Objective: To help develop criteria to identify older patients suitable for admission to medical short-stay units, by determining predictors of length of stay (LOS) of 3 days or less. Methods: The data were prospectively collected from consecutive older patients admitted from the emergency department of a university hospital to an acute geriatric medicine service. Data included active medical diagnoses, the Modified Barthel Index (MBI), the Timed Up and Go (TUG) test, and demographic information. Logistic regression was used to model the probability of LOS of 3 days or less (short LOS). Results: Among 2036 patients discharged alive from hospital (mean age, 82 years; median LOS, 7 days), 398 had a short LOS (median, 2 days), while 1638 had a long LOS (median, 9 days). In logistic regression analysis, the main independent predictors of short LOS were an MBI score > 15/20 (OR, 2.98; 95% CI, 1.97-4.49), ability to perform the TUG test (OR, 2.08; 95% CI, 1.34-3.24) and absence of delirium (OR, 2.66; 95% CI, 1.56-4.54). Patients without infection, anaemia, gastrointestinal disorder and stroke were also more likely to have a short LOS in multivariate analysis (all P < 0.05). Conclusion: Preserved function, measured using the MBI and TUG, and the absence of delirium are strong predictors of short LOS. In conjunction with early, skilled clinical evaluation, these criteria could be used to select older patients presenting to the emergency depart- ments for admission to short-stay units.THE INCREASING NUMBER and proportion of older persons is having profound consequences and far-reaching implications, especially for health care systems. Older persons visit the emergency department and are admitted to hospital in greater numbers than their proportion in the total population.1 Due to escalating demands on a limited health workforce and budget, pressures on emergency departments and hospitals are growing, 2 leading to increasing scrutiny and evaluation of systems of care.