Computerized population-based hospital discharge data in New Jersey offer new opportunities for surveillance of serious work-related injuries. This database was evaluated for its potential in identifying selected injuries that occurred at work during 1985 and 1986. Hospital discharge data were compared with data collected by telephone interview of discharged patients. A total of 1,575 unique hospital discharge records for the selected injuries included finger amputation (1,041), thumb amputation (209), crush injury of the lower limb (208), toxic effects of heavy metals (69), and eye burns (48). Of 809 study subjects sent letters, 445 (55%) could be contacted and 289 (36%) were interviewed for the study. Sixty-one percent (175) said their injury was work related. A comparison was made between self-reported injury at work, and the presence of workers' compensation payer codes on the discharge database. The agreement beyond chance (Kappa) was 0.78 (95% CI = 0.67, 0.89). The sensitivity of this indicator of work relatedness was 83%; specificity was 98%. These data suggest that workers' compensation payment on the hospital discharge database may be a good to excellent proxy indicator of the work relatedness of these injuries. However, this proxy indicator will underestimate the number of work-related injuries by about 20%. Only 11% of hospital discharge records had external cause of injury codes (E-codes), which reduces the utility of the database for understanding the causal mechanisms of work-related injuries.