Although it is known that autopsies often disclose unexpected findings, few studies have been published that address the effect of institutional setting, selection bias, and length of hospitalization. Records of medical autopsies from 3 institutional settings were studied for discrepancies between clinical and autopsy findings. The settings were university hospital (n = 85); community hospital (n = 146); and private autopsy (n = 60), which were referred from various community hospitals and paid for by family members. The same prosector performed the autopsies in the community and private settings. The overall rate of major discrepancy that involved the cause of death was 17.2%. Factors that increased the likelihood of missed diagnoses were private setting (P = .0005), community setting (P = .02), and short hospital stay before death (P = .02). Additional major findings were present in 28.5% of autopsies. Length of hospital stay before death, institution, and selection bias all affect the rate of major unexpected findings in hospital-based autopsies.