Online Well-Being Group Coaching Program for Women Physician Trainees: A Randomized Clinical Trial

Key Points

Question  Can a 4-month, online, group coaching program reduce burnout, moral injury, and impostor syndrome and increase self-compassion and flourishing among a sample of women physician trainees across multiple sites?

Findings  In this randomized clinical trial of 1017 women trainee physicians, participants randomly assigned to a 4-month group-coaching program had a statistically significant reduction in all scales of burnout, moral injury, and impostor syndrome, as well as improved self-compassion and flourishing, compared with the control group.

Meaning  These findings suggest that an online, multimodal, group coaching program is an effective intervention to decrease distress and improve well-being for women physician trainees.

Importance  Physician burnout disproportionately affects women physicians and begins in training. Professional coaching may improve well-being, but generalizable evidence is lacking.

Objective  To assess the generalizability of a coaching program (Better Together Physician Coaching) in a national sample of women physician trainees.

Design, Setting, and Participants  A randomized clinical trial involving trainees in 26 graduate medical education institutions in 19 states was conducted between September 1, 2022, and December 31, 2022. Eligible participants included physician trainees at included sites who self-identified as a woman (ie, self-reported their gender identity as woman, including those who reported woman if multiple genders were reported).

Intervention  A 4-month, web-based, group coaching program.

Main Outcomes and Measures  The primary outcomes were change in burnout (measured using subscales for emotional exhaustion, depersonalization, and personal achievement from the Maslach Burnout Inventory). Secondary outcomes included changes in impostor syndrome, moral injury, self-compassion, and flourishing, which were assessed using standardized measures. A linear mixed model analysis was performed on an intent-to-treat basis. A sensitivity analysis was performed to account for the missing outcomes.

Results  Among the 1017 women trainees in the study (mean [SD] age, 30.8 [4.0] years; 540 White participants [53.1%]; 186 surgical trainees [18.6%]), 502 were randomized to the intervention group and 515 were randomized to the control group. Emotional exhaustion decreased by an estimated mean (SE) −3.81 (0.73) points in the intervention group compared with a mean (SE) increase of 0.32 (0.57) points in the control group (absolute difference [SE], −4.13 [0.92] points; 95% CI, −5.94 to −2.32 points; P < .001). Depersonalization decreased by a mean (SE) of −1.66 (0.42) points in the intervention group compared with a mean (SE) increase of 0.20 (0.32) points in the control group (absolute difference [SE], −1.87 [0.53] points; 95%CI, −2.91 to −0.82 points; P < .001). Impostor syndrome decreased by a mean (SE) of −1.43 (0.14) points in the intervention group compared with −0.15 (0.11) points in the control group (absolute difference [SE], −1.28 (0.18) points; 95% CI −1.63 to −0.93 points; P < .001). Moral injury decreased by a mean (SE) of −5.60 (0.92) points in the intervention group compared with −0.92 (0.71) points in the control group (absolute difference [SE], −4.68 [1.16] points; 95% CI, −6.95 to −2.41 points; P < .001). Self-compassion increased by a mean (SE) of 5.27 (0.47) points in the intervention group and by 1.36 (0.36) points in the control group (absolute difference [SE], 3.91 [0.60] points; 95% CI, 2.73 to 5.08 points; P < .001). Flourishing improved by a mean (SE) of 0.48 (0.09) points in the intervention group vs 0.09 (0.07) points in the control group (absolute difference [SE], 0.38 [0.11] points; 95% CI, 0.17 to 0.60 points; P < .001). The sensitivity analysis found similar findings.

Conclusions and Relevance  The findings of this randomized clinical trial suggest that web-based professional group-coaching can improve outcomes of well-being and mitigate symptoms of burnout for women physician trainees.

