Changing Disease Prevalence, Incidence, and Mortality Among Older Cohorts: The Health and Retirement Study

Abstract

Background

This article investigates changes in disease prevalence, incidence, and mortality among four cohorts of older persons in the Health and Retirement Study.

Methods

We examine two cohorts initially aged 51 to 61, whom we call younger cohorts, and two older cohorts aged 70 to 80 at the start of observation. Each of the paired cohorts was born about 10 years apart. We follow the cohorts for approximately 10 years.

Results

The prevalence of cancer, stroke, and diabetes increased in later-born cohorts; while the prevalence of myocardial infarction decreased markedly in both later-born cohorts. The incidence of heart disease, myocardial infarction, and stroke decreased among those in the later-born older cohort; while only the incidence of myocardial infarction decreased in the later-born younger cohort. On the other hand, diabetes incidence increased among those in both later-born cohorts. Death rates among those with heart disease, cancer, and diabetes decreased in the later-born cohorts. The declining incidence of three cardiovascular conditions among those who are over age 70 reflects improving population health and has resulted in stemming the increase in prevalence of people with heart disease and stroke.

Discussion

While these results provide some important signs of improving population health, especially among those over 70; trends for those less than 70 in the United States are not as positive.

Research reporting disease trends for older populations has been fairly consistent in showing an increase over recent decades in the prevalence of major diseases such as heart disease, cancer, stroke, and diabetes in older populations (1–5). Because increased prevalence stems from either an increase in the incidence of disease or an increase in the length of time lived with disease, it is important to determine how incidence and mortality among people with diseases have changed, in order to identify the determinants of changes in observed prevalence as well as the implications of the trends for future health change (6,7). Increases in incidence would imply worsening health of the population. Conversely, decreases in incidence may reflect the success of interventions to prevent and delay disease onset; while increases in the length of time lived with disease would imply better survival among those with disease resulting from increasing effectiveness of medical treatment; however, this could also have the seemingly negative effect of increasing disease prevalence in a population.

Prevalence of diseases in an older population can reflect the onset of conditions occurring at any time in the life cycle; incidence of diseases reflects onset in a given time period among those who do not have the condition and thus can better characterize recent conditions. Recent studies of hospitalizations examining heart disease and stroke incidence among selected samples of older persons have indicated that incidence of these conditions may have begun to decline after the year 2000 (8–14). The incidence of myocardial infarction has been reported to be declining since the 1990s in a multi-site sample (15). Surveillance, Epidemiology, and End Results (SEER) data also suggest some decline in overall cancer incidence (16). This would suggest that after a long period of increase in the prevalence of many conditions, we may be seeing changes in the process of disease onset that may reflect improvement in the health of older persons. On the other hand, numerous studies have shown an increase in diabetes incidence over time and project further increase in the future (17,18).

Due to advances in early detection and treatment as well as improvements in modifiable risk factors, reductions in mortality from major chronic diseases have been observed in the United States over past decades. Cardiovascular mortality has been decreasing since the late 1960s and has been the major contributor to increasing life expectancy among older Americans (19). From 2000 to 2010, age-adjusted mortality decreased by about 30% for heart disease and almost 40% for stroke (20,21). Deaths attributed to acute myocardial infarction also showed a substantial downward trend over recent decades, reflecting improvements in the quality of care (22). Cancer mortality has also declined in recent decades, but at a slower rate than cardiovascular mortality (20). From 1980 to 2014, mortality from all cancers combined declined by 20%, and the overall decline has been driven by decreasing death rates from specific cancers (eg, lung cancer and breast cancer) (23). Falling death rates among individuals with diabetes have also been observed. National data from 1988 to 2015 showed death rates declined by 20% every 10 years among U.S. adults with diabetes, a faster decline than among those without diabetes (24).

It is important to note that there is evidence that the very significant declines in heart disease, stroke, and diabetes mortality have slowed, or been eliminated in the last couple of years (19–21). For instance, the annual decline in heart disease mortality slowed from 3.9% during the 2000–2010 period to 1.4% during the 2010–2013 period (19). In fact, very recently the national vital statistics showed significant increases in age-adjusted rates for many leading causes of death, with the exception of cancer (25).

