Data Sheet 2_Integrating inflammatory and coagulation biomarkers for surgical risk stratif

A dataset of 1,060 patients with Crohn's disease used to develop machine learning models for predicting long-term surgical risk. The dataset includes inflammatory and coagulation biomarkers, and disease-related features, and was used to build a gradient boosting model achieving a C-index of 0.816. The data was published by Kailing Xie on figshare in May 2026 under a CC-BY-4.0 license.

Use Cases

• Predicting surgery-free survival in Crohn's disease patients based on inflammatory and coagulation biomarkers.
• Assessing the benefit of biologic therapy across different risk strata using inverse probability weighting.
• Interpreting machine learning model predictions for surgical risk using SHAP analysis on biomarker contributions.
• Comparing the performance of Cox regression and gradient boosting models for clinical risk stratification.

Strengths

• Includes data from 1,060 patients with Crohn's disease.
• The best-performing gradient boosting model achieved a C-index of 0.816 and time-dependent AUC values up to 0.851.
• SHAP analysis indicates inflammatory and coagulation markers contributed approximately 70% of the model's explainability.
• Model performance was validated using five-fold cross-validation.

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 temporal or source bias inherent to the single study cohort.

Provenance

Source

figshare

Collection Method

Clinical cohort study of 1,060 patients; predictors selected using LASSO-Cox regression and the Boruta algorithm.

Freshness

Last updated 2026-05-07 04:45:25; freshness should be verified.