Addressing the physician burnout epidemic with resilience curricula in medical education: a systematic review

A variety of stressors throughout medical education have contributed to a burnout epidemic at both the undergraduate medical education (UGME) and postgraduate medical education (PGME) levels. In response, UGME and PGME programs have recently begun to explore resilience-based interventions. As these interventions are in their infancy, little is known about their efficacy in promoting trainee resilience. This systematic review aims to synthesize the available research evidence on the efficacy of resilience curricula in UGME and PGME. We performed a comprehensive search of the literature using MEDLINE, EMBASE, PsycINFO, Educational Resources Information Centre (ERIC), and Education Source from their inception to June 2020. Studies reporting the effect of resilience curricula in UGME and PGME settings were included. A qualitative analysis of the available studies was conducted in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Risk of bias was assessed using the ROBINS-I Tool. Twenty-one studies met the inclusion criteria. Thirteen were single-arm studies, 6 quasi-experiments, and 2 RCTs. Thirty-eight percent (8/21; n = 598) were implemented in UGME, while 62 % (13/21, n = 778) were in PGME. There was significant heterogeneity in the duration, delivery, and curricular topics and only two studies implemented the same training model. Similarly, there was considerable variation in curricula outcome measures, with the majority reporting modest improvement in resilience, while three studies reported worsening of resilience upon completion of training. Overall assessment of risk of bias was moderate and only few curricula were previously validated by other research groups. Findings suggest that resilience curricula may be of benefit to medical trainees. Resilience training is an emerging area of medical education that merits further investigation. Additional research is needed to construct optimal methods to foster resilience in medical education.

The burnout epidemic among medical trainees was only recently acknowledged. In the 1990s, the "Triple Aim" of healthcare was established by healthcare institutions and organizations in the US. This new initiative emphasized patient satisfaction, quality of care and cost reductions [51]. It was not until the 2000s that staff and healthcare provider satisfaction was formally considered in the rebranded "Quadruple Aim" of healthcare [52]. In 2015, physician groups in both the United States and Canada began to adopt formal campaigns recognizing the resident burnout epidemic. In the United States, the Accreditation Council for Graduate Medical Education (ACGME) amended its accreditation requirements to address trainee well-being and resilience more comprehensively across all residency and fellowship programs [53]. In addition to providing self-care resources, the campaign helped residents find meaning in their work, enhanced communication and professional relationships, evaluated and promoted safety in the working and learning environment, and provided education and resources to identify and treat burnout, depression, substance abuse and other challenges [54]. In Canada, the 2015 CanMEDS Physician Competency Framework outlined that, in order to be a professional, one must have a commitment to self [55]. In response, an increasing number of residency programs have restructured their training to provide and foster specific skills and dispositions towards work-life balance and self-care [56,57].
The ACGME's formal recognition of the resident burnout epidemic paved the way for new wellness initiatives. Institutions have begun exploring a vast array of healthpromotion programs including mindfulness [58][59][60], yoga [61,62], self-hypnosis protocols [63], small group debriefing and stress-management programs [64], curricular changes [65], evaluation changes [66,67], time management programs [68], reflective writing sessions [69,70], and self-development groups [71]. However, recent literature on therapeutic stress management programs for medical students and residents have revealed mixed and inconsistent results, with overall unclear long-term benefits. In response, medical schools and residency programs have begun exploring alternate wellness interventions, aiming to prevent burnout long-term by fostering trainee resilience [72][73][74].
Resilience is broadly defined but can be conceptualized as the ability to face adversity forthrightly and intentionally instead of aiming merely to survive through hardships. Resilient individuals re-frame challenges as opportunities for growth and thus willingly engage with the harsh realities of life in a healthy manner that ultimately achieves goals at a minimal physical and psychological cost [75]. A systematic review on resilience identified five themes used to define resilient individuals: rising above adversity, adapting and adjusting, resilience as a dynamic process, "ordinary magic" (i.e., resilience is an inherent trait in all people) and mental illness as a marker of resilience [76]. Resilience was initially regarded as an inherited, static character trait [77,78], however research has identified it as a dynamic and transient quality [78]. Importantly, resilience education requires a foundation of self-awareness and the ability to self-monitor [75]. This necessitates that the individual willingly accepts their limits and uncertainties, and uses their insight to recognize errors and problem solve [77]. Initial studies have indicated that physicians who exhibit high-resilience personality traits have an objectivelyelevated sense of overall well-being, provide better quality patient care and ultimately contribute to an overall decrease in healthcare costs [79].
These initial observations support the implementation of resilience-based interventions in medical training to prepare trainees for the inevitable hardships of clinical practice [75,80]. Although such interventions have been developed and integrated into medical school and residency programs, little is known about their efficacy in promoting resilience. To address this paucity in the literature, we conducted a systematic review to synthesize the available research evidence on the efficacy of resilience curricula in undergraduate and postgraduate medical education.

