(1423) Evaluating Nephrotoxic Medications in Predicting Acute Kidney Injury: A Machine Learning Study

Introduction: On average, critically ill patients receive over 30 medications during their hospital stay. Antibiotic-induced nephrotoxicity is responsible for up to 60% of acute kidney injury (AKI) cases and is associated with a doubled risk of mortality during hospitalization. While the literature extensively explores antibiotic-related injury, less is known about the impact of other agents on the incidence of AKI in critically ill patients. This study aims to develop a machine learning (ML) model to predict AKI by incorporating data on various classes of nephrotoxic medications.

Methods: A retrospective analysis was performed using electronic health records of 319 patients (> 18 years of age). AKI status was determined according to the 2012 Kidney Disease Improving Global Outcomes guidelines. Descriptive statistics were conducted using an independent t-test, chi-square test, or fisher’s exact test. Twenty-two classes of nephrotoxic medications were evaluated using several ML algorithms, including Logistic Regression, Random Forest, and XGBoost, to predict AKI. The important predictors were ranked to evaluate their influence in AKI predictions.

Results: Among 319 patients, 153 (48%) met AKI criteria with a median age of 65 + 13 years. Sepsis and septic shock syndrome were found to be statistically significant (p < 0.001) between AKI and non-AKI individuals (31.4% and 30.7% vs. 17.5% and 12.7%), respectively. AKI patients more commonly received proton pump inhibitors (71.2%) and diuretics (47%). Beta-lactams (56.9%), carbapenems (26.1%), and glycopeptide (54.2%) antibiotics were found to be statistically significant (p < 0.01).The XGBoost ML model had the highest prediction accuracy of AKI (81%; IQR: 70-89) with an average sensitivity of 80% compared to other models. Body mass index, sepsis with shock, and the use of diuretics and proton pump inhibitors significantly influenced AKI prediction across the three models.

Conclusions: The incidence of AKI in the ICU remains high and is associated with poorer clinical outcomes. Our study is unique in that it incorporates less well-known nephrotoxic medications and their influence on predicting AKI. These findings may help clinicians identify at-risk patients early, enabling the proactive implementation of preventative care strategies during hospitalization.