The object of this article was to validate the predictive value of Charlson's comorbidity index for the prediction of short-term mortality or morbidity in elderly people. The design was a cohort study comparing survival and hospitalization in institutionalized elderly people with different levels of comorbidity at baseline. The setting was 16 Flemish nursing homes for the elderly. The subjects were 2,727 inhabitants of which full data were available for 2,624. The outcome measures were hazard ratios resulting from Cox regression analysis, comparing 6 months survival in patients with moderate and a high level to low level of comorbidity. Odds ratios resulting from multiple logistic regression analysis comparing the occurrence of at least one hospitalization during the follow-up period in surviving patients of the same groups. Mortality adjusted for age group was significantly increased in patients with a moderate (HR = 2.00) and even more in those with a high level (HR = 3.62) of comorbidity. Hospitalization was more frequent in both groups (OR = 1.54 and 2.19, respectively), with statistical significance only being reached for the highest group. Adjustment for age, gender, mobility status, and disorientation did not change the general picture. Charlson's comorbidity index is a predictor of short-term mortality in institutionalized elderly patients and, to a lesser extend, also of hospitalization. These results support its use as a measure for introducing comorbidity as a covariable in longitudinal studies with a geriatric population.