Estimating institutional physician turnover attributable to self-reported burnout and associated financial burden: a case study

The increasing prevalence of burnout among physicians and its negative effects on health care quality are well documented [1, 2]. Prevention programs implemented at the organizational level are likely to be more effective in reducing physician burnout than programs that only target individual physicians (e.g. time management, stress reduction and resilience building skills) [3]. Accurate information on the economic cost of physician burnout to health care organizations may help motivate systematic organizational level efforts to improve physician wellbeing [4]. Medical errors, malpractice suits, physician turnover, reductions in clinical work hours, and lower patient satisfaction are some the consequences of physician burnout that result in increased organizational costs [4‐7]. Many factors that have been identified as contributing factors to burnout- such as low professional satisfaction, medical-legal, health, and financial issues, lack of alignment between personal and organizational values and challenging practice arrangements [8‐10] are also factors that contribute to physicians’ turnover or early retirement [1, 11‐13].

Although many reports have demonstrated associations between burnout and intent to leave, [7, 14‐23] only one study has examined the association between physician burnout and actual subsequent turnover in an academic medical center in the US [6]. In terms of estimated cost of physician burnout, there is only one study that has estimated the cost of departure, based on intent to leave data, attributable to burnout in Canada in 2014 [17].

The current gold standard to assess burnout out is the Maslach Burnout Inventory-Human Services Survey (MBI-HSS), with 22 items measuring three common symptoms of burnout: emotional exhaustion, depersonalization, and lack of personal accomplishment. Another commonly used measure of burnout in physicians is a validated non-proprietary one-item self-reported burnout scale [24‐27] which has a strong positive correlation with the emotional exhaustion sub-scale of the MBI [26, 27]. It is the emotional exhaustion subscale of MBI that has been shown to be the strongest predictor of physician attrition [6] and intent to leave the institution [12, 19, 28]. Burnout indicated by the one-item self-reported scale is also associated with physicians’ intent to leave, [15, 23] while no associations have been found between depersonalization and physician attrition [6] or intent to leave [19, 28]. Emotional exhaustion correlates with anxiety and depression in physicians [29, 30]. Previous reports examining the effect of physician burnout on intent to leave and/or actual departure have not adjusted for the effects of anxiety and depression.

The primary objective of our study is to examine the associations of physician self-reported burnout with intent to leave and actual turnover, using a recent longitudinal data case example, both before and after adjusting for anxiety and depression. The second objective is to estimate the cost of physician turnover attributable to burnout. The third objective is to explore the relationship between intent to leave and actual turnover, and thus to assess the feasibility of using intent to leave data as a proxy for turnover data when data on actual turnover is not available.

Our baseline study sample includes 472 physicians (Fig. 1). Sixty-one (13%) physicians in the baseline sample in 2013 had left Stanford by 2015. As baseline 26% of physician reported experiencing burnout and 133 (28%) reported intent to leave within the next 2 years (Table 1). The majority of physicians in our sample reported working over 50 h per week. Based on the PROMIS guidelines, the closest T-scores and standard errors corresponding to the raw score of sleep-related impairment falls between 50.3 ± 2.7 and 51.6 ± 2.6, [37] for anxiety falls between 48.0 ± 3.6 and 51.2 ± 3.1, [38] and for depression falls between 49.0 ± 3.2 and 51.8 ± 2.7 [37, 39].

The results of our analysis indicate that 1) physicians who are experiencing burnout are more than twice likely to leave their practice, and that the effect of burnout on turnover is independent of personal factors such as anxiety or depression; 2) the economic cost of turnover attributable to physician burnout is high, and; 3) intent to leave is a strong predictor of actual departure. Those who reported intent to leave the institution were more than three times as likely to leave within 2 years, compared to those who did not report intent to leave.

