Cohort studies have great potential in healthy ageing research

Most studies of healthy ageing have used surveys of older adults. Although these data provide snapshots of the health status of older individuals, any study of patterns or determinants of health in older age is inevitably hampered by mortality selection. Such a selection could alternatively be called health-care selection. For example, if a health-care system is unable to prevent deaths among individuals with ill-health, then the people who are in the sample tend to be healthy. Conversely, if a health-care system succeeds in preventing deaths, then the sample tends to be unhealthy. Without adjusting for mortality selection, the first scenario might cause a population to be labeled as attaining healthy ageing while, in fact, it has not.

Although macro-level statistics explicitly account for both morbidity and mortality, such as disability-adjusted life years (DALY), such a concept is not defined at the individual level as required by research on determinants of health status. One approach to correct the mortality selection is to explicitly model the mortality process to derive suitable weights, which requires data on causes of death. However, the best way to mitigate this selection bias is to use a preselection population, as Tuija M Mikkola and colleagues have done in their recent paper, published in The Lancet Healthy Longevity.¹ In their study, they analysed longitudinal data from the Helsinki Birth Cohort Study, which comprises 13 345 individuals born in Helsinki, Finland between 1934 and 1944, who were still alive in 1971. The study modified the concept of healthy ageing by including mortality and coined a new term, healthy survival, defined as being alive and free of chronic diseases that are likely to have an impact on functioning. This is a valuable concept that can potentially be adopted in other studies. Many longitudinal studies in the family of Health and Retirement Studies recruited respondents from middle age onwards. For example, individuals were aged 45 years or older in the China Health and Retirement Longitudinal Study.² Studying individuals when at younger age in these cohort studies and following them through to older age will permit the study of healthy survival.

The most important contribution of the study by Mikkola and colleagues,¹ which is difficult to replicate in the aforementioned longitudinal studies that started in middle age, is that it assessed the association of variables at birth or in early childhood with health outcomes in older age. The authors found that lower maternal BMI in late pregnancy was associated with healthy survival in men. Additionally, years of education and socioeconomic position in childhood were associated with healthy survival in both men and women, and shorter height at 7 years was associated with healthy survival in women. Such results support Barker’s hypothesis, which states that early-life circumstances have long-term effects.³ Many previous studies have shown that childhood conditions affect adult health, education, marriage, and employment (economic) outcomes,⁴ but few cohorts have followed up people into older age. Previous studies that examined the effects of childhood conditions on outcomes in older age have generally used self-reported childhood health^5,6 or hunger experiences,⁷ with childhood information provided in retrospective life history studies. Although life history surveys are useful supplements to existing longitudinal studies that started from midlife onwards, and have the advantage of not needing to wait 70 years, the gold standard is birth cohorts followed over the life course. As respondents in other birth cohorts worldwide, such as the 1946 British national birth cohort study,⁸ enter into older age, we can expect further research output in this area.

The third innovative aspect of the study by Mikkola and colleagues¹ is the linking of participants with administrative records. The ability to link administrative data is especially important for cohort studies that do not have frequent follow-ups or that have high attrition rates typical of birth cohort studies into older age. In this analysis of the Helsinki Birth Cohort, dates of death were obtained from the National Death Register maintained by Statistics Finland, and dates of diagnoses of chronic diseases were obtained from the Finnish Care Register for Health Care, which has high coverage of the Finnish population. The health register has the added value of alleviating the common problem of respondents misreporting their health conditions due to unawareness or cognitive impairment. However, the administrative linkage can not completely overcome the underdiagnosis issue, given that clinical diagnosis depends on patient contact with a doctor. An ideal but more expensive design would be to conduct clinical assessments of all surviving cohort members. The authors used clinical assessment data that were available for a subset of the cohort, but because the subset comprised of survivors, healthy survival could not be studied in isolation for this subset.

Birth cohort studies are a highly valuable resource for studying life-cycle aspects of population health. Maintaining such studies requires dedication and investment; however, research such as this Article shows that eventually the investment is worthwhile.