Narrative medicine workshops for emergency medicine residents: Effects on empathy and burnout

Emergency medicine (EM) residents face unique affective challenges and barriers to interpersonal connection in their clinical environment which may contribute to decreased empathy and increased burnout. Narrative medicine (NM) might address these barriers and has had beneficial impacts in various populations but has never been studied in EM residents. In this study, we sought to evaluate the effect of NM workshops on burnout and empathy and to assess resident perceptions of the workshops.


INTRODUC TI ON
2][3] Some of these challenges are shared across specialties-navigating confrontations with patients and providers, witnessing trauma, and finding a sense of identity and maintaining self-worth. 1 Additionally, EM residents face unique conflicts that arise from their clinical environment, one in which the acuity, volume, and brevity of patient interactions hinders the development of empathic, meaningful relationships with patients. 2,4[7][8][9] One field that shows promise in addressing the affective challenges faced by physicians is narrative medicine (NM).Developed and championed by physician and literary scholar Rita Charon, NM promotes the integration of narrative competence, the "ability to listen to the narratives of the patient, grasp and honor their meanings, and be moved to act," into clinical practice. 10Charon herself has described narrative competence as an ability that promotes empathic engagement on the behalf of the physician. 10She proposes that the traditional form of knowledge embraced within medicine, logicoscientific knowledge, needs to be supplemented by narrative knowledge: while the former "transcends the particular" to frame a patient in the context of epidemiology and statistics, the latter "reveals the particular" and allows a physician to understand the inherent uniqueness to each patient. 10She and others have argued that this dual knowledge is critical for clinicians because it allows for a fuller and more biopsychosocially sound understanding of individual patients' stories and facilitates empathetic, trustworthy, professional practice. 11,12e mechanisms by which narrative competence transforms the physician and boosts empathy are complex.Certain theories propose that the study of narratives introduces learners to a form a knowledge that allows them to "regard patients holistically and to feel the emotions of compassion towards patients and themselves." 13More specifically, Kleinman 14 argues that NM teaches physicians to contextualize patient encounters and to "make sense of and give value to the experience."This skill may be especially useful in the ED, where the pace and intensity of the clinical environment make a nuanced understanding of patients' stories nearly impossible.
By highlighting personal narratives and bolstering empathetic connection, narrative competence may reduce depersonalization and emotional exhaustion, two key components of burnout. 15,16ditionally, NM may benefit residents by allowing them to explore personal changes as they process their role in an overwhelming, traumatic, and complicated health care system. 10In one examination of the experiences of narratively trained medical students, five students name, recognize, and confront the "painful alienation" they experienced in medical training and, because of their narrative skill, were able to filter through these emotions to find the meaning and beauty in their work. 17 has been evaluated in various settings, with outcomes that demonstrate positive effects on empathy, communication skills, personal growth, and burnout. 18,19While these positive effects have been described in interventions with medical students, [20][21][22][23][24][25][26][27][28] faculty, [29][30][31][32][33][34] and interprofessional teams, 35,36 few studies have examined NM within GME settings and none have been performed within EM.In this report, we describe the implementation and evaluation of NM workshops in an EM residency and assess their effects on resident empathy and burnout as well as perceptions of the workshops.

Setting and participants
We performed a quasi-experimental study from July 2020 to October 2020 at two EM residencies: University of Chicago (UC, intervention site) and Emory University (control).UC is a home site of some of the study authors; Emory was sought as the second site given its similarities to UC: both university-affiliated residencies are 3 years long, serve an underserved and urban population, and have a diverse resident body.Enrollment in the study was voluntary and offered to all residents via email or in person.

Protocol
Residents at the intervention site participated in two NM workshops.In response to restrictions related to the COVID-19 pandemic, the first session was conducted virtually using Zoom Video Communications software and the second was held in person.The first workshop was held in the beginning of academic year in June 2020 during mandatory class orientations.The second was held in early October during mandatory resident retreat, about 3 months apart.Residents at the control site participated in their already scheduled didactics, which did not include narrative workshops.

Residents at both sites completed two instruments during this study: an abbreviated version of the Interpersonal Reactivity
Index (IRI) 37 and a single-item measure of burnout. 38,39Baseline demographic data including postgraduate year (PGY) and gender were gathered at the first time point.A postintervention feedback survey was given to UC residents to assess resident perceptions of the NM intervention at the second time-point.All data were managed using Research Electronic Data Capture (REDCap) hosted at UC; this software also allowed for the anonymization of survey data. 40,41e anonymous surveys were collected at two times: just prior to the initial workshop and 1 week after the second workshop.
Residents were emailed the surveys and received weekly reminders up to three times.][45] Enrollment in the study was voluntary and was offered to residents in PGY-1 to -3.Residents were offered enrollment either via email or in person before the first survey was administered.

