Findings

Who’s next?

Trying to predict physician burnout and turnover.

Alex Eben Meyer

Alex Eben Meyer

View full image

Rates of physician turnover—driven in large part by burnout—are on the rise, creating disruptions in patient care, burdening families and coworkers, and imposing significant costs on the health care industry. In a new study, Yale researchers used a machine learning model to identify the most significant turnover risks. They factored in elements such as physicians’ demographics, like age and length of practice; patterns of electronic health record (EHR) use; and the characteristics of a doctor’s “patient panel” (the patients assigned to a particular doctor), such as the number of patients and complex cases.  

EHRs, besides collecting patients’ clinical information, also generate real-time, work-related evidence. The researchers analyzed nearly three years of data from 319 physicians in multi-specialty practices across the northeastern US. (The paper was published in PLOS ONE.)

Their model predicted, with 79 percent accuracy, the likelihood of a physician leaving a practice. They also gathered other findings. The risk of departure was highest for physicians more recently hired and for those with longer tenures—but lower for those in mid-career. Departure risk was higher for those up to 44 years old, lower for physicians 45 to 64, and higher again for those 65 or older. Time spent on EHR activities reduced departure risk for physicians who had worked for under ten years, but increased departure risk for those employed longer.

Andrew Loza ’18MD, a fellow at the Yale/VA clinical informatics program and a coauthor of the study, cautions that “We’ve learned about factors associated with physician departure, not what causes those departures.” The next phase, he says, will include performance testing and ethical implementation of the model.

The team has designed a dashboard to monitor variables contributing to physician turnover. “Ultimately,” Loza says, “we need to understand what human interventions can improve physician well-being and reduce turnover risk. Those answers will come from conversations with people who understand the issues.”

The comment period has expired.