Sources of hospital-level variation in major bleeding among patients with non-ST-segment elevation myocardial infarction: A report from the national cardiovascular data registry (NCDR)

Background-Major bleeding has received increasing attention as a target for quality improvement in care of patients with acute myocardial infarction. However, little is known about variation in bleeding across hospitals and whether variation is attributable to quality of hospital care, treatments, or case mix. Methods and Results-We characterized hospital variation in major bleeding events (an absolute hemoglobin drop =4 g/ dL, intracranial hemorrhage, retroperitoneal bleed, or transfusion) among 99 200 patients with non-ST-segment elevation myocardial infarction in the National Cardiovascular Data Registry (NCDR) Acute Coronary Treatment and Intervention Outcomes Network Registry-Get With the Guidelines (ACTIOM Registry-GWTG) between January 2007 and June 2010. A total of 9566 (9.6%) patients experienced a major bleeding event during hospitalization. The median of the estimated distribution of major bleeding rates across hospitals was 9.4% (interquartile range, 7.5%-11.7%), with some hospitals having bleeding rates >2.3 times higher than others (10th-90th percentile, 6.1%-14.2%). Multivariable hierarchical models revealed that differences in case mix explained 19.2% of the hospital variation in bleeding complications, where anticoagulation and antiplatelet strategies explained an incremental 9.9% and 6.8%, respectively. Together, 32.3% of hospital variation in major bleeding rates was attributable to differences in patient case mix and identifiable differences in treatment strategies in patients with non-ST-segment elevation myocardial infarction. Conclusions-In-hospital major bleeding rates varied widely across hospitals. Although patient factors and treatments explained less than one third of hospital-level variation, "70% of bleeding variation remains after adjustment. A better understanding of causes for substantial hospital-level bleeding variations is needed to help target high-risk patients or practices and to optimize care.