Because of the inadequacies of surveillance data for occupational disease in the United States, this study was undertaken to determine the feasibility of using physicians' and hospital records for occupational disease surveillance. The study was based on Rutstein's sentinel health event (occupational) classification and the use of sentinel physicians. Data were gathered on nonmalignant dermatologic and respiratory diagnoses prospectively at five sites and retrospectively at four of these sites. The combined approaches identified 49 patients whose illnesses were probably work-related. Because the prospective approach took less than 10% of the time for each case identified compared with the time required for the retrospective approach, and prospective data are more timely and easier to access, the prospective approach appears to be better.