Self-reported questionnaires are frequently used to assess health status in epidemiological studies. The Cornell medical index is one such tool used to determine the presence of physical and psychiatric illness but its accuracy and value have been questioned. In this study we have assessed the ability of the CMI to predict health status in two separate patient populations (n = 101, 88) by comparison to a structured medical assessment based on the SENIEUR protocol by two physicians. There was good agreement between medication use reported on the CMI and on medical assessment (k = 0.79; CI: 0.70-0.88). Accuracy of prediction of the CMI for specific medical conditions was good 89-99%. A threshold score from the CMI was not predictive of health as determined by the SENIEUR protocol. In our older populations, we conclude that the CMI accurately predicted health status. The determination of normal health by a threshold score was poorly predictive of heath status. Self-reported medication use was the best predictor of health status.