We examine changes over about a decade in disease onset and survival across cohorts of the Health and Retirement Study (HRS): two cohorts representing change at the ages 51 through 71 of this nationally representative panel and two representing change at ages 70 through 90 or 91. We examine a set of major chronic diseases in this analysis as disease is the focus of medical intervention and of prevention. A cohort rather than a period formulation of the data is used as this most appropriately examines health change over the lifecycle of a set of individuals born at different times and is the best way to link time changes in prevalence, incidence, and mortality to the aging process within individuals. It is better to examine changes in the lives of real, rather than synthetic cohorts as is often done when age differences are examined within cross-sectional samples; earlier health change experienced by individuals affects the current health state and the likelihood of later change within cohorts. Our analysis allows us to examine change over about a 10-year span in disease prevalence across cohorts along with changes in disease incidence and mortality among those with diseases within these cohorts.

Methods and Data

Data

Our source of data is the HRS, a longitudinal panel study of Americans over age 50 and their spouses, which was begun in 1992 and has collected data approximately every 2 years up to the present. When the study began in 1992, it included sample members 51–61 years of age; in 1993, a sample of persons 70+ was added; in the ensuing years sample members were added from missing cohorts and the sample was refreshed with 51- to 56-year-olds every 6 years. This structure of the data is central to how we select our comparison cohorts. To represent persons as they approach old age and in the earlier older ages, we use data for two cohorts aged 51 to 61 from the HRS, the 1931 to 1941 birth cohorts who were first interviewed in 1992 (N = 10,645) and the 1943 through 1953 cohorts who were first interviewed at age 51 to 61 in 2004 (N = 7,896). Each of these cohorts is then followed for 10 years or until they were aged 61 to 71: Cohort 1, 1992/1993, 1994/1995, 1996, 1998, 2000, and 2002; Cohort 2, 2004, 2006, 2008, 2010, 2012, and 2014. We refer to these as younger cohorts. Figure 1 shows the age of the cohort at each observation and the year of observation.

Figure 1.

Cohorts in Health and Retirement Study: cohort years of birth, years of observation, age at observation.

To represent older persons, we use cohorts aged 70 to 80 at the start of observation. The first cohort was born in 1913–1923 and first interviewed in 1993; a second cohort represents those born in 1924–1934, and we began their observation in 2004. These cohorts are followed for 11 and 10 years: Cohort 1, 1993, 1995, 1998, 2000, 2002, 2004; Cohort 2: 2004, 2006, 2008, 2010, 2012, 2014. The first cohort is followed until they are aged 81 to 91 and the second cohort is observed until age 80 to 90. We refer to these as older cohorts.

At the first interview, and at each later interview, respondents report the presence of diseases diagnosed by a doctor. We examine five conditions that represent leading causes of death and important causes of disability for older persons: cancer (excluding non-melanoma skin cancer), heart disease, myocardial infarction (heart attack), stroke, and diabetes. We assume that once respondents report a disease, they always have it and it is included in the prevalence estimate until they die. This is shown in Figure 2.

Figure 2.

Prevalence of diseases (%) by age in two younger cohorts (age 51–71) and two older cohorts (age 70–90).

Disease incidence reflects the reporting of a disease onset after the baseline observation among those who did not have the condition at the beginning of the interval. People who already had a disease at baseline are excluded from the incidence analysis. We assume that disease onset occurs at a person’s age at the midpoint of the survey interval. People who do not have a disease are at risk for incidence and once they report a condition they are no longer at risk. In our figures, we show 2-year moving averages for incidence in order to make the pattern clearer. We examine mortality among people with a disease using mortality records collected by the HRS from the Social Security Administration, from designated survivors, and from interviewer reports. Mortality reporting in the HRS has been shown to be effectively complete (26).

Statistical Methods

We examine the cohort differences in disease prevalence at the first time of observation using logistic regressions for each disease, where being in the later cohort is included as a dummy variable with age and sex. For incidence of disease, we use hazard models to examine interval age-specific incidence over 10 or 11 years among those who do not have a condition at the beginning of the interval in equations with age, sex, and cohort. Each equation includes a variable indicating the exposure time in the interval. For death, we examine death using a hazard model at each later wave over the 10 or 11 years among those who have a disease at the first wave. The equations include age, sex, and an indication of being in a later-born cohort.