Workshops
Workshop design, text selection, and teacher's guides were created by two authors (ZM and MB).MB is a nonclinical scholar with extensive knowledge of health humanities pedagogy and NM.ZM is an EM physician with a background in medical education.
We implemented two similar workshops at the different time points.The residents were grouped in these workshops according to PGY level; this was an intentional choice to help foster an atmosphere of openness and to allow for a conversation around struggles that are specific to training level (e.g., the PGY-1 experience of being a doctor for the first time).
The workshops were composed of three main parts: close reading of a literary text, reflective/creative writing, and discussion.
Participants were emailed the text before the workshops; this allowed them the option to engage with it beforehand if they chose.

Since medical education experts encourage educators to address
preconceptions prior to learning and we anticipated some resistance to this different form of learning, the first session opened with a brief introduction to NM, an explanation of its utility, and an establishment of the group as a safe space. 46This was followed by an out-loud, group reading of the selected literary text.This portion of the workshop required residents to listen actively to a complex narrative.
After the reading, there was a brief, guided, discussion analyzing the text's central themes and the narrator's perspective, which encouraged residents to identify points of narrative importance and to empathize with the narrator.Residents then moved onto a 15-min writing exercise guided by prompts that were crafted by study authors to encourage an examination of personal and emotional challenges within the context of being a physician.Workshops closed with a group discussion of residents' writings and, optionally, with readings of what they had written.Here, residents shared-and shared in-their struggles and explored common themes in their writings.
This general format is the most common one used to teach NM as noted in a review of NM educational interventions. 18Furthermore, it is consistent with traditional curricular design as framed by Thomas et al. 46 : we had planned to target affective educational objectives that are most adequately matched by discussion and reflection, both of which are included in our workshop. 46Additionally, following the principles laid out by Thomas, the workshops employed multiple educational methods (i.e.reading, writing, discussion) to help "maintain learner interest and provide … reinforcement of learning." Workshops were facilitated by a single EM faculty, all of whom had or were completing a masters in medical education but had no specific training in NM.Facilitators were given teacher's guides that were crafted by the authors (Appendix S1) and instructed to follow the best practices for leading NM workshops as outlined by Winkel. 47 No faculty facilitators were in leadership positions; we hoped this would eliminate any desire to impress or not be forthcoming.
In the first workshop, residents read "Hug or Ugh," a story by EM physician Jay Baruch about a physician struggling to find a balance between humanist and professional impulses after an unhoused patient with substance use disorder asks for a hug. 48In the second workshop, learners read an excerpt of "Emergency Room Notebook, 1977," a story by Lucia Berlin told from the perspective of an ED nurse confronting the cynicism that one feels after repeatedly seeing death and the repercussions of societal ills. 49These pieces, chosen because of their focus on the stressors and culture within EM, offered many discussion points that could be tied to a young EM physician's experiences.

Instruments
The IRI, created by Davis et al., 37 is a 28-item, self-survey instrument that can be broken down into four, seven-item, independent, subdomains: perspective taking (PT), empathetic concern (EC), personal distress, and fantasy.Each item on this survey is assessed using a 5-point Likert scale and outcomes are typically reported as means of these subdomain scores considered as continuous variables.Experts in the field have concluded that this instrument considers both the cognitive and the affective aspects empathy, which is consistent with the conceptualization of empathy as outlined by empathy researchers. 50,513][54][55][56][57][58][59] Of the subdomains, many authors have chosen to abbreviate the IRI by evaluating only the EC and PT domains as these measure the affective and cognitive aspects of empathy, respectively.These subdomains have also been shown to be reliable and reproducible measures of sensitivity to others' views and emotions. 5,37,57,60,61e most commonly used measure of burnout is the Mashlach Burnout Inventory (MBI), a well-validated, 22-item, self-survey instrument. 62,63The MBI is broken down into three subscales: emotional exhaustion (EE), depersonalization (DP), and diminished personal accomplishment.5][66][67] West et al. 68,69 were able to demonstrate validity across differing levels of practice (medical student, resident, practicing physician) of single-item measures of both EE and DP.However, there are financial barriers to use of the MBI that make routine use cost-prohibitive, as it is a proprietary instrument.In response to this fact, researchers have evaluated the psychometric properties of a nonproprietary, single-item instrument and have determined that this instrument is a reliable substitute for the MBI:EE. 38,39,70,71This item has demonstrated validity across differing fields within the health care setting (physicians [American and Australian], nurse practitioners, physician associates, nurses, clinical associates, and administrative clerks) but has not specifically been validated in resident physicians.However, we still chose to use this nonproprietary instrument to evaluate burnout given its relatively robust validity and issues related to cost and time.
To evaluate resident perceptions, we designed a postintervention survey that sought feedback on workshop structure and explored resident reactions to them, following general survey design principles as outlined by Fowler 72 (Appendix S2).This survey underwent critical review by experts in survey design and was piloted by two education researchers before use.

