Association of Variation in Contrast Volume With Acute Kidney Injury in Patients Undergoing Percutaneous Coronary Intervention.

Importance: Acute kidney injury (AKI) after percutaneous coronary intervention (PCI) is common, morbid, and costly; increases patients' mortality risk; and can be mitigated by limiting contrast use. Objective: To examine the national variation in AKI incidence and contrast use among US physicians and the variation's association with patients' risk of developing AKI after PCI. Design, Setting, and Participants: This cross-sectional study used the American College of Cardiology National Cardiovascular Data Registry (NCDR) CathPCI Registry to identify in-hospital care for PCI in the United States. Participants included 1 349 612 patients who underwent PCI performed by 5973 physicians in 1338 hospitals between June 1, 2009, and June 30, 2012. Data analysis was performed from July 1, 2014, to August 31, 2016. Main Outcomes and Measures: The primary outcome was AKI, defined according to the Acute Kidney Injury Network criteria as an absolute increase of 0.3 mg/dL or more or a relative increase of 50% or more from preprocedural to peak creatinine. A secondary outcome was the mean contrast volume as reported in the NCDR CathPCI Registry. Physicians who performed more than 50 PCIs per year were the main exposure variable of interest. Hierarchical regression with adjustment for patients' AKI risk was used to identify the variation in AKI rates, the variation in contrast use, and the association of contrast volume with patients' predicted AKI risk.
Results: Of the 1 349 612 patients who underwent PCI, the mean (SD) age was 64.9 (12.2) years, 908 318 (67.3%) were men, and 441 294 (32.7%) were women. Acute kidney injury occurred in 94 584 patients (7%). A large variation in AKI rates was observed among individual physicians ranging from 0% to 30% (unadjusted), with a mean adjusted 43% excess likelihood of AKI (median odds ratio, 1.43; 95% CI, 1.41-1.44) for statistically identical patients presenting to 2 random physicians. A large variation in physicians' mean contrast volume, ranging from 79 mL to 487 mL with an intraclass correlation coefficient of 0.23 (interquartile range, 0.21-0.25), was also observed, implying a 23% variation in contrast volume among physicians after adjustment. There was minimal correlation between contrast use and patients' AKI risk (r = -0.054). Sensitivity analysis after excluding complex cases showed that the physician variation in AKI remained unchanged. Conclusions and Relevance: Acute kidney injury rates vary greatly among physicians, who also vary markedly in their use of contrast and do not use substantially less contrast in patients with higher risk for AKI. These findings suggest an important opportunity to reduce AKI by reducing the variation in contrast volumes across physicians and lowering its use in higher-risk patients.