782-P: Machine Learning to Identify Diabetes Patients Initiating SGLT2i at High-Risk of Acute Kidney Injury

One of the main safety concerns associated with SGLT2 inhibitors (SGLT2i) is the increased risk of acute kidney injury (AKI), likely attributed to their diuretic effect and the induced hypovolemia. Yet, little is known about what patients are at highest risk of developing AKI when using SGLT2is. We applied machine learning techniques to predict AKI in type 2 diabetes (T2D) patients treated with SGLT2i and identify risk factors associated with AKI. Using a 5% random sample of Medicare claim data, we identified 17,694 T2D patients who filled ≥1 prescription for canagliflozin, dapagliflozin or empagliflozin prescriptions between 04/01/2013 and 12/31/2016. The cohort was split randomly and equally into training and testing sets. We measured 65 predictor candidates using claims data from the year prior to SGLT2i initiation and applied 3 machine learning models, i.e., random forests (RF), elastic net, least absolute shrinkage and selection operator (LASSO). The incidence rate of AKI was 1.1 % over a median 1.5-year follow up. Among 3 machine learning methods applied, RF produced the best prediction (C- statistic =0.72 ,95% CI 0.68, 0.76), followed by LASSO (C-statistic=0.69 95% CI 0.65,0.73). Among individuals classified in the top 10% of the RF risk score (i.e., high risk group), the actual incidence rate of AKI was as high as 3.7%. RF and LASSO selected similarly sets of important predictors, such as age, Medicaid eligibility, and loop diuretics use. In the logistic regression including 14 predictors selected by LASSO, use of loop diuretics [adjusted OR: 3.72 (2.44, 5.76)] and history of AKI [adjusted OR: 2.78 (1.06,7.3)] had the strongest associations with incidence of AKI. Our machine learning model efficiently identified patients who are at high risk of AKI after SGLT2i initiation. Strikingly, we identified an almost four-fold increased risk of AKI associated with background loop diuretic use among SGLT2i users. Disclosure L. Yang: None. N. Gabriel: None. I. Hernandez: Consultant; Self; Bristol-Myers Squibb Company, Pfizer Inc. A. G. Winterstein: Research Support; Self; Merck Sharp & Dohme Corp. S. Kimmel: Consultant; Self; GE Healthcare, GlaxoSmithKline plc. J. Guo: None.