long-term decline in chronic disability to be attributed to the 1999 to 2004/2005 period. The disability declines are thus distributed over a longer time period but with a consistent acceleration of the decline of disability prevalence. The 0.6% per annum rate of change in 1982–1984 is similar to the 0.6% per annum rate estimated by Fogel (4) and Costa (9). This per annum rate of prevalence rate declines then accelerates over the 22-year period of study (1.1% 1984–1989; 1.3% 1989–1994; 1.8% 1994–1999; 2.2% 1999–2004/2005).

Changes in specific disability levels 1982–2004/2005 are also documented in Table 1. The age-adjusted prevalence of institutional residence declined from 7.5% in 1982 to 4.0% in 2004/2005 (a 46.7% relative decline). The decline is smaller, 3.5% to 3.2%, for severely disabled community residents with five or six ADLs, probably because large numbers of persons who did not enter nursing homes are likely to be found in this disability category. Recent changes in the nature of institutionalization (e.g., the emergence of assisted-living options 1994–1999, which continued to 2004) and in reimbursement of Medicare hospice, home health assistance (HHA) and SNF benefits as well as improvements in functional rehabilitation and, more recently, regenerative medicine likely have helped reduce the proportion of the elderly and the very elderly in LTC institutions. Combining the three most highly disabled groups (i.e., persons with three or more ADL impairments in the community and persons in institutions), the decline is from 13.9% in 1982 to 11.1% in 2004/2005, or a relative drop of 20.1%, which is quite large given an absolute increase in the community-resident population with three or four ADLs.

Age-Specific Changes in Chronic Disability Prevalence.

In Table 2, we examine disability changes within three age groups (65–74, 75–84, and 85+). Between 1982 and 2004/2005, there is a 6.2% relative increase, from 85.8% to 91.1%, in the proportion nondisabled at ages 65–74. At ages 75–84, the relative increase is 12.6%, from 69.3% to 78.1%. At ages 85+, the relative increase is larger (32.6%), from 37.9% to 50.3%. Each of the age-specific increases is also highly significant (e.g., at age 65–74, t = 11.3, P < 0.0001; at ages 85+, t = 8.6, P < 0.0001).

Table 2.

Percentage of disability group estimates by age, NLTCS 1982–2004/2005

Age and disability level	1982	1984	1989	1994	1999	2004/2005	Relative changes, %
Age 65–74
Nondisabled	85.8	86.7	88.1	88.2	89.3	91.1	6.2
IADL only	4.3	4.1	3.0	3.2	2.5	1.8	−58.7
One or two ADL	4.1	4.0	3.8	3.7	3.4	3.1	−25.6
Three or four ADL	1.8	1.8	1.7	1.7	2.0	1.6	−8.4
Five or six ADL	2.0	1.7	1.5	1.6	1.4	1.5	−22.8
Institution	2.0	1.7	1.9	1.6	1.4	0.9	−54.6
Age 75–84
Nondisabled	69.3	70.2	70.6	73.8	76.6	78.1	12.6
IADL only	7.0	7.5	5.8	5.2	3.6	2.5	−63.9
One or two ADL	8.2	8.1	8.6	7.5	8.0	6.7	−18.6
Three or four ADL	3.4	3.5	4.5	4.1	4.2	4.5	33.5
Five or six ADL	3.9	3.6	3.5	3.0	3.4	4.0	2.4
Institution	8.1	7.1	7.0	6.3	4.3	4.1	−48.8
Age 85+
Nondisabled	37.9	34.1	38.6	41.5	44.4	50.3	32.6
IADL only	7.5	9.4	6.8	7.1	5.5	4.2	−43.9
One or two ADL	13.3	14.6	11.9	11.6	12.9	12.1	−9.2
Three or four ADL	6.2	6.6	8.9	7.7	9.2	10.2	64.9
Five or six ADL	7.8	8.6	7.7	7.6	8.5	7.6	−3.4
Institution	27.2	26.6	26.1	24.6	19.5	15.6	−42.7

The larger relative rate of improvement at higher ages is related to large declines in institutional use at ages 75+. The decline, from 8.1% to 4.2%, is almost 49% at ages 75–84. At ages 85+ the decline in institutionalization, 27.2% to 15.6%, is almost 43%.

At all ages there are also large declines in the IADL-only and one-to-two ADL-impaired groups. A possible explanation is the increased availability of assistive devices and modifications in the built environment, although some of these declines may also be because of earlier and better intervention in disability-related disease processes (e.g., better interventions in diabetes management; ref. 16). Only the three-or-four ADL-impaired group shows increases for age groups 75–84 and 85+. Further analyses should focus on why this specific level of disability should be resistant to change at ages 75+.

