Acute Kidney Injury (AKI) is a common complication in critically ill patients with sepsis, often associated with poor prognosis. This study aims to construct and validate an interpretable prognostic prediction model for sepsis-associated AKI (S-AKI) patients using machine learning (ML) methods. The training cohort data was sourced from the MIMIC-IV database, and the external validation cohort data was obtained from the First Affiliated Hospital of Zhejiang University School of Medicine. The recursive feature elimination (RFE) method was used to identify mortality prediction factors. Prognostic prediction models for ICU admission were established for 7, 14, and 28 days using random forest, XGBoost, multilayer perceptron classifier, support vector machine classifier, and logistic regression. The prediction performance was evaluated using receiver operating characteristic (ROC) curves and decision curve analysis (DCA). The SHAP method was used to interpret the ML models.
A total of 2599 S-AKI patients were included in the study. Forty variables were selected for model development. Based on the area under the ROC curve (AUC) and DCA results from the training cohort, the XGBoost model achieved F1 scores of 0.847, 0.715, and 0.765 for the 7-day, 14-day, and 28-day groups, respectively, with AUC (95% CI) values of 0.91 (0.90, 0.92), 0.78 (0.76, 0.80), and 0.83 (0.81, 0.85), demonstrating excellent performance. It also showed good discrimination in the external validation cohort, with AUC (95% CI) values of 0.81 (0.79, 0.83), 0.75 (0.73, 0.77), and 0.79 (0.77, 0.81) for the 7d, 14d, and 28d groups, respectively. Global and local interpretations of the XGBoost model were provided using SHAP summary plots and force plots.
The study concludes that ML is a reliable tool for predicting the prognosis of S-AKI patients. The SHAP method is used to interpret the intrinsic information of the XGBoost model, which may prove useful in clinical settings and assist clinicians in making refined management decisions.
References: Fan Z, Jiang J, Xiao C, Chen Y, Xia Q, Wang J, Fang M, Wu Z, Chen F. Construction and validation of prognostic models in critically ill patients with sepsis-associated acute kidney injury: interpretable machine learning approach. J Transl Med. 2023 Jun 22;21(1):406.
Translator: Xu Jing
Editor: Zhang Shaolei
Reviewer: Mao Zhengrong
