Sociology Plus
Cohort Study Definition
A cohort study is a research design that follows and re-investigates a group of people who share some common characteristic, tracking them over time to observe how outcomes develop. Most strongly associated with medical research, it is treated by social scientists as a special case of the longitudinal study.
Origins of the Term
The name draws on two traditions. A cohort was a Roman army unit of roughly 600 men; the medical profession borrowed the classical term to describe studies of groups of patients undergoing particular treatments. In both senses the underlying question is the same: how many survived.
The Core Design
In its strongest health-research form, a cohort study begins with two matched groups. After initial assessment, the experimental group receives a measurable treatment and its outcomes are followed through repeated testing, ideally until recovery; the control group receives no treatment, so that any emerging differences can plausibly be attributed to the intervention. The design thus combines features of an experiment (though matching cannot control every intervening factor) with those of a longitudinal study, in which causal claims are supported by temporal order — the treatment precedes, and so can be argued to produce, the recovery.
Medical and Social Science Usage
Doctors mainly speak of cohort studies, while social scientists use “cohort study,” “longitudinal study,” “trend study,” and “time series” almost interchangeably; the terminology owes more to disciplinary convention than to fundamental differences of method. Medical usage typically carries stronger notions of causality, whereas social-science usage is more exploratory and historical and involves more variables. Strictly, cohort and longitudinal studies re-study the same people, while trend studies and time series draw fresh samples at each point in time. The design’s popularity in medicine reflects its simplicity, its compatibility with basic statistics, and easier sample retention: medical studies are shorter than lifelong social projects, patients have a self-interested reason to stay in treatment, hospitals supply recruits and record-keeping infrastructure, and the status gap between doctor and patient encourages compliance in a way academic researchers cannot match. It is telling that most large social-science cohort studies, such as the UK National Child Development Study (NCDS), originated in public health.
Variations
Retrospective (historical) designs use existing records to assemble a sample sharing prior characteristics — the Australian Gulf War Veterans Health Study, set up a decade after the conflict, was almost uniquely successful in recruiting every veteran and nearly all controls. Concurrent or prospective designs use larger samples without a control group, sometimes adding members as the study proceeds: the US Multicenter AIDS Cohort Study began with 4,954 men in 1985 and recruited a further 668 between 1987 and 1991.
Strengths and Weaknesses
The design’s chief attraction is its time dimension, which reveals interactions among variables far better than one-off cross-sectional surveys. Time-series studies of successive independent samples are cheaper and easier to run, but some variation in their findings may stem from sampling itself; they are nonetheless greatly preferable to comparing respondents of different ages within a single sample, a practice that can seriously mislead — as in the flawed picture of social mobility drawn by the Nuffield College study. Against these strengths stand high costs and the long wait for results: research on how childhood experience shapes adult disease or social outcomes can take a lifetime, during which priorities shift and better techniques emerge. Attrition is the major defect. After fifty years, three successive host organisations, and six data-collection agencies, the NCDS retains only about two-thirds of its original sample, and the Aberdeen Birth Cohort Study of people born in 1921 could contact only six in ten members seventy years on. If those who drop out differ socially from those who remain, findings cease to be generalisable to the wider population.