1.1 What epidemiology studies

Three words in the definition carry the subject. Distribution asks who becomes ill, where and when; answering it is descriptive epidemiology. Determinants asks why: which exposures, behaviours, genes or circumstances change the chance of becoming ill. Answering it is analytic epidemiology, and almost all of this course is about it. Populations says the unit of interest is a group, not a patient. A clinician asks whether this patient has the disease; an epidemiologist asks how often the disease occurs in a group and whether it occurs more often in one group than another.

The word exposure is used for any characteristic whose effect on disease is being studied: a drug, an occupation, smoking, a gene, a water supply. The outcome is usually the occurrence of a disease or death, coded \(D\) (disease) and \(\bar D\) (no disease); exposure is coded \(E\) and \(\bar E\). Both may be more than two-valued, but the binary case carries every idea and is used throughout.

The statistical problems of the subject come from the fact that in most epidemiological work the investigator cannot assign the exposure. Nobody can be told to smoke for twenty years or to drink from a contaminated well. The data are observational: people chose, or were given by circumstance, their exposures, and whatever led them to it may also affect their disease. Separating the effect of the exposure from the effects of everything that travels with it is the central task, and the methods of the later sections (standardisation, stratification, the Mantel–Haenszel procedure, regression) are all tools for it.