Early (1982–1994) and Late (1994–2004) Changes in Moderate and Severe Disability.

We compare rates of decline in the light-to-moderately (i.e., those with up to four ADLs, including persons with impairment in at least one IADL) and severely (five or more ADLs or in nursing home residence) disabled groups for 1982–1994 and 1994–2004/2005 in Table 3. The selection of 1994 as a division point is motivated partly because it is roughly halfway through the current NLTCS series and partly because it may reflect certain turning points in specific health and health care processes. For example, in 1990 cancer mortality rates started to decline in the U.S., partly because of treatment innovation for specific solid tumors (e.g., for early onset, aggressive female breast cancer) and partly because of public health interventions (e.g., cohort-related reductions in male smoking rates). There also emerged an economically viable U.S. biotechnology sector during this time period, and biomedical research spending has been increased significantly.

Table 3.

Declines in low and high disability levels, NLTCS 1982–1994, 1994–2004/2005

Disability levels	Disabled at survey years, %			Disability decline, %		Per year decline, %
	1982	1994	2004/2005	1982–1994	1994–2004/2005	1982–1994	1994–2004/2005
Low (up to four ADL)	15.5	13.9	11.8	1.6	2.2	0.9	1.7
High (five plus ADL)	11.0	9.2	7.2	1.8	2.0	1.4	2.4

The rate of improvement in the light-to-moderately impaired group increased between the two time periods (0.9% to 1.7% per annum) but not as rapidly in either time period as declines in the severely impaired groups (five or six ADLs or institutionalized), where the per annum rate of improvement increased from 1.4% to 2.4%. Thus, the rate of improvement for the most severely disabled groups is more rapid in both time periods than for light-to-moderately disabled groups. This finding contradicts the view that long-term declines in chronic disability prevalence are primarily because of changes in the built environment and improvements in assistive devices, and it suggests that biomedical innovations in the management of diseases leading to serious disability are increasing in importance post-1994. These estimates are statistically significantly different both over time and across disability level based on a cross-time comparison of roughly 30,000 NLTCS cases followed from 1982 to 1994 (i.e., 20,000 cases in the 1982–1984 NLTCS plus aged-in cohorts in 1989 and 1994) vs. 30,000 cases followed 1994–2004/2005, or roughly 288,000 person-years of experience 1982–1994 vs. 200,000 person-years of experience observed between 1994 and 2004/2005.

Changes in the Distribution of Medicare Expenditures over Disability Levels and Age.

Table 4 shows temporal changes (1982–2004) in the average level of Medicare Part A per capita costs, adjusted for the medical care consumer price index, over disability level. We examine Part A Medicare expenditures because they reflect acute hospitalization, HHA, and SNF use; i.e., more medically intense types of Medicare services. In addition, Medicare Part A data are complete with diagnoses back to 1984 when the acute hospital prospective payment system was instituted. Diagnoses for Part B services, which are heavily weighted to physician or outpatient services, are only available since 1991. We performed analyses both with Part A and total (A + B) Medicare expenses and found similar trends. Part A data, representing use of more medically intense services, tended to show clearer trends and patterns. In addition, Part A funding directly involves the Medicare Trust Fund, and it has less flexibility to respond to the cost of Medicare service use increases than Part B, which is funded from premiums (supplementary medical insurance) and general revenue.

Table 4.

Average Medicare Part A expenditures by disability category and year (in 2004 dollars)

Age and disability level	1982	1984	1989	1994	1999	2004
Total
Nondisabled	2,238	2,212	1,967	1,881	1,745	1,584
IADL only	5,190	4,440	2,973	3,802	2,904	3,590
One or two ADL	5,733	5,369	4,580	5,307	4,403	4,849
Three or four ADL	9,884	7,855	6,439	10,956	7,712	6,476
Five or six ADL	15,236	13,706	11,880	17,716	13,558	11,971
Institution	8,528	7,084	7,853	8,550	10,070	9,788
Age 65–84
Nondisabled	2,208	2,189	1,926	1,873	1,683	1,521
IADL only	5,291	4,603	3,059	4,058	2,765	3,631
One or two ADL	5,852	5,466	4,679	5,103	4,461	5,338
Three or four ADL	10,901	7,961	6,976	9,017	8,363	7,284
Five or six ADL	17,359	14,896	11,679	10,479	14,922	12,454
Institution	9,808	8,159	9,314	6,574	12,821	10,795
Age 85+
Nondisabled	2,893	2,742	2,804	2,015	2,593	2,296
IADL only	4,531	3,556	2,524	2,694	3,448	3,459
One or two ADL	5,267	5,022	4,222	4,496	4,231	3,678
Three or four ADL	6,278	7,482	5,044	8,528	6,236	5,050
Five or six ADL	8,415	10,512	12,391	17,359	11,081	10,929
Institution	6,675	5,654	6,016	7,076	7,316	8,816

