A comparison of three prediction models for acute kidney injury requiring renal replacement therapy after coronary artery bypass graft surgery at St. Luke’s Medical Center, QC
Background: Acute kidney injury (AKI) following cardiac surgery is associated with increased post- operative morbidity and mortality. Scoring systems to predict acute kidney injury requiring renal replacement therapy (RRT) among patients undergoing cardiac surgery have been developed to assess risk pre-operatively and assist clinicians on the management post-operatively. These predictive models have good discriminative value. The significance of this study is to determine the most predictive model by comparing the 3 models to be able to be utilized and applied in our setting.
Objective: To compare the Cleveland Score by Thakar, Simplified Renal Index (SRI) by Wijeysundera, and Simplified Bedside Risk tool by Mehta in predicting acute kidney injury requiring renal replacement therapy among patients who underwent cardiac surgery.
Design, Setting, and Participants: Cross sectional analytic study of 427 patients who underwent coronary artery bypass graft surgery from St. Luke's Medical Center, Quezon City from January 2009-October 2014.
Primary Outcome: Acute Kidney Injury requiring Renal Replacement Therapy after cardiac surgery.
Results: A total of 427 patients who underwent coronary artery bypass graft surgery. Acute kidney injury was documented in 25.5% (n=109) of subjects, 13.3% (n=57) underwent post-operative renal replacement therapy (RRT), either intermittent hemodialysis or continuous renal replacement therapy. Discrimination for the prediction of RRT was good for the three scoring models using areas under the receiver operating characteristic curve (AUROCs): 0.94 (95% CI, 0.916 to 0.963) using Mehta; 0.92 (95% CI, 0.890 to 0.944) using Thakar, and 0.90 (95% CI, 0.867 to 0.926) using SRI. Mehta showed the highest predictive value, with significant difference with SRI (P = 0.0053).
Conclusion: The Bedside Tool for Predicting Risk of Postoperative Dialysis by Mehta, et al., provides very good discriminative power and is relevant in most patients undergoing cardiac surgery. Compared to the Thakar, Mehta does better by 2% and is not statistically significant. Compared to SRI, it does better by 4% and is statistically significant. Although Mehta has the highest predictive value, any of the three predictive models is sufficient to predict acute kidney injury requiring renal replacement therapy in patients undergoing cardiac surgery.