Chronic Nonhealing Wound Infection Prediction Model with Six Variables

500 patient records from a primary cohort and 300 from an external validation cohort were used to develop a random forest model predicting infection risk in chronic wounds. The model, created by Yang Jiang and published on figshare in 2026, uses six predictors derived from routine electronic medical record data. It achieved an AUROC of 0.884 in testing and 0.855 in external validation.

Use Cases

• Predicting infection risk in chronic wounds based on smoking status, diabetes duration, and wound depth.
• Evaluating model performance and calibration using metrics like AUROC and decision curve analysis.
• Investigating feature importance for infection prediction using SHAP values, focusing on inflammatory markers and hypoalbuminemia.
• Testing a deployed web calculator for clinical application of the prediction model.

Strengths

• Model was developed using 500 patient records and externally validated on 300 records from a separate cohort.
• Performance metrics are reported with confidence intervals, including AUROC of 0.884 (95% CI 0.841–0.928) in testing.
• The model is based on six readily available clinical predictors, enhancing pragmatic utility.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Data may reflect geographic bias inherent to the two tertiary hospitals in China.

Provenance

Source

figshare

Collection Method

Two-center retrospective study using electronic medical record data.

Freshness

Last updated 2026-05-19 05:44:09; freshness should be verified.

Geography

Two tertiary hospitals in China.