There are two general sets of expenditure trends: 1982–1989 and 1994–2004. The highest average level of Medicare costs was incurred for persons with five or six ADLs, and it occurs in all years and for both age strata. Table 4 shows that costs at high levels of disability decreased from 1982 to 1989 [possibly because of the effects of the 1983–1984 introduction of the diagnosis-related group/prospective payment system (DRG/PPS), which reduced acute hospital lengths of stay and costs, and imposed restrictions on postacute HHA and SNF use], increased in 1989–1994 (with HHA and SNF benefit expansion because of the Duggan vs. Bowen and Fox vs. Bowen decisions), and then declined in 1999 and 2004 after passage of the 1997 BBA and 1999 BBRA, which enacted and fine-tuned a PPS for HHA and SNF benefits.

Medicare costs for community-dwelling persons with five or six ADLs declined in 1994–2004, whereas costs for persons in nursing homes increased or were stable, which may reflect a shifting of the most severely impaired cases into postacute care institutions and a movement of less disabled persons out of institutions into community residences because of declines in the rate of discharge from acute-stay hospitals to SNFs. One of the striking, consistent trends was the large decline in inflation-adjusted Part A costs for the nondisabled population that was evident across the period 1982–2004.

Table 4 also contains costs stratified on age. Two groups (for persons aged 65–84 and those aged 85+) showed different cost trends for the 1982–1999 period. It is significant that for persons with three or more ADLs the average adjusted Part A costs declined for persons aged 65–84. The Medicare Part A costs for persons older than 85 showed declines for persons with one to six ADLs who were community residents. The cost changes in 1999–2004 for persons age 85+ appear to reflect refinements made to the PPS system for HHA and SNF made in the 1999 BBRA. Thus, Medicare Part A cost dynamics in 1982–2004 specific to disability level and age seem to have beneficial properties that interact with the long-term changes in disability prevalence rates.

Discussion

The 1982–2004/2005 data show several clear trends. The rate of decline in chronic disability prevalence accelerates from 1982 to 2004/2005, starting at 0.6% per annum in 1982–1984 and increasing to almost four times that level (2.2%) by 1999–2004/2005. The rate of decline for persons with five or more ADLs accelerates more in 1994–2004/2005 than 1982–1994.

A major health care dynamic in 1982–2004/2005 has been a change in the management of severely disabled persons, especially in long-term institutional care. Before the 1980s, institutional LTC was primarily conceived of as residential and housing services and maintenance care, which was not well developed medically or for rehabilitation (17). In the 1960s and 1970s, many elderly persons, especially those with depression and severe dementia and without significant socioeconomic resources, were in state mental hospitals. With the development of pharmaceutical treatments (e.g., Thorazine) for select mental health problems and the deinstitutionalization movement in the 1960s, 1970s, and 1980s, younger and middle-aged psychiatric patients were increasingly treated in the community, allowing numerous state mental hospitals to be closed.

Many elderly patients with depression and dementia and without viable private community residence options were discharged to nursing homes reimbursed by Medicaid. The residential use of nursing homes was facilitated by the 1955 National Housing Act and the 1965 implementation of Medicare and Medicaid, both of which promoted nursing home construction. The peak number of nursing homes in the U.S. was reached in 1985. From 1985 to 1995, the nursing home population grew more slowly than the U.S. elderly population, in part because of more stringent regulation of HHA and SNF use after the institution of PPS in 1983–1984 for acute-care hospital reimbursements.

In 1988, the Catastrophe Care Act was instituted and then rescinded. In response, Medicare SNF and HHA regulations were loosened in 1989 (in part because of the Duggan vs. Bowen court decision in 1988), which allowed rapid increases in Medicare SNF and HHA use from 1989 to 1996. Because Medicare SNF care is intended to be postacute care and strongly limited in duration (100 days), the increased availability of postacute care tended to reduce traditional nursing home admissions.

Medicare expenditures on SNF and HHA service grew rapidly until 1997, when those benefits were restricted by the BBA. For example, the BBA instituted a PPS system for SNFs with reimbursements weighted by how much rehabilitation was undertaken. Associated with changes in Medicare policy was the improved economic status of the elderly, which gave them additional private payment options such as the use of assisted-living facilities.

