Methodologic issues in injury case-control studies.

In this paper some methodological problems particularly relevant to case-control studies of injury are illustrated by reference to previous childhood injury case-control studies. In contrast to studies of disease, where 'person time' constitutes the observational experience of interest, in injury studies person time engaged in a particular activity to often more appropriate. The implications for the definition of the study base are discussed. The potential for hospital admission bias in injury case-control studies is considered along with potential strategies for avoiding it. The importance of errors in exposure measurement, including those arising from inappropriate induction time assumptions, are illustrated. Finally, the potential for bias resulting from the combination of etiologically unrelated injury outcomes into a single outcome measure is illustrated and discussed.