The CLASS Project: A Proof-of-Concept Machine Learning-Driven Complexity Level Algorithm for Surgical Scheduling in...

Area of Science:

• Dermatology
• Surgical Oncology
• Health Systems Engineering

Background:

• Mohs surgery workflow coordination presents inherent complexities.
• Effective case complexity grading is crucial for efficient scheduling.
• Precise scheduling can reduce patient wait times and optimize resource allocation.

Purpose of the Study:

• To predict Mohs surgery case complexity by estimating the number of stages and reconstruction type.
• To develop data-driven scheduling recommendations by integrating model insights and workflow analysis.

Main Methods:

• Machine learning models were trained using pathology report data to predict Mohs stages and complex reconstruction.
• A health systems engineering approach was employed for workflow analysis to identify efficiency improvements.
• Scheduling recommendations were developed by combining machine learning predictions and workflow analysis findings.

Main Results:

• Machine learning models demonstrated feasibility in predicting Mohs case complexity.
• Models achieved an AUC-ROC of 0.83 for complex reconstruction prediction and 0.62 for number of stages.
• Anatomical location, particularly midface, was a key predictor of higher complexity and longer surgical duration.

Conclusions:

• The study demonstrated the feasibility of predicting Mohs surgery case complexity.
• Integrating machine-derived and human-derived knowledge offers complementary insights for workflow enhancement.
• These combined insights can significantly improve Mohs surgery scheduling and overall workflow efficiency.