Trial Registration  ClinicalTrials.gov Identifier: NCT05222685

Physician burnout is highly prevalent in the US; is disproportionately experienced by physician trainees and women; and is associated with substance abuse, job turnover, higher rates of medical errors, and patient mortality.^1-5 In 2022, the US Surgeon General declared physician burnout a crisis deserving of a multipronged approach to bring about “bold, fundamental change,”⁶ yet little is known about scalable, effective interventions to mitigate burnout risk.^1,7-9

Professional coaching is a promising intervention to reduce burnout.^10-13 The 2022 Surgeon General’s Advisory emphasized building a culture of well-being in training institutions and included coaching as a recommended tool.^6,14 Coaching, unlike therapy, does not diagnose or treat, and instead uses inquiry and metacognition (ie, thinking about one’s thinking) to guide self-progress.^15,16 Evidence supporting physician coaching is growing, but predominantly describes individual coaching led by nonphysician or noncertified faculty coaches in small studies.^11,12,17-20 Literature on outcomes of coaching for physician trainees is sparse, limited to small samples and single specialties, and primarily includes programs of short duration.^{10,11,17,21,22}

An online group coaching program, Better Together Physician Coaching (BT), was piloted in response to high physician trainee burnout.^10,13 Because women trainees are disproportionately affected by burnout,^3-5 BT was initially evaluated among women resident physicians at the University of Colorado in a pilot, single site, randomized clinical trial (RCT), which indicated that online group coaching improved burnout.¹⁰ Building on previous work, the objective of this multisite RCT was to evaluate the generalizability of the 4-month online group coaching program to reduce distress and improve well-being in a national sample of women physician trainees.

This RCT follows the Consolidated Standards of Reporting Trials (CONSORT) reporting guideline²³ and was approved by the University of Colorado institutional review board (see the Trial Protocol in Supplement 1). The study was conducted from September 1, 2022, to December 31, 2022, at 26 graduate medical education (GME) institutions across 19 states (eTable 1 in Supplement 2). Sites of different geographic locations, sizes, and focuses (ie, community vs academic) were recruited and included academic, county, Veterans Health Administration, and community-based hospitals and clinics. Recruitment emails were initially sent to institutional leadership, and video conference calls were held to confirm partnership in this study. Participant enrollment was voluntary, and all participants provided written informed consent. Data were collected and managed with The University of Colorado Research Electronic Data Capture (REDCap). Participating sites were not involved in research and could not access identifiable data.

All GME physician trainees who self-identified as a woman (ie, self-reported their gender identity as a woman, including those who reported woman if multiple genders were reported) at participating sites were eligible. Trainees were recruited through a series of 3 emails to a GME electronic mailing list. After enrollment, participants were randomly assigned to the intervention (access to online group coaching) or control group (no access). Randomization was stratified on the basis of the site. Intervention participants were not given protected time for the intervention and carried the same clinical schedules as control participants. All participants were asked to complete baseline (prior to randomization) and 4-month (end of intervention) surveys containing self-reported demographic questions and validated instruments measuring dimensions of distress and well-being. Participants had the ability to select multiple options for race and ethnicity and gender identity questions. Response options for gender identity included man, woman, nonbinary, not sure, not listed, and prefer not to say. Race and ethnicity categories included Asian, Black, Latinx, multiracial, Native American, Native Hawaiian and Other Pacific Islander, White, and other (defined as unknown or any other race or ethnicity not otherwise specified); race and ethnicity were included to ensure generalizability of the intervention. The control group was offered the intervention after the study. To account for benefits accruing from expectations rather than the intervention, both groups were emailed alternative online wellness resources.^24-28

BT is a 4-month, online, group coaching program developed by 2 professional physician coaches (T.F. and A.M.) and was delivered by a cohort of physician coaches (including author A.S.) who were all certified by The Life Coach School. Coach selection is described in the eMethods in Supplement 2. The program incorporates facets of user engagement from Short et al²⁹ (eg, self-monitoring, reminders, and aesthetics), and the Cole-Lewis framework³⁰ for behavior change (ie, modular, course-like, weekly introduction of content). More information on the foundational framework, program modalities, and curriculum can be found in the eMethods in Supplement 2. Participants had access to the following services housed on a members-only, password-protected website: (1) 3 to 4 live group coaching calls per week via video teleconference (Zoom Video Communications), (2) unlimited anonymous written coaching, and (3) weekly self-study modules on pertinent topics. Video teleconference sessions were recorded for asynchronous listening via private podcast. Topics were selected on the basis of an informal trainee needs assessment done prior to the pilot and were iteratively refined on the basis of rapid content analyses of coaching requests. The program required approximately 5 hours per week of total coach time.