More fine-grained analyses of the specific biological mechanisms underlying the long-term disability decline will be necessary to determine which biomedical research and treatment innovation investments will best sustain the rate of declines in functional disability at late ages in the future. Continuing progress will require significantly greater biomedical research expenditures on fundamental mechanisms (i.e., at the cellular and molecular level) of chronic disease and human senescence (K.G.M., G. Lowrimore, D. Tolley, and A. Ullian, unpublished work). Such fundamental discoveries will help enhance basic rehabilitation efforts at later ages through improved understanding of neural and muscle function and of metabolism at the molecular, cellular, and mitochondrial levels (e.g., 18, 19). This endeavor will involve looking in greater detail at the genetic and molecular basis of disease processes and senescence to develop effective interventions at advanced ages (e.g., ages 95+ in the post-2015–2020 period). It is thus heartening that the rate of disability reduction was found to be higher at advanced ages (e.g., ages 85+) in the NLTCS data series and that NLTCS data linked to diagnostic records in Medicare files showed that cognitive impairment caused by circulatory events (especially stroke) declined significantly to 1999 (20). Reductions in cognitive impairment may set the stage for further physical rehabilitation because of the greater compliance capacity in cognitively intact persons.

Further analyses also need to be done with the 1982–2004/2005 NLTCS linked to Medicare (and possibly Medicaid) service use, diagnoses, and expenditure data. With the six surveys spanning 1982–2004, and continuous Medicare history files (i.e., services and expenditures recorded on a daily basis), more detailed longitudinal analyses of individual and cohort disability and health dynamics can be undertaken. Knowledge of those dynamics will also be important in translating biomedical and rehabilitation innovations that can reduce chronic disability at more advanced ages in the U.S. population.

Materials and Methods

The NLTCS was conducted in 1982, 1984, 1989, 1994, 1999, and 2004/2005. The NLTCS sample is drawn from Medicare enrollment files covering >97% of the U.S. elderly population. Each survey record is linked to continuous Medicare service use and cost data from the date that each NLTCS sample person passed age 65 or was included in the NLTCS sample above age 65 in 1982.

Interviewing is done in two phases. The first is a screen that determines whether the person is chronically disabled (i.e., for 90+ days) on at least one of seven ADLs and nine IADLs. The presence of nursing home or institutional residence (and starting in 1994, assisted-living residence) is determined on the screen instrument. The NLTCS definition of institutional residence requires that medical services be available at the facility on a 24-h basis. The NLTCS sample size is roughly 20,000 persons at each wave, with 15,000 persons continuing from earlier waves and ≈5,000 persons drawn in a supplementary sample, aged 65–69, who became Medicare-eligible between NLTCS waves. The sample supplement aged 65–69 maintains the national representativeness of the NLTCS for the entire elderly Medicare beneficiary population, and it replaces long-term (5-year) sample attrition caused by death. In 1994, 1999, and 2004/2005, oversamples of persons aged 95+ were drawn. These oversamples may be critical for designing interventions to maintain disability rate declines in the long run. Medicare vital statistic records provide the exact date of death for all Medicare enrollees selected into the NLTCS sample.

The second phase is a detailed community or institutional interview that generates more comprehensive data on disabilities, LTC not covered by Medicare, residential characteristics, socioeconomic traits, and family and informal LTC caregiver characteristics. This phase takes roughly 55–60 min. Additional risk factors [e.g., smoking, body mass index (BMI), and BMI change] and nutritional measures have been added to the NLTCS instrument. In each NLTCS, innovations have been made in the community interview content (e.g., next-of-kin surveys 1982–1984 and 1999–2004/2005; informal caregiver surveys in 1982, 1989, 1999, and 2004/2005; a health insurance module in 1989), some involving ancillary respondents. Rosters are made of caregivers, children, and family members. Small additions have also been made to the screening instrument, e.g., adding veteran status in 2004.

The NLTCS began in 1982, when, not knowing the prevalence of chronic disability in 1982, the Census Bureau selected 55,000 persons over age 65 from Medicare enrollment lists. This sample was organized into roughly 100 reduction sets, which were to be drawn until >6,000 persons chronically disabled (90+ days) in at least one of seven ADLs and nine IADLs were identified on the screen. After 35,789 persons were screened, 6,393 with chronic disability were identified, with an additional 1,992 identified as living in institutions.

