The aim of this study was to evaluate factors predicting length of stay in hospital of geriatric patients. Study participants were 402 patients (183 males and 219 females) consecutively admitted to four geriatric wards located in Chieti, Perugia, Pescara and Prato, Italy. Information on potential predictors of length of stay was collected; in particular, we assessed the presence and severity of specific chronic medical conditions, level of physical function, cognitive status, and depressive symptoms. Moreover, information on family and social support was obtained. In general, participants were old, often cognitively impaired and physically disabled. Average length of stay ranged from 9.4 +/- 3.3 days (Perugia) to 14.1 +/- 7.2 days (Chieti), and was statistically different across centers (p < 0.001). None of the specific medical diagnoses was associated with different length of stay. However, higher comorbidity score (p < 0.001), living alone (p < 0.01), lower MMSE score (p = 0.03), and poor functional status (p = 0.05) were all associated with longer length of stay. When these variables were included in a multivariate model predicting length of stay, differences between centers were no longer statistically significant. Findings of this study show that specific medical diagnoses are not adequate instruments to estimate length of stay in geriatric units. Other assessment systems based on extension of the social network, comorbidity, and the cognitive as well as the functional level need to be developed.