The 22-item Maslach Burnout Inventory (MBI) is considered the benchmark standard to measure burnout and was used under license.³¹ The MBI contains 3 subscales: emotional exhaustion (EE; score range, 0-54), where a higher score indicates worse EE; depersonalization (DP; score range, 0-30), where a higher score indicates worse DP; and personal accomplishment (PA; score range, 0-48), where a higher score indicates better feelings of PA.³² We used established threshold definitions of high EE (subscale score ≥27), high DP (subscale score ≥10), and low PA (subscale score ≤33), and considered those with high EE or DP to have at least 1 manifestation of burnout and meet the definition for positive burnout.^19,20,33

The Young Impostor Syndrome scale³⁴ is an 8-item instrument with yes-or-no scoring and a range of 0 to 8. A score of 5 or greater indicates the presence of impostor syndrome. The Moral Injury Symptom Scale–Healthcare Professionals³⁵ is a 10-item instrument with scores ranging from 10 to 100. Higher scores indicate greater moral injury.

The Neff Self-Compassion Scale–Short Form³⁶ is a 12-item instrument (total scale range, 12-60). Scores of 12 to 30 are considered low, 30 to 42 are considered moderate, and 42 to 60 are considered high. The Secure Flourish Index³⁷ is a 12-item instrument assessing 5 domains of flourishing. Scores range from 0 to12 and higher scores indicate greater flourishing.

We assumed no change in preintervention scores and postintervention scores for the control group and used estimates for the SD of the change in MBI scores from our pilot data.¹⁰ The power calculation for a minimum power of 80% (2-sided α = .05) to detect a standardized effect size (SD = 1) of 0.2, corresponded to a difference in mean (SD) values between the groups of 1.7 (8.6) in EE, 1.7 (8.6) in DP, and 1.0 (4.9) in PA, resulting in an enrollment target of 1000.

Descriptive statistics were computed for characteristics of the whole cohort and by intervention, with comparisons made using Wilcoxon rank sum tests for continuous covariates and Fisher exact or χ² tests for categorical covariates. Characteristics of final survey responders and nonresponders were compared. Descriptive statistics were adjusted for multiple comparisons using the Benjamini-Hochberg procedure.³⁸ P values from regression models were reported unadjusted.

An intent-to-treat analysis was performed on all participants regardless of postsurvey completion using linear and logistic mixed-effects models including the main effects of the time period (baseline vs postintervention), treatment (intervention vs control), the interaction between time period and treatment, and a random intercept for participants estimated using restricted maximum likelihood. Mean change from baseline within each group and the difference in mean change between groups and their 95% CIs were reported. Mixed-effects logistic regression models were used for EE, PA, DP, Young Impostor Syndrome Scale, Moral Injury Symptom Scale-Healthcare Professionals, Self-Compassion Scale-Short Form, Secure Flourish Index and for the binary outcomes of burnout and impostor syndrome.

Linear and logistic regression models were used to estimate mean change from baseline and odds ratios (ORs) adjusted for baseline value and their 95% CIs. The binary outcomes of burnout and impostor syndrome at follow-up were separately modeled using logistic regressions with treatment group and baseline values as variables. Effect sizes (risk difference, number needed to treat [NNT], and relative risk) for burnout and impostor syndrome were estimated from models using the same counterfactual framework for marginal effects. The 95% CIs were estimated with simulation.

