Machine Learning Models for Predicting Postoperative Outcomes following Skull Base Meningioma Surgery

Area of Science:

• Neurosurgery
• Medical Informatics
• Health Services Research

Background:

• Predictive analytics are increasingly used for meningioma outcomes.
• Limited research exists on machine learning (ML) for skull base meningioma prognostication.
• Skull base meningiomas present unique surgical challenges and potential for complex postoperative courses.

Purpose of the Study:

• To develop and validate ML models for predicting postoperative outcomes in skull base meningioma patients.
• Specifically aimed to predict prolonged hospital length of stay (LOS), nonroutine discharge disposition, and high hospital charges.
• To assess the utility of ML in enhancing high-value healthcare and optimizing surgical workflows.

Main Methods:

• Retrospective study of 265 patients undergoing skull base meningioma surgery (2016-2019).
• Elastic net logistic regression algorithms were trained on 70% of data and validated on 30%.
• Outcomes included prolonged LOS (>4 days), nonroutine discharge, and high hospital charges (>$47,887).

Main Results:

• Models achieved areas under the receiver operating characteristic curve of 0.798 for prolonged LOS, 0.752 for nonroutine discharge, and 0.592 for high charges.
• All developed models demonstrated adequate calibration (p >0.05).
• The cohort was predominantly female (77.7%) and Caucasian (63.4%).

Conclusions:

• Developed and validated ML models for predicting key postoperative outcomes in skull base meningioma surgery.
• These models can assist surgeons in identifying at-risk patients and optimizing resource allocation.
• ML offers a valuable tool for improving healthcare value and streamlining clinical processes in neurosurgery.