Hospital records as a data source for occupational disease surveillance: a feasibility study.
To assess the feasibility of using hospital records for occupational disease surveillance and to evaluate the quality of the industry/occupation (I/O) information available in these records, the computer file of all discharge diagnoses from a large health maintenance organization during 1985 was reviewed. The frequencies of discharge diagnoses previously listed as Sentinel Health Events (Occupational), or SHE (O), were calculated and three possible SHE(O) diagnoses--lung cancer, bladder cancer, and toxic hepatitis--were selected for further review. Outpatient charts of patients discharged for each diagnosis were abstracted with regard to I/O information and the discharged patients were interviewed by telephone to obtain a lifetime occupational history. The accuracy of the I/O information obtained from the hospital chart was compared to that obtained by patient interview by number of digits matched on standard classification codes. The frequencies of matches for occupation and industry were greater for "usual" than for "last" categories with both cancer diagnoses, but were similar for "usual" and "last" categories with toxic hepatitis. To assess the proportion of each possible SHE(O) diagnosis that was related to workplace exposures, the I/O information obtained by interview was rated in a blinded fashion by an experienced occupational medicine physician. The highest probability ratings for work-relatedness were noted for lung cancer, primarily due to asbestos exposure. The results of this study suggest that hospital records can be used to identify possible SHE(O); if adequate I/O information is available, then work-relatedness can be assessed. However, the accuracy of I/O obtained from hospital charts is relatively low. The efficient and accurate collection of I/O information from hospital records will require the use of a simple, easily coded instrument to be routinely administered on admission.