Professional burnout is a growing epidemic with symptoms affecting over 500,000 US physicians at any given time and carries significant adverse consequences for physician mental health, health care value, quality of care, and patient safety. There is an urgent need to develop reliable methods to proactively identify work environments at high risk for physician burnout, and tailor process improvements accordingly. The objective of this proposal is to develop and validate a real-time prediction model that uses operational data in primary care practices to identify high-risk clinics for physician burnout, enabling timely and tailored process improvements. The central hypothesis is that routinely-collected electronic health record (EHR) usage metrics coupled with practice-specific metrics can predict physicians? risk for burnout and inform process improvement. The rationale for the proposed research is that early identification of high-risk clinics will allow organizations to tailor interventions to those clinics before burnout and its individual or health system consequences arise. This would be an innovative approach that prevents burnout rather than reacting to it. This project capitalizes on routinely-collect data from Stanford primary care clinics to create a database encompassing EHR usage metrics and practice-specific metrics.
The specific aims are:
Aim 1 : Develop a prediction model to quantify risk for physician burnout. The working hypothesis is that real-time metrics of practice efficiency tracked by the EHR and other practice-specific metrics can predict physician burnout using a machine learning approach.
Aim 2 : Refine the prediction model using qualitative methods. The working hypothesis is that qualitative assessment will inform refinements to the prediction model created in Aim 1, and will demonstrate the face validity of the model.
Aim 3 : Validate the use of the prediction model to identify high-risk clinics. The working hypothesis is that quality of care metrics will demonstrate the concurrent validity, that subsequent routinely administered burnout surveys will demonstrate the predictive validity of the prediction model created in Aims 1 and 2, when aggregated at the clinic level. This proposal is significant because physician burnout is a growing problem with important implications for patient safety. It is also innovative in deploying machine learning and mixed methods to identify physicians at increased risk for burnout which will enable testing interventions to reverse this trend. In combination with formal training in quantitative and qualitative methods, expert mentorship, and participation in selected scholarly activities at Stanford, the experience gained through this project will facilitate progress toward a long-term goal to become an academic leader advancing evidence-based reform of the health care delivery system to optimize human factors that improve quality and safety.
There is an urgent need to develop reliable methods to proactively identify and improve work environments with high risk for physician burnout, and doing so will have broad public health relevance by reducing risk for burnout and its negative effects on quality and safety of care. Addressing physician burnout in primary care carries relevance for AHRQ?s priority populations of women, children, the elderly, and low-income patients, who are particularly dependent on high quality primary care. This project will support AHRQ?s mission of producing evidence to make health care safer and higher quality by developing and validating a prediction model to identify high-risk primary care clinics for physician burnout in order to 1) inform resource allocation to clinics with the highest need, 2) tailor process improvements to reduce risk for burnout and poor quality of care, and 3) lay the groundwork for expanding this model to other high-risk practice settings and provider types.