Data analysis
For our primary outcomes, we evaluated differences in IRI subdomain scores (PT and EC) and burnout scores as measured by the single-item instrument over the two time points.We treated IRI subdomain scores as continuous variables and fitted linear regression models; we treated burnout as an ordinal variable and fitted a conditional logistic regression model.Each model included group (intervention vs. control) and time point predictors as well as the interaction between group and time.We evaluated resident perceptions of the workshops using summary statistics.This study qualified for exemption by the institutional review board at UC.
We had intended for nearly all residents at the intervention site to participate in both NM workshops, but logistical challenges brought on by the COVID-19 pandemic and routine scheduling made this impossible.At the intervention site, 34 of 48 (70.8%) residents provided information about how many sessions they attended; seven of these respondents were unable to attend any session, 12 attended one session, and 15 attended both sessions.Of the 15 who attended both a virtual and an in-person workshop, nine preferred or greatly preferred the in-person format and six had no preference; no one preferred the virtual format.

Burnout
While all residents demonstrated worsening burnout over time, residents at the intervention site exhibited less severe increases.
Baseline burnout scores were significantly higher at the intervention site (p = 0.005) and increased significantly over time on average (p = 0.001), but the increase over time was significantly slower in the intervention than control (interaction p < 0.001; Figure 1).Burnout was significantly higher for PGY-2 and PGY-3 residents than PGY-1 residents across sites and time points (p = 0.005 and p < 0.001, respectively).Gender had no association with burnout (p = 0.34).

Empathy
Baseline PT and EC scores were similar to those reported elsewhere among residents and did not change significantly over time. 5,57,58ere were no significant main effects or interactions of site and time point on empathy subscores, suggesting that the workshops did not affect resident empathy (Table 2).Men had lower mean EC subscores than women (men 19.2, 95% confidence interval [CI] 17.8-20.5;women 22.4, 95% CI 21.1-23.7;p < 0.001).

Acceptability and feasibility
Twenty-four of the 48 (50%) residents in the intervention group completed the postintervention evaluative survey (Figure 2).Most (n = 20, 83.3%) agreed that the workshops should be a standard part of EM residency didactics.The majority were satisfied with the hour-long length of the workshops (n = 17, 70.8%) and the 3-month interval between workshops (n = 19, 79.2%).Free-text feedback to the workshops was largely positive; the little constructive feedback we received suggested that we incorporate more teaching on how to read, interpret, and respond to prompts.We encountered no major hurdles during the implementation of the workshops; after we created teachers' guides, we simply sent these to the recruited EM faculty and coordinated times for the workshops.

DISCUSS ION
In this study, the first to evaluate the impact of NM in EM, residents at a site that participated in NM workshops exhibited slower burnout but demonstrated no change in empathy compared to a group of residents at a control site.The workshops were designed following the most common format for NM and were adopted for an EM audience; they were easily implemented and well received.
In a review of NM as an educational tool, Milota et al. 18 found two quantitative papers describing NM's effects on resident burnout, both by Winkel et al. 73,74 in obstetrics and gynecology.In their first single-site study, authors noted an increase in burnout over time for all and a trend toward greater increases in burnout among nonparticipants. 73In their follow-up, multisite study, they noted worsening burnout over time for all but found that high workshop attendance correlated with a significant decrease in the emotional exhaustion subdomain of burnout. 74Similarly, our findings suggest that NM workshops confer some protection against worsening of the emotional exhaustion component of burnout.
These same authors also examined empathy and found no differences in changes in empathy between high-and low-attendance subjects, again, similar to our findings. 74This result is in contrast to literature from outside GME, where NM has improved learner empathy. 18,19,43,75This may be explained by differences in learner population: residency is more grueling than other periods of training so residents may be more recalcitrant to efforts to cultivate empathy.Further, EM and obstetrics and gynecolory are uniquely stressful fields that have the highest and third-highest rates of burnout in medicine, respectively; this may make empathybuilding more challenging in these populations. 76Interestingly, we also noted that respondents felt closer to each other as opposed to closer to their patients after participation.This finding, in combination with the lack of effect on empathy, suggests that our workshops' effects on community-building might be stronger than effects on empathy.This may reflect our workshop design, which largely focused on narratives by health care providers and writing prompts that encouraged an examination of the self as a clinician.Some proponents of NM advocate for interprofessional groups and a focus on patient narratives and experience 42 ; while we adhered to most principles of NM in our workshops, some may challenge our use of narratives that focus on provider perspectives in a space with only physicians.Note: Data are reported as mean (±SD).
We had anticipated potential barriers to adoption of the workshops by residents including hesitancy to share and skepticism or discomfort with nonmedical content.However, the design and implementation of the workshops seemed to mitigate any resistance, as most residents actively participated and we received no overtly negative feedback.In fact, we believe the feedback demonstrates a success of the workshops as the majority felt that they should be standard in residency training.

