Methodological and conceptual problems in the study of mortality in psychiatry.

The study of mortality has a long tradition in Psychiatry. The specificity required to define risks associated with particular disorders and to monitor the effects of changing patterns of treatment necessitates rigorous attention to methodology. The strengths and limitations of particular study designs and how these must be considered when interpreting results are reviewed. Sampling of the study population involves deciding whether to define a group retrospectively or prospectively, the type of psychiatric population to be utilized (eg inpatients or outpatients), and whether cases are to be chosen randomly as to diagnosis or with selection for certain profiles. Ascertainment of death is critical in mortality studies. Once the index population is defined, cases lost to final ascertainment significantly weaken the study. Cause of death as determined by coroner's verdict is readily obtained but of questionable accuracy. Results may be influenced by the method of psychiatric assessment and diagnosis, especially whether case record or interview data is used, and whether multiple or only single diagnoses are determined. With computerization improved statistical analysis are now possible. There is no perfect mortality study design. Any particular design involves a trade-off of advantages and disadvantages.