In sensitivity analyses, to assess the potential influence of missing follow-up survey data on study outcomes, multiple imputation was used for missing score values via iterative chained random forest³⁹ with predictive mean matching. The imputation model included baseline characteristics, treatment assignment, and baseline scores.

All P values are from 2-sided hypothesis tests and statistical significance was assessed at the α = .05 level. All analyses were performed using R statistical software version 4.2.3 (R Project for Statistical Computing).

Of the 1017 participants (mean [SD] age, 30.8 [4.0] years; age range, 24.0-54.0 years; 540 White participants [53.1%]) enrolled, 502 were randomized to the intervention and 515 were randomized to the control group (Figure 1). All 26 sites had at least 1 participant in the intervention group (eTable 1 in Supplement 2). There were no significant baseline differences in demographics or outcome scores between groups (Table 1). Among all participants, 959 (94.3%) self-reported their gender identity as woman and 843 (88.0%) self-reported their sexual orientation as heterosexual. A total of 207 physician trainees (20.7%) were in postgraduate year (PGY) 1, 198 (19.8%) in PGY 2, and 595 (59.5%) in PGY 3 or beyond, with 186 of 999 physician trainees (18.6%) in a surgical subspecialty.

At baseline, the mean (SD) score was 30.6 (10.8) for EE, 11.8 (6.4) for DP, and 33.3 (7.3) for PA. A total of 570 participants scored positively for burnout, which represents 73.0% of the 781 participants who completed the full MBI. Of 784 participants who completed the YIS, 598 (76.3%) scored positively for impostor syndrome. The mean (SD) score was 45.9 (13.7) for moral injury, 31.6 (7.3) for self-compassion, and 6.2 (1.3) for flourishing. There were 411 participants (40.4%) who responded to the follow-up survey, with more participants in the control group (263 participants [51.0%; 95% CI, 47%-55%]) than the intervention group (148 participants [30.0%; 95% CI, 26%-34%]) (P < .05). Postsurvey respondents and nonrespondents differed by race, gender identity, and PGY (eTable 2 in Supplement 2).

Intervention participants experienced significant improvement compared with the control group on all MBI subscales. For EE, the estimated change in score for the intervention group (mean [SE] −3.81 [0.73] points; 95% CI, −5.24 to −2.38 points) vs the control group (mean [SE] 0.32 [0.57] points; 95% CI, −0.79 to 1.43 points) resulted in an absolute difference (SE) in EE of −4.13 (0.92) points (95% CI, −5.94 to −2.32 points; P < .001).For DP, the estimated change in score for the intervention group (mean [SE] −1.66 [0.42] points; 95% CI, −2.49 to 0.83 points) vs the control group (mean [SE] 0.20 [0.32] points; 95% CI, −0.43 to 0.84 points) resulted in an absolute difference (SE) of −1.87 (0.53) points (95% CI, −2.91 to −0.82 points; P < .001). For PA, the estimated change in score for the intervention group (mean [SE] 2.08 [0.47] points; 95% CI, 1.15-3.00 points) vs control group (mean [SE] 0.43 [0.36] points; 95% CI, −0.28 to 1.13 points) resulted in an absolute difference (SE) of 1.65 (0.59) points (95% CI, 0.48 to 2.81 points; P = .006) (Table 2 and Figure 2).

Intervention participants experienced an 18% (95% CI, 5%-31%) reduction in probability of burnout, with an OR of 0.47 (95% CI, 0.28-0.78; P = .008) compared with the control (ie, intervention participants had 53% lower odds of experiencing burnout at follow-up), with an NNT of 11 (95% CI, 7.1-22.4) to go from positive to negative burnout (eTable 3 in Supplement 2).

