Abstract
A comparative study was conducted to assess empirically the suspected bias associated with occupational health studies based only on information obtainable from employers' records. Standardised mortality ratios were extracted from five occupational cohorts for which extensive follow up had been conducted and were compared with proportional mortality ratios based on deaths known (KTE) and unknown (UTE) to the employer, and on total deaths. For the five cohorts examined the suspected bias for the cause of death categories all cancer, respiratory cancer, heart disease, and accidents was generally found to be minimal unless the number of UTE deaths was in excess of 75%. Furthermore, a strong inverse relationship was observed between the percentage of UTE deaths and the level of agreement between proportional mortality ratios based on KTE deaths and total deaths. In partial support of these findings it was also observed that workers who left the industry in three of the five cohorts subsequently experienced similar overall mortality compared with those who continued employment. These findings should serve as a rough indication of the potential value of estimating disease risks from KTE deaths and serve as a guide for other investigators using limited data.