Yale University researchers have developed a machine-learning model to predict physician turnover, providing healthcare organizations with valuable insight to improve retention rates. Physician turnover is a significant concern for healthcare organizations, leading to the disruption of patient care continuity, affecting clinical outcomes, and causing a considerable financial burden on the industry. According to recent studies, each physician’s departure can result in healthcare organizations losing up to $1 million in both direct and indirect expenses, including recruitment, onboarding costs, and lost revenue. The cost of physician turnover is a staggering $4.6 billion annually at the national level, with a significant contributor being professional burnout.
Burnout has been identified as a significant driver of physician turnover and closely linked to physicians’ intention to leave their current positions, reduce their professional effort, and eventually leave their practice. Burnout in healthcare professionals is a complex issue caused by several factors, including EHR inefficiencies, organizational culture, and documentation burden. Poor usability and frustration with electronic health records (EHR) have been linked to burnout, as have organizational leadership culture and the excessive inbox messages and notifications that physicians receive. The burden of documentation also causes healthcare professionals to spend a significant amount of time on EHR-related tasks.
As an increasing number of part-time and younger physicians prioritize work-life balance, practice leaders must adopt effective recruitment and retention strategies for talented clinicians. However, currently, practice leaders rely on self-reported data through physician surveys to identify physician burnout, job satisfaction, and intention to leave. Unfortunately, self-reported data are not always reliable, leading to limitations such as response fatigue and bias.
Ted Melnick, the associate professor of emergency medicine and co-senior author of the new study, noted the difficulty with surveys; physicians often feel obliged to respond, leading to low response rates. “Surveys can tell you what’s happening at that moment,” he added, “but not what’s happening the next day, the next month, or over the following year.”
To address this issue, the PLOS ONE study proposed the use of EHR audit logs to track physician burnout objectively. The study utilized EHR use metrics based on time to standardize ambulatory physician EHR use measurement, and the study found that physician productivity and several core EHR use metrics were associated with physician departure. Interestingly, the study revealed that physicians who spent less time on the EHR (especially on inbox tasks) had higher rates of departure, which was unexpected. In response to these findings, the study authors developed a predictive model that included new metrics recently linked to burnout, extending the study through the Covid-19 pandemic. Monthly data was collected from 319 physicians across 26 specialties for 34 months, including EHR usage, clinical productivity measures, and physician characteristics. The study used different data portions to train, validate, and test the machine learning model.
“We wanted something that would be useful on a personalized level,” said Andrew Loza, a lecturer and clinical informatics fellow at Yale School of Medicine and co-senior author of the study. “So if someone were to use this approach, they could see the likelihood of departure for a position as well as the variables contributing most to the estimate in that moment, and intervene where possible.”
According to the study, the model exhibited a high degree of accuracy, with a 97% success rate, and it identified the significant factors that contribute to physician turnover risk, including how they interact with each other and which factors change when a physician moves from a low-risk to a high-risk of departure. The top four factors contributing to physician turnover risk were the length of employment, age, case complexity, and service demand. Interestingly, the study found that the probability of departure was less for physicians with intermediate tenures and greater for those who had recently been appointed or had longer terms in office. Additionally, the study found that the risk of departure shifted over the 34-month study period, which covered both pre-pandemic and pandemic periods, and linked COVID-19 waves to a change in the risk of physician departure. By monitoring variables that contribute to physician burnout and turnover, healthcare organizations can take timely and targeted action to prevent physician departure and maintain continuity of care for patients.
The researchers created a dashboard that can display this valuable information, and health care leaders have shown interest in this type of analysis. With physician burnout becoming increasingly recognized, healthcare organizations, hospitals, and larger groups must prioritize the well-being of physicians and other clinicians who care for patients. Robert McLean, New Haven Regional Medical Director of Northeast Medical Group, said that wellness officers and committees could collect and analyze the data to develop conclusions and implementation plans for changes and improvements.