Intervention participants experienced a significant decrease in impostor syndrome with an absolute difference (SE) of −1.28 (0.18) points (95% CI, −1.63 to −0.93 points; P < .001); the mean (SE) change in score for the intervention group was −1.43 (0.14) points (95% CI, −1.70 to −1.15 points) and the mean (SE) change in score for the control group was −0.15 (0.11) points (95% CI, −0.36 to 0.07 points) (Figure 3). Intervention participants had a 34% (95% CI, 16% to 52%) reduced probability of impostor syndrome (OR, 0.36; 95% CI, 0.21 to 0.62; P < .001) with an NNT of 9 (95% CI, 6.8 to 17.8 to go from positive to negative impostor syndrome (eTable 3 in Supplement 2).

Intervention participants experienced a significant decrease in moral injury with an absolute difference (SE) of −4.68 (1.16) points (95% CI, –6.95 to −2.41 points; P < .001). The mean (SE) change on the Moral Injury Symptom Scale for the intervention group was −5.60 (0.92) points (95% CI, −7.40 to −3.81 points) vs −0.92 (0.71) points (95% CI, −2.31 to 0.46 points) for the control group (Figure 3).

There was a significant improvement in self-compassion with an absolute difference (SE) of 3.91 (0.60) points (95% CI, 2.73 to 5.08 points; P < .001). The mean (SE) change on the Self-Compassion Scale–Short Form was 5.27 (0.47) points (95% CI, 4.34 to 6.20 points) for the intervention group vs 1.36 (0.36) points (95% CI, 0.65 to 20.80 points) for the control group (Figure 3).

There was a significant improvement in flourishing with an absolute difference (SE) of 0.38 (0.11) points (95% CI, 0.17 to 0.60 points; P < .001). The mean (SE) change on the Secure Flourish Index was 0.48 (0.09) points (95% CI, 0.31 to 0.65 points) for the intervention group vs 0.09 (0.07) points (95% CI, −0.04 to 0.23 points) for the control group (Figure 3).

Similar outcome results were obtained using multiple imputation and when baseline scores were carried forward for missing follow-up scores. See eTable 4 in Supplement 2 for more information on the sensitivity analysis.

In this large, national RCT, women physician trainees who received online group coaching over 4 months had substantial reductions in multiple dimensions of professional distress (burnout, moral injury, and impostor syndrome) and improvements in well-being (self-compassion and flourishing). The improvement in burnout was significant and likely meaningful, with past studies^40,41 showing that even a 1-point increase in EE has been associated with a 7% increase in suicidal ideation and a 5% increase in self-reported major medical errors. Additionally, intervention trainees reported less moral injury, lower odds of impostor syndrome, and increased flourishing and self-compassion.

We have previously demonstrated online group coaching reduced burnout among women trainees across specialties at a single institution.¹⁰ The digital platform and group nature of this study allowed a 10-fold expansion of participants and an addition of 25 sites with only 30 additional hours of direct coaching (1-2 more calls per week compared with the pilot study).¹⁰ The group and online delivery of the intervention supported greater scalability, accessibility, and lower cost compared with individual coaching. Furthermore, the magnitudes of improvement in scores described here were generally higher than other coaching and wellness interventions^{17,19,20,42-44} and higher than that of our single-site pilot RCT.¹⁰ Our findings show effectiveness in a broader population and match national data for trainees that self-report as underrepresented in medicine.⁴⁵ We hypothesize that the greater impact may be due to iterative refinements, the maturity of the program, and the addition of nationwide participants who created community and normalized otherwise isolating challenges in medicine.¹³

In a recent advisory,⁴⁶ the US Surgeon General strongly recommends integrating social connection into wellness programs focused on burnout. A major theme that arose from a qualitative analysis¹³ of participants’ experience of online group coaching was an increased sense of connection, even during a time of profound social isolation during the COVID-19 pandemic. Given the digital platform, online group coaching is easily accessed by rural and less resourced programs. Online group coaching is an example of an institutionally provided, individually harnessed tool to build a culture of connection necessary to heal physician burnout.

