Curricula for empathy and compassion training in medical education: A systematic review

Background Empathy and compassion are vital components of health care quality; however, physicians frequently miss opportunities for empathy and compassion in patient care. Despite evidence that empathy and compassion training can be effective, the specific behaviors that should be taught remain unclear. We synthesized the biomedical literature on empathy and compassion training in medical education to find the specific curricula components (skills and behaviors) demonstrated to be effective. Methods We searched CENTRAL, MEDLINE, EMBASE, and CINAHL using a previously published comprehensive search strategy. We screened reference lists of the articles meeting inclusion criteria to identify additional studies for potential inclusion. Study inclusion criteria were: (1) intervention arm in which subjects underwent an educational curriculum aimed at enhancing empathy and/or compassion; (2) clearly defined control arm in which subjects did not receive the curriculum; (3) curriculum was tested on physicians (or physicians-in-training); and (4) outcome measure assessing the effect of the curriculum on physician empathy and/or compassion. We performed a qualitative analysis to collate and tabulate effects of tested curricula according to recommended methodology from the Cochrane Handbook. We used the Cochrane Collaboration’s tool for assessing risk of bias. Results Fifty-two studies (total n = 5,316) met inclusion criteria. Most (75%) studies found that the tested curricula improved physician empathy and/or compassion on at least one outcome measure. We identified the following key behaviors to be effective: (1) sitting (versus standing) during the interview; (2) detecting patients’ non-verbal cues of emotion; (3) recognizing and responding to opportunities for compassion; (4) non-verbal communication of caring (e.g. eye contact); and (5) verbal statements of acknowledgement, validation, and support. These behaviors were found to improve patient perception of physician empathy and/or compassion. Conclusion Evidence suggests that training can enhance physician empathy and compassion. Training curricula should incorporate the specific behaviors identified in this report.

curricula, which can be implemented during medical training, as well as help inform currently practicing physicians. The first step in developing evidence-based curricula is to identify the specific skills and behaviors that ought to be taught and how best to transfer this knowledge to the learner.
The objectives of this systematic review are to collate the world's literature on empathy and compassion training in medical education to determine (1) the specific skills and behaviors that should be taught (i.e. have been demonstrated to enhance patient perception of compassion), and (2) the methods of training that are most effective. [32] The results of this report will help inform the development of evidence-based curricula for empathy and compassion training in medical education.

Protocol and registration
We developed and published a systematic review protocol [32] in accordance with the Cochrane Handbook for systematic reviews of interventions, [33] and the Preferred Reporting Items for Systematic Reviews and Meta-Analysis Protocols (PRISMA-P) statement.
[34] Our final results are reported according to the PRISMA guidelines. [35] This systematic review was registered in the PROSPERO international prospective register of systematic reviews (registration number CRD42018095040).

Search for and identification of studies
Our electronic search included databases generally considered to be the most important sources: [33] CENTRAL, MEDLINE, EMBASE, and CINAHL. The search strategies were established using a combination of standardized terms and key words (including empathy, compassion, and derivations thereof), and the fully reproducible search strategy was previously published. [32] We also performed recommended techniques for systematic reviews of complex evidence: we reviewed reference lists of the included articles to identify additional studies for potential inclusion, used electronic citation tracking, and consulted experts in the field. [36] The final search was performed on Feb 1 st , 2019.

Eligibility criteria
We included all clinical studies of educational curricula that were described as either empathy training or compassion training. We included both on the grounds of the inter-relatedness and inter-dependence of these constructs as described above, and the fact that training to improve empathy (i.e. the understanding component) typically also improves compassion (i.e. the action component), and training to improve compassion would likely require improving empathy. Further, it would not be possible to perform a rigorous systematic review of one without the other, in that most training programs in this domain involve enhancement of both understanding patients' emotions and taking action with behaviors toward patients.
As stated in our previously published protocol the inclusion criteria for studies were: (1) an intervention arm in which subjects clearly underwent an educational curriculum aimed at enhancing empathy or compassion; (2) a clearly defined control arm in which subjects did not receive the curriculum (e.g. wait-list, before/after, standard training); (3) the curriculum was tested on physicians, or physicians-in-training; and (4) an outcome measure assessing the effect of the curriculum on physician empathy or compassion. [32] We included outcomes measured from any perspective, including physician self-assessment as well as assessment by patients, standardized patients, or third party observers. We did not exclude studies based on language or publication type or date. We excluded secondary reports of previously published trials, reviews, correspondence, and editorials; however, we screened the reference lists of review articles to identify further studies for inclusion.

