A Causal and interpretable machine learning framework for postcranioplasty risk prediction and surgical decision...

Area of Science:

• Neurosurgery
• Medical Informatics
• Machine Learning

Background:

• Cranioplasty procedures carry a significant risk of postoperative complications.
• Accurate prediction of these complications is crucial for patient management.

Purpose of the Study:

• To develop and validate a machine learning-based clinical decision-support tool for predicting postoperative complications after cranioplasty.
• To identify modifiable intraoperative variables influencing complication risk.

Main Methods:

• A multicenter study utilizing a random forest model trained on nine selected features.
• Model validation using internal cross-validation and external geographical and temporal cohorts.
• Causal inference methods (double machine learning, T-learner) to analyze intraoperative variables.

Main Results:

• The random forest model achieved high predictive performance (AUROC 0.930-0.949).
• Subgroup analyses confirmed consistent accuracy across age and sex demographics.
• Subcutaneous negative-pressure drainage and titanium mesh showed protective effects against complications.

Conclusions:

• The developed tool offers a practical framework for real-time risk stratification in cranioplasty.
• Findings support optimizing intraoperative decisions to mitigate postoperative complication risks.
• An accessible web-based tool is available for clinical decision-making.