Research Letter
Physician Work Environment and Well-Being
Turnover and Burnout Among Family Physicians
Dhruv Khullar, Lawrence P. Casalino, Richard G. Kronick, et al
JAMA Intern Med 2026;186;(5):635-637. doi:10.1001/jamainternmed.2026.0271
Physician burnout may be associated with leaving practice or moving to a new practice, resulting in clinician shortages, especially in primary care.1 Evidence suggests that patients who lose a primary care physician experience higher emergency department use and spending and lower care satisfaction and that a physician’s departure may have substantial financial consequences for health care organizations.2,3
Although leaving or changing practice has been associated with burnout within individual health systems,1 the extent to which burnout affects turnover nationally is not well understood. We linked repeated, cross-sectional national surveys of board-certified family physicians to Medicare claims data to examine the association between burnout and turnover.
Methods
We used the 2016 to 2020 American Board of Family Medicine (ABFM) surveys to measure burnout. ABFM surveys are completed by physicians applying to continue their board certification or in the last year of a certification stage. Similar to prior work,4 we constructed a physician turnover measure and extracted physician and practice characteristics using the 2015 to 2022 Medicare Data on Provider Practice and Specialty (MD-PPAS) files and a 20% sample of the Medicare Carrier file. We excluded physicians who treated fewer than 4 Medicare enrollees in each year and who billed more than 2.5% of service line items to a third Tax Identification Number not in MD-PPAS. The Weill Cornell Medical College Institutional Review Board deemed this cross-sectional study exempt from review and informed consent because deidentified data were used. We followed the STROBE reporting guideline.
Physician burnout was defined as reporting either of the following at least weekly: (1) “I feel burned out from my work,” or (2) “I have become more callous toward people.”5 Physicians were considered to have moved practices if they billed at least 4 months in 1 practice followed by at least 4 months in a different practice within 12 months of survey completion. Physicians were considered to have left practice if they did not bill any practice for 12 months, beginning in the 12 months after survey completion (eMethods 1-5 in Supplement 1).
We constructed unadjusted and adjusted shares of physicians who changed practices, left practice, and either changed or left practice by burnout status. Shares were adjusted using logit regressions, controlling for physician demographic characteristics and primary practice size, geography, share of patients with low income (proxied by Medicare and Medicaid dual eligibility), and health system affiliation (using the Compendium of US Health Systems6).
Regressions used robust SEs clustered at the physician level. Two-sided P < .05 indicated statistical significance. Analysis was conducted from July to December 2025 using Stata 16 (StataCorp).
Results
We included 19 929 physicians (10 440 males [52.4%]; mean [SD] age, 46.9 [11.1] years) who completed the burnout portion of the ABFM surveys. Of these physicians, 43.5% (8676) reported burnout (Table).
Table. Physician and Practice Characteristics by Burnout Status
| Characteristic | Physicians with burnout, No./total No. (%) |
|---|---|
| Overall No. | 8676/19 929 (43.5) |
| Year of survey | |
| 2016 | 528/1164 (45.4) |
| 2017 | 1016/2248 (45.2) |
| 2018 | 949/2147 (44.2) |
| 2019 | 3891/8904 (43.7) |
| 2020 | 2292/5466 (41.9) |
| Survey | |
| Continuous Certification Questionnaire | 1412/3308 (42.7) |
| National Graduates Survey | 2580/5821 (44.3) |
| Practice Demographic Survey | 4684/10 800 (43.4) |
| Physician covariates | |
| Age, y | |
| 28-34 | 1466/3241 (45.2) |
| 35-44 | 2665/5896 (45.2) |
| 45-54 | 2466/5384 (45.8) |
| 55-64 | 1660/3988 (41.6) |
| ≥65 | 419/1420 (29.5) |
| Sex | |
| Female | 4668/9489 (49.2) |
| Male | 4008/10 440 (38.4) |
| Ruralitya | |
| Nonrural | 7278/16 731 (43.5) |
| Rural | 1398/3198 (43.7) |
| US Census region | |
| Northeast | 1407/3045 (46.2) |
| Midwest | 2554/5753 (44.4) |
| South | 2825/6774 (41.7) |
| West | 1890/4357 (43.4) |
| Practice covariatesb | |
| Practice size, No. of physicians | |
| 1-2 | 1098/2850 (38.5) |
| 3-9 | 1127/2572 (43.8) |
| 10-20 | 674/1604 (42.0) |
| 21-50 | 898/2020 (44.5) |
| 51-199 | 2039/4659 (43.8) |
| 200-499 | 1507/3222 (46.8) |
| ≥500 | 1333/3002 (44.4) |
| % Of dual-eligible visits, quartilec | |
| First, lowest | 2172/4985 (43.6) |
| Second | 2311/4986 (46.3) |
| Third | 2180/4982 (43.8) |
| Fourth, highest | 2013/4976 (40.5) |
| Health system affiliationd | |
| No | 3297/7964 (41.4) |
| Yes | 5379/11 965 (45.0) |
Overall, 8.4% of physicians (1674) changed practices or stopped practicing. Among these physicians, in adjusted analyses, 10.2% reported burnout compared with 7.1% who did not report burnout (difference, 3.1 [95% CI, 2.3-3.9] percentage points; P < .001) (Figure). Physicians reporting burnout were more likely to leave practice (5.4% vs 3.7%; difference, 1.7 [95% CI, 1.1-2.3] percentage points; P < .001) and move practices (4.8% vs 3.4%; difference, 1.4 [95% CI, 0.8-2.0] percentage points; P < .001).
Figure. Bar Graph of Adjusted Shares of Turnover by Burnout Status
Share of turnover was adjusted using logistic regressions where the dependent variable was an indicator of whether a physician turned over. Separate regressions were used for overall turnover, leaving practice, and moving practices. The primary covariate was whether the physician was burned out. Additional covariates are listed in the Table, including physician and practice covariates. Differences were statistically significant at P < .001 (eMethods 1-5 in Supplement 1 provide details).
Discussion
To our knowledge, this is the first national-level study examining the association between burnout and observed turnover. We found that family physicians with reported burnout were nearly 1.5 times more likely to leave or change practices than physicians without reported burnout.
Study limitations included use of surveys preceding the COVID-19 pandemic, inability to observe physicians who stopped seeing Medicare patients but continued to practice, and noncausal design. However, burnout and observed turnover rates were similar to prior estimates.1,4 These findings suggest that physician burnout warrants sustained attention, not only because of its adverse relationship with professional well-being and patient safety but also because it compromises the physician workforce stability