In this study sample, physicians had modestly higher scores for sleep-related impairment, and similar to median scores for anxiety, and depression to median scores from PROMIS reference populations (the 2000 General US Census and a clinical sample for sleep-related impairment, and the 2000 General US Census for anxiety and depression) [37‐39]. Although depression was a significant predictor of intent to leave even after adjusting for other variables, only burnout (an indicator of work-specific distress), rather than depression and anxiety (indicators of general distress) was a significant predictor of actual physician turnover. Similar to other studies, [40, 44] our results suggest that intent to leave is a strong predictor of physician turnover. In our baseline sample, 28% reported intent to leave the institution within 2 years, and after 2 years 13% of physicians had left Stanford. Those who reported intent to leave in 2013 were more than three times as likely to have left by 2015. Intent to leave may be a valid proxy indicator of turnover to calculate organization-specific financial burden of physician turnover attributable to burnout, when data on actual turnover is not available. However, intent to leave attributable to burnout is much higher than actual subsequent turnover attributable to burnout. An adjustment factor for this difference will improve accuracy when estimating financial burden from intent to leave data where turnover data is not yet available. The adjustment factor indicated by the present study is to divide intent to leave attributable to burnout by 3.7 to estimate actual subsequent turnover attributable to burnout. Additional data from a heterogeneous set of health care organizations will help clarify the variance in difference between intent to leave versus actual subsequent turnover attributable to burnout.

Future investigation from a heterogeneous set of health care organizations will also help clarify the variance in economic cost of physician turnover attributable to burnout. Physicians working in academic medical centers often have unique challenges (e.g. academic productivity expectations) and benefits (e.g. opportunities to teach and conduct research) that may affect the relationship between burnout and turnover. The only other published data—to our knowledge—on the association between burnout and turnover indicates an effect of burnout on physician turnover similar in magnitude [6] to the present study finding, which provides some across-organization validity. However, this collaborating data is also from an AMC, whereas most health care organizations in the United States are not AMCs. It is not clear that these findings will generalize accurately in attempts to estimate cost of turnover attributable to burnout in non-AMC organizations. The only published article to date [4] that attempts to use turnover attributable to burnout to build an economic argument for investing in physician well-being relies on unpublished data from the present study [45‐47].

The present study is limited to observation of the relationship between burnout and subsequent turnover, which provides necessary but far from sufficient evidence of a causal relationship. In addition to an AMC focus, the present study is also limited to secondary data analysis of completely de-identified evaluation data, without demographic data that might be helpful in assessing generalizability of results. With these limitations in mind, the findings of the present study, coupled with previous evidence of effective strategies to mitigate burnout, [48] suggest feasibility of longitudinal study to determine the effect of interventions to mitigate burnout on cost of physician turnover.

Average turnover rates observed at Stanford (13% in 2 years) are similar to national averages. The Cejka Search and American Medical Group Association Physician Retention Survey results show that the annual rate of physician turnover was about 6% in 2010 and 2011 [49]. If interventions can reduce burnout to half of the current rate, Stanford Medicine may be able to save $7,772,000 to $27,753,000 in physician recruitment costs alone during the subsequent 2 years. These figures based on recruitment cost at Stanford, which are comparable to national averages reported elsewhere [4, 50].

Our estimate of the high cost of turnover attributable to burnout does not include either the loss of revenue due to physician faculty position vacancy or the diminished economic productivity associated with burnout among those who remain employed—such as a reduction in work-hours [51]. Nor does this estimate include a broader range of costs associated with burnout, such as the disruption in the quality and continuity of direct patient care or the reduced health of physicians [9, 52]. Investment in interventions to reduce burnout may pay dividends in all of these domains.

Our results suggest that the economic cost of physician faculty attrition attributable to burnout is high. This finding may help motivate further investigation to determine the effect of strategies to mitigate burnout on physician turnover and associated economic burden. Academic medical centers offer physicians rewarding careers, including opportunities to develop new frontiers in patient care, create and advance new knowledge, and to educate and mentor trainees. In turn, these institutions benefit immensely from talented physicians who drive high standards in patient care, research, and education. Investing in strategies that prevent or reduce burnout may render economic benefits associated with reduced physician attrition. Perhaps more compelling for AMC-employed physicians and the patients they serve are the benefits that reduced burnout may render to physicians’ personal and professional wellbeing and to the quality of clinical care.