Protocol and registration
A systematic review protocol was prospectively developed and registered in the International Prospective Register of Systematic Reviews (PROSPERO, CRD42020191511). This systematic review was performed in concordance with the Cochrane Handbook for systematic reviews of interventions and the Preferred Reporting Items for Systematic Reviews and Meta-Analysis Protocols (PRISMA-P) statement (Additional file 1) [81][82][83].

Search method
A systematic search of the literature was performed by an information specialist (K.F.) in MEDLINE(R) ALL (Ovid, 1946 to June 15, 2020), Embase (Ovid, 1947 to 2020 June 15), APA PsycInfo (Ovid, 1806 to June Week 22,020), ERIC (Ovid, 1965 to March 2020), Education Source (EBSCOHost, 1880-2020) from database inception to June 16, 2020 using a combination of subject headings terms and keywords (Additional file 2): "medical education", "medical student", "resilience", "curriculum", and "training". In addition, a manual search of the reference lists of the retrieved articles was conducted to capture all relevant studies for potential inclusion.

Eligibility criteria
We included all peer reviewed primary research articles that implemented resilience curricula in UGME and PGME settings and reported efficacy outcome measures. As per our review protocol published a priori, the inclusion criteria for studies were: (1) both single-and double-arm studies of either qualitative or quantitative nature in which the participants received a clearly defined curriculum which included at least one concept of resilience including, but not limited to, coping skills, self-efficacy, goal-setting, and emotional regulation skills; (2) study participants including undergraduate and postgraduate medical learners; (3) an outcome measure assessing the effect of resilience curricula through quantitative or qualitative research instruments. Both randomized controlled trials and quasi-experimental studies were included. No restrictions were placed on the study subjects' age, gender, country, level or specialty of training, and length of curricula and follow-up period.
We excluded review articles, editorials, letters, commentaries, theoretical articles (e.g., curricular development), and non-English articles; however, a manual search of the reference lists of review articles was conducted to ensure broad and comprehensive inclusion of the available literature. For the purpose of this review, studies were excluded if their primary focus was therapeutic in nature (e.g., cognitive therapy, counselling, and mind-body skills). Finally, studies that only described participant satisfaction or those that only described the implementation of resilience curricula were also excluded.

Study selection and data abstraction
After identifying all citations and removing duplicates, a two-step screening process was employed to determine the relevancy of the identified articles. First, each study title and abstract was independently screened for eligibility in accordance with the aforementioned eligibility criteria by two reviewers (C.S. and M.C.). An inter-rater calibration test was performed using 20 randomly identified articles prior to formally commencing the screening process to ensure reliable screening accuracy. Following the title-abstract screening, the same two reviewers independently retrieved the full text of relevant articles and determined their eligibility for inclusion using the preset inclusion criteria. Any discrepancies between the two reviewers were resolved by consensus. When no consensus could be reached, a third reviewer was consulted.
A standardized data collection table was created by the review team prior to formally commencing the extraction process. Extracted variables included: (1) study characteristics (e.g., country of origin, study design, target population, sample size); (2) curriculum description (e.g., name, curricular content, delivery mode, duration and frequency); (3) resilience outcome scales; and (4) secondary curriculum outcome measures. Two reviewers (C.S. and M.C.) independently extracted the data and compared the results for verification. Any disagreements between the two reviewers were resolved by consensus after consulting the entire review team.

Analysis of studies
A primarily qualitative appraisal of the literature was conducted following the methodological guidelines for qualitative reviews outlined in the Cochrane Handbook [83]. When appropriate, descriptive statistics were utilized to present the overall trend or effect of results. Due to the heterogeneity in both curricula and outcome measures across the included studies, a pooled analysis of the effects of curricula using a meta-analytic methodology was not possible.

