Diagnostic Accuracy of Simple Postoperative AKI Risk (SPARK) Classification and General Surgery AKI (GS AKI) Index in Predicting Postoperative Acute Kidney Injury among Patients Undergoing Non-Cardiac Surgery at a Tertiary Hospital in the Philippines

Abstract

AbstractBackgroundPostoperative AKI is a significant postoperative complication. Clinical risk prediction models are lacking for patients undergoing non-cardiac surgery. SPARK Classification and GS AKI Index are tools that have shown fair discriminative ability to predict post-operative AKI in non-cardiac surgery and have external validation in their original cohorts. There is no study that compares the diagnostic accuracy of both tools.ObjectivesThis study aims to compare the diagnostic accuracy of SPARK Classification vs GS-AKI Risk Index in predicting post-operative AKI among patients who will undergo non-cardiac surgery at a tertiary hospital in the Philippines.MethodsThis is a cross-sectional study, including adult patients who underwent non-cardiac surgeries from January 2019 to July 2021. The individual risk of post-operative AKI for both models were determined. Descriptive data was described using t-test and logistic regression. Measures of accuracy were described using sensitivity, specificity, positive and negative predictive value, positive and negative likelihood ratio, and discriminative ability using concordance (c) statistic.ResultsOf the 340 patients in this study, 77 (22.65%) developed post-operative AKI and 24 (7.06%) developed critical AKI. Based on demographic data, older age, pre-existing renal disease, longer duration of surgery, anemia, hypoalbuminemia, and hyponatremia were associated with higher incidence of post-operative AKI. SPARK had a sensitivity ranging from 17-43% and specificity ranging from 58-93% for Class B to C. GS AKI had a sensitivity ranging from 10-26% and specificity ranging from 61-97% for Class I to V. SPARK had a discriminative power (c statistic) ranging from 0.46 to 0.61 while GS AKI had a discriminative power ranging from 0.41 to 0.54.ConclusionBased on this study, there is an association between higher risk classification in both SPARK and GS AKI and postoperative AKI. However, both clinical prediction models demonstrate poor discriminative power to predict post-operative AKI.