A machine learning approach to predict surgical learning curves.

Area of Science:

• Medical Education Technology
• Machine Learning in Surgery
• Surgical Skill Acquisition

Background:

• Current surgical training relies on repetitive motor tasks, lacking personalization and efficient progress tracking.
• Individual differences in initial skill and learning capacity create variability in training duration and outcomes.
• A predictive model is needed to tailor surgical training protocols based on early learning data.

Purpose of the Study:

• To develop and validate a machine learning model for predicting surgical learning curve characteristics.
• To identify key factors that define trainee learning ability and enable classification.
• To demonstrate the feasibility of personalizing surgical training regimens for enhanced efficiency.

Main Methods:

• Trained a multivariate supervised machine learning model using surgical learning curve data.
• Extracted a single factor, the 'learning index,' to characterize trainee learning ability.
• Employed an unsupervised machine learning model for classifying trainees based on learning patterns.

Main Results:

• The model accurately predicted the number of trials to achieve proficiency (R²=0.72) and final performance level (R²=0.89) using data from the first 10 trials.
• A single 'learning index' factor effectively described the learning process and classified learners.
• Early performance data in surgical tasks contains sufficient information to predict future learning trajectories.

Conclusions:

• Machine learning models can predict surgical learning curve characteristics from initial trials, offering a novel approach to training.
• A single learning index factor can capture complex individual learning behaviors in surgical trainees.
• Personalized surgical training regimens, enabled by these predictive models, promise increased efficiency and cost-effectiveness.