Study selection and data abstraction
As described in our previously published protocol two independent reviewers screened the titles and abstracts of identified studies for potential eligibility. After completion of the relevance screen, the two reviewers compared exclusion logs to determine whether there was disagreement and used the Kappa statistic to quantify the inter-observer agreement. In cases of disagreement, the full text was reviewed for inclusion. For all studies deemed potentially relevant the full manuscripts were reviewed for inclusion. Two reviewers independently abstracted data on all study populations, interventions tested, outcome measures, and effect of interventions on outcome measures compared to control groups, using a standardized data collection form. Any disagreements in these processes were resolved by consensus with a third reviewer. [32]

Assessment of study bias
Study quality was assessed using the Cochrane Collaboration's tool for assessing the risk of bias evaluating six domains (selection, performance, detection, attrition, reporting, and other biases). [33] Analysis We performed a primarily qualitative analysis of the literature in accordance with the recommended methodology for qualitative reviews published in the Cochrane Handbook. [33] In table format, stratified by individual publication, we collated and summarized the following: (1) study design; (2) population sampled (i.e. medical student, resident, attending physician); (3) sample size; (4) specific skills (e.g. identifying compassion opportunities) and behaviors [both verbal (e.g. compassionate statements) and non-verbal (e.g. eye contact, body position)] taught by the curriculum; (5) training methods utilized (e.g. lecture, small groups sessions, simulated experiential learning); (6) assessment methods for outcome measures; and (7) effect of curriculum on outcome measures compared to control groups. We determined the interventional curriculum of each study to be effective if the study identified a statistically significant difference in an empathy or compassion outcome measure in favor of the study curriculum group compared to the control group.
We were unable to use a meta-analytic approach to quantitatively analyze the data secondary to the heterogeneity in both interventions and outcome measures. review of references we identified 11 additional potential articles. A full manuscript review was performed on 333 papers, resulting in 52 papers included for final analysis with a total of 5,316 subjects.

Study characteristics
The 52 studies were published over 42 years . Study characteristics for all 52 studies are displayed in Table 1. The majority of studies (54%) were published in the last five years (Fig 2). The most common study design was before/after [44% (23/52), n = 1,977], followed by randomized control trial [29% (15/52), n = 1,286], and prospective cohort study [27% (14/52), n = 2053]. The Cochrane tool for assessing risk of bias identified some concern for risk of bias for all included studies (i.e. "high risk" or "unclear risk") (S1 Table). There were no adverse events related to the study interventions reported.

Study populations
Forty-six percent (24/52, n = 3120) of studies tested the training curriculum among medical students and 38% (20/52, n = 882) tested the curriculum among residents. Only eight studies (15%, n = 1314) tested the curriculum among practicing physicians.  Poole, et al [85] 1980 Prospective cohort 2 nd year medical students 25:20 Commercial training program: "Tune-In, Empathy Training Workshop" [86] Eight 1.5 to 2 hour audiotape-led sessions Improvement in the Accurate Empathy Scale as rated by a third party observer during a patient interview three years after the intervention compared to preintervention, as well as compared to controls.

Training methods
Duration of the included training curricula varied considerably from a single one-hour session to multiple sessions over three years. The majority of the study curricula involved more than one session [75% (39/52), n = 3323). The majority of the study curricula incorporated small group sessions as a part of the curriculum [63%, 33/52, n = 3791] and 46% (24/52, n = 2544) incorporated didactic lectures. Thirty (58%, n = 2679) studies incorporated practicing learned  skills through role-playing (16 studies, n = 1246), standardized patient interviews (9 studies, n = 1118), or real patient interactions (6 studies, n = 343). One included study incorporated both role-playing and standardized patient interviews (n = 28). Four studies (n = 386) incorporated video recording of interviews, on which subjects reviewed and received feedback.

