Are the marginal returns of coronary artery surgery smaller in high-rate areas?

Background Population-based rates of surgery vary within and between health-care systems, causing concern that case selection is less appropriate in high-rate areas. This inverse relationship has not been shown with appropriateness criteria generated by expert panels. We applied a trials-based measure of the potential survival benefit of coronary artery bypass graft surgery (CABG) to patients in a provincial registry, to determine the relationship between survival gains and rates of CABG. Methods We did a population-based retrospective review of linked registry and administrative datasets. 5058 patients in the linked datasets underwent isolated CABG in Ontario between April 1, 1992, and March 31, 1993. Potential survival benefit of surgery was scored with an algorithm derived from a published overview of trials comparing CABG to medical treatment, analysed by county and by referral regions. Findings Overall, case selection was appropriate whether assessed clinically (96•3% had either severe disease as judged on the coronary arteries affected or moderate to severe angina) or on the basis of survival benefit scores (94•0% predicted to obtain moderate or high benefit). There was significant variation in benefit scores across referral regions (p<0•001). Benefit scores correlated inversely with county surgical rate (r=20•49, p<0•005) and the proportion of low-benefit cases increased with rates (r=0•50, p<0•005). Referral regions served by high rate surgical centres had lower mean benefit scores. Interpretation Most patients undergoing CABG in Ontario are in the high-survival benefit category. Surgery is defensible for patients with low survival benefit on the grounds of symptom relief, but the proportion of cases with low benefit rises with higher local rates of surgery. The inverse relationship between surgery rates and appropriateness of case selection may be better understood as diminishing marginal returns for specific outcomes with rising local use of procedures.