To determine whether emergent readmissions within 30 days of discharge are potentially preventable, we prospectively studied all readmissions to the medical service of a university teaching hospital during a 4-month period. The 327 readmissions, including 42 patients who were readmitted more than once during the study period, accounted for 12% of medical admissions. Nearly 75% of readmissions were due to previously diagnosed medical conditions; complications of drugs or therapeutic procedures caused 29%, and patient compliance contributed to 11%. Three reviewers judged 28 (9%) readmissions to be potentially preventable. These 28 readmissions constituted 1% of all admissions to the medical service and did not result in any fatalities, indicating that the quality of care was generally very good. One third of the potentially preventable readmissions were thought to be related to medical system failures, one third to an unfulfilled hope that the patient would improve after discharge, and one third to other suboptimal judgements in evaluation or treatment. Of the 28 potentially preventable readmissions, 89% occurred within 10 days of discharge. We conclude that potentially preventable readmissions would nearly always be detected by review of readmissions within 10 days of discharge and that many may be amenable to systematic interventions such as standardized predischarge assessment and better coordination of post-discharge follow-up.