Although it has been shown that protein-energy malnutrition is a predictor of adverse outcome in geriatric patients, it is unclear whether this is due to underlying disease or disability, or whether malnutrition is an independent outcome predictor. To clarify the predictive role of malnutrition, we analysed the 4.5-year mortality and living location follow-ups of 219 geriatric patients admitted to a geriatric assessment unit. Prevalence of anthropometric and serological malnutrition indicators were between 13.7% and 39.8% at hospital admission. In bivariate models, prealbumin, subnormal arm muscle area, and subnormal body weight were predictors of mortality and survival at home. On the other hand, albumin, transferrin, and triceps skin-fold thickness did not predict these outcomes. In multivariate models the hazard ratio (HR) of 4.5-year mortality remained significant with an HR of 1.8 (95% CI 1.3-2.6) for subnormal arm muscle area, and 1.6 (95% CI 1.0-2.6) for subnormal body weight. Prealbumin was the strongest serological outcome predictor (multivariate mortality HR 1.9, 95% CI, 1.3-2.8). In these models, subnormal cognitive function, impaired physical function, and creatinine clearance < 30 ml/min were also associated with increased mortality. Malnutrition did not predict hospital discharge location, but among patients discharged home, those with initial malnutrition had a decreased length of survival at home. Our findings indicate that certain protein-energy malnutrition indicators are independent risk factors predicting decreased length of overall survival and survival at home in geriatric patients.(ABSTRACT TRUNCATED AT 250 WORDS)