Application of Advanced Bioinformatics to Understand and Predict Burnout Among Surgical Trainees

Area of Science:

• Medical Education
• Surgical Training
• Psychiatry

Background:

• Physician burnout is a significant issue, particularly among surgical trainees.
• Existing burnout prediction tools have limitations in scope and strength.
• Identifying high-risk subpopulations is crucial for effective interventions.

Purpose of the Study:

• To apply bioinformatics and machine learning methods to meta-analyze existing burnout predictive tools.
• To improve the identification of surgical resident subpopulations at highest risk of burnout.
• To develop a composite predictor for burnout in surgical residents.

Main Methods:

• A composite survey was administered to surgical residents, including measures of burnout, grit, fatigue, financial well-being, leadership perceptions, and attitudes toward robotic surgery.
• Data were analyzed using k-means clustering and supervised/unsupervised clustering techniques.
• Hierarchical clustering and k-means analysis were employed to identify distinct responder clusters.

Main Results:

• Unsupervised hierarchical clustering revealed heterogeneous resident response patterns.
• K-means clustering identified 3 discrete clusters of responders with differential burnout risk (p=0.021).
• The highest-risk cluster exhibited the lowest grit, low interest in innovation/leadership, higher financial stress, and highest rates of anxiety and burnout.

Conclusions:

• The limited scope of traditional burnout prediction tools necessitates novel approaches.
• Machine learning, specifically cluster analysis, can organize complex data to predict outcomes like burnout.
• This approach identifies specific subgroups of residents at risk, allowing residencies to allocate resources effectively for improved well-being.