To develop and externally validate an explainable machine learning framework to predict VTE in patients with septic shock.
Approach:
Study Design: A retrospective cohort study was conducted including adult septic shock patients admitted at Shaanxi Provincial People’s Hospital from January 2020 to December 2025.
Key Findings:
VTE incidence in the development cohort was 17.74%.
Six predictors identified: FDP, PT, WBC, APTT, D-dimer, and CRP.
Random Forest algorithm showed superior performance with an AUC of 0.9718, an F1 score of 0.7917, and a sensitivity of 0.7037 in the validation cohort.
Interpretation:
SHAP analysis revealed that increased thrombo-inflammatory markers and shortened coagulation intervals significantly influenced VTE risk.
Limitations:
The study is retrospective and may have inherent biases.
External validation was conducted at a single site, which may limit generalizability.
Conclusion:
A highly accurate RF-based predictive model for VTE in septic shock patients was established.
