Trying Times: Waiting to Learn What Is Happening Now in American Premature Mortality

Over time, the effect of social and health policies across populations, including multiple racial/ethnic groups, can be especially challenging to track. Long latency periods between shifts in health care delivery, broad environmental exposures, and outcomes including mortality often limit the inferences one can draw from what become complex observational studies of health.

In this issue of AJPH, Roy et al. (p. 530) present mortality data after 32 years of follow-up in the Coronary Artery Risk Development in Young Adults (CARDIA) study to offer insights in health trajectories relative to race/ethnicity, education, and common chronic health conditions. Overall, 5115 participants enrolled in 1985 to 1986 as young adults (aged 18–30 years; mean age = 25 years) in four diverse American locations: Birmingham, Alabama; Chicago, Illinois; Minneapolis, Minnesota; and Oakland, California. On the basis of years of potential life lost, the primary findings identified important differences in mortality rates and causes of mortality, by race/ethnicity and education, and supported education as an explanatory variable for racial/ethnic differences in mortality.

Fundamentally, this serves as a follow-up to an earlier report on the cohort’s mortality after 16 years of follow-up.¹ From the 1985 to 1986 enrollment period through 2001, predictors of early mortality then included male sex, Black race, diabetes, liver and kidney disease, cigarette smoking, low education level, and other factors relating to social distress. The original sampling strategy for CARDIA created a cohort with 40% completing 12 years of education or less, balanced across racial/ethnic groups. Relatedly, education in CARDIA did not have a specific effect on one race/ethnicity over another; an overrepresentation of mortality was found in low education groups in both non-Hispanic Black and non-Hispanic White cohort members. Through 2001, 55.1% of the deceased and 39.3% of the living cohort members had completed 12 years of education or less. Currently, after 32 years of follow-up, the cohort has gradually become more educated, with 21.0% overall having a high school education or less. Yet the absolute differences in mortality were similar: by 2017, 36.3% of the deceased had a high school education or less, compared with 19.7% of the living, supporting a lasting effect of education on mortality across the life span.

CONFOUNDING DEATH

Interestingly, this updated analysis offers a window into shifts in mortality from early to mid adulthood. Perhaps as expected, some changes in mortality reflect changes in disease exposures in different stages of life, as well as changes in disease incidence and prevalence over time. With most of the cohort now aging into their sixth decade of life, cancer and cardiovascular disease have jumped to the top two causes of death, constituting 33.8% of all deaths, an increase from 15.0% after the 16-year follow-up. At that time in 2001, the two leading causes of death in the cohort were AIDS and homicide. Taking a deeper look into each of those two diseases is informative. Of the 47 AIDS-related deaths now identified after 32 years of follow-up, 35 of these had occurred through 2001, by which time 127 cohort members had died. Thus, 27.6% of deaths in the initial follow-up period were due to AIDS, but since 2001, only 12 (5.4%) of the subsequent 222 deaths with known cause were due to AIDS. This shift in mortality reflects a cohort that has been followed up through nearly the entire arc of AIDS, from the first description of HIV and its initial treatments to its current state as a potentially manageable chronic disease. Despite the passage of time, in both the 16- and the 32-year follow-up reports, AIDS remained the most common cause of death for White men (50.0% through 2001 and 27.8% through 2017), and homicide remained the most common cause of death for Black men (28.8% through 2001 and 19.3% through 2017, by which time mortality due to homicide and cardiovascular disease became equal). Clearly, stark differences in mortality causes and rates by race/ethnicity persist, with shifts by stage of life and age (Figure 1).

FIGURE 1—

Proportionate Mortality in the Coronary Artery Risk Development in Young Adults (CARDIA) Study Among the Top Six Underlying Causes of Death as Reported Through 2001 and 2017, by Race/Ethnicity

Note. Black = non-Hispanic Black; CVD = cardiovascular disease (includes coronary heart disease, stroke, and other heart disease); injury = unintentional injury; White = non-Hispanic White.

In the current CARDIA study, education is considered an explanatory variable or mediator in relations between race/ethnicity and mortality. These findings support many other studies that consistently identified education as a strong determinant of health in multiethnic American populations with regard to early- and late-life morbidity, mortality,² and cognitive aging.³ Years of education is likely an overly simplified metric. It does not capture the quality of education received, which historically has been variable across communities, and is among other especially strong determinants of health in racial/ethnic minorities.⁴ It is also recognized that education may be representative of many more complex psychosocial and health variables across a life span,⁵ including diet, exercise, income and other markers of social status, and health beliefs and attitudes. Similarly, race itself represents a host of deep-seated social and health inequities, including direct and indirect effects of racism. Thus, for this study, given the complex relations between race/ethnicity, education, and long-term health, a host of important residual confounders may explain these relations, including when diverse geographic cohorts are considered in aggregate.

THE LIMITATIONS OF TIME

The changes in mortality in CARDIA highlight the critical importance of considering time in cohort studies: time in history when cohorts are enrolled and studied, amount of time the cohort spends facing chronic diseases and their determinants, and time in life the cohort faces these exposures. Three separate effects (age, period, and cohort effects) are collectively termed “secular trends” and are important to consider in understanding changes in mortality over time.⁶ Most obviously, age effects in mortality are naturally expected to have increased risk and potentially different causes with advancing age. Period effects, or the calendar time during which outcomes occur, affect mortality as evidenced by AIDS in this cohort. Cohort effects represent the period when an individual was born, or entered a study, and provide an index for potential generational effects.⁶ Examples include major shared societal experiences, including episodic ones (e.g., wars and economic depressions), and general educational and social experiences in early life and health care across the life span.

A final way to reflect on time is to question whether enough time has passed to understand a cohort’s mortality when many remain alive. The primary measure of years of potential life lost can be helpful beyond simple proportions of mortality and differences by race/ethnicity and education. At the oldest, the cohort is aged to about 62 years, which precedes a host of disorders associated with aging, including (more) cancer and cardiovascular disease, as well as neurodegenerative disorders, all of which disproportionately affect disadvantaged and minority populations. Having enrolled participants in young adulthood, the CARDIA study does offer a distinct advantage over many studies of aging that often begin enrollment in the sixth or seventh decade of life—an age after a significant number have died from competing causes before becoming elderly. Although we will have to wait for the CARDIA cohort to age further, trends in mortality might change as the cohort ages (particularly those who begin to outlive their life expectancy). New treatments for neurodegenerative disorders in particular could dramatically increase life expectancy, as has occurred recently in cancer care. Making matters even more complicated is that American life expectancy continues to change, declining since 2014, particularly among young and middle-aged adults⁷—ages the CARDIA cohort is beginning to live past.

However, without trying time too much more, one must ask: How much longer do we have to wait to understand and act on what is happening in health care right now?