Cisneros M, Sevenster M, Forsberg T, Lee ME, Nathan S, Liu P, Mukta D, Chang PJ, Spencer KT, Blair JEA
Objective: CIN is the most common non-cardiac complication following coronary angiography, however riskstratification for CIN can be cumbersome. We hypothesize that a computer-based Kidney Injury Risk Tool (KIRT) would assess contrast-induced nephropathy (CIN) risk accurately and may out-perform standard clinical estimation. Methods: This was a prospective study of all-comers undergoing coronary angiography at a single institution. KIRT is based upon an established risk model and risk factors were derived from the electronic medical records using phenotyping rules. Operator reported and KIRT-derived risk factors were compared against adjudicated ground truth obtained by a blinded investigator through chart review and accuracy of the risk model outputs were compared. Operator’s assessment of risk without risk model vs. KIRT assessment were compared and analyzed (Wilcoxon-test, and Spearman's correlation). Results: A total of 132 patients consented for the study, 127 patients were included. KIRT-derived risk factors out-performed or matched the operator-reported for most riskfactors (sensitivity and specificity>0.86). KIRT output accuracy was higher than operator output: 79% vs. 76%. Mean operator-estimated CIN risk was lower than KIRT's estimate: 9% vs. 17% (P<0.001, paired Wilcoxon test), and held true for both high and low risk patients. Conclusion: KIRT has high accuracy in determining individual risk factors for CIN and identification of high-risk patients, and operator-based risk for CIN over-estimated risk compared to KIRT