LI M ITATI O N S
Our most significant limitation is the low response rate at the control site, which makes comparisons between sites difficult.There are also potential confounders for which we have not accounted: the two residency programs may have different hidden curricula or cultures, experiences during the study period could have varied by state or institution, and policies, especially as they relate to 19, could have been different between sites and had effects on resident well-being.Additionally, this intervention was a short-lived one and a more longitudinal intervention may result in different outcomes.We also have no data to comment on longterm outcomes or assess impacts beyond the immediate postintervention period.
As a result of social gathering guidelines during the COVID-19 pandemic, we were forced to hold one session virtually and elected to have the second session in person.We elected to have the inperson session because we believed that the group discussions, which were a cornerstone of the activity, would be better suited for in-person interactions.However, we are unable to say whether the virtual or live sessions were more impactful and it is possible that the effects that we measured would be more robust had we been able to hold two live sessions.
There are a few other limitations to this study which are primarily related to outcomes.Although the decision to evaluate empathy and burnout was made after an in-depth evaluation of the literature around both constructs, these outcomes may be challenged.
Empathy specifically can be difficult to measure and assess.
Although the majority of empathy research in health care uses quantitative measures, the instruments used are variable in their assumptions and definitions of empathy, with some defining empathy as entirely cognitive, entirely affective, or as a combination. 45,50The IRI, which we chose to use in our study, weighs these facets differently than other instruments such as the Jefferson Scale of Physician Empathy (JSPE). 52,77ditionally, empathy can be measured using first-person measures (self-report) and second or third-person measures (patient or observer report), both of which have benefits and pitfalls.While selfassessment is logistically easier, it is also riddled with issues of bias.
Second-and third-person measures attempt to rate one's empathy via observation or patient evaluation, but some question the utility of measuring empathy as an observable trait or behavior. 44,45,50,51,78nally, the short interval between our assessments may not have been long enough to capture a meaningful change in empathy.
Similarly, the tool we chose to measure burnout may be questioned as since most consider the MBI, a longer and costlier tool, to be the standard to measure burnout in health care.The single-item tool that we used correlates with the EE domain of the MBI; while EE has shown to predict overall burnout, the single-item tool that we used has not been demonstrated specifically to predict overall burnout.Further, the drivers of overall physician wellness and unwellness are quite complex and may not adequately be assessed with quantitative survey data alone; thus our primary outcomes may insufficiently evaluate what NM has to offer. 50,79,80

CON CLUS IONS
Our study establishes the feasibility and acceptability of incorporating narrative medicine into emergency medicine residency education and provides a framework for doing so.we build on existing work suggesting that while narrative medicine may not affect resident empathy, it may provide residents with tools and a safe space for sharing that may decelerate the emotional exhaustion

CO N FLI C T O F I NTER E S T S TATEM ENT
The authors declare no conflicts of interest.

F I G U R E 1
Predicted proportion of burnout scores on a five-point, single-item measure by group, PGY level, and time point.Higher scores correspond with worse burnout.TA B L E 2 Mean scores for PT and EC on the IRI by time point and site.

F I G U R E 2
Proportion of responses (n = 24) to postintervention evaluation survey items on a 5-point Likert scale.
component of burnout.Future work might expand on our investigation through the use of qualitative methodologies that could further clarify narrative medicine's benefits to residents and help guide the development of narrative medicine curricula.AUTH O R CO NTR I B UTI O N S Zayir Malik contributed to study concept and design, acquisition of the data, analysis and interpretation of the data, drafting of the manuscript, critical revision of the manuscript for important intellectual content, statistical expertise, and study supervision.James Ahn contributed to study concept and design; acquisition of the data; critical revision of the manuscript for important intellectual content; administrative, technical, or material support; and study supervision.Alan Schwartz contributed to study concept and design, analysis and interpretation of the data, critical revision of the manuscript for important intellectual content, and statistical expertise.Michael Blackie contributed to study concept and design; analysis and interpretation of the data; drafting of the manuscript; critical revision of the manuscript for important intellectual content; administrative, technical, or material support; and study supervision.

TA B L E 1
Survey response rates and demographics by intervention site.