Trajectories of Nonagenarian Health: Sex, Age, and Period Effects

Abstract

The US population aged 90 years or more is growing rapidly, and there are limited data on their health. The Cardiovascular Health Study is a prospective study of black and white adults aged ≥65 years recruited in 2 waves (1989–1990 and 1992–1993) from Medicare eligibility lists in Forsyth County, North Carolina; Sacramento County, California; Washington County, Maryland; and Pittsburgh, Pennsylvania. We created a synthetic cohort of the 1,889 participants who had reached age 90 years at baseline or during follow-up through July 16, 2015. Participants entered the cohort at 90 years of age, and we evaluated their changes in health after age 90 years (median duration of follow-up, 3 years (interquartile range, 1.3–5)). Measures of health included cardiovascular events, cognitive function, depressive symptoms, prescription medications, self-rated health, and functional status. The mortality rate was high: 19.0 per 100 person-years (95% confidence interval : 17.8, 20.3) in women and 20.9 per 100 person-years (95% confidence interval: 19.2, 22.8) in men. Cognitive function and all measures of functional status declined with age; these changes were similar by sex. When we isolated period effects, we found that medication use increased over time. These estimates can help inform future research and can help health-care systems meet the needs of this growing population.

Persons aged 90 years or more are the fastest growing segment of the US population (1). Life expectancy has progressively increased over the last century and continues to increase, such that in the present era, a person aged 90 years is expected to live nearly another 5 years (2). Despite this impressive growth, there are few data on trajectories of health in this population. Understanding the changes in health status that affect quality of life and the need for long-term care will better allow for planning for services needed to support older adults, such as in-home caregiving or nursing home care.

The majority of the data on the population aged 90 years or more have come from outside the United States (3–6). Some US studies have examined survival to age 90 years (7–10), but there are few data on trajectories of health once individuals reach this milestone. The 90+ Study, initiated in 2003, includes rich measures of neurological function and physical performance; this predominantly white and well-educated population represents the surviving residents of a retirement community in Orange County, California (11). Results from this cohort study demonstrate a greater burden of disability and cognitive impairment at older ages, and investigators have observed that women have more disability and dementia than men (12, 13). Additional descriptions of the trajectories of health status among persons aged 90 years or more will help us to understand better this growing population. Moreover, studies that have enrolled participants from a wide range of birth cohorts can better disentangle age associations from secular trends.

The Cardiovascular Health Study (CHS) represents a unique opportunity to gain insight into the health of the growing 90-and-over population in a community-based sample. CHS investigators enrolled participants aged 65–101 years in 1989, so participants have aged into our synthetic 90+ cohort over the past 28 years. Additionally, we have shown that survival rates in the CHS are similar to those in the overall US population, which supports the representativeness of this sample (14). Only 1 previous analysis in CHS participants has focused specifically on the nearly 2,000 participants who reached age 90 years in the CHS, and that study was limited to cross-sectional data (15). In this study, we determined the cardiovascular event rates and trajectories of health status in this population. Specifically, we quantified the changes in cognitive function, depressive symptoms, medication use, self-rated health, and functional status after participants reached age 90 years.

METHODS

Study population

The CHS is a community-based study of black and white adults who were aged 65 years or more at baseline. Its primary aim is to evaluate risk factors for the development and progression of cardiovascular disease in older adults (16). In 1989–1990, CHS investigators recruited persons from Medicare eligibility lists in Forsyth County, North Carolina; Sacramento County, California; Washington County, Maryland; and Pittsburgh, Pennsylvania. Additional black participants were actively recruited during a supplemental enrollment process in 1992–1993; they constitute 15% of CHS participants. To be considered eligible, persons had to be aged 65 years or more and not institutionalized, to expect to remain in the current community for 3 years or longer, to not be under active treatment for cancer, and to provide informed consent without requiring a proxy respondent.

Participants completed study visits at enrollment and annually through 1999; these visits included an interview, health questionnaire, physical examination, and collection of blood specimens. Participants were invited to return for an additional follow-up visit in 2005–2006, which included the same components. Follow-up contacts for functional status, cognitive function, health status, depressive symptoms, medication use, hospitalizations, and other health outcomes were conducted by telephone every 6 months. Details on the timing of health status measures are given in Web Table 1 (available at https://academic.oup.com/aje).