This study has limitations. Voluntary participation may have created selection bias. This study enrolled only participants who identify as women, so outcomes among individuals who identify as men is unknown. There could have been selection bias of sites, a process that was dependent on leadership buy-in, which is vulnerable to bias. We did not use cluster randomization given our interest in individual rather than group-level changes in outcomes, and we also wanted to avoid loss of statistical power. This study is unique in that the intervention was administered digitally by a centralized entity and not separately at each site as in more traditional clinical trials. However, we may consider cluster randomization in future studies. Contamination was possible because each site had participants in control and intervention groups, although mitigation efforts were implemented (eg, password access to intervention materials).

We had substantial loss to follow-up. Control participants were significantly more likely to respond to the survey than intervention participants, perhaps due to email fatigue (the intervention group received 2emails weekly), or control participants may have been more motivated in anticipation of receiving the intervention. There were racial, gender identity, and PGY differences between follow-up survey responders and nonresponders. This finding may be related to previously established observations that those with higher time constraints (ie, participants in PGY 1) may feel marginalized, and those who are underrepresented in medicine due to their race and/or gender identity may feel less socially connected and less inclined to complete surveys.^47,48 We attempted to assess the effect of missing follow-up survey data on study outcomes with the sensitivity analysis, a statistical technique to evaluate the effect of the missing data (eTable 4 in Supplement 2), which did not find significant outcome differences. Because participants did not use the same login for each coaching call, downloaded podcast, or curriculum module, we were unable to measure engagement or correlate it with outcomes. Additionally, the study team and participants could not be masked, and outcomes could have accrued in part from participant expectations despite providing both groups access to well-being resources as a plausible alternative. Furthermore, we did not evaluate the postintervention effect, which warrants future study.

Compared with GME training as usual, an online 4-month group-coaching program for women physician trainees delivered by certified physician coaches resulted in significant improvement in professional distress and well-being, with an NNT of 11 to mitigate burnout. Group coaching interventions may have a role in mitigating physician trainee burnout and improving well-being along with system-level interventions to improve the work and learning environment.

Accepted for Publication: August 21, 2023.

Published: October 4, 2023. doi:10.1001/jamanetworkopen.2023.35541

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2023 Mann A et al. JAMA Network Open.

Corresponding Author: Tyra Fainstad, MD, Division of General Internal Medicine, Department of Medicine, University of Colorado Anschutz Medical Campus, 13001 E 17th Pl, Aurora, CO 80045 (tyra.fainstad@cuanschutz.edu).

Author Contributions: Dr Fainstad had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Mann, Thibodeau, Dyrbye, Jones, Fainstad.

Acquisition, analysis, or interpretation of data: Mann, Shah, Thibodeau, Dyrbye, Syed, Woodward, Thurmon, Dunbar, Fainstad.

Drafting of the manuscript: Mann, Shah, Thibodeau, Syed, Woodward, Fainstad.

Critical review of the manuscript for important intellectual content: Mann, Shah, Thibodeau, Dyrbye, Woodward, Thurmon, Jones, Dunbar, Fainstad.

Statistical analysis: Mann, Syed, Woodward.

Obtained funding: Mann, Fainstad.

Administrative, technical, or material support: Mann, Shah, Thibodeau, Dyrbye, Syed, Thurmon, Fainstad.

Supervision: Mann, Dyrbye, Jones, Fainstad.