Results

Before we describe our results in detail, we should note that the direction of the effect when comparing later-born and earlier-born cohorts indicated by all three parameters (ie, for prevalence, incidence, and mortality) was the same for the younger and older age groups. In looking at the detailed results, we first examine cohort differences in prevalence of major diseases; then we will examine incidence, and mortality among those with disease. Prevalence results are shown in Figure 2 and Table 1. Prevalence significantly increases in later-born cohorts for 3 out of 5 diseases in the younger age group and 4 out of 5 diseases in the older age group. The prevalence of MI decreases in the later-born cohorts for both the younger and older age groups (Table 1). The decline in the relative likelihood of having had a myocardial infarction was quite large in both the later-born cohorts, that is, the younger and older cohorts: 68% and 67%, respectively. On the other hand, cancer prevalence is higher in the later-born cohorts. The odds ratios indicate a 29% increase in cancer among those in their 50s and 60s and a 46% increase among those 70 to 80. The relative increases in diabetes are even larger in the later-born cohorts: 48% and 67%. There was also an increase in the likelihood of having had a stroke in the later-born cohorts: 29% among the younger and 15% among the older. There was no change in the prevalence of heart disease in the younger cohorts but an increase in the likelihood of 9% in the older group.

Table 1.

Odds Ratio or Hazard Ratio (with 95% Confidence Interval) for Effect of Being in Later-Born Cohort on Prevalence, Incidence, and Death Among Those with Disease—Older and Younger Cohorts

		Younger Cohorts	N	p	Older Cohorts	N	p
		Effect of Being Born in 1943–1953 Compared to 1931–1941			Effect of Being Born in 1924–1934 Compared to 1913–1923
Prevalence	Cancer	1.29 (1.12–1.48)	15,386	.0003	1.46 (1.33–1.60)	10,438	<.0001
	Heart disease	1.02 (0.93–1.13)	15,386	.6892	1.09 (1.02–1.18)	10,409	.0161
	MI	0.32 (0.26–0.38)	15,229	<.0001	0.33 (0.27–0.40)	10,265	<.0001
	Stroke	1.29 (1.06–1.58)	15,387	.0106	1.15 (1.02–1.31)	10,324	.0264
	Diabetes	1.48 (1.33–1.64)	15,385	<.0001	1.67 (1.52–1.83)	10,452	<.0001
Incidence	Cancer	1.10 (0.98–1.23)	13,867	.1021	1.00 (0.90–1.11)	8,016	.9415
	Heart disease	0.94 (0.85–1.03)	12,826	.1560	0.88 (0.81–0.95)	6,612	.0016
	MI	0.42 (0.35–0.50)	13,937	<.0001	0.52 (0.45–0.59)	8,810	<.0001
	Stroke	0.92 (0.79–1.07)	14,295	.2857	0.70 (0.63–0.78)	8,624	<.0001
	Diabetes	1.54 (1.40–1.70)	12,979	<.0001	1.21(1.08–1.37)	7,785	.0012
Mortality	Cancer	0.69 (0.51–0.94)	933	.0189	0.82 (0.73–0.92)	1,626	.0011
	Heart disease	0.76 (0.62–0.93)	1,972	.0076	0.88 (0.81–0.96)	3,198	.0022
	MI	0.92 (0.64–1.31)	656	.6267	0.80 (0.62–1.02)	459	.0722
	Stroke	0.89 (0.64–1.23)	484	.4718	0.87 (0.76–1.01)	846	.0665
	Diabetes	0.77 (0.63–0.94)	1,872	.0088	0.77 (0.69–0.85)	1,889	<.0001
	None of the diseases	0.85 (0.73–0.98)	11,117	.0229	0.85 (0.79–0.93)	4,985	.0002

Now we examine changes in incidence. Cancer incidence did not change significantly in either the younger or older cohorts (Figure 3 and Table 1). Diabetes incidence increased in both the younger and older cohorts: with the hazard increasing 54% and 21%, respectively. The incidence of the three cardiovascular-related diseases declined in the later-born cohort of older persons; the decrease in the hazard ratio for heart disease was 12%, 48% for myocardial infarction, and 30% for stroke. The hazard of stroke and heart disease did not change among the younger cohort while the hazard of having a myocardial infarction decreased by 58% among the later-born younger cohort.

Figure 3.

Incidence of disease (% with onset among those without condition at beginning of interval) by age (2-year moving averages) in two younger cohorts and two older cohorts.

