Original Research
Sunday, October 13, 2024
12:45 PM–02:15 PM
Abstract
Introduction:
Acute kidney injury (AKI) associated with the use of vancomycin and teicoplanin is crucial for severely infected patients.
Research Question or Hypothesis: This study aimed to develop and validate clinically prognostic machine learning models for predicting vancomycin-associated AKI (VA-AKI) and teicoplanin-associated AKI (TA-AKI).
Study Design: This study analyzed the data of patients receiving intravenous vancomycin or teicoplanin therapy between February 2010 and December 2020 in the Taipei Medical University Clinical Research Database.
Methods: AKI was determined using serum creatinine criteria of the Kidney Disease Improving Global Outcomes (KDIGO). Features were selected from 198 variables encompassing demographics, admission diagnoses, comorbidities, medication profiles, and laboratory results through recursive feature elimination using feature importance and SHapley Additive exPlanations (SHAP) importance. Twelve models were constructed using two machine learning algorithms: eXtreme Gradient Boosting (XGBoost) and Light Gradient Boosting Machine (LightGBM). Model performance was compared using eight evaluation metrics, including the area under the receiver operating characteristic curve (AUROC).
Results: Among 9342 patients, 19.70% (1383/7020) of patients in the training set, 18.58% (326/1755) in the internal validation set, and 20.5% (116/567) in the external validation set developed AKI. The XGBoost model, leveraging time-series data, demonstrated optimal performance in internal (AUROC 0.798, 95% confidence interval (CI) 0.791–0.804) and external (AUROC 0.779, 95% CI 0.767–0.791) validation. Days on medication, BUN, eGFR, number of concomitant nephrotoxic drugs, and total bilirubin were the top 5 features of AKI. SHAP analysis revealed the feature-feature interaction and force plots accessed the individual patient’s real-time risk.
Conclusion: This model can serve as a tool for clinical decision-making to reduce VA-AKI and TA-AKI. The individualized risk assessment devised personalized adjustment plans. Future prospective studies incorporating the predicting model into the Clinical Decision Support System are warranted.
Presenting Author
Hsiang-Yin Chen Pharm.D, M.S.Wan-Fang Hospital, Taipei Medical University
Authors
Fan-Ying Chan M.S.
Department of Clinical Pharmacy, School of Pharmacy, Taipei Medical University
Pin-Yi Lo M.S.
Department of Clinical Pharmacy, School of Pharmacy, Taipei Medical University