Quality appraisal
A risk of bias assessment of each included study was performed by two independent reviewers (C.S. and M.C.) using the ROBINS-I tool [84,85]. Risk of bias was assessed across seven domains (confounding, selection, classification, intervention, missing data, measurement, reporting). Results of quality assessments were compared, and when there were any disagreements, the entire review team was consulted which were then resolved by consensus. Figure 1 depicts the study selection process in a PRIS MA flow diagram. The initial electronic search of the databases identified 6101 citations. Two thousand one hundred one duplicate records were removed using Covidence (Veritas Health Information, Melbourne, Australia), which left 4000 references for the screening phase. Following the title-abstract screening, 3886 studies were deemed to be irrelevant for the present review based on the exclusion criteria. Inter-rater reliability for the title-abstract screening was excellent (k > 0.90). Of the 114 studies assessed for eligibility in the full-text screening, 93 studies were further excluded as they did not meet the inclusion criteria. No additional studies were identified after a manual screening of reference lists of the included studies. In total, 21 studies that fully satisfied the previously stipulated eligibility criteria were included in this review [57,[86][87][88][89][90][91][92][93][94][95][96][97][98][99][100][101][102][103][104][105].

Study characteristics
The characteristics of the included studies are collated and summarized in Table 1. Of the 21 studies included in this review, 13 studies (62%, n = 984 total participants) used a single-arm design that evaluated the effect of resilience curricula pre-and post-intervention, 2 studies (9.5%, n = 100) used a randomized controlled design, and 6 (28.5%, n = 292) used a quasi-experimental design. Sample sizes of studies varied between 12 and 258, but none of the studies performed an a priori sample size determination. As shown in Fig. 2, the earliest study on resilience curricula in medical education was published in 1982, with the majority of studies (90.5%, 19/21) published in the last decade.

Study populations
Thirty-eight percent (8/21, n = 598 total participants) of the included studies implemented their resilience curricula in a UGME setting: 3 studies were implemented in pre-clerkship, 2 in clerkship, 2 in both pre-clerkship and clerkship, and 1 study did not specify the level of UGME  participants. Sixty-two percent (13/21, n = 778) implemented the curricula in a PGME setting. Of the studies that tested their curricula among residents, 5 were in internal medicine, 2 in surgery, 2 in family medicine, 2 in pediatrics, 1 in psychiatry, 1 in anesthesiology, 1 in pathology, and 1 in internship. Geographic distribution of the included studies was heavily concentrated in the United States (n = 16), with only 1 study in Canada, China, Brazil, South Africa and Australia, respectively.

Training methods
Considerable heterogeneity existed in the duration and frequency of resilience curricula and ranged from a 1-h single session to 1.5-h weekly sessions for 16 consecutive weeks. 76% (n = 16) of the studies implemented inperson training over at least 4 sessions. The majority of studies incorporated more than one component to facilitate resilience, including didactic teaching sessions (n = 16) and skill-building exercises around the themes of mindfulness, relaxation, and stress reduction (n = 16). Of the 5 resilience curricula that were based on previously developed training models, 2 used the Stress Management and Resilience Training Program [95,98], and 3 used the Penn Resilience Program [90], Coping with Work and Family Stress Intervention [89], and Energy Leadership Well-Being and Resiliency Program [99], respectively. Two studies implemented non-conventional, outside-of-classroom curricular design. In a study by Shapiro et al. (2019), the authors sought to promote emotional well-being and resiliency in graduate medical trainees through individualized or group spiritual care sessions wherein residents were given an opportunity to openly discuss their religious, cultural, and spiritual beliefs with Hospital Spiritual Care representatives. In another study by Orr et al. (2019), senior internal medicine residents participated in a single, 4-h workshop at the Philadelphia Museum of Art that focused on the medical humanities and artful thinking exercises to enhance resilience and ameliorate burnout. There was a considerable variety in the background of instructors which included residents, psychologists, social workers, chaplain, and staff physicians. Only two studies were facilitated by trained resilience instructors [97,105].

Outcome measures
All studies relied on self-report questionnaires to assess curricular outcomes. The majority of studies evaluated the effects of the resilience curriculum using previously validated outcome instruments (90.5%, 19/21), while 2 studies used a new or non-proprietary measurement tool [91,102]. Most commonly used outcome measure was Maslach Burnout Scale (9/21), followed by Connor-Davidson Resilience Scale (4/21), Perceived Stress Scale (4/21), Professional Quality of Life Scale (3/21), and Spielberger Trait Anxiety Inventory (3/21). Eleven studies assessed the effects of the curriculum immediately pre-and post-intervention with no follow-up. Only 4 studies reported long-term effects of their curriculum (i.e., after at least 12 months of delivering the curriculum). As shown in Table 1, there was considerable heterogeneity in the efficacy of resilience curricula. While most studies suggested a modest improvement in their efficacy measures upon the completion of resilience curricula, a few studies reported no significant impact, and in some, even worsening of resilience and associated measures. Interestingly, despite positive feedback from students who also supported continuation of the resilience program, Bird et al. (2017) reported that both resilience and burnout measures significantly worsened post-

Feasibility and acceptability
Participant satisfaction and acceptability of resilience curricula were reported in 12 of the 21 studies [87, 88, 91, 94, 95, 97-99, 101, 103-105]. Participants reported generally positive experience and acceptability of resilience curricula, however, the methodology of program assessment was largely heterogeneous and unsystematic.
In Pereira et al. (2015), 90% of the participants reported that the course was helpful, and 76.3% indicated that they had incorporated new coping mechanisms into daily life as a result of the training. Similarly, in Chaukos et al. (2018), 87% considered the resilience skills taught during training to be helpful. While some participants suggested that resilience training should be part of mandatory curricula [91], others found it counterproductive and unnecessary [95].
In terms of curriculum duration, some participants indicated that resilience curricula should be implemented across all years of residency, as resilience skills and techniques may be more instrumental in later years with added stressors and clinical responsibilities [105]. Similarly, 75% of participants stated that the resilience curricula should continue in the following year [94]. 19% reported that training sessions were too short in duration and infrequent [98]. Need for greater availability of resilience training and resources was noted by Tucker et al. (2017) as the authors quote responses from two participants: "I almost wish that there were more opportunities to talk about burnout during the year. I would have really appreciated that." and "Back in 1st and 2nd year the student affairs office had these wellness checks. I wish we had that in third year because we can all sign up for one but a lot of us are too busy to even think of it." Participants appreciated protected time for resilience curricula and having a safe space to discuss resilience topics: "[T] here was something special about having an opportunity to sort of vent about my experiences... I don't feel like there's anyone else in the program that I'd be willing to reveal that stuff to. And talking to each other is helpful, but it doesn't have that degree of separation" [105]. Similar trend of having a venue to openly discuss personal issues with peers was repeated in Pereira et al. unique opportunity to build a community of shared experiences that allowed validation of participants' worries and hardships. Understandably, 8% reported feeling vulnerable in sessions [98].

Risk of bias assessment
To assess potential risk of bias, we used the ROBINS-I scoring system to evaluate bias across seven domains (Additional file 3): confounding bias, selection bias, classification bias, intervention bias, attrition bias, measurement bias, and reporting bias. Overall, the risk of bias was moderate for all studies, which, in light of the nonrandomized nature of the studies and reliance on selfreporting measures, is understandable. While confounding bias and classification bias were deemed to be low in most of the included studies, selection bias was moderate in 9 (43%) and measurement bias was moderate in all (100%) studies. Due to the high attrition rate in most studies, bias due to missing data was low in 6 studies (29%), moderate in 12 studies (57%), and unknown in 3 studies (14%). To date, there have been only two randomized controlled trials that assessed the curricular efficacy of a resilience program, however these studies were assessed with the same bias scoring system to maintain coherence with other included studies.

