With growing awareness of medical fallibility, researchers need to develop tools to identify and study medical mistakes. We examined the utility of hospital readmissions for this purpose in a prospective case-control study in a large academic medical center in Israel. All patients with nonelective readmissions to 2 departments of medicine within 30 days of discharge were interviewed, and their medical records were carefully examined with emphasis on the index admission. Patient data were compared to data for age- and sex-matched controls (n = 140) who were not readmitted. Medical records of readmitted and control patients were blindly evaluated by 2 senior clinicians who independently identified potential quality of care (QOC) problems during the index admission. Inhospital and late mortality was determined 6 months after discharge.Over a period of 3 months there were 1988 urgent admissions; 1913 discharges and subsequently 271 unplanned readmissions occurred (14.1% of discharges). Readmissions occurred an average of 10 days after discharge, and readmitted patients were sicker than controls (mean, 4.3 vs. 3.3 diagnoses per patient), although their length of stay was similarly short (3.4 +/- 2.8 d). Analysis of all readmissions revealed QOC problems in 90/271 (33%) of readmissions, 4.5% of hospitalizations. All were deemed preventable. Interobserver agreement was good (83%, kappa = 0.67). Among matched controls, only 8/140 admissions revealed QOC problems (6%, p < 0.001) (k = 0.77). The preventable readmissions mostly involved a vascular event or congestive heart failure; they occurred within a mean of 10 +/- 8 days of the index admission, and their inpatient mortality was 6.7% vs. 1.7% among readmissions that had no QOC problems (odds ratio, 4.1; 95% confidence interval, 1.0-16.7). The main pitfalls identified during the index admission included incomplete workup (33%), too short hospital stay (31%), inappropriate medication (44%), diagnostic error (16%), and disregarding a significant laboratory result (12%). In many patients more than 1 pitfall was identified (mean, 1.5 per patient). Risk factors for preventable readmission include older age and living in an institution (p < 0.05). Almost two-thirds of the readmitted patients with QOC problems were discharged after spending 2 days or fewer at the hospital. In conclusion, unplanned readmissions within 30 days of discharge are frequent, more prevalent in sicker patients, and possibly associated with increased mortality. In a third of readmitted patients a QOC problem can be identified, and these problems are preventable. Thus, readmission may be used as a screening tool for potential QOC problems in the department of medicine. Routine monitoring of all readmissions may provide a simple cost-effective means of identifying and addressing medical mistakes.