Risk Factors and Machine Learning Model for Proton Pump Inhibitor Related Acute Kidney Injury

Recent evidence concerns acute kidney injury (AKI) following proton pump inhibitor (PPI) application. Few actual studies have compared the incidence, risk factors, and predictive models of AKI associated with PPI. The present study was a single-center retrospective study. The researchers retrospectively analyzed data from patients who received PPI medications between January 2018 and December 2020. PPI drugs included omeprazole, esomeprazole, rabeprazole, and pantoprazole. The primary outcome of the study was AKI, as defined by kidney disease: improving global outcomes (KDIGO). Secondary outcomes included length of hospital stay, hospital costs, and continuous renal replacement therapy. Independent risk factors associated with AKI were identified by univariate analysis and multifactorial logistic regression analysis (P < 0.05). Logistic regression models were constructed based on the variables obtained from the analysis. Internal validation of the model was performed by the ten-fold cross-validation method. Model discriminatory power was assessed by the area under the curve (AUC) of the receiver operating characteristic curve (ROC). The study aims to develop a PPI-related AKI prediction model based on an electronic medical record system that can be used to predict AKI in hospitalized patients and contribute to the early prevention, diagnosis and treatment of AKI, ultimately reducing morbidity and improving prognosis.