Who Benefits Most? A Multisite Study of Coaching and Resident Well-being

Abstract

Background

Coaching has been shown to improve resident well-being; however, not all benefit equally.

Objective

Assess predictors of changes in resident physician well-being and burnout in a multisite implementation of a Professional Development Coaching Program.

Design

Pre- and post-implementation surveys administered to participant cohorts at implementation sites in their intern year. Effect size was calculated comparing pre- and post-intervention paired data.

Participants

In total, 272 residents in their intern year at five internal medicine residency programs (Boston Medical Center, University Hospitals Cleveland Medical Center, Duke University, Emory University, Massachusetts General Hospital). Analyses included 129 residents with paired data.

Interventions

Interns were paired with a faculty coach trained in positive psychology and coaching skills and asked to meet quarterly with coaches.

Main Measures

Primary outcomes included Maslach Burnout Inventory depersonalization (DP) and emotional exhaustion (EE) subscales, and the PERMA well-being scale. Key predictors included site, demographics, intolerance of uncertainty, hardiness-resilience, gratitude, and coping. Program moderators included were reflection, goal setting, and feedback.

Key Results

Well-being (PERMA) changed from baseline to follow-up in all participants; females showed a decline and males an increase (−1.41 vs. .83, p = 0.04). Self-reflection was associated with positive change in PERMA (mean positive change 1.93, p = 0.009). Burnout (EE) declined in non-Hispanic white residents vs. Black/Asian/Hispanic/other residents (−1.86, p = 0.021). Burnout improved with increased goal setting.

Conclusion

Coaching programs should consider tailored approaches to support residents whose well-being is impacted by gender and/or race, and who have higher intolerance of uncertainty and lower resilience at baseline. Coaching skills of goal setting and reflection may positively affect interns and teach coping skills.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11606-021-06903-5.

INTRODUCTION

The stresses and challenges of medicine residency threaten learners’ well-being and burnout risk is high. Concerns about resident well-being have driven the National Academy of Medicine’s Committee on Systems Approaches to Improve Patient Care by Supporting Clinician Well-Being call to action, where they urged healthcare organizations to adopt and apply principles to improve the work environment and balance job demands and job resources.^{1, 2} A key recommendation included professional development of faculty to promote positive role model behaviors.^{3, 4}

Well-documented challenges affecting residents include excessive workload, inefficient practice, work-life integration, lack of autonomy and control, loss of meaning in work, disconnection with and structural inequities within organizational culture, and high stress environment.^5–8 These challenges affect the learning environment, as well as resident motivation, decision-making, quality of care, and communication at work.^{9, 10} In response, residents use a variety of coping strategies, often without prior experience or reflection to help them understand what may or may not work for them. Physicians with lower resilience and who report lacking the skills to help them function well under stress are more likely to experience burnout.^{10, 11}

Learning effective coping strategies may improve burnout. Prior research indicates that both individual-focused and systems-based strategies can improve well-being.^{4, 12} Few examples exist of programmatic interventions that combine both approaches as levers of change. Coaching is an example of such a resource to support residents in managing stress and adapting to change through development of coping behaviors, increasing positive emotions, and fostering positive relationships with faculty.^{5, 13–15} Coaching is a relationship between an individual coachee and a coach with a specific purpose to help the coachee design and achieve professional or personal goals.^{14, 15} Coaching allows for both structured and unstructured conversation, emphasizes active listening, and enhances strengths.^13–16 The coaching literature suggests that the process of coaching is highly complex, and multiple moderating and mediating factors in the individual and the approach contribute to the success of a coaching intervention (coachee motivations and expectations, the coaching relationship, using an internal vs. external coach, coachee self-efficacy).¹³ To our knowledge, the exploration of these mechanisms is limited in the literature.

By engaging faculty trained in positive psychology and coaching skills, we endeavored to develop and disseminate a systems-based solution focused on individuals to improve physician well-being in the Professional Development Coaching Program (PDCP). Our initial work reported on a single-center positive psychology coaching program and demonstrated that in those residents who participated and reported a positive coaching experience, their opportunities to reflect increased, as did coping and relationship skills.¹⁷ Those residents with poor or fair opportunities to reflect exhibited higher emotional exhaustion scores on the Maslach Burnout Inventory (MBI).^{18, 19} While it was clear that some benefitted from the program more than others, program size limited a more comprehensive assessment of other factors that might be used to predict who came into the program with better coping and resiliency practices and might benefit more or less from the program.

In 2016–2017, we expanded the PDCP to five institutions to assess the generalizability of a positive psychology coaching program. In this study, we explored how coachee attributes and behaviors at baseline contribute to their well-being and impact their experience with a coaching program. In baseline data of participants, there was no correlation between burnout (using MBI) and a well-being scale known as PERMA (Positive Emotion, Engagement, Relationships, Meaning, and Accomplishments).^{20, 21} This prompted exploration of the associations between participant characteristics and changes in well-being and burnout in a multivariate model to assess how to maximize the benefit of the coaching program. Using primary outcome variables of MBI and PERMA, and controlling for resident coachee demographics and program site, our predictors included skills the PDCP was designed to enhance through the intervention (intolerance of uncertainty, hardiness/resilience, gratitude, and coping) and previously identified outcomes of the coaching relationship (self-reported practice of goal setting, reflection, and feedback).

METHODS

Design

This was a prospective, pre-post study of 272 interns in internal medicine training programs at five institutions (Boston Medical Center, University Hospitals Cleveland Medical Center, Duke University, Emory University, and Massachusetts General Hospital) who were enrolled in the previously described PDCP during the 2016–2017 academic year.¹⁸ Participants were surveyed in October of intern year to assess baseline burnout and well-being, as well as modifiers/mediators and key skills necessary to achieve well-being. October was chosen as interns would have 3+ months of work experience to ensure measurement reflected their internship baseline, rather than July which might mistakenly reflect well-being and burnout from medical school. They were then surveyed with the same tools in May of intern year.

Survey data are available for 235 interns at baseline; paired data are available for 129 interns. Effect size was calculated comparing pooled baseline samples to pooled end of year samples and calculated comparing pre- and post-intervention paired data.

Intervention

The PDCP has been previously described.¹⁸ All interns were paired with a faculty coach at their institution and outside their field of specialty interest. Coaches were provided a 3-h skills-based training in positive psychology and coaching principles and given curricular guides for 1:1 meetings with their coachees. Each coach was assigned 1–2 coachees; pairs were expected to meet quarterly. A typical coaching meeting was 45–60 min in length. Themes and skills to practice reflection, processing feedback, and goal setting were woven throughout the program. Coaching meetings focused on strengths evaluation, resiliency, and constructing a framework of positive well-being (PERMA).¹⁸ Appendix Figure 1 outlines the curriculum. Survey and program participation were voluntary; there was no consequence for not meeting with their coach. Dr. Palamara led the faculty coach trainings at each site. Each site had an appointed champion(s) present for trainings, creating a “train the trainer” model for sustainability beyond the pilot. Champions were in regular communication with Dr. Palamara to oversee the implementation and administration of the program. Site-specific curricular materials were created and stored in a cloud-based system for coaches to access for meetings.

Surveys were administered by a research team at MGH. Sites did not have access to the data but were asked to encourage survey completion at their site. Sites determined the use of incentives for survey participation; the amount and type varied.

Measurements

Survey metrics were grouped into 3 categories: (1) study outcomes—MBI score for burnout (7-point frequency [0–6], 9 items for EE and 5 items for DP, range 0–54 for EE and 0–30 for DP); PERMA score as a metric of well-being (5-point frequency [1–5], 15 items, range 0–75); (2) baseline predictors (modifiers)—program site and resident demographics, the Intolerance of Uncertainty Scale (IUS, 5-point frequency [1–5], 12 items, range 0–60); HRS or hardiness-resilience scoring (DRS-15, 4-point frequency [0–3], 15 items, range 0–45), gratitude (GR-2, 7-point frequency [1–7], 2 items from the GQ-6 scale, range 0–14) and a 3-item measure of coping (Measurement of Current Status; MOCS3, 5-point frequency [1–5], 3 items, range 0–15), self-reported reflection and goal setting; and (3) post program predictors (mediators)—self-reported reflection, goal setting during follow-up.^{19, 21–26}

Statistical Analysis

Participant characteristics were summarized for all respondents to the baseline survey grouped by completion of follow-up survey. We then restricted the remaining analysis to those who answered both baseline and follow-up surveys. Due to considerable variation in resident survey response and participation level by site, we controlled for program site in all analyses. For the primary outcomes of burnout and well-being, a positive change in PERMA indicates improvement in well-being. For EE and DP, a negative change indicates improvement in burnout. For independent variables, scoring measures that were continuous (HRS, IUS, MOCS, and GR2) were divided the scores into quartiles and dichotomized further as top quartile vs. lower quartiles.

We performed bivariate analysis for the primary outcomes to measure three differences in scores outcomes for participants who answered both baseline and follow-up surveys. For each predictor, we performed a one-way ANCOVA test while adjusting for baseline score of each respective outcome. The sample size limited our ability to use all possible predictors in our models; we used our pre-specified predictors based on prior knowledge and bivariate analyses to perform multiple linear regression modeling for the three outcomes separately to examine predictors including baseline score, demographics (gender, race, site), goal setting, HRS, GR-2, IUS at baseline, and reflection at follow-up. A two-sided p-value <0.05 was considered statistically significant. All analyses were conducted using SAS version 9.4 (SAS Institute, Cary, NC).

RESULTS

A total of 235 participants answered the baseline survey. Of those, 129 responded to the post-survey. In Table 1, we compared participant characteristics and survey responses of responders with non-responders. To examine selection bias, we compared demographics between responders and non-responders and found no significant differences in baseline characteristics. As previously noted, we controlled for site in all multivariate analyses.

Table 1.

Comparison of Participant Characteristics Between Those Who Did Not Complete Follow-up Survey and Those Who Did Complete the Follow-up Survey

	Lost to follow-up (pre only)		Respondents (pre and post)		p
Response	N	%	N	%
All	106	100.0	129	100.0
Gender					0.19
Male	45	47.4	72	56.3
Female	50	52.6	56	43.8
Race/ethnicity					0.17
Non-Hispanic white	53	57.0	81	63.3
Black	4	4.3	5	3.9
Asian	32	34.4	28	21.9
Other	2	2.2	7	5.5
Hispanic	2	2.2	7	5.5
Site					<0.0001
A	22	20.8	21	16.3
B	29	27.4	16	12.4
C	12	11.3	22	17.1
D	15	14.2	63	48.8
E	28	26.4	7	5.4
Goal setting					1.00
Never/rarely	38	40.0	50	40.0
Sometimes/frequently	57	60.0	75	60.0
Opportunity to reflect					0.49
Poor/fair	24	23.3	22	17.1
Good	56	54.4	75	58.1
Excellent	23	22.3	32	24.8
Receiving feedback					0.64
Poor/fair	21	20.6	32	24.8
Good	65	63.7	81	62.8
Excellent	16	15.7	16	12.4
Hardiness-resilience score					0.81
Low (<Q4)	66	75.9	92	77.3
High (Q4)	21	24.1	27	22.7
Intolerance of uncertainty score					0.094
Low (<Q4)	62	69.7	98	79.7
High (Q4)	27	30.3	25	20.3
MOCS: Coping score					0.37
Low (<Q4)	72	76.6	89	71.2
High (Q4)	22	23.4	36	28.8
Gratitude score					0.72
Low (<Q4)	71	75.5	97	77.6
High (Q4)	23	24.5	28	22.4
Burnout					0.25
No	46	63.9	66	55.5
Yes	26	36.1	53	44.5
PERMA overall score					0.10
Low (36–59)	36	39.1	60	50.4
High (60–75)	56	60.9	59	49.6
Meetings					<0.0001
0–1	28	26.4	28	21.7
2–3	21	19.8	73	56.6
4–5	3	2.8	26	20.2
Missing	54	50.9	2	1.6

Abbreviations: LTFU, lost to follow-up; Q, quartile; MOCS, Measurement of Current Status; PERMA, Positive Emotion, Engagement, Relationships, Meaning, and Accomplishments

In Table 2 and Figure 1, we show the bivariate analysis to assess associations of each predictor with the differences in scores between baseline and follow-up measures of primary outcomes PERMA and MBI subscales of EE and DP. Among all respondents, there were declines in PERMA, MBI, and EE scores, and a small increase in DP. The change is calculated from subtracting baseline score from the follow-up score. A positive change in PERMA and a negative change in EE or DP indicate improvement.

Table 2.

Examination of Change (Post-Pre) in Outcome Scores for PERMA, Emotional Exhaustion, and Depersonalization for Each Predictor. This Analysis Is Limited to the Number of Persons Providing Complete Data

	Diff* in PERMA			Diff* in emotional exhaustion			Diff* in depersonalization			Diff* in MBI** score
	N	Mean	p***	N	Mean	p***	N	Mean	p***	N	Mean	p***
All	116	−0.16		118	−0.92		117	0.22		116	−0.73
Gender			0.040			0.50			0.95			0.73
Male	65	0.83		64	−0.50		63	−0.06		63	−0.68
Female	51	−1.41		53	−1.62		53	0.49		52	−1.08
Race			0.28			0.021			0.64			0.10
Black/Asian/Hispanic/other	43	−0.47		43	0.47		41	0.22		41	0.59
Non-Hispanic white	73	0.03		74	−1.86		75	0.17		74	−1.66
Site			0.23			0.81			0.70			0.75
A	20	−1.70		21	−1.38		21	0.38		21	−1.00
B	12	1.42		14	−1.36		14	0.43		14	−0.93
C	18	1.72		20	−0.30		20	−0.20		20	−0.50
D	60	−0.57		61	−0.66		60	0.43		59	−0.29
E	6	0.33		2	−7.00		2	−5.00		2	−12.00
Opportunity to reflect (pre)			0.95			0.57			0.42			0.57
Poor/fair	20	0.00		20	−1.70		21	0.52		20	−1.05
Good	68	0.07		68	−0.66		66	−0.18		66	−0.94
Excellent	28	−0.82		30	−0.97		30	0.90		30	−0.07
Opportunity to reflect (post)			0.009			0.10			0.21			0.090
Poor/fair	18	−1.72		18	0.78		19	0.16		18	1.06
Good	68	−0.66		69	−0.49		67	0.64		67	0.06
Excellent	30	1.93		31	−2.84		31	−0.65		31	−3.48
Goal setting (pre)			0.77			0.056			0.099			0.046
Never/rarely	47	0.60		47	−3.45		47	−1.34		46	−4.80
Sometimes/frequently	69	−0.67		71	0.76		70	1.27		70	1.94
Goal setting (post)			0.29			0.96			0.45			0.69
Never/rarely	20	−1.05		22	−1.59		22	0.14		22	−1.45
Sometimes/frequently	96	0.03		96	−0.76		95	0.24		94	−0.56
Hardiness-resilience score			0.29			0.034			0.38			0.074
Low (<Q4)	89	−0.27		87	−0.55		86	0.22		86	−0.42
High (Q4)	24	0.17		25	−2.28		26	−0.23		25	−2.44
Gratitude score			0.31			0.62			0.081			0.29
Low (<Q4)	88	−0.01		91	−1.19		90	0.29		89	−0.96
High (Q4)	28	−0.61		27	0.00		27	0.00		27	0.00
Intolerance of uncertainty score			0.79			0.002			0.096			0.005
Low (<Q4)	93	−0.18		93	−1.57		93	0.18		92	−1.36
High (Q4)	23	−0.04		23	1.13		22	0.05		22	0.91
Coping score			0.90			0.26			0.86			0.53
Low (<Q4)	81	0.31		84	−1.25		82	−0.26		82	−1.60
High (Q4)	35	−1.23		34	−0.09		35	1.34		34	1.35

*Difference is calculated post-pre

**MBI score is sum of emotional exhaustion and depersonalization

***p-value adjusted for baseline score

Figure 1.

Graphical display of selected predictors that are associated with change in PERMA, emotional exhaustion, and depersonalization scores. PERMA, Positive Emotion, Engagement, Relationships, Meaning, and Accomplishments; HRS, Hardiness-Resilience Scale; IUS, intolerance of uncertainty; GR2, Gratitude Scale. For PERMA, males indicated an average positive change (0.83), and females a negative change (−1.41). Participants who reported their post-program opportunity to reflect was “Excellent” showed mean positive change in PERMA (1.93) while those who answered “Poor/Fair” and “Good” had mean decreases (−1.72, −0.66). Examining burnout in terms of EE, those who were non-Hispanic white had a mean decrease of EE (−1.86). Higher HRS and lower IUS at baseline are associated with decreases in EE on average (−2.28, −1.57). Those who reported that they “Never/Rarely” practiced goal setting at baseline had improved DP with average reduction of −1.34. Lower GR2 and lower IUS at baseline were associated with increases in DP (0.29, 0.18).

PERMA changed from baseline to follow-up in all participants, with significant differences between male and female coachees—males indicated a small positive change, females a negative change in PERMA (males 0.83 vs. females −1.41, p = 0.04). Of several other predictors, few were significantly associated with a change in PERMA except for participant-reported self-reflection at follow-up, a key program skill. Participants who reported their post-program opportunity to reflect was “Excellent” showed mean positive change in well-being (PERMA) (1.93, p = 0.009).

We also examined bivariate associations of changes in the MBI and its subscales. Examining burnout, those who were non-Hispanic white had a mean decrease of EE (−1.86, p = 0.021). Those who reported that they “Never/Rarely” practiced goal setting at baseline had improved EE and DP during the program year with a respective average reduction of −3.45 (p = 0.056) and − 1.34 (p = 0.099). Also, for EE, higher HRS and lower IUS scores at baseline were associated with lower burnout on average (−2.28, p = 0.034; −1.57, p = 0.002).

Given our hypothesis that program site, participant characteristics, and other factors might influence participant experience and receptivity to program mediators, we conducted multiple linear regression modeling to examine these multiple factors in predicting our primary outcomes (see Table 3). Covariates included in these models included baseline scores, gender, race (Black/Asian/Hispanic/other vs. non-Hispanic white), site, post-opportunity to reflect, baseline goal setting, HRS, gratitude score, and IUS score, all of which were significant in bivariate analyses. In general, if a participant had higher well-being (PERMA) at baseline, there was less improvement seen (−0.407, p = <0.001). Those whose resilience improved as measured by higher HRS score (vs. lower at baseline) also experienced a positive change in well-being (3.474, p = 0.008). However, controlling for all factors, females see a significant negative change (−2.475, p = 0.012) in well-being (PERMA) compared to males. Participants with improved post-reflection (report “Excellent” vs. poor/fair reflection at baseline) experienced a positive change (4.025, p = 0.011).

Table 3.

Modeling of Change (Post-Pre) in Outcome Scores for PERMA, Emotional Exhaustion (EE), and Depersonalization (DP) for Pre-specified Predictors

	Outcome
	Diff in PERMA		Diff in EE		Diff in DP		Diff in MBI
Predictors	Estimate	p-value	Estimate	p-value	Estimate	p-value	Estimate	p-value
Baseline score	−0.407	<0.0001	−0.393	<0.0001	−0.476	<0.0001	−0.430	<0.0001
Gender
Male (REF)	-	-	-	-	-	-	-	-
Female	−2.475	0.012	−0.706	0.56	0.201	0.81	−0.395	0.83
Race
Black/Asian/Hispanic/other	−0.870	0.40	2.608	0.046	0.056	0.95	2.643	0.17
Non-Hispanic white (REF)	-	-	-	-	-	-	-	-
Site
A	−1.051	0.44	−1.194	0.49	−0.924	0.43	−2.149	0.40
B	2.139	0.20	0.289	0.88	0.292	0.83	0.586	0.84
C	1.777	0.20	−0.133	0.94	−0.482	0.68	−0.597	0.82
D (REF)	-	-	-	-	-	-	-	-
E	−0.040	0.99	−5.933	0.20	−5.753	0.068	−11.58	0.093
Opportunity to reflect (post)
Poor/fair (REF)	-	-	-	-	-	-	-	-
Good	0.898	0.52	−1.058	0.55	−0.346	0.77	−1.378	0.60
Excellent	4.025	0.011	−3.821	0.052	−1.858	0.16	−5.715	0.049
Goal setting (pre)
Never/rarely	−0.067	0.95	−2.228	0.094	−1.598	0.076	−3.750	0.058
Sometimes/frequently (REF)	-	-	-	-	-	-	-	-
Hardiness-resilience score
Low (<Q4, REF)	-	-	-	-	-	-	-	-
High (Q4)	3.474	0.008	−4.469	0.003	−1.021	0.30	−5.579	0.012
Gratitude score
Low (<Q4, REF)	-	-	-	-	-	-	-	-
High (Q4)	1.573	0.21	0.495	0.74	−1.865	0.065	−1.438	0.52
Intolerance of uncertainty score
Low (<Q4, REF)	-	-	-	-	-	-	-	-
High (Q4)	−1.039	0.41	6.022	0.0002	2.555	0.025	8.626	0.0005

Participants experiencing higher burnout at baseline, measured by higher baseline EE and DP scores, were more likely to improve and experience lower burnout in the program year (−0.393, p = <0.001; −0.476, p = <0.001). Those who are Black/Asian/Hispanic/other were significantly more likely to have an increase in EE (2.608, p = 0.046) compared to non-Hispanic white participants. Higher HRS score at baseline is significantly associated with improvement in EE (−4.469, p = 0.003), whereas higher IUS score at baseline is significantly associated with more EE (6.022, p = 0.0002) and DP (2.555, p = 0.025) when compared to lower scores in HRS and IUS, respectively. Respondents with higher gratitude scores vs. lower scores at baseline are more likely to report an improvement in DP (−1.865, p = 0.065).

Finally, in another aspect of the survey, we included direct measures of the residents’ own perceptions of whether coaching had helped them cope with key stressors in internship year (work-life balance, cultural competence, working relationships, work hour restrictions, administrative burdens, self-confidence, and information processing). Appendix Table 1 reports whether interns “definitely” or “somewhat” agreed or disagreed with the following statement: “In previous research, the following have been noted as major challenges for trainees. For each one, please indicate if you believe the coaching program has improved your ability to cope.” The strongest agreement was seen with self-confidence (89.3%), working relationships (89.1%), information processing (84.7%), cultural competence (75.7%), work hour restrictions (73.9%), and work-life balance (73.8%). The lowest agreement was with administrative burdens (60.5%).

DISCUSSION

These data are taken from a multisite implementation of the PDCP, a program targeted at residents in medicine to improve well-being and decrease burnout through positive psychology coaching and the introduction of new skills in goal setting, reflection, and receiving feedback. These analyses reveal the complex relationships that are likely to exist in the lives of new physicians and in the global measures of burnout and well-being. If we look at the entire cohort, those outcomes do not change very much during this program year, though the overall minor changes mask significant differences by gender and race, as female residents are more likely to experience decreased well-being and residents who are Black/Asian/Hispanic/other are more likely to experience burnout (EE and DP). Those with higher scores in resilience and gratitude at baseline do better in our primary outcomes, and those with high intolerance of uncertainty experience significantly higher EE and DP.

The positive psychology component of this program allows for the opportunity to explore the modifiable traits that can impact an individual’s well-being and burnout, including effective response to stress and tolerance of uncertainty. Goal setting, reflection, and receiving feedback are core components of a coaching relationship. In our bivariate and multivariate analyses, the increased practice of reflection was significantly associated with improved well-being and goal setting practices with reduced burnout. This finding is consistent with our prior research, which linked high EE with reflecting less often.¹⁷ Reflection is a core coaching skill emphasized in trainings; which helps coachees explore their experiences, motivations, positive emotions, and strengths.^{13, 14} Based on this analysis, we propose that a coaching program can offer deliverables to support resident well-being, including time, space, and tools for setting goals, reflecting, and receiving feedback.

Interns in this study showed high resilience at baseline, similar to West et al.’s study.¹¹ The finding of better EE and PERMA in those with the highest resilience suggests learning skills to help cope with stress effectively may be useful to improve well-being for interns. Previously, the PDCP has been shown to provide coping skills in areas of greatest stress for interns, such as work-life balance, working relationships, information processing, work hours, and self-confidence.¹⁷ Greater than two-thirds of respondents agreed that this program helped them with coping skills to respond to stressors in these areas, suggesting that this program can be useful to develop skills to improve resilience, which is shown here to be strongly protective against burnout. Managing uncertainty, a key skill throughout physician practice, and developing coping skills through reflection opportunities are target areas for further evolution of the coaching program, and we believe this finding is useful in helping residency programs address well-being.

In previous evaluations of this coaching program, our ability to analyze gender, race, and ethnicity was limited. In this analysis, we learn that Black/Asian/Hispanic/other residents experience higher burnout than their non-Hispanic white peers, and male residents experience higher well-being than female residents. This warrants further exploration in future studies to discover if different coaching approaches are needed to address structural racism and gender bias within the coaching intervention or within organizations. In recent years, multiple studies have explored resident experience of bias.^27–30 These experiences can lead to increased stress, lower well-being, impaired coping, and increased burnout.^{31, 32} The coaching relationship may provide an important layer of support for residents to address the role implicit bias, imposter phenomenon, and microaggressions may have in daily intern experiences.

We acknowledge several limitations in this multisite study. This study is not a randomized controlled trial (RCT). This study is also limited by a low survey response rate that varied across sites. Sites are heterogenous in their experiences and local culture, which may impact generalizability in understanding and applying these results. While coaches at all sites were trained by the same subject matter expert, the implementation and support at each site were variable. Based on this experience, we recommend institutions wishing to adopt such a program have an appointed faculty champion with administrative support and incentives to improve participation in measurements. We also do not have a means to evaluate the efficacy of the coach themselves, understanding that this is high-level communication skill that we are teaching over the course of a 3-h workshop. Future studies on the efficacy of the coach as well as the role the coach’s well-being may play in the relationship would be useful. Additionally, in this study, we address differences in outcomes based on race and ethnicity though we are limited by the low number of interns who are under-represented in medicine (URiM). A future study with greater number of URiM interns can help explore these associations and outcomes further.

A potential limitation to these findings might be the perceived impact that small differences in these scales have on physician burnout or well-being. We believe the subtle nuances within these parameters may provide insight into the multifactorial nature of physician well-being. While this cannot be answered in a prospective cohort study such as this, we anticipate that an RCT might further elucidate these subtleties and the impact of coaching. One hypothesis that could be explored in an RCT is whether no change is actually an improvement compared to someone who is not receiving coaching. Given the many ways that internship can change an individual in this formative time, it may be that stability is in fact success, given what is known about increasing rates of burnout across the trajectory of a physician’s career.^{3, 4, 11}

The costs of this program were minimized for sustainability. Coaching materials have been published to aid residencies in onboarding a similar program.¹⁸ Still, we recognize that not all programs can implement a program like this. For these programs, we believe that incorporating coaching into existing advising and mentorship programs would benefit resident well-being. The question of the difference between coaching and mentoring is often raised, and why coaching is needed if well-structured mentoring programs exist. It should be noted that this coaching program was initiated in residencies with existing mentorship programs, who were seeking additional ways to support trainee well-being. Coaching provides a non-evaluative, safe space where the personal and professional journeys of these trainees can co-exist and be explored.

Residency programs should consider formal assessments of well-being and predictive factors early in the intern year to help target support for residents who may not arrive at internship with coping, resilience, and other skills. Organizational investments in faculty education in coaching and communication skills may help residents build in time and space for goal setting and reflection. Organizations also need to measure and understand factors in their culture that may contribute to disparities in the resident experience by race and gender. Providing support, coaching, or otherwise, that targets tolerance of uncertainty, resilience, and reflection, may improve resident burnout and well-being.

Abbreviations

PDCP

Professional Development Coaching Program

DP

Depersonalization

EE

Emotional exhaustion

MBI

Maslach Burnout Inventory

PERMA

Positive Emotion, Engagement, Relationships, Meaning, and Accomplishments

MGH

Massachusetts General Hospital

IUS

Intolerance of uncertainty

HRS

Hardiness resilience score

DRS-15

Dispositional-Resilience Scale-15

GR-2

2-question Gratitude Questionnaire

GQ-6

6-question Gratitude Questionnaire

MOCS3

Measurement of Current Status

RCT

Randomized controlled trial

URiM

Under-represented in medicine

Funding

The Arthur Vining Davis Foundation and The Physicians Foundation each provided $100,000 of funding to support the investigation and dissemination of this program.

Data Availability

The datasets during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethical Approval

This program and its evaluation were exempt by the Institutional Review Board with exemption #45 CFR 46.101(b) (1) Research conducted in established or commonly accepted educational settings, involving normal educational practices.

Conflict of Interest

The authors declare that they do not have a conflict of interest.