In this analysis, participants were included if they were aged 90 years or more at the time of enrollment (n = 41) or turned age 90 years during active follow-up (n = 1,848) and had data on 1 or more of the health measures. Baseline for this study began for each participant at the first study visit that occurred at or after age 90 years. Follow-up was censored at the end of the 25-year follow-up visits, on July 16, 2015. There were 38 participants alive who had not reached age 90 years by this time. Deaths were identified by consulting obituaries, medical records, death certificates, household contacts, and the National Death Index (17); 100% follow-up for ascertainment of mortality status was achieved.

Outcomes

Cognitive function was assessed using the Modified Mini-Mental State Examination (3MSE); scores range from 0 to 100, with scores less than 80 indicating cognitive impairment. Beginning in 1996, participants who did not attend the in-person visit were contacted by phone and asked to complete the Telephone Interview for Cognitive Status; the Informant Questionnaire on Cognitive Decline in the Elderly was administered to a proxy if needed. An internal validation equation has been developed to convert Telephone Interview for Cognitive Status and Informant Questionnaire on Cognitive Decline in the Elderly scores to 3MSE scores (18). Depressive symptoms were assessed by means of a modified 10-item Centers for Epidemiologic Studies Depression Scale; scores on this scale range from 0 to 24, and a score of 8 or higher indicates the presence of depressive symptoms (19). Participants were asked to rate their health as excellent, very good, good, fair, or poor. We considered participants’ health to be “good” if they selected good, very good, or excellent and “not good” if they selected fair or poor. Participants were asked to bring in prescription medication containers at all clinic visits and were asked annually about their medication use during semiannual phone calls. Participants were asked whether they had any difficulty walking a half-mile (0.8 km) or 5–6 blocks or walking up 10 steps. Participants were asked to report whether they had difficulty performing or were unable to perform any of 6 instrumental activities of daily living (IADL): using a telephone, shopping, preparing food, performing light household work, performing heavy household work, and managing money. Participants were asked to report whether they had difficulty performing or were unable to perform any of 6 basic activities of daily living (ADL): bathing, eating, dressing, using the toilet, getting out of bed or a chair, and walking around at home.

Coronary heart disease was defined as a history of myocardial infarction, angina pectoris, or revascularization; cerebrovascular disease was defined as a history of transient ischemic attack or stroke. “Any cardiovascular disease” included coronary heart disease, cerebrovascular disease, claudication, or heart failure. CHS participants were classified with regard to heart disease at baseline using a combination of hospital records and physician confirmation, and a CHS outcome-assessment committee adjudicated all incident cardiovascular disease events (20).

Statistical analysis

Descriptive statistics were stratified by sex and are presented as mean values and standard deviations or as numbers and percentages. Baseline measurements of clinic site, sex, race, self-rated health, depression score, 3MSE, prescription medications, difficulty walking half a mile, difficulty walking up 10 steps, IADL, and ADL were obtained on the basis of the first study visit held after the participant had turned 90 years of age.

Graphical summaries using stacked bar charts were computed at each half year of age throughout follow-up for self-rated health (Figure 1), cognitive function (Figure 2), and difficulty walking a half mile (Figure 3). Not all measures were assessed semiannually in all years, so missing values were imputed using the last-value-carried-forward method for measures assessed up to 1 year prior. The first nonmissing observation closest to each half year was used to produce the stacked bar charts. Percentages were computed relative to the total number of participants who reached age 90 years while under active surveillance. Beyond age 90 years, the percentages of persons who were missing, censored at last follow-up (i.e., last visit in 2015), and deceased were included.

Figure 1.

Longitudinal data on the self-rated health of women (A) and men (B) aged 90 years or more, Cardiovascular Health Study, 1989–2015. Bars represent participants with excellent, very good, or good self-rated health (orange); participants with fair or poor self-rated health (blue); participants with missing data (light gray); participants who were censored at the last follow-up (medium gray); and participants who were deceased (dark gray).

Figure 2.

