Patients at risk of malnutrition and related morbidity and mortality can be identified with the Nutritional Risk Index (NRI). However, this index remains limited for elderly patients because of difficulties in establishing their normal weight. Therefore, we replaced the usual weight in this formula by ideal weight according to the Lorentz formula (WLo), creating a new index called the Geriatric Nutritional Risk Index (GNRI). First, a prospective study enrolled 181 hospitalized elderly patients. Nutritional status [albumin, prealbumin, and body mass index (BMI)] and GNRI were assessed. GNRI correlated with a severity score taking into account complications (bedsores or infections) and 6-mo mortality. Second, the GNRI was measured prospectively in 2474 patients admitted to a geriatric rehabilitation care unit over a 3-y period. The severity score correlated with albumin and GNRI but not with BMI or weight:WLo. Risk of mortality (odds ratio) and risk of complications were, respectively, 29 (95% CI: 5.2, 161.4) and 4.4 (95% CI: 1.3, 14.9) for major nutrition-related risk (GNRI: <82), 6.6 (95% CI: 1.3, 33.0), 4.9 (95% CI: 1.9, 12.5) for moderate nutrition-related risk (GNRI: 82 to <92), and 5.6 (95% CI: 1.2, 26.6) and 3.3 (95% CI: 1.4, 8.0) for a low nutrition-related risk (GNRI: 92 to < or =98). Accordingly, 12.2%, 31.4%, 29.4%, and 27.0% of the 2474 patients had major, moderate, low, and no nutrition-related risk, respectively. GNRI is a simple and accurate tool for predicting the risk of morbidity and mortality in hospitalized elderly patients and should be recorded systematically on admission.