Send to

Choose Destination
Med Care. 2016 Jun;54(6):623-31. doi: 10.1097/MLR.0000000000000536.

Practice Variation Among Hospitals in Revascularization Therapy and Its Association With Procedure-related Mortality.

Author information

*Departments of Quantitative Health Sciences and Outcomes Research, Cleveland Clinic †Harrington Heart and Vascular Institute, Case Western Reserve University/University Hospitals Case Medical Center ‡Department of Outcomes Research, Cleveland Clinic, Cleveland, OH §Duke Clinical Research Institute, Duke University, Durham, NC ∥Department of Quantitative Health Sciences, Cleveland Clinic ¶Center for Healthcare Research and Policy, Case Western Reserve University/MetroHealth Medical Center, Cleveland, OH.



While substantial practice variation in coronary revascularization has been described and deviation from clinical practice guidelines has been associated with worse outcomes, the degree to which this is driven by flawed decision making and/or appropriate deviation associated with comorbid conditions is unknown. We evaluated heterogeneity in procedure use, and the extent to which hospital-level practice variation is related to surgical mortality.


We analyzed data on 554,563 inpatients undergoing either percutaneous coronary intervention or coronary artery bypass grafting at 391 centers in 6 states. Procedure-specific risk models were developed based on demographics and comorbidities, allowing for differential effects of comorbidities for each sex. For each patient, the revascularization procedure that minimized predicted probability of inhospital mortality was designated as the model-preferred procedure.Hospital-level discordance rates-the proportion of cases in each hospital for which the opposite from the model-preferred procedure was performed-were calculated. Hierarchical linear models were used to analyze the relationship between HDRs and hospital-level risk-standardized mortality ratios (RSMRs).


Comorbidities and demographics alone explained between 68% and 86% of overall variation in inhospital mortality (corresponding C-statistics of 0.84-0.93). The mean (SD) HDR was 26.3% (9.6%). There was a positive independent association between HDRs and inhospital mortality, with a 10% increase in HDR associated with an 11% increase in RSMR (P<0.001).


Variance in procedure use according to model preference was strongly associated with worse outcomes. A systematic approach to incorporating comorbidity as part of the decision-making process for coronary revascularization is needed.

[Indexed for MEDLINE]
Free PMC Article

Supplemental Content

Full text links

Icon for Wolters Kluwer Icon for PubMed Central
Loading ...
Support Center