The chance of a person with osteoarthritis of the knee receiving a knee replacement is highly variable. To understand better the reasons for this variation, all practicing orthopaedists in Indiana were surveyed about their management of severe knee osteoarthritis and their perception of tricompartmental knee replacement as a therapeutic option. Their perceptions of indications and outcomes of knee replacement were compared with the self reported annual number of patients for whom they performed (or referred to other surgeons for) tricompartmental knee replacements. A completed survey was returned by 220 (79%) of the 280 orthopaedists surveyed; analyses were limited to the 188 respondents who had cared for at least one patient with osteoarthritis of the knee in the prior 2 weeks (mean = 13). These surgeons reported performing (or referring patients for) a mean of 31 knee replacements in the prior year (SD 45, median 21, range 0-480 knee replacements). There was strong agreement (> 95%) among respondents for seven (21%) of 33 surgical indications and contraindications, and more general agreement (> 60%) for 21 (64%). In the five factors (15%) for which there was disagreement, there was no consistent relationship between opinions and self reported knee replacement performance rate. Surgeons reporting more knee replacements had significantly higher estimates of pain relief and functional improvement following surgery, and lower estimates of prosthesis infection and failure rates. When all responses were considered together, four decision factors correlated independently with the performance of more knee replacements, but these four factors explained only 24% of the variation in self reported knee replacement performance.(ABSTRACT TRUNCATED AT 250 WORDS)