Unexplained variation in length of stay (LOS) following acute myocardial infarction (AMI) has been observed among American hospitals. We explored this phenomenon in the universal hospital care system of Ontario, Canada's largest province, analyzing general hospital discharge abstracts for all patients with a primary diagnosis of AMI. Case homogeneity was increased by excluding inter-hospital transfers, in-hospital deaths, patients with revascularization during the index admission and patients with severe comorbid conditions. This left 11,411 records of patients in 187 hospitals from April 1, 1990 to March 31, 1991. The mean length of stay was 9.9 days with standard deviation of 3.8. Available patient and hospital characteristics explained only 12% of the individual variation in LOS. Interinstitutional variation remained highly significant after controlling for patients' characteristics within the 87 hospitals admitting more than 50 cases per annum; these hospitals accounted for 84% of the eligible provincial admissions. The grand mean length of stay for 87 hospitals was 10 days, ranging from 6.6 to 12.9 days. Stepwise multiple linear regression analyses showed that lower caseload was associated with an increased length of hospitalization. Thus, despite Ontario's uniform system of hospital funding and medical insurance, a large amount of unexplained variation in length of stay exists for patients hospitalized with AMI, affecting thousands of bed-days per annum.