Commentary: The longitudinal perspective and cohort analysis

Cohorts are closed populations defined and bounded by their timepoints of entry, often but not necessarily at birth. In this issue, a classic article by Kermack, McKendrick and McKinley reprinted from the Lancet of 1934 illustrates an early use of cohort analysis. Such analyses follow successive generations of entrants through the life course. The object is to link the pattern of specified outcomes to the particular previous experience defined by membership of a generation. The outcome of interest to Kermack et al. was overall mortality at successive ages; their significant discovery was to demonstrate the potentially large and continuing contribution of experience in the earliest years to rates of death throughout the lifetime of each generation. This newly-invented approach cast bright light on the longitudinal perspective of health and disease through the life course. It was a latecomer in the history of relevant methods. Halley’s life table of 1693 is credited as the first valid numerical approach. A major figure in astronomy and mathematics, he used it to challenge the claim of astrologers that certain years in a man’s life were predictably hazardous. 1 William Farr, a founder of quantitative epidemiology, made good use of the life table method 2 and other approaches to longitudinal analysis. During 40 years from 1839 as Compiler of Medical Statistics for England and Wales, Farr invented the cohort life table, a method widely used in our times to follow the institutional careers of mental patients. He also devised a cohort study i.e. longitudinal follow-up of outcomes in a closed population defined by a specific point of entry or exposure (not that he named them as such). In the latter instance, he constructed a retrospective analysis of records of admissions and deaths in a mental institution. With that, he demonstrated positive benefit from John Connolly’s ‘moral treatment’, a therapeutic revolution of its time in psychiatry. In the institution Connolly directed, mortality was decidedly less than in others. The advent of cohort analysis, however, had to wait until the later 1920s in the actuarial world (for two papers presented to the Faculty of Actuaries by Derrick and by Davidson and Reid 3 ), and until 1930 in epidemiology (if one can be allowed to skip a scarcely instructive 1927 paper by Korteweg). In the latter year, Andvord published a paper in which he examined the outcome of tuberculosis in successive generations. 4 The next significant