Outcome measures
The majority of studies tested the effects of the training curriculum on self-assessed outcomes [56% (29/52), n = 3643] (i.e. trainees assessment of their own empathy or compassion). The most commonly used self-assessment outcome measure was the Jefferson Scale of Empathy [72% (21/29), n = 3258). Twenty-five studies (48%, n = 2002) measured empathy or compassion as rated by a third party observer, seven studies (13%, n = 805) as rated by standardized patients, and eight studies (15%, n = 1132) as rated by actual patients. The majority of studies used a previously validated measurement tool [73% (38/52), n = 4342], while 17 studies (33%, n = 1098) incorporated a new measurement tool. Only two studies evaluated long-term effects of the training curriculum (i.e. at 12 months after completing training). [74,84]

Study results
The majority of studies found the tested training curriculum improved physician empathy or compassion as measured on at least one outcome measure [75% (39/52), n = 4532]. Success rates among studies involving medical students, residents, and physicians were, 87% (21/24), 65% (13/20), and 63% (5/8) respectively. Success rates among studies using self-assessment outcomes, third party raters, standardized patient raters, and actual patient raters were, 45% (13/29), 88% (22/25), 57% (4/7), and 75% (6/8) respectively. We found training methods involving actual patients (six studies), as well as video recording of interviews (four studies), had the highest success rate with 100% of these curricula demonstrating improvement on at least one outcome measure. Success rates for other training methods are displayed in Fig 3. We found 77% (30/39) of curricula involving more than one session had improvement on at least one outcome measure compared to 69% (9/13) of curricula involving a single session.

Clinical skills and behaviors
All study curricula incorporated teaching some aspect of taking time to listen and/or having awareness of the patient's emotional state. Skills and behaviors that demonstrated an increase in real patient perception of compassion included (1) sitting (versus standing) during the interview;[50] (2) detecting patients' facial expressions and non-verbal cues of emotion; [66] (3) recognizing and responding to opportunities for compassion; [70] (4) non-verbal communication of caring [i.e. employing non-verbal caring behavior (e.g. body position facing the patient, eye contact, tone of voice, and appropriate hand and arm movements), as opposed to avoidant or aggressive behavior]; [67] (5) incorporating statements of support (e.g. "I'm here for you. Let's work together"), worry (e.g. "What concerns you most?"), acknowledgement (e.g. "This has been hard for you"), patient's perspective (e.g. "How does it disrupt your daily activity?"), emotion naming (e.g. "You seem sad"), and validation (e.g. "Most people would feel the way you do"). [48] Of the two studies that measured outcomes out to 12 months, one curriculum incorporated Balint group training (i.e. a method of exploring the dynamics of patient interactions, and gaining insight into personal reactions to patients, in an effort to more effectively meet the biopsychosocial needs and challenges of patients), and did not find a difference in Jefferson Scale of Physician Empathy at 12 months. [74] The second curriculum taught specific skills including verbal explanation, small talk, listening, calming, compassionate response, encouragement, questioning, nodding, smiling, laughing, eye contact, and supportive touching, and found an increase in supportive behaviors at 12 months as measured by a third party observer. [84] Table 1 displays the clinical skills and behaviors taught among the included studies, along with outcome measures and results.

Discussion
In this report, we collated the current biomedical literature on empathy and compassion training curricula for physicians and physicians-in-training. Our objective was to qualitatively describe the specific skills and behaviors that have previously been demonstrated to improve physician empathy and compassion, and the methods of training that are most effective at transferring this knowledge to the learner.
Consistent with previous reports, we found that among the 52 studies meeting criteria for inclusion the preponderance of evidence indicates that training curricula are effective for enhancing physician empathy and compassion. This report further advances this field of research in that we have tabulated the specific skills and behaviors, which have been demonstrated to enhance (or failed to enhance) physician empathy and compassion. Thus, we have developed an evidence-based framework from which researchers and educators can develop and test future training curricula. Specifically, we identified the following behaviors may improve patient perception of provider empathy and compassion: (1) sitting (versus standing) during the interview; (2) detecting patients' facial expressions and non-verbal cues of emotion; (3) recognizