Impact of a Wellness Leadership Intervention on the Empathy, Burnout, and Resting Heart Rate of Medical Faculty

Abstract

Objective

To evaluate the efficacy of a wellness leadership intervention for improving the empathy, burnout, and physiological stress of medical faculty leaders.

Participants and Methods

Participants were 49 medical faculty leaders (80% physicians, 20% basic scientists; 67% female). The 6-week course was evaluated with a 15-week longitudinal waitlist-control quasi-experiment from September 1, 2021, through December 20, 2021 (during the COVID-19 pandemic). We analyzed 3 pretest-posttest-posttest and 6 weekly survey measurements of affective empathy and burnout, and mean=85 (SD=31) aggregated daily resting heart rates per participant, using 2-level hierarchical linear modeling.

Results

The course found a preventive effect for leaders’ burnout escalation. As the control group awaited the course, their empathy decreased (coefficient_Time=−1.27; P=.02) and their resting heart rates increased an average of 1.4 beats/min (coefficient_Time=0.18; P<.001), reflecting the toll of the pandemic. Intervention group leaders reported no empathy decrements (coefficient_Time=.33; P=.59) or escalated resting heart rate (coefficient_Time=−0.05; P=.27) during the same period. Dose-response analysis revealed that both groups reduced their self-rated burnout over the 6 weeks of the course (coefficient_Time=−0.28; P=.007), and those who attended more of the course showed less heart rate increase (coefficient_{Time∗Dosage}=−0.05; P<.001). In addition, 12.73% of the within-person fluctuation in empathy was associated with burnout and resting heart rate.

Conclusion

A wellness leadership intervention helped prevent burnout escalation and empathy decrement in medical faculty leaders during the COVID-19 pandemic, showing potential to improve the supportiveness and psychological safety of the medical training environment.

Before the advent of the COVID-19 pandemic, physician occupational burnout was recognized as a global problem, with studies showing an average burnout rate of 20%-30% of physicians.¹ Occupational burnout is a serious syndrome of ill-being, comprising emotional exhaustion, cynicism, and lowered professional efficacy; it takes months or years for physicians to develop clinically elevated burnout.² During 2020, research found 50% of health care professionals displaying symptoms of depression, 20%-40% symptoms of anxiety, and 30% symptoms of insomnia.³ US physicians surveyed in late 2021, almost 2 years into the COVID-19 pandemic, reported 40% higher emotional exhaustion and 60% higher depersonalization than that in a similar sample in 2020.⁴ Once physicians reach burnout, their distress is severe enough to predict subsequent high cholesterol, type 2 diabetes, coronary heart disease, hospitalization, and mortality before the age of 45 years^5,6 and to contribute to motor vehicle accidents and suicide.⁷

Although there are geocultural variations in the nature of pressures on physicians,⁸ some common systematic aspects of medical care are found to contribute to physician burnout.⁹ Excessively high workload; long on-call periods; night shifts; time pressure; low work group cohesion; reduced professional autonomy; lower financial reward; and working in the contact-intensive subspecialties of critical care, emergency, oncology, and internal medicine^1,7 are associated with greater physician burnout. These conditions deplete physicians’ cognitive, emotional, and physical resources, and the sheer physical work hours (often in excess of 60 hours per week¹⁰), and excessive administrative duties block physicians from taking the time for restorative activities and recovery.11, 12, 13

The crisis of physician burnout threatens not only their well-being but also the well-being of their patients and the efficacy of the broader health care system. The cognitive depletion and personal withdrawal associated with burnout can lead to increased minor and major medical errors.^10,14, 15, 16, 17 The final stage of burnout is reduced professional efficacy,^18,19 which in turn leads to lower objectively measured work effort²⁰ and intentions to leave the profession, which then sparks a vicious cycle of short-staffing and expertise drain within the health care system.

Recognizing the burnout crisis among physicians, there have been calls for a shift in the culture of medical training. Studies have shown that burnout first emerges during medical school,²¹ with physical cellular aging occurring at 6 times the normal rate during the residency/internship year.²² At the same time, empathy—the critical ability to connect with one’s supervisor, peers, and patients—plummets during medical school^23,24 and residency.²⁵ In fact, residents show declines in empathy and the quality of communication with patients during a single long-call shift (24-30 hours²⁶).

Scholars have argued that physician burnout is a system problem that needs system solutions.²⁷ A “wellness leadership” approach to medical training would emphasize skills in caring for people and relationships, including oneself, patients, and colleagues, and inspiring change.²⁸ This new model of medical education would aim to instill both work habits and supervisory relationships that would prevent burnout in learners and future physicians. Regarding work habits, leaders’ own stress management is critical and trainable. Well-being interventions focused on cognitions and behaviors, including self-care, reduce physician leaders’ own burnout²⁹ and that of their team.³⁰ Tailored training on work stress recovery techniques has been particularly effective31, 32, 33 even up to 6 months later.³¹ Regarding supervisory climate, research shows that social support from supervisors³⁴ and peers³⁵ can prevent stress escalating to burnout and reduce physician intentions to leave the profession.36, 37, 38 Empathy in particular is a strong predictor of supportive attitudes and behaviors and positive emotional responses from colleagues³⁹ and is a trainable competency.⁴⁰

On the basis of this research, we delivered an evidence-based wellness leadership training intervention course to medical faculty leaders. Our course trained 5 specific elements to cultivate leader capacity to focus on relationships and inspire change: leader core values,⁴¹ self-care,²⁸ work stress recovery,³¹ empathic leadership skills,⁴⁰ and psychological flexibility⁴¹ (Table 1). We tested whether the course can (1) decrease leaders’ burnout, (2) improve their empathy, and (3) reduce their general stress responsivity, in comparison with a waitlist-control group and those who did not actively participate. We were ultimately most concerned about the implications of burnout and empathy for the medical education culture, with resting heart rate (RHR) not having been previously reported as a stress intervention outcome and, therefore, more exploratory.

Table 1.

Wellness Leadership Intervention Elements

Program Week	Program Elements	Program Content
		Workshops: 75 min/online: small group lecture, activities, discussion	Coaching: 30 min/online: one-on-one with content expert
1	•Core values •Self-care	ACT Matrix and wellness leadership introduction •Identifying leader and team identity and core values •Leading a wellness culture shift •Behaviors moving towards and away from core values
2	•Self-care •Work stress recovery	Charging batteries as a leader •Executive cognitive function, stress, and wellness •Identifying recharge behaviors •Goals for and barriers to recharge behaviors •How to promote system change toward a culture of self-care
3	•Self-care •Work stress recovery	Work stressors and intervention plans •System causes of burnout in medicine; wellness culture shift •Mapping individual stress triggers and responses •ACT Matrix for leader stress intervention plan •Fostering stress recovery as a leader
4	•Empathic leadership •Psychological flexibility	Empathic and supportive leadership •Empathy, medicine, and burnout: evidence •Fostering psychological safety and social support •Empathy style feedback •Psychological flexibility to modify empathic behavior
5	•Empathic leadership •Psychological flexibility	Building a village •Asking for and accepting help as a leader •Attachment styles and trauma-informed leadership •Supportive leadership skills •System-level changes to promote psychological safety
6	•Empathic leadership •Psychological flexibility	Leader conflict handling •Conflict handling to support relationships and wellness •Identifying shared purpose in conflict •ACT Matrix for conflict style feedback •Tools for continuing a wellness leadership culture shift

Participants and Methods

We used a nonrandom waitlist-control quasi-experimental study conducted from September 1, 2021 through December 20, 2021.⁴² Medical faculty leaders self-selected into the intervention (trained first) vs waitlist-control (trained second) groups. In week 1, participants received a Fitbit device and completed a baseline survey; in weeks 2-7, the intervention group completed the course; in week 8, all participants completed a posttest survey; in weeks 9-14, the waitlist-control group completed the course; in week 15, all participants completed a final survey. Complete participation included 3 full-measure surveys (pretest-posttest-posttest) and 6 brief weekly surveys (during their active course).

Study Participants and Recruitment

Participants included 49 faculty leaders from 19 of the 21 departments in science and medicine (n=10 or 25.64% scientists, n=39 or 79.60% physicians; n=33 or 67.35% female; n=42 or 85.71% White) within a Faculty of Medicine in Canada (intervention group: n=22 or 44.90%; waitlist-control group: n=27 or 55.10%). The Assistant Dean of Faculty Wellness invited participation through an email to faculty with broad influence on the learning environments of physicians and scientists. The university researchers handled scheduling, inquiries, and reminders. Fitbits were an incentive, but research participation was voluntary. Ten (of the 49) participants did not agree to include their Fitbit data (3/10 scientists and 7/39 physicians). Those opting in vs out of Fitbit data inclusion did not differ on any study variables (Supplemental Tables 1 and 2, available online at http://www.mcpiqojournal.org). The study received clearance through the Research Ethics Boards of Dalhousie University (2021-5763) and Saint Mary’s University (21-033), and electronic informed consent was required before each survey.

Materials and Procedure

Intervention

The intervention was grounded in the Acceptance and Commitment Therapy (ACT) Matrix, a validated tool for behavior change⁴¹ (Table 1).

Measures

Self-report surveys were hosted on the Qualtrics online platform at pretest (week 1), posttest 1 (week 8), and posttest 2 (week 15) and during workshops.

Pretest Survey

Participants reported demographic variables and completed a validated measure of conflict handling styles⁴³ and baseline self-report measures of burnout and empathy.

Posttest Surveys

Participants completed validated scales of job burnout (Maslach Burnout Inventory General Survey²) and empathy styles (Interpersonal Reactivity Index⁴⁴). We analyzed a burnout composite score, which was the average of the emotional exhaustion, cynicism, and professional (in)efficacy subscales, provided to participants at the end of each full survey as feedback.^18,45 Although all 4 subscales of empathy were used for feedback, the intervention focused specifically on boosting affective empathy, measured by the 7-item Empathic Concern subscale.⁴⁴

In-class Surveys

In-class surveys presented a single-item version of each of the 4 Interpersonal Reactivity Index empathy style subscales⁴⁶ and a 1-item self-assessment of physician burnout.⁴⁷

Resting Heart Rate

Resting heart rate during sleep was tracked with Fitbit Inspire 2 devices, as an objective measure of daily stress (higher RHRs indicate greater stress⁴⁸). The device emits light on the skin and measures light absorption. An online application was programed to extract heart rate data aggregated to average daily RHR.

Dosage

We tracked workshop (6 total; mean=4.10, SD=1.96, range=0-6) and coaching session (6 total; mean=2.22, SD=1.77, range=0-6) completion. We added these to create a composite measure of total course completion (dosage: mean=6.33, SD=3.34, range=0-12).

Analytic Approach

Analyses were conducted using 2-level hierarchical linear modeling⁴⁹ with time nested within participants, specifying random intercepts and fixed slopes (random slopes results summarized in Supplemental Table 3, available online at http://www.mcpiqojournal.org). Our models predicted the outcomes (empathy, burnout, and RHR) from, for example, the linear effects of time, intervention, and a time∗intervention interaction; that is, the moderation effect of intervention group (situated at the person, level 2) on the slope of change in the outcome variable over time (situated at the time level, level 1⁵⁰).

Results

Descriptive statistics and intercorrelations at pretest are presented in Supplemental Table 4 (available online at http://www.mcpiqojournal.org). Because we were unable to randomly assign participants to intervention vs waitlist-control conditions, we tested the groups for equivalence at pretest using SPSS between-group t tests and χ² tests of association (Table 2). Only faculty role and age differed; the waitlist-control group had proportionally higher representation of basic scientists and was marginally older than the intervention group. However, adding age and faculty role as covariates in statistical models did not change any intervention effects, so our reported models exclude these for parsimony.

Table 2.

Participant Demographic Characteristics and Pretest Measures by Study Group

Characteristic	Experimental (n=22)	Control (n=27)	Equivalence Test (χ2 or t)b
Sex, n (%)
Male	6 (27.3)	10 (37)	χ2(1)=0.52, P=.47
Female	16 (72.7)	17 (63)
Relationship status, n (%)
Single	1 (4.5)	3 (11.1)	χ2(1)=0.67, P=.40
Coupled	21 (95.5)	24 (88.9)
Ethnicity, n (%)
White	20 (90.9)	22 (81.5)	χ2(1)=0.88, P=.35
NonWhite	2 (9.1)	5 (18.5)
Role, n (%)
Physician	21 (95.5)	18 (66.7)	χ2(1)=6.19, P=.01
Scientist	1 (4.5)	9 (33.3)
Age (y), mean (SD)	44.18 (7.51)	48.30 (7.03)	t(47)=1.98, P=.054
Dependents, mean (SD)	1.95 (1.36)	1.70 (0.95)	t(47)=−0.76, P=.45
Years in occupation, mean (SD)	10.98 (8.14)	14.15 (8.39)	t(47)=1.55, P=.19
Hours worked past week, mean (SD)	44.18 (7.51)	48.30 (7.03)	t(47)=1.72, P=.09
Dosage (0-12), mean (SD)	6.09 (3.52)	6.52 (3.20)	t(47)=0.45, P=.66
Resting heart rate week 1 (beats/min), mean (SD)	66.15 (9.55) (n=18)	61.56 (7.96) (n=21)	t(33)=−1.52, P=.14
Empathic concern week 1 (0-28), mean (SD)	21.50 (4.77)	21.92 (3.46)	t(46)=0.36, P=.72
Burnouta week 1 (MBI composite), mean (SD)	2.41 (1.08)	2.39 (1.00)	t(45)=−.06, P=.95

^aMBI, Maslach Burnout Inventory.

^bP values in Equivalence Test column show the exact probability of the test statistic differing from a null effect.

Statistical Analyses

Although our study spanned 15 weeks in total, we analyze targeted comparisons (intent-to-treat and dose-response analysis) rather than the entire 15-week trajectory.

Intent-to-Treat Analysis

Our intent-to-treat analyses⁵¹ compared assigned intervention participants with waitlist-control participants during the quasi-experimental trial, that is, from pretest (week 1) to posttest 1 (week 8).

Affective Empathy

Empathic Concern scores showed a buffering effect of the intervention (pretest-posttest∗intervention group interaction, P=.045; small effect, pseudo-R²∼2%). As presented in Figure 1 and Table 3, affective empathy decreased on average from pretest to posttest (main effect of time, P=.02). Simple effect tests indicate this was only for the waitlist-control group (coefficient=−1.27; P=.02), which means the control group reported a decrement of 1.27 points (on a scale from 0-28) in their empathy as they awaited the wellness leadership course, equivalent to the mean decrement in empathy shown by physicians-in-training over their internship year.⁵³ The intervention group reported no decrement in empathic responding (simple effect coefficient=0.33; P=.59). Therefore, change (decrement) for control was 3.85 times the size of change in the intervention group.

Figure 1.

Intervention effect on empathic concern: time (pretest-posttest). ∗Group, intent-to-treat analysis. Multilevel linear model (Table 2) was solved and plotted for each study group coded as follows: intervention=0 and waitlist-control=1.⁵²

Table 3.

Multilevel Models Testing Impact of Wellness Leadership Intervention: Intent-to-Treat Analysis^a

Model Component	Empathy, nobservations=95			Burnout (MBI), nobservations=93			Resting Heart Rate, nobservations=1204
	B	SE	P	B	SE	P	B	SE	P
Intercept	23.35	0.88	<.001	2.35	0.20	<.001	61.75	1.86	<.001
Intervention groupb	−2.17	1.59	.179	−0.08	0.30	.79	3.86	2.70	.16
Time (pretest-posttest or week)	−1.27	0.50	.015	−0.01	0.10	.96	0.18	0.04	<.001
Groupb × time	1.59	0.77	.045	−0.29	0.15	.06	−0.23	0.06	<.001
ICC	0.78			0.87			0.95
Pseudo-R2, % variance	0.0173, 1.73			0.013, 1.3			0.03, 3
Deviance (−2LL)	481.37			206.1			5172.81

^aICC, intraclass correlation; MBI, Maslach Burnout Inventory; SE, standard error.

^b0=control group, 1=experimental group. Group mean centering. ICC indicates how much of the total variance in the outcome variable varies between vs within individuals. Conducting multilevel analysis is appropriate when the ICC is >0.05, suggesting at least 5% of the variability in level 1 variable is due to the presence of level 2 predictors. P column shows the exact probability of t values testing whether model coefficients differ from zero.

Burnout

The time∗group interaction effect approached significance on burnout (pretest-posttest∗intervention group, P=.06; small effect, pseudo-R²=1.3%) (Table 3). Probing this trend, we find that burnout scores improved (reduced) for the intervention group (simple effect coefficient=−0.29; P=.02) but not for the waitlist-control group (coefficient=−0.01; P=.89). This means the intervention participants reported, on average, about one-third of a point improvement on a 0- to 6-point scale from pretest to posttest, equivalent also to one-third of a standard deviation.

Resting Heart Rate

Resting heart rate increases were buffered by the intervention (study weeks 1-8∗group interaction, coefficient=−0.23; P<.001; small/medium effect, pseudo-R²=3%). As presented in Figure 2 and Table 3, although daily RHRs increased on average across each study week (coefficient=0.18; P<.001), simple effect tests reported that this occurred only for the waitlist-control group (simple effect coefficient=0.18; P<.001). Their RHRs increased by approximately 1.4 beats/min as they awaited the intervention. Meanwhile, intervention participants reported no escalation in RHR (simple effect coefficient=−0.05; P=.27).

Figure 2.

Daily resting heart rate during sleep during 6-week trial period—Fall 2021: intent-to-treat analysis. Multilevel linear model (Table 3) was solved and plotted for each study group coded as follows: intervention=0 and waitlist-control=1.⁵²

Dose-Response Analyses

Next, we pooled both study groups from the 6 weeks of their active course participation (overlaying weeks 2-7 from intervention and weeks 9-14 from waitlist-control) (Table 4).

Table 4.

Multilevel Models Testing Impact of Wellness Leadership Intervention: Dose-Response Analysis

Model Component	Empathy (Single Item), nobservations=202			Burnout (Single Item), nobservations=201			Resting Heart Rate, nobservations=1363
	B	SE	P	B	SE	P	B	SE	P
Intercept	16.71	1.85	<.001	2.33	0.11	<.001	64.21	4.05	<.001
Dosage (0-12)	0.67	0.25	.01	0.003	0.04	.95	−0.05	0.51	.93
Time (Study Weeks 1-6)	0.98	0.98	.32	−0.28	0.10	.007	0.54	0.09	<.001
Dosagea × time	−0.14	0.13	.30	0.04	0.04	.39	−0.05	0.01	<.001
ICC	0.66			0.67			0.94
Pseudo-R2, % variance	0.0645, 6.5			0.066, ∼1			0.06, 6
Deviance (−2LL)	432.95			406.29			5377.03

^aGroup mean centering. ICC indicates how much of the total variance in the outcome variable varies between vs within individuals. t values test whether model coefficients differ from zero. P column shows the exact probability of t values testing whether model coefficients differ from zero.

Affective Empathy

The only significant effect for the weekly (in-class) single-item measure of empathy was a main (linear) effect of the dosage variable (coefficient=0.67; P=.01). That equates to empathy scores being (on average) 4.02 points higher (on a 28-point scale, ∼1 SD), with each 6 program elements completed (of the 12). Given this effect was not qualified by a significant dosage∗time interaction, this likely indicates a selection effect (those with higher empathy attended more sessions).

Burnout

The only significant effect for the 1-item self-assessment of burnout was a burnout reduction across the 6 intervention weeks (coefficient=−0.28; P=.007, small effect, pseudo-R²=1%), which did not vary by dosage. In other words, for each week of the intervention, regardless of the number of sessions attended, participants’ burnout reduced by 0.28 on average. This equates to an average 1.68-point decrease on a 5-point scale over the 6-week course, equivalent to moving “definitely burning out” to “some stress, not burned out⁴⁷”.

RHR in Sleep

Heart rates reported a significant linear main effect, worsening (increasing) by an average of approximately half a beat per minute (coefficient=0.54; P<.001) per week (an average 3.25-beat increase over 6 weeks). However, stress escalation was moderated by dosage (study week∗dosage interaction coefficient=−0.05; P<.001; medium effect, model pseudo-R²=6%), such that those with high course participation reported a buffering effect (no escalation) (Figure 3). Probing this interaction, participants attending 6-course elements (conditional effect coefficient=0.24) had a smaller average increase of 1.44 beats/min over the 6-week course (cutting the decrement approximately in half), and those attending all 12 elements (conditional coefficient=−0.06) reported no heart rate increase on average.

Figure 3.

Daily resting heart rate during sleep, pooled (intervention first + waitlist-control): dose-response analysis. Multilevel linear model (Table 3) was solved and plotted for 0, 6, and 12 sessions completed.⁵²

Associations Among Weekly Empathy, Burnout, and RHR

We conducted an exploratory analysis of within-person associations (see Supplemental Material for full results, available online at http://www.mcpiqojournal.org). This found that these fluctuated together within leaders, for example, empathy is uniquely associated with both RHR (coefficient=−0.09; P=.002) and burnout (coefficient=−0.20; P=.02), accounting for (pseudo-R²=12.73% or .095 of 0.746 variance) a medium-large effect.

Discussion

The medical training environment is a highly demanding one in which increased burnout^1,17,21 and reduced empathy^8,24,25 are common and potentially devastating for both educators and learners.^5,6,10,14 Our study simultaneously tracked empathy, burnout, and physiological stress (RHR) among medical and basic science faculty in response to a wellness leadership–centered intervention²⁸ focused on improving leaders’ self-care and empathic leadership. The intervention group reported lowered burnout and prevention of worsened RHR and empathy. In the same period, the waitlist-control group reported no improvement in burnout but worse (increased) RHRs and (decreased) empathy.

Furthermore, RHRs also showed a preventive dose-response effect. Participants who did not engage actively with the course reported RHR escalation, but those who attended all sessions reported a decreasing (improving) RHR trend. Our course was delivered in the fall of 2021, approximately 18 months into the COVID-19 pandemic, when many physicians peaked in their exhaustion.⁹ When work energy is drained, preventing escalated burnout is a successful program outcome—our pattern is similar to that among oncology ward teams undergoing a burnout prevention program.³⁷ Together these findings add further credence to the value of faculty leaders being trained on self-care and stress management to reduce their own burnout.²⁹

Our results also have downstream implications for the medical training environment. Although the intervention group did not reduce their burnout significantly more than the waitlist-control group in the first phase of our study, dose-response analyses pooling the groups during their wellness leadership course found a combined significant week-to-week burnout reduction. Leaders with lower burnout and preserved empathic capacities should, according to past evidence, engender feelings of support³⁹ and, hence, help buffer their trainees from future burnout.³¹ We noted that the strength and direction of our findings did not differ between medical and basic science faculty leader participants, although we would have had lower power to detect such differences. In comparison with the extant research literature on physician and medical faculty burnout and wellness, very little attention has been paid to this important group who also influence the broader basic science training environment. Our results suggest the wellness leadership approach may also be beneficial for this unique faculty group.

Previous studies have reported positive effects from workplace interventions with one component in common with our wellness leadership course, such as the ACT Matrix,⁵⁴ psychological safety climate,³⁶ social support, and improving empathy.^55,56 However, this study is the first to test a comprehensive intervention, directed at leaders who can bring a wellness-centered leadership approach to the broader medical training environment.²⁸

Although our effect sizes are modest (3%-12% of variance), these can represent meaningful effects on well-being. For RHRs, the majority (94%) of variance exists between individuals—that is, heart rates change very little from night to night. As shown in Figure 3, people attending zero course sessions averaged a 3-beat-per-minute heart rate escalation in 6 weeks’ time. This is a great deal larger than the 1-beat-per-minute increase among people with high job strain (compared with those with low job strain⁵⁷) and larger than the 2-beat-per-minute reduction in RHR found from a 12-month physical activity intervention.⁵⁸ Furthermore, leader empathy changed with burnout and RHR—they fluctuated together—making our study the first to demonstrate a dynamic relationship among these variables. This finding strikingly found the benefits of boosting empathy for the wellness of faculty themselves (not just their patients and learners) and confirmed it is modifiable with training and coaching interventions.

Limitations

We acknowledge some important limitations. First, our sample size was small because the course was resource-intensive, and our effects sizes were generally small as well. However, we note that these small effects should be realistic or even conservative and expected to generalize because the wellness leadership course was piloted under realistic work conditions at a busy time of year and during a pandemic. In future research, we aim for a larger-scale evaluation study. Second, we could not randomly assign participants to experimental groups Indeed, more scientists took the course later, relegating them to the waitlist-control group. However, when we controlled for faculty role statistically, our results were not diminished. Furthermore, our dose-response analysis pools both groups and shows buffering of RHR increases with greater course participation. Finally, our strongest results are for objectively measured RHRs, for which we analyzed ∼40 daily data points per participant. This level of repeated measurement is similar to time series designs, which approach the internal validity of randomized experiments.⁴¹ This gives us confidence in the intervention-outcome link. Third, although our goal was to affect the broader medical training environment—ultimately reducing burnout among young physicians—we measured outcomes only for leaders themselves, preventing demonstration of wider effects. We developed this course for medical faculty in the second year of its COVID-19 response, and health system capacity was strained to ask superiors and subordinates to provide ratings. This is an important extension for future research, although a strong association between leadership on follower burnout has been previously reported.³⁴

Implications

Our work has implications for how medical faculty can remain engaged and avoid burning out. Conservation of resources theory posits that the key to remaining engaged is to maximize positive energies for effort invested—referred to as positive resource spirals.⁵⁹ Resilient physicians recharge from the draining aspects of their work and protect the energizing and meaningful aspects.^7,60 On the flip side, loss spirals lead to burnout. A vicious cycle of faculty burnout and insomnia could result—work leaves them sleep deprived, prompting more negative interactions and poorer decisions, resulting in worry and even less sleep.⁶¹ Our intervention focused on stress and recovery plans, protecting positive resources and recovering from losses.

The second implication of our work is for the relevance of affective empathy as a protective factor against burnout. Although personal distress (in response to others who are suffering) drains energy and confers risk of burnout, our study confirms that empathic concern is a positive energetic resource and prevents burnout.62, 63, 64 Empathy changed within-person in our study, confirming its trainability, and related to lower RHRs and burnout. This is an important finding for health care, that increasing empathy with learners and patients may also translate to greater physical and emotional well-being for scientists and physicians.

Conclusion

A wellness leadership intervention for medical faculty leaders helped prevent empathy from decreasing and burnout and RHR from increasing during the COVID-19 pandemic. Given these leaders’ direct role in medical education and training, their capacity to lead empathically and encourage a new generation of physicians to do the same should be protected. This intervention shows potential to improve the supportiveness of the medical training and work environment for wellness even in the face of ongoing stressors, such as a long pandemic.

Potential Competing Interests

The authors report no competing interests.