Quality of the Last Year of Life of Older Adults: 1986 vs 1993 FREE

When people survive to old age, fear of death is often joined by fears of disability or institutionalization. Morbidity and disability before death have a tremendous impact on such persons, their families, and the health care delivery system. Health care use increases substantially as people draw closer to death.¹ In 1993, 73% of all deaths in the United States occurred among persons 65 years and older and 23% among those 85 years and older.² It is therefore important to have a clear description of the health status and quality of life before death for this population. The theory of compression of morbidity³ posits that most persons would survive to the end of their biologically endowed life span and die of natural causes. The prevalence of disability and the proportion of life lived with disability would therefore decrease. Although this may be an ideal personal goal that could also reduce the burden on the health care system, other commentators predict that in the face of continued increases in total life expectancy, an extension of the period of disability prior to death may be an inevitable consequence of living to an advanced age.^{4 - 5} United States residents may instead be living a longer life with worsening health.⁶

Although no consensus exists, many studies have described the pattern of change in morbidity and disability among the general US population of older adults.^{6 - 11} Quality of life prior to death has been studied less thoroughly.¹² This study used data from the 1986 and 1993 National Mortality Followback Surveys to examine the changes in the use of formal medical care, disability, and overall quality of life in the year before death among older persons living in the United States and to examine how this change varied by sex and age at death.

The National Center for Health Statistics periodically conducts the National Mortality Followback Surveys on probability samples of all deaths and obtains in-depth information from decedents' next of kin. The 1986 and 1993 surveys were the fifth and sixth such surveys^{13 - 15} to gather disability, activities of daily living (ADL), lifestyle, and other health information in the last year of life.

Death certificates were drawn from the current mortality sample, a 10% systematic sample of the death certificates of the 50 states, the District of Columbia, and the independent registration area of New York City. Because of respondent consent requirements, Oregon and South Dakota were not included in the 1986 and 1993 survey, respectively. From these certificates, a complex sampling design was used to ensure that enough decedents from certain strata would be selected. Some populations (eg, blacks) and some diseases (eg, cardiovascular disease and external causes) were oversampled. The final sample represents a national 1% probability sample of all deaths in the survey-specified age ranges, ie, 25 years and older in 1986 and 15 years and older in 1993.

The next of kin, identified on the decedent's death certificate, was contacted by personalized letter and asked to participate in the survey. For cases with no identified next of kin on the death certificate, letters were sent to funeral directors asking for the next of kin. The US Bureau of the Census served as the data collection agent and subsequently contacted the next of kin respondents either by telephone or in person. When no next of kin was identified or could not be located, another person familiar with the decedent's life history was contacted. The overall response rate for the informant questionnaire was 89% and 83%, respectively, for the 2 surveys; and 90% and 88%, respectively, for decedents 65 years and older.

Proxy respondents were asked to report the decedents' total lengths of overnight stay in each hospital or nursing home during the last year of life. Information was obtained on the decedent's ability to perform 5 self-care activities: walking, bathing or showering, using the toilet (including getting to and from the toilet), dressing, and eating. If the decedent had difficulty in doing any of the activities, respondents were asked to indicate the length of time the decedent experienced difficulty for each activity. Cognitive function in the last year of life was evaluated by asking whether the decedent had trouble understanding where he/she was, remembering what year it was, and recognizing family members or good friends. The impairment of each function was classified by 4 categories: never or hardly ever, only in last few hours or days before death, some of the time, and all or most of the time.

Decedents included in this analysis were those 65 years and older at death. Analyses were performed by sex and by 2 age groups, those aged 65 through 84 years, and those 85 years and older. Use of health care services and numbers of disabilities in the last year of life of decedents were compared between the 2 survey years with several approaches. First, we compared the length of hospital or nursing home stay and the proportion of decedents with limitation in each of the 5 ADLs and in each of the 3 cognitive measures. Then we constructed 2 opposite-extreme categories for each of the following indicators: use of service, less than 1 week of total hospital and nursing home stay (minimum use) vs 6 months or longer combined hospital and nursing home stays (extensive use); ADL, no difficulty in all 5 ADLs or limited for less than 30 days if any (minimally restricted) vs difficulty in at least 2 ADLs for 6 months or longer (severely restricted); cognitive function, no impairment in all 3 domains of function or in only the last few hours or days before death if any (normal) vs having impairment in at least 2 functions all or most of the time (severely impaired); and overall quality of life in the last year, minimal use of service and minimal restriction in ADL and normal cognitive function (good quality) vs at least 1 indicator described above that was in the unfavorable extreme (severely impaired).

