Problems in a statistical study of disease based on death certificates.

Data obtained from death certificates are used to establish the frequency of diseases for vital statistics. In Morgan County, Georgia, 5,787 death certificates registered between 1927 and 1979 were examined for this study. Data referable to age, sex, race, and the cause of death were recorded, with primary emphasis on cancer. The cause of death in these certificates was based primarily on clinical diagnoses. Some certificates cited neoplasms in which the site of origin was not given, and in others the type of neoplasm was difficult to correctly code. In some, a tumor was recorded but was not coded. The format of these certificates was changed by the state several times during this period, apparently to improve their credibility. Review of these registered death certificates, used for vital statistics in both Georgia and Washington, indicates that the data would contribute no scientific information on the frequency of cancer.