Integrating Real-Time Clinical Activity, Physiological Sensors, and Behavioral Responses for Predicting Physician Burnout (IGNITE)
The Opportunity
Physician burnout is increasingly widespread, now estimated to affect 50% of physicians in practice and 70% of trainees. The challenge with understanding the scale of the physician burnout problem is the way it is currently measured – having physicians fill out standardized surveys, which busy people are less likely to do.
Our Approach
The team set out to develop a screening tool that required no active participation on the physician’s end, rather they hypothesized it could be done using data from the activity log of the Electronic Health Record (EHR). The team believed they could monitor physician workload and its association with burnout and medication errors, and then develop advanced machine learning algorithms to predict physician burnout.
A longitudinal study of 75 interns (72% enrollment) was conducted, with a 76% survey completion rate. The team found burnout to be highly variable month to month. They developed measures for total EHR time, after-hours EHR time, note-writing/chart review/inbox time, number of ordering sessions, and patient load. A high percentage of burnout was found to be associated with increased total EHR time, patient load, and chart review time.
The Solution
The team found that elevations in burnout might be avoided by alternating high workload rotations (inpatient) with lower workload rotations (outpatient, elective) and that interventions to improve provider efficiency might improve burnout, such as EHR usability and more support staff.
