We conducted a retrospective cohort study based on autopsy reports of 311 patients who underwent full postmortem examinations from January 1, 1999, to December 31, 2002. Clinically unsuspected diagnoses were categorized as follows: class I, major clinically unsuspected diagnoses that were responsible for death; class II, major clinically unsuspected diagnoses that were not directly responsible for death but if left undiagnosed may have resulted in patient death; and class V, no clinically unsuspected diagnoses. Two multivariate analyses were performed using 4 variables to predict class I diagnoses. Both analyses included the variables sex, race, and age; the fourth variable included umbilicus pannus size or body mass index (BMI). Only BMI (P = .006) and umbilicus pannus size (P = .037) were independent predictors of class I diagnoses. Obese patients were 1.65 times more likely (relative risk, 1.65) to have a class I diagnosis than the normal weight and underweight groups combined (confidence interval, 1.04-2.64). Patients with obese-level BMIs seem to be at increased risk for clinically significant unsuspected diagnoses compared with underweight and normal weight populations.