Real-world data-driven early warning system for risk-stratified liver injury in hospitalized COVID-19 patients - Machine learning models for clinical decision support

Aug 5, 2026

Objective: develop and validate a real-world evidence-driven early warning system to predict COVID-19–associated hepatic dysfunction in hospitalized patients using interpretable machine learning models for clinically actionable decision support. Methods: a retrospective single-center cohort used high-resolution EHRs from 983 COVID-19 inpatients; 13 ML algorithms were benchmarked after SMOTE to address class imbalance, and SHAP was used for transparency. Results: SMOTEENN resampling with random forest and extra trees achieved AUC around 0.998 and 0.997, with glutathione and key hepatic enzymes as top predictors. Conclusion: the ensemble models enable real-time, risk-stratified monitoring and support intensified care and hepatoprotective optimization.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages15

Reading Time38 min

AccessFree to download

Objective

Methods

Study design and data source

Modeling strategy

Results

Predictive performance

Interpretability and key predictors

Conclusion

Keywords

TYPE Original Research PUBLISHED 03 September 2025 DOI 10. 3389/fpubh.2025.1566260 OPEN ACCESS EDITED BY Rawad Hodeify, American University of Ras Al Khaimah, United Arab Emirates REVIEWED BY Vibhuti Gupta, Meharry Medical College, United States Vijendra Ramlall, Memorial Sloan Kettering Cancer Center, United States *CORRESPONDENCE Xianxi Guo [hustgxx@163.com](hustgxx@163.com)[ ](hustgxx@163.com)Ge Yang [yangge@tmmu.edu.cn](yangge@tmmu.edu.cn)[ ](yangge@tmmu.edu.cn)Yue Wu [xiong0810541216@163.com](xiong0810541216@163.com) †These authors have contributed equally to this work and share ﬁrst authorship RECEIVED 17 February 2025 ACCEPTED 30 May 2025 PUBLISHED 03 September 2025 CITATION Xiong Y, Cai X, Lai X, Wang Y, Xin H, Song W, Lv F, Guo X, Yang G and Wu Y (2025) Real-world data-driven early warning system for risk-stratiﬁed liver injury in hospitalized COVID-19 patients—Machine learning models for clinical decision support. Front. Public Health 13:1566260 . doi: 10.3389/fpubh.2025.1566260 COPYRIGHT © 2025 Xiong, Cai, Lai, Wang, Xin, Song, Lv, Guo, Yang and Wu. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY) . The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms. Real-world data-driven early warning system for risk-stratiﬁed liver injury in hospitalized COVID-19 patients—Machine learning models for clinical decision support Yuanguo Xiong1†, Xu Cai1†, Xin Lai2 , Yuwen Wang3 , Hao Xin4 , Wei Song1 , Feng Lv1 , Xianxi Guo1*, Ge Yang5* and Yue Wu1* 1 Department of Pharmacy, Renmin Hospital of Wuhan University, Wuhan, China, 2 Department of Pharmacy, The First Affiliated Hospital of Nanchang University, Nanchang, China, 3 Department of Clinical Pharmacy, People’s Hospital of Macheng, Huanggang, China, 4 Department of Pharmacy, Qingdao Third People’s Hospital Affiliated to Qingdao University, Qingdao, China, 5 Department of Pharmacy, First Affiliated Hospital of Army Medical University, Chongqing, China Objective: To develop and validate a real-world evidence-driven early warning system for the risk-stratiﬁed prediction of coronavirus disease 2019 (COVID-19) -associated hepatic dysfunction in hospitalized patients, leveraging interpretable machine learning models to provide clinically actionable decision support for timely intervention. Methods: A retrospective single-center cohort study was conducted utilizing high-resolution electronic health records (EHRs) from 983 hospitalized COVID- 19 patients. Clinical features (e.g., laboratory results, medication exposures, and disease progression markers) were systematically analyzed. To mitigate class imbalance, we employed the Synthetic Minority Oversampling TEchnique (SMOTE) prior to model development. Thirteen distinct machine learning (ML) algorithms were trained and benchmarked to construct an optimal risk stratiﬁcation framework. Model performance was rigorously evaluated using metrics, including accuracy, precision, recall, F1-score, and area under the receiver operating characteristic curve (AUC) . SHapley Additive exPlanations (SHAP) analysis was employed to enhance clinical interpretability and provide transparent insights for decision-making. Results: The SMOTE-edited nearest neighbors (ENN) technique (SMOTEENN) resampling strategy, combined with random forest (RF) and extra trees (ET) models, demonstrated superior predictive performance, achieving AUC values of 0 .998 ± 0.002 (RF) and 0 .997 ± 0.002 (ET), respectively. The SHAPbased interpretability analysis identiﬁed glutathione administration and hepatic enzymes (e.g., gamma-glutamyltransferase [GGT] and alanine aminotrans

FAQs

What is the purpose of the proposed early warning system?

To predict risk-stratified COVID-19–associated liver dysfunction in hospitalized patients and provide clinically actionable decision support for timely intervention.

How was the model developed and validated?

A retrospective single-center cohort of 983 patients used high-resolution EHRs; SMOTE addressed class imbalance, 13 ML algorithms were benchmarked, and model interpretability was assessed using SHAP.

Which factors were most influential for predicting liver injury risk?

SHAP identified glutathione administration and hepatic enzymes such as gamma-glutamyltransferase (GGT) and alanine aminotransferase (ALT) as the most influential predictors.

Aug 5, 2026

Upload by Evangeline