Malnutrition is often over-seen in elderly acute medical patients. It is a need for a simple and robust screening tool.The aim was to evaluate, with regard to validity, the Mini Nutritional Assessment-Short Form (MNA-SF) as a screening tool for malnutrition in elderly acute medical patients.This is an observational study where a nurse's scoring of MNA-SF is compared to comprehensive assessment by a clinical nutritionist (gold standard). Sixty-nine patients aged 70 years and older and admitted to a general medical department in year 2000 and 2001 were included.The mean MNA-SF score was 7.8 (SD +/- 2.88, range 2 to 12). Fifty-one patients (74%) scored positive for malnutrition or risk of malnutrition (MNA-SF < 11), whereas only 21 (30%) were scored to have malnutrition by the nutritionist. Sensitivity of the MNA-SF was 1.0 and specificity 0.38, giving 0.57 correctly classified subjects. Best subset logistic regression showed BMI < 23 to be the only item explaining the gold standard. When using BMI < 23, 32 (46%) subjects screened positive for malnutrition (sensitivity 0.86, specificity 0.71), giving 0.75 correctly classified subjects.When screening elderly acute medical patients in general wards for malnutrition or risk of malnutrition, the MNA-SF have a high sensitivity and can be useful. The sole use of BMI < 23 may be equally effective, but will give no information leading towards an explanation. We recommend that a score of BM < 23 should be followed by MNA-SF when the aim is to identify poor nutritional status in elderly acute medical patients.