Longitudinal data on the cognitive function of women (A) and men (B) aged 90 years or more, Cardiovascular Health Study, 1989–2015. Bars represent participants with preserved cognitive function (Modified Mini-Mental State Examination score ≥80, orange); participants with cognitive impairment (Modified Mini-Mental State Examination score <80, blue); participants with missing data (light gray); participants who were censored at the last follow-up (medium gray); and participants who were deceased (dark gray).

Figure 3.

Longitudinal data on the self-reported mobility limitations of women (A) and men (B) aged 90 years or more, Cardiovascular Health Study, 1989–2015. Bars represent participants with no difficulty walking half a mile (0.8 km) (orange); participants with difficulty walking half a mile (blue); participants with missing data (light gray); participants who were censored at the last follow-up (medium gray); and participants who were deceased (dark gray).

The incident cardiovascular and mortality event rates were calculated nonparametrically by dividing the number of events by total person-time at risk. Participants who died or were lost to follow-up were censored.

We next applied multiple imputation by chained equations for all outcomes. Using the imputed data, we fitted multilevel mixed-effects linear regression models separately for each outcome variable of interest: 3MSE score, depression score, number of medications, self-rated health (scored as 1 = excellent, 2 = very good, 3 = good, 4 = fair, and 5 = poor), IADL, and ADL. We fitted multilevel mixed-effects logistic regression models for the binary outcomes: difficulty walking half a mile and difficulty walking up 10 steps. We adjusted for sex and for grouped time period categories (defined on the basis of when health status was measured: 1989–1994 (referent), 1995–1999, 2000–2004, 2005–2009, and 2010 onwards) and modeled age linearly, centered at age 90 years. We evaluated the P value for trend across time periods. We assessed effect modification by age and sex using a robust Wald test. Second-order trends in centered age were also assessed in a similar fashion. The independent correlation structure was assumed for each cluster of patient data for continuous centered age at the time when the measurement was made. As a sensitivity analysis, we reran the multilevel analyses after excluding participants who reported having stayed at least 1 night in a nursing home since the last call or visit and those who used a proxy respondent, since these were exclusion criteria at enrollment.

All statistical analyses were performed and graphical displays created using STATA/IC 14.0 (StataCorp LLC, College Station, Texas) and R 3.2.3 (R Foundation for Statistical Computing, Vienna, Austria).

RESULTS

Of 5,888 CHS participants, 2,062 (35%) lived to age 90 years or more and 564 (10%) were alive as of July 16, 2015. Of the 2,062 participants who lived to age 90 years or more, 1,889 (92%) had available data on demographic factors, chronic disease, and health status; of these, 1,237 (65%) were women. About one-third of women and men had cognitive impairment, defined as a 3MSE score less than 80 (Table 1). Scores on the Center for Epidemiologic Studies Depression Scale indicated low levels of depressive symptoms. Women were taking a mean of 6.0 (standard deviation, 4.0) prescription medications, and men were taking 5.5 (standard deviation, 3.7). The majority of women (59%) and men (62%) reported being in good or better health. The majority of women (75%) and men (61%) reported having difficulty with 1 or more IADL; fewer reported difficulty with 1 or more ADL (52% and 37%, respectively). The majority also reported difficulty walking half a mile (76% of women and 59% of men), although fewer reported difficulty walking up 10 steps (61% and 43%, respectively) (Table 1).

Table 1.

Characteristics of Participants at the First Study Visit at Which They Were Aged 90 Years or More, Cardiovascular Health Study, 1989–2015

Characteristic	Possible No. of Participants	Women (n = 1,237)				Men (n = 652)				P Value
		No. Availablea	No. of Persons	%	Mean (SD)	No. Availablea	No. of Persons	%	Mean (SD)
Cognitive impairment (3MSE score <80)	1,273	718	269	37		376	114	30		0.02
Depression score (CES-D)	976	440			5.2 (4.9)	281			4.9 (4.7)	0.44
No. of prescription medications	1,889	1,182			6.0 (4.0)	625			5.5 (3.7)	0.007
Good, very good, or excellent self-rated health	1,889	1,222	721	59		635	394	62		0.20
≥1 IADL limitation	1,802	1,116	836	75		593	359	61		<0.001
≥1 ADL limitation	1,802	1,096	573	52		587	219	37		<0.001
Difficulty walking half a mile (0.8 km)	1,802	1,113	841	76		594	350	59		<0.001
Difficulty walking up 10 steps	1,802	1,084	657	61		584	249	43		<0.001

Abbreviations: ADL, activities of daily living; CES-D, Center for Epidemiologic Studies Depression Scale; IADL, instrumental activities of daily living; 3MSE, Modified Mini-Mental State Examination; SD, standard deviation.

