Purpose: To determine the relation of bleeding to prothrombin times and important remediable lesions in outpatients treated with warfarin. Patients and Methods: An inception cohort of 565 patients starting outpatient therapy with warfarin on discharge from a university hospital was assembled. Detailed records of outpatient prothrombin times were obtained for 103 of 130 case subjects with major or minor bleeding and for 117 control patients without bleeding. A nested casecontrol design was used to evaluate the association of bleeding with temporally related prothrombin times; odds ratios were estimated using multivariate logistic regression analysis to control for known predictors of major bleeding. The relation of bleeding to important remediable lesions was determined in all 130 cases of bleeding. Results: For each 1.0 increase in the prothrombin time-to-control ratio, the odds ratio for major bleeding during the week after a prothrombin time measurement increased 80% the odds ratio for minor bleeding increased 50%. These odds ratios were lower during the first month of therapy and higher thereafter. Bleeding was related to important remediable lesions in 49 of 130 cases (38%), but these lesions were unknown before bleeding in only 22 cases (17%). The mean prothrombin time rose sharply at the time of bleeding in patients without important remediable lesions, but not in patients with lesions. New, previously unknown lesions (including nine malignancies) were discovered in 20 of 59 case subjects (34%) with gastrointestinal bleeding or hematuria, but in only two of 71 case subjects (3%) with other bleeding (p < 0.001). Conclusions: Our results provide a valid quantitative basis for estimating the odds of bleeding in relation to the prothrombin time and the yield of diagnostic evaluation in patients with bleeding. © 1989 Reed Publishing USA.