Conflict of Interest Disclosures: Dr Mann reported receiving personal fees from Community Physician Coaching for consulting in development and delivery of a coach training curriculum for COMPASS Guides at the University of Colorado; personal fees and honoraria from the Department of Obstetrics and Gynecology at the University of Tennessee Chattanooga, Johns Hopkins University School of Medicine, University of California Davis School of Medicine, and Vanderbilt School of Medicine; nonfinancial support from the Accreditation Council for Graduate Medical Education which included travel and lodging for a workshop and lectures in 2023; and working as a certified professional coach through an individually owned and operated LLC outside the submitted work. Dr Shah reported working as a professional coach outside their academic role. Dr Dyrbye reported receiving personal fees from Med Ed Solutions; royalties from the Mayo Clinic; grants from the Physicians Foundation, and personal fees and honoraria from the Denver Medical Society, Asante Medical Group, American Medical Association, Accreditation Council for Graduate Medical Education, American College of Emergency Physicians, Midwest Center for Occupational Health and Safety (MCOHS) at the University of Minnesota School of Public Health, Legacy Health, Mayo Clinic, Society of General Internal Medicine, America’s Physician Groups, University of Colorado, Estes Park, Coalition for Physician Well-being, Society of General Internal Medicine, Norwegian National Safety Conference, University of California Davis Health and Leadership in Academic Medicine, University of Rochester School of Medicine and Dentistry, Stanford School of Medicine, McMaster University, Johns Hopkins Medicine, and the University of Michigan outside the submitted work. Dr Woodward reported receiving personal fees from the University of Colorado during the conduct of the study; and working as a professional coach outside the submitted work. Dr Thurmon reported working as an unpaid professional coach with Better Together outside of the submitted work; receiving personal fees from the University of Kansas Department of Obstetrics and Gynecology, the University of Chicago Division of Urology, and the University of Colorado Department of Surgery outside the submitted work. Dr Fainstad reported receiving personal fees from Community Physician Coaching Contract work for developing and implementing an educational coaching curriculum; personal fees and honoraria from Johns Hopkins University, Oklahoma State University, Yale University, Vanderbilt University, University of California at Davis, the American Medical Association, and the Accreditation Council for Graduate Medical Education outside the submitted work. No other disclosures were reported.

Funding/Support: Funding for this study was provided by the University of Colorado Department of Medicine Program for Academic Clinician Educators (to Drs Mann and Fainstad), Division of Hospital Medicine Small Grant (to Drs Mann and Fainstad) and Division of General Internal Medicine Small Grant (to Drs Mann and Fainstad). This material is also based upon work partially supported by a National Science Foundation award (grant number 2041339 to Dr Dyrbye) and the National Institute of Nursing Research (grant number R01NR020362-01 to Dr Dyrbye).

Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors and do not necessarily reflect the views of the National Science Foundation or National Institute of Nursing Research.

Data Sharing Statement: See Supplement 3.

Additional Contributions: We thank the trainees who participated in the study as well as the following certified coaches who provided unpaid contributions to Better Together Physician Coaching in the form of curricular content and/or live coaching: Peter Baum, DO (San Ysidro Health), Elisa Boden, MD (Oregon Health and Science University), Gretchen Bruno, MD (SCL Health, Kaiser Permanente, Denver), Rebecca Caldwell, MD (Spectrum Health, MI), Tonya Caylor, MD (University of Washington Family Medicine Residency), Yashika Dooley, MD, MS (Uniformed Services University of the Health Sciences), Michael Hersh, MD (Northwest Community Healthcare), Tricia James, MD (Providence Portland), Kanapa Kornsawad, MD (University of Texas Health San Antonio), Ursula Lang, MD, PhD (University of California at San Francisco), Kara Loewentheil, JD (KL Coaching, Inc.), Angie Mathai, MD (East Carolina University), Megan R Mistry, DO (Massachusetts General Brigham Newton-Wellesley Hospital), Junaid Niazi, MD, MPH (Prosperous Life MD), Saidie Rodriguez, MD (Emory University), Wendy Schofer, MD (Children's Hospital of the King's Daughters), Rachel Swigris, DO (University of Colorado), Shannon Weinstein, DO (Broward Health), Carole Warde, MD (University of California at Los Angeles), Danielle Wilhour, MD (University of Colorado), Sonia Wright, MD (Midwest Radiology). Mark Earnest MD, PhD (University of Colorado) provided support, mentorship, and sponsorship.

Additional Information: The Better Together (BT) curriculum is owned and delivered by the University of Colorado, a not-for-profit school of medicine.