Measuring the accuracy of vital status data in cohort studies.

To measure the quality of vital status data in a retrospective cohort study of mortality among former servicemen of the Vietnam Conflict era, test subjects of independently determined vital status were included among study subjects during vital status ascertainment procedures. This allowed for differentiation between vital status "unknown" and incorrect assignment of vital status, and enabled measurement of the quality of both live and deceased vital status data. Four parameters based on sensitivity and specificity were used to express the quality of vital status data. The deceased specificity rate was 100 per cent, the deceased sensitivity rate was 95.7 per cent, the live specificity rate was 98.5 per cent, and the live sensitivity rate was 95.4 per cent. Using models of misclassification, the estimated death rate was found to be most sensitive to changes in the deceased specificity rate, indicating that emphasis should be given to minimizing incorrect ascertainment of truly alive subjects as deceased when developing vital status ascertainment procedures.