^a Sample size available for measure.

After age 90 years, the event rates were high and similar by sex. The mortality rate was 19.0 per 100 person-years (95% confidence interval (CI): 17.8, 20.3) in women and 20.9 per 100 person-years (95% CI: 19.2, 22.8) in men (Web Figure 1). The cardiovascular event rates were as follows for women and men, respectively: 2.7 per 100 person-years (95% CI: 2.3, 3.3) and 2.4 per 100 person-years (95% CI: 1.9, 3.2) for stroke; 2.6 per 100 person-years (95% CI: 2.2, 3.2) and 3.1 per 100 person-years (95% CI: 2.4, 4.1) for myocardial infarction; and 6.0 per 100 person-years (95% CI: 5.3, 6.9) and 7.3 per 100 person-years (95% CI: 6.1, 8.6) for heart failure. The mortality rates increased with age in both women and men, whereas rates of the other events were relatively constant or increased and then decreased with age (Web Figures 2 and 3).

The high death rate is also illustrated by the steep drop in the proportion of participants with longitudinal data at age 90 years or more (Figures 1–3). Only a small proportion of surviving participants had missing data on cognitive function, ADL, and mobility (Figures 1–3). Censored participants were alive, but their follow-up was censored at the most recent CHS visit (ending on July 16, 2015). When followed forward over time, women and men reported being in good or better health well into their 90s and even early 100s (Figure 1). Fewer who completed the 3MSE had preserved cognitive function at age 90 years or more, and the proportion of those with preserved cognitive function was lower in woman than in men (Figure 2). Few nonagenarians and centenarians reported having no difficulty walking a half mile; the proportion was smallest among older women (Figure 3).

The estimates from the multilevel mixed models demonstrated significantly worse cognitive function, depressive symptoms, and self-rated health among women than among men, although the increase with age was similar for both sexes (P values for interaction of change after age 90 years with sex were 0.07, 0.78, and 0.41 for cognitive function, depressive symptoms, and self-rated health, respectively) (Web Table 2). Women also had worse functional status when it was measured by IADL and ADL limitations, although the rates of change were not different (P values for interaction of change after age 90 years with sex were 0.94 and 0.27 for IADL and ADL limitations, respectively). Interestingly, the rate of worsening mobility limitation was greater in men; men had a 1.5% (95% CI: 0.5, 2.5) increase in the risk of having difficulty walking half a mile and a 1.1% (95% CI: 0.0, 2.3) increase in the risk of having difficulty walking up 10 steps per year of greater age after age 90 years. Medication use was not different in men and women (P = 0.67), and the change after age 90 years did not differ by sex (P for interaction = 0.13). Cognitive function and all measures of functional status worsened with age among women and men, and the number of medications used decreased slightly with age. There were secular trends for some of the measures. Levels of depressive symptoms appeared to decline in recent years. The number of medications used increased substantially in all periods of the study. All of the measures of functional status were worse in later years.

Findings were largely similar in a sensitivity analysis excluding participants who had stayed at least 1 night in a nursing home since the last call or visit and those who had a proxy respondent (Web Table 3).

DISCUSSION

Persons surviving to age 90 years or more have, on average, considerable impairment in cognitive and functional health, yet they remain positive about their health status. Although women had worse cognitive function, more depressive symptoms, poorer self-rated health, and worse functional status than men did, the fact that the rates of change appeared similar above age 90 years suggests that survival to very old age is a sex equalizer. The age-related declines in cognitive and functional health are important considerations for planning for the future care needs of this population.

