The pitfalls of the denominator: towards an accurate estimation of a population at risk.

Observations in epidemiological, operational or sociomedical research often need to be related to a denominator to obtain a measure of rate. Defining the denominator for a particular population at risk is a difficult task. The basis of this study was a census track study combined with a survey questionnaire which included every household in a part of Nes municipality, in Norway. This meant it was possible to define the population at risk nearly perfectly. The results were compared with official records and the records of general practitioners. Some of the pitfalls in deciding on the population at risk are discussed: the newborn and the dead, the migrants, the institutionalized, the incomplete census tract, patients in contact with other physicians, the non-attenders, including 'the forgotten'. Eight different approaches to the estimation of the denominator are briefly commented on.