OBJECTIVE: The systematic use of regular hospital utilization reviews has proved costly, particularly in countries with short average lengths of stay. This study examines the performance of three tests based on routinely collected data when screening inappropriate hospital days. DESIGN: The Appropriateness Evaluation Protocol was used to set the gold standard. The first screening test was simply based on the comparison of an observed length of stay with a target value; the second test additionally made allowances for surgical and intensive care procedures while the third added the amount of required nursing workload to these data. SETTING: The neurology and general surgery departments of a Swiss university hospital. PARTICIPANTS: Every day of care for all inpatients stays was reviewed to assess the appropriateness of hospital use and submitted to the screening algorithm (9,000 hospital days). MAIN OUTCOME MEASURES: Receiver-operating characteristics curves were compared to optimize the performance of the screening tests. The best test was applied to all units of the hospital and rates of inappropriate days were computed using a Bayesian approach. RESULTS: The first and the second tests have a sensitivity of 66-80% and a specificity of 66-67%. Nursing workload data yield no significant improvement of the screening test. An unbiased estimate of the rate of inappropriate days may be computed. CONCLUSION: The present study provides some evidence that a screening approach is useful, feasible and efficient for detecting inappropriate hospital days. [Authors]