Discussion
In response to the growing awareness of the high risk of burnout and stress in health professionals, there has been growing interest in resilience in medical education over the past decade. While physicians as a group have significantly higher resilience than the general public [106], efforts to overcome systemic challenges in the clinical care environment are truly needed to curve physician burnout and foster well-being. To the best of our knowledge, this is the first systematic review to provide a comprehensive summary of published literature on the effect of resilience curricula in undergraduate and postgraduate medical education. Our results highlight the considerable heterogeneity in content, delivery, and outcomes of resilience curricula implemented to date in medical education. In particular, in spite of the modest improvement in resilience and well-being reported in many studies, a number of other studies [94,95,98] suggested that the implementation of the program resulted in detrimental effects to the participants' overall resilience. Notably, all three studies delivered their curricula over 6 months to a year, and thus their negative results may have been confounded by an interaction between trainees' resilience level and the progression or level of training. True to the significant heterogeneity seen in the resilience curricula in this review, currently no standardized, efficacy-proven resilience curriculum exists around the world that could be implemented by undergraduate and postgraduate medical programs. This present review clearly calls for the development of more systematic and evidence-based programs and resources which are desperately needed in the fight to reduce physician stress and burnout.
The strengths of this review include its systematic methodology; prospectively registered with a clearly defined objective, comprehensive and detailed search strategy, and the screening of each study's risk of bias and methodological rigor. In spite of this, the findings of the present review should be interpreted in light of several limitations. Firstly, all included studies had moderate risk of bias according to the ROBINS-I tool for assessing the risk of bias in nonrandomized studies. Therefore, the results presented in this review must be interpreted in light of these potential biases. Secondly, most of the included studies did not assess long-term resilience outcomes of resilience programs. As such, it remains unclear whether the resilience curricula successfully fostered lasting resilience in medical students and residents, or if the reported benefits were transient. As previous research has shown resilience is an acquired skill and thought pattern that requires continual practice to achieve mastery, an assessment methodology that incorporates long-term follow-up on curriculum participants should be considered in future studies. In parallel, resilience curricula may result in long-term resilience building by providing strategies that trainees can then choose to continuously implement throughout their training and future practice. By providing appropriate tools, long-term sustainable resilience may be achievable. Next, all studies relied on self-reporting psychometric instruments and, as such, it is unclear whether the reported changes in the indices of resilience curricula have meaningful objective implications for personal and professional development. Furthermore, although an increasing number of resilience curricula are being offered throughout UGME and PGME, the results of this review demonstrate vast heterogeneity of curricular methods, ranging from one-time didactic lectures to a mix of lecture-based and practical sessions over many months. Consequently, a standardized approach to resilience curricula development does not currently exist. However, the lack of a common foundation between resilience programs is not surprising given the early nature of resilience-centered initiatives. Additionally, the diversity of medical trainees and trainee stressors may necessitate a resilience curriculum tailored specifically to their unique beliefs, values, and stressors. The broad range of curricula makes inter-institutional comparison difficult. More research is necessary to determine whether a standardized approach to resilience is feasible and beneficial, or if medical programs should develop unique curricula adapted to their trainees.
Despite these limitations, this systematic review provides a guidance for future resilience curricula development and related studies. Future research on resilience in medical education should include clear operationalization of resilience and curricular components to enable reproducibility and accurate comparison of outcomes. In addition, it would be beneficial to compare conventional in-class delivery of resilience curricula, the most common mode of delivery among the included studies, with other educational methods such as online learning and electronic interventions (e.g., smartphone apps) to explore their efficacy in different populations and contexts. Future research must assess the effect of varying duration and frequency of resilience curricula on participants' resilience and personal wellness; thereby elucidating the optimal curriculum delivery timeline. Lastly, future investigations must incorporate rigorous and robust research methodology to accurately determine the true effects of resilience training. A recent systematic review revealed that there is currently no gold standard outcome measure of resilience, thus rendering it difficult to assess criterion validity of various measures [107]. Further research should aim to develop a common instrument for objective resilience measurement.
The development and implementation of resilience education is a new phenomenon highlighted by an upstroke in publications during the past decade. Following recommendations by the ACGME in the United States and CanMEDS in Canada to improve trainee well-being, UGME and PGME programs in Canada and the US began to develop their own curricula to better support their trainees. Although medical trainee burnout has been extensively reported internationally, the implementation of resilience curricula has remained mostly limited to the US. Results of these preliminary studies can help to inform resilience education worldwide.
The benefits of resilience curricula are relevant to medical education, and at large, the medical workforce in today's society. Longitudinal studies have responded to the growing concern of the highly prevalent burnout and stress in the medical field and have demonstrated the predictive value of self-report resilience scales on future mental health problems in their career [108]. Yet, there remains hesitancy and lack of enthusiasm for routine provision of resilience training in medicine, where the belief that "doctors are invincible" remains fixed [109]. It is time to address the stigma, recognize vulnerabilities and push for cultural change.