The second NLTCS was conducted in 1984 to do a prepost analysis of the consequences of the hospital DRG/PPS for the Medicare elderly population. Cost constraints required that the sample be reduced by using roughly 45% (12,100) of the 25,000 persons not disabled in the 1982 sample of 35,789. The prepost analyses were successful in identifying which chronically disabled groups were most susceptible to PPS-induced changes in acute hospital use (e.g., decreased length of stay).

The NLTCS sample is followed longitudinally, but it can also be assessed cross-sectionally. Supplemental samples of persons reaching age 65 between surveys are cross-sectionally constructed for each NLTCS, but they are followed longitudinally through later NLTCS waves. This sample replenishment design was used in 1984, 1989, 1994, 1999, and 2004/2005. Six surveys now follow 22 years of disability and 23 years of mortality changes and Medicare service use in a total of 49,000 sample persons, and they cover (to 2005) almost 30,000 deaths (with Medicare mortality and service records updated on a quarterly and an annual basis, respectively). The NLTCS now represents roughly 500,000 person-years of experience, which is sufficient to support subgroup analyses and accurately portray overall, cohort, and period-specific disability trends.

The 1989 and 1999 NLTCS had small sample weight anomalies because each sample was necessarily designed before the 1990 and 2000 decennial census counts were available. Specifically, both the 1989 and 1999 NLTCS samples were based on 9-year projections of population components (e.g., by age, sex, race, and institutional status), with exact estimates known with precision only from the decennial census after April 1, the date of the sample draw. As a consequence, there were changes in sample weight calculations after the 1989 and 1999 NLTCS field periods. In 1999, the Census Bureau institutional population estimates for poststratification weighting based on the 9-year projections did not reflect the emergence of the assisted-living phenomenon or the decline in the size of the nursing home population, which were detected in the NLTCS because the sample was based on a list of individual Medicare enrollees.

In 2004/2005, there was an almost 6-month delay in interviewing the population sample, which had two effects. First, the proportion of the sample drawn on April 1, 2004, who died before the interview attempt increased from 5.9% (in 1999) to 9.5% (in 2004/2005). Mortality between the screen interview and the detailed interview was nearly unchanged between 1999 (0.26%) and 2004 (0.34%). Second, persons aged 65–74 were far more difficult to locate in 2004 than in past NLTCS despite an enhanced tracking procedure. An analysis of the difficult-to-locate 65- to 74-year-olds revealed that they had very low Medicare costs, especially for Part A services. These highly mobile 65- to 74-year-olds with low medical expenditures and service use were the major cause of the response rate decline from 95% (in all prior waves) to 91% in 2004/2005.

To deal with potential bias in health and functional trait estimates caused by the higher nonresponse rates in 2004/2005, we compared Medicare Part A expenditures in nonrespondents with Part A expenditures in the total sample. Part A was used rather than total Medicare expenditures because persons using Part A services (inpatient hospital care, HHA, and SNF) are generally more ill than persons using Part B services, which involve predominantly outpatient and physician care. Part A expenditures were used to generate a ratio for solving a simple equation to estimate the mix of healthy and nonhealthy persons (stratified by age) in the larger 2004/2005 nonresponse (nonlocatable) group. Knowledge of this ratio and Part A costs in the disabled and nondisabled groups allowed us to impute the health and functional status of nonrespondents by using an independent data source (Medicare Part A service use files) with 100% sample coverage. This analysis indicated that in the 2004/2005 population aged 65–74, Part A costs for nonrespondents are lower than for the total population, suggesting that 65- to 74-year-old nonrespondents are healthier than the total 65- to 74-year-old population.

To keep our results consistent over time, we used Medicare Part A data to adjust for nonresponse bias in all five prior NLTCS. This adjustment had a larger effect than age-specific sample weight adjustments of the 1989 and 1999 NLTCS, and it produced a smoother (more consistent over time) set of cross-temporal results than in previous analyses because additional cross-sectional sampling artifacts were removed (see Table 1 and ref. 8). Disability status for the 2004/2005 sample was coded in a fashion fully consistent with earlier years. The screener was used to assess disability prevalence for the screened-in sample, and the detailed interviews were used to assess disability prevalence and LTC use for the longitudinal sample.

Abbreviations

ADLs

activities of daily living

BBA

Balanced Budget Act

BBRA

Balanced Budget Refinement Act

DRG

diagnosis-related group

GDP

gross domestic product

IADLs

instrumental activities of daily living

LTC

long-term care

NLTCS

National Long-Term Care Survey

PPS

prospective payment systems

SNF

skilled nursing facility.