Little is known about how physicians make decisions when the evidence is incomplete or controversial. While thrombolysis improves survival following acute myocardial infarction (AMI), conflicting evidence exists as to any specific agent's superiority, particularly if cost-effectiveness is considered. Using a Bayesian hierarchical model, the authors examined the patient, physician, and hospital characteristics that are related to the decision-making process concerning the choice of thrombolytic agent in a prospective registry of 1,165 AMI patients receiving thrombolysis. Tissue plasminogen activator (t-PA) was administered to 432 patients (31.8%) and streptokinase (SK) to the remainder. The presence of an anterior infarction, a previous myocardial infarction, low blood pressure, a cardiologist decision maker, younger age, and receiving treatment within six hours after the start of symptoms were independent predictors of receiving t-PA. The levels of importance that physicians accorded to these patient characteristics differed according to their practicing institutions. Generally, they followed evidence-based medicine and reasonably targeted high-risk patients to receive the more expensive t-PA. However, they also preferentially treated younger patients, where only a small absolute advantage appears to exist.