Predicting physician burnout using clinical activity logs: Model performance and lessons learned

Sunny S. Lou, Hanyang Liu, Benjamin C. Warner, Derek Harford, Chenyang Lu, Thomas George Kannampallil

Abstract: Highlights•Burnout is a significant public health concern that affects more than half of the healthcare workforce.•Clinical workload is often characterized as a significant contributor to physician burnout.•We developed a machine learning-based digital phenotype of burnout using audit log files.•Machine learning models relying on EHR-based workload metrics had limited discriminative performance, highlighting the multi-factorial and highly-individualized nature of physician burnout.