When one discusses the dynamic changes in human health over time, one innately conceptualizes time from the three different, but related perspectives - age, period, and cohort. To determine their separate contributions to health outcomes, age-period-cohort analyses have been used for the past 80 years. This commentary aims to provide some insight into this analytical method by distinguishing the concept of time in terms of composition and context. To demonstrate, the author uses hypothetical nested data structures of age-period-cohort analyses in the two types of individual-level data, i.e., repeated cross-sectional survey and longitudinal data on the same individuals. The conceptual distinctions between composition and context have profound implications of hypothetical interventions in age-period-cohort analyses. Age is a compositional variable, and a hypothetical intervention to change age is at the individual level. By contrast, both period and cohort are contexts, and thus two distinct types of hypothetical interventions can be envisaged to examine their contextual effects. On a related issue, the author also discusses manipulability of time. Although time is a significant context in biomedical science, it is not the only context. In this commentary, context is proposed to be classified into three fundamental dimensions - relational, spatial, and temporal. Inattention to the contextual triad leads to a biased and precarious knowledge base for public health action, and the continuing flow of performance over time is an intrinsic component of improving our understanding of multilevel causal inference in the new era of eco-epidemiology.
- Age-period-cohort analysis
- Multilevel analysis
ASJC Scopus subject areas
- Health(social science)
- History and Philosophy of Science