It is well established that women tend to live longer than men, but with greater disability (21). Advanced age appears to attenuate the sex gap in longevity (22); however, evidence points towards a disability sex gap that persists even until the end of life (22–24). Our study demonstrates that although the sex gap in disability and cognitive function persists in nonagenarians, the rate of change in physical and cognitive function with age in women and men is similar, and the gap is not widening. In fact, the significant age interaction with sex for the mobility outcomes suggests that the sex gap for mobility disability may attenuate in very old age. The 90+ Study demonstrated that women have a higher burden of disability and dementia, and the prevalence of both increased with age (12, 13). As in our study, sex was not associated with incident disability among those aged 90 years or more (25). However, our estimates of ADL difficulty or dependence were slightly lower than those in the 90+ Study, where investigators reported that 75% of women and 59% of men aged 90–94 years had difficulty with 1 or more ADL (12). In contrast, we observed a greater burden of disability than was found in a population aged 90 years or more in Stockholm, Sweden, where investigators reported that 28% of women and 15% of men aged 90–94 years had ADL disability (26)—although in that study, disability was defined as needing assistance with 2 or more ADL. These differences could also be due to the differences in populations and the differing periods of examination; participants in the 90+ Study were examined beginning in 2003, whereas those in the Swedish study were examined in 1993.

Cognitive function and all measures of functional status declined with age. Although this finding is not surprising, these data provide some estimates of the magnitude of this decrease. As the nonagenarian population in the United States continues to grow, it is important to consider these age-related declines when planning for the health-care resources needed to support this population. Interestingly, depressive symptoms and self-rated health did not worsen with age. These data suggest that although health may be declining, mental health and the perception of health are relatively consistent in older age. We observed apparent secular trends of worsening functional health, although these trends were attenuated in the analysis restricted to participants who had not stayed in a nursing home or used a proxy. It is possible that these estimates were affected by healthy participant bias, such that participants turning 90 in the earlier years were more likely to be healthier than their peers. If a period effect in worsening functional status is validated in other populations, this could also increase the long-term care resources needed to support older adults. Future research on the mechanisms driving this secular trend are needed. One potential explanation for this finding is the bourgeoning obesity epidemic. Obesity in early and midlife has been demonstrated to be an important predictor of late-life cognitive and functional decline (27).

Nonagenarians in our sample were on an average of 5.9 medications when they turned 90, and medication use was greater in the later visit years. The high burden of medication use is likely to be a consequence of the high prevalence of multiple chronic conditions and increased availability of medications to treat these conditions. Although increased awareness and treatment of disease may be beneficial for some older adults, the balance of benefits and potential harms of multiple-drug regimens in nonagenarians is often unclear. Many evidenced-based guidelines for the management of disease are not tailored to account for persons with multiple chronic conditions, so people with high morbidity are at increased risk of polypharmacy (28). Of concern is the high burden of morbidity, disability, and cognitive impairment in this highly medicated population, which could increase their vulnerability to adverse drug events, falls, and worsening physical or cognitive functioning (29).

The primary strength of this study is that we provided a comprehensive description of trajectories of health and function among nonagenarians in a representative sample of community-dwelling older adults in the United States. However, our study had some limitations that should be considered when interpreting the results. Although the CHS used a community-based sample and had limited study dropout, it is likely that those participants in the worst health were more likely to have missing data, and this could have biased our results towards a more optimistic view of nonagenarian health. Therefore, we used multiple imputation to account for missing data in our analysis of sex, age, and period effects. Additionally, these measures were all self-reported by the participant or his or her proxy, and there could be reporting bias or measurement error (30). This may be of special concern for self-reported ability to walk half a mile or climb 10 stairs if participants do not routinely complete these tasks. There could also have been learning effects or differential misclassification by visit type (in-person vs. telephone), which may have biased the period effects, as telephone visits were more frequent in the later visit years. There was a small number of black participants (15%) achieving age 90 years in our sample, which precluded our ability to evaluate racial differences, and our findings may not be generalizable to nonwhite racial/ethnic populations.

In summary, we characterized changes in health status among nonagenarians in a community-based US sample. These estimates will be important for research and planning to meet the needs of this growing population and ameliorate secular trends in worsening health. Policies that accommodate the high burden of morbidity and disability among nonagenarians are needed to enable long-lived adults to make the most of their lives.

Abbreviations

ADL

activities of daily living

CHS

Cardiovascular Health Study

CI

confidence interval

IADL

instrumental activities of daily living

3MSE

Modified Mini-Mental State Examination.