[Mortality analysis: when is single evaluation of the basic cause of death allowable, when should multi-causality be assessed?].

Data quality is often a critical point in mortality studies. The purpose of the present report is to present criteria for assessing the value of death-certificate-based mortality studies. For this purpose all 57,454 Swiss death certificates of the year 1979 were analysed. Reliability of the diagnosis listed on the death certificate was investigated by comparing for each case of a linked sample of 12,478 deaths the cause of death with medical information available from the hospital record. Retrieval rates (percentage of cases for which the given diagnosis appears in both registries) were calculated for the primary diagnoses named in each data set. These can be considered as measures of reliability of diagnoses. The graphs given indicate a high reliability for cancers and accidents. Reliability was lower for other causes of death such as cardiovascular diseases, diabetes mellitus, rheumatic diseases. Restriction to the primary cause of death can be accepted for most cancers, accidents and violent deaths. For other causes of death, decisions must be made individually and multicausal analysis may be indicated. In addition, knowledge of the reliability of the diagnoses of interest is necessary for the interpretation of results derived from death certificate-based mortality studies.