The hazard of dying among those with none of these diseases at baseline was reduced by 15% for the later-born cohorts in both age groups (Table 1). Among the younger old, the hazard of dying was reduced for those with cancer (31%), heart disease (24%), and diabetes (23%); older cohorts experienced significant reductions in the likelihood of dying among those with the same three conditions: cancer (18%), heart disease (12%), and diabetes (23%) (Figure 4 and Table 1). There was no significant cohort change in the hazard of dying among those with a stroke or myocardial infarction for the younger or older cohorts, although the decrease in both is close to significant among the older group.

Figure 4.

Mortality by years after interview among those with disease at baseline for two younger cohorts and two older cohorts.

Discussion

Understanding the meaning of trends in health in an older population characterized by the presence of chronic disease requires knowledge of several indicators of the process of health change. The routinely observed increase in the prevalence of disease among the older population has left researchers unable to clearly state whether health among the older population is improving, or not, over time. In this nationally representative data set where health change has been observed over more than a 20-year period, we observe increases in the prevalence of three diseases: cancer, diabetes, and stroke among people over age 50 and up to age 90. We also observe an increase in heart disease among those over 70. These increases in prevalence, however, result from a set of different processes across the diseases.

Cancer incidence does not change significantly across these cohorts but cancer mortality is lower among the later-born cohorts. We can assume that the increase in cancer prevalence reflects the fact that people are living longer once they have been diagnosed with cancer; an indicator of improving treatments and possibly earlier diagnosis. The increase in diabetes prevalence reflects both an increase in incidence and a decrease in mortality. The increase in incidence is largely the result of increasing obesity, while the decrease in mortality probably reflects better diagnosis and treatment (27). Both of these trends work to increase the prevalence of diabetes in the population; one reflects improvement in health and one deterioration. Trends for stroke portray an even more complex pattern of change. The incidence decreases across the older cohorts, indicating health improvement; however, decreases in mortality among this age group, while not significant, tended to work to increase the prevalence in the population. At the younger ages, the changes in incidence and mortality both indicate lower levels in the more recent cohort, but neither change is significant.

The overall heart disease picture is relatively positive with no significant prevalence increase among the young old and only a modest increase among the old; the major decreases in both prevalence and incidence of myocardial infarction indicate major health improvements for more recent cohorts. Finding this in a national population sample is an important addition to the findings from more localized studies. Declines in incidence of myocardial infarction may reflect emphasis on prevention including increasing treatment and effectiveness of treatment of cardiovascular risk factors such as hypertension and high cholesterol and reductions in smoking.

Researchers examining trends in disability among the old have tended to find more positive trends among the older segments of the group and less improvement among the younger old, especially those born in the baby-boom (28). We find differentials by age in the pattern of significant change with more reduction in the incidence of all three cardiovascular conditions, heart disease, myocardial infarction and stroke, among the older cohorts.

There are some limitations to our analysis. We use self-reports from respondents of their being told by a doctor that they have a condition. These may be sensitive to use of medical care and changes in medical care usage could affect self-reports. Reporting of diseases can also be sensitive to changes in diagnosis and reporting to patients. One potential result from our analysis, a large increase in stroke prevalence among the later-born younger cohort with small decreases in mortality and incidence, could reflect more reporting of stroke in this group. The increasing education of cohorts over time could result in more informed patients who are more likely to report diseases. Unfortunately, we cannot control for these changes. It would have been preferable to follow two individual cohorts from their 50s through their 90s, rather than having four cohorts. It is possible that the outcomes for earlier-born and later-born cohorts both represent the 20-year period of observation and we cannot assume change in the future for the younger cohort will be reproduced.

While we recognize the limitations, we feel the results of our analysis of data from a unique nationally representative sample further our understanding of health trends in the older population. Trends are influenced by a set of complex processes. In general, researchers have not recognized the feedback role of declining mortality in affecting the subsequent health of older populations. The outcome we have sought to achieve, lower mortality, can feed back into the population and produce indicators which appear to indicate worsening health. Adding incidence of disease to our analysis of a national population database has allowed us to see some very positive trends in cardiovascular health and to understand why cancer might be increasing in prevalence.

The study of individual diseases also adds to our understanding of health trends and needs for intervention. Some trends are clearly adverse, for example, the rise in diabetes and prevention needs to target this condition. Prevention of myocardial infarction appears to be progressing, but cancer is not being prevented, and this will be a challenge for the future.

Funding

This paper was published as part of a supplement sponsored and funded by AARP. The statements and opinions expressed herein by the authors are for information, debate, and discussion, and do not necessarily represent official policies of AARP.

Conflict of Interest Statement

None reported.