Background: Insurance companies use medical information to classify the mortality risk of applicants. Adding genetic tests to this assessment is currently being debated. This debate would be more meaningful, if results of present-day risk prediction were known. Therefore, we compared the predicted with the observed mortality of men who applied for life insurance, and determined the prognostic value of the risk assessment. Methods: Long-term follow-up was available for 62,334 male applicants whose mortality risk was predicted with medical evaluation and they were assigned to five groups with increasing risk from 1 to 5. We calculated all cause standardized mortality ratios relative to the Dutch population and compared groups with Cox's regression. We compared the discriminative ability of risk assessments as indicated by a concordance index (c). Results: In 844,815 person years we observed 3,433 deaths. The standardized mortality relative to the Dutch male population was 0.76 (95 percent confidence interval, 0.73 to 0.78). The standardized mortality ratios ranged from 0.54 in risk group 1 to 2.37 in group 5. A large number of risk factors and diseases were significantly associated with increased mortality. The algorithm of prediction was significantly, but only slightly better than summation of the number of disorders and risk factors (c-index, 0.64 versus 0.60, P,0.001). Conclusions: Men applying for insurance clearly had better survival relative to the general population. Readily available medical evaluation enabled accurate prediction of the mortality risk of large groups, but the deceased men could not have been identified with the applied prediction method.