Finally, we created a summary sickness score for each decedent jointly considering the severity and length of each service used, disability, and impairment. We began by assigning scores for each group of indicators: service use, ADL, and cognitive function. Each indicator consisted of a various number of disability items, and different disabilities may not represent the same extent of sickness (eg, difficulty in eating vs walking). It was therefore necessary to weight different items and to standardize the scores of the 3 indicators. For service use, the score is the total days of all hospital and nursing home stays. For ADL and cognitive function, assuming severity of a dysfunction and its prevalence are negatively related, the score was calculated as ΣW_iT_i/ΣW_i, where W_i is the inverse of the prevalence rate of disability _i in the combined data, T_i is the duration (days) of disability. We assigned 0, 5, 30, and 270 days for the 4 categories of impairment in cognitive function, respectively. The scores for service use, ADL, and cognitive function were then summed up to derive a single measure of overall sickness. This score ignores the interrelationships among 3 indicators but should correctly order decedents according to severity of disabilities. A higher score (with day as the unit) represents a worse quality of life compared with a lower score.

The National Mortality Followback Survey is a stratified random sample that allows for projections of national estimates from the sample. Data were weighted by SUDAAN software¹⁶ to be representative of the age, sex, and racial composition noted on all death certificates and for factors determining the probability of selection. The response rate of informants was also adjusted in the weighting process. The t test was used to compare the age-specific difference between the 2 survey years and the difference in change from 1986 to 1993 between the 2 age groups. Nights of hospital or nursing home stay were log-transformed in the statistical tests. Multivariate regression analysis was performed to adjust for changes between the 2 survey years in the distribution of the relationship between proxy respondents and decedent (spouse, child/parent, sibling/other relative, or other) and the living arrangement of the decedents before death (lived alone or lived with spouse, or nonspouse). To examine whether the changes in sickness score were significantly different by age, the interaction term year × age was tested in the regression model.

Basic demographic data for the US decedents in 1986 and 1993 are shown in Table 1. The available survey sample sizes were 9179 in 1986 and 6735 in 1993, representing about 1.5 and 1.6 million decedents 65 years and older who died during the 2 survey years, 1986 and 1993. Age and race distributions were comparable between the 2 years within the same age groups (65-84 years; ≥85 years). Among younger women (aged 65-84 years) the percentage of widows decreased from 55.3% in 1986 to 50.4% in 1993 (P<.05). No significant distribution change in marital status was found in other groups. The proportion of decedents who lived alone in the last year of life decreased from 1986 to 1993 among women in both age groups (P<.05). The proportion of spouses completing the survey decreased in 1993, which was statistically significant for men aged 65 through 84 years. A greater proportion of the respondents were the decedents' children among men aged 65 through 84 years (P<.05).

The year before death, hospital stays for men did not change from 1986 to 1993, but significantly more women in both age groups had no hospital stays in the year before death in 1993 than in 1986. More men and women aged 65 through 84 years were hospitalized than those 85 years and older (Table 2). Nursing home stays decreased significantly in older men and in both age groups of women. Older men and women had large increases in never staying in a nursing home and declined in the number of nursing home stays than the younger group (P<.05).

Table 3 presents percentages of decedents with difficulty carrying out the 5 ADLs regardless of the length of disability. Walking and bathing were the most common disabilities. For the same age group women had a greater proportion with disability than did men. Older men and women experienced fewer disabilities in 1993 than in 1986, although most differences were significant only for women.

Percentages of subjects with impaired cognitive function in at least some portion of the last year of life are shown in Table 4. The rates of cognitive impairments reported in both sexes declined from 1986 to 1993 but were not statistically significant among older men.

Individuals at the extremes of health measures are compared in Table 5. Among the oldest men and both age groups of women, percentages requiring minimal use of hospital and nursing home services increased significantly, while percentages of extensive users decreased significantly from 1986 to 1993. Among the parameters that changed significantly from 1986 to 1993, all trends were in the direction of improved function and quality of life for those 85 years and older.

Table 6 presents the changes in mean overall sickness score (1993 minus 1986), constructed as an additive index of health in the last year of life. The composition of respondents and living arrangements before death were not entirely comparable between 1986 and 1993 in some sex and age groups (Table 1). A lower sickness score was found when the spouse was the proxy respondent or when the person lived alone before death (data available on request). The effects of changes in these 2 potential confounders on the changes in the sickness score over time were considered in the multivariate regression analyses. A significant reduction of the adjusted mean score was demonstrated in both men and women 85 years and older. The reduction was also statistically significant for women aged 65 through 84 years in the full adjusted model. When analyses were restricted to responses from proxies who lived with and were relatives of the decedents, greater declines in sickness scores were found than those with all decedents included in the analyses (Table 6). The impact of age on the changes in quality of life was examined by including year × age interaction terms in the regression models. The estimates of the interactions were statistically significant for both men and women in all models (P<.01).

Data from this large national representative sample demonstrate that women in their last year of life in 1993 used less in-patient medical care (hospital and nursing home stay) than those in 1986. The oldest women also experienced less disability in ADL and decline in cognitive function. The overall quality of life in their last year of life improved among all women but was most prominent in those 85 years and older. Similar patterns of changes were also observed in the oldest men but not in men aged 65 through 84 years.

In the general US population, between 1986 and 1993, the discharge rate in nonfederal short-stay hospitals decreased from 367 to 342 per 1000 persons 65 years and older.¹⁷ The average length of stay decreased from 8.5 to 7.8 days. Older patients are also not entering nursing homes at the same rate or staying as long as they did in previous years.¹⁸ The decline in the use of formal health care services in the last year of life found in this study parallels trends in the general population. This likely reflects, in part, improvement in health of those older adults, although it could also be influenced by changes in the health care system and in the Medicare/Medicaid reimbursement policy. The availability of alternative forms of care, such as home health care, have contributed significantly to diverting older persons from nursing home care, which can and may be because they prefer to receive care at home.¹⁸ Medical technology developed for use in the home with the use of assistive devices and housing modifications has sharply increased. Both the levels of care and the mix of long-term care services available concomitant with the improving health and increased use of assistive devices, social conditions, and desire for independence among older adults may shift care for older patients to a community-based approach.¹⁹

Earlier trend reports on morbidity and disability in the elderly US population showed that life expectancy gains before the 1980s were entirely lost in disabilities.⁷ A reversal of this trend was observed between 1980 and 1990.⁹ Data from the National Long Term Care Surveys⁸ showed a decline in disability prevalence, in terms of ADL and instrumental ADL, from 1982 to 1989 in the elderly population, which continued through 1994. A large decline in limitation of physical functioning from 1984 to 1993 was also reported.¹¹ However, the analyses of the 1984-1990 Longitudinal Survey on Aging and the 1982-1993 National Health Interview Survey¹⁰ yielded ambiguous results. Our study demonstrates that the burden of disability in the last year of life was reduced from 1986 to 1993 in the oldest men and in all women. This finding indicates that the decline in hospital and nursing home use is partly due to better health.

One of the unique aspects of current aging trends is the rapid growth of the oldest old.²⁰ Studies of only healthy survivors to advanced ages suggest that physiological function of many types either declines much more slowly than previously estimated,²¹ or that many types of declines were concentrated in a relatively short period just before death.²² National data show that a surprisingly large percentage of the oldest old not only manage without daily personal assistance but also are physically robust.²³ These improvements suggest that changes in health and the natural history of disease processes may be occurring concurrently.²⁴ This study further demonstrates that the overall quality of life in the last year is improving at a greater rate in the oldest-old persons than in the younger-old. Consistent with an earlier report in the 1980s,²⁵ those dying at the oldest ages generally had more disability but spent less time in the hospital prior to death than those dying at younger ages. Data showed that Medicare payments in the last years of life decreased as age at death increased.^{1 ,26} One explanation is that care of the functionally impaired can be supportive, rather than aggressive.²⁷ The very old and frail elderly use long-term care more often than expensive, acute care. It appears that those in the best functional condition consumed the greatest amount of expensive "high-tech" care, countering arguments for rationing care to elderly persons with chronic illnesses and bolstering arguments for supportive care.²⁸

The information provided in these 2 surveys was obtained from proxies, which is a potential limitation; however, some carefully designed studies have found no evidence of a difference between reports by proxies and those by patients themselves.^{29 - 31} A major objective of the 1993 National Mortality Followback Survey was to replicate critical components of the 1986 survey to examine the secular changes among the US decedents. To this end, considerable effort was made to maintain comparability of the 2 surveys, including design and data collection. Imprecision as a result of proxy report is subject to large uncertainty for a given year but less variability in measuring changes over time. This is because many errors in the estimates are likely consistent across respondents and hence tend to "cancel" in comparisons. We have used multivariate analysis techniques and subgroup analysis to account for the shifts in respondent compositions and living arrangements in the 2 survey years. A minor (2%) change in age-specific response rate between the 2 surveys was adjusted in the analyses as well.

In summary, these data from national samples demonstrate that compared with 1986, elderly decedents in 1993—at least in the oldest men and all women—experienced a better overall quality of life in the last year before death. Medical care costs are disproportionately concentrated in the years prior to death in most of the persons. If the observed favorable trend continues and extends to all groups, it could portend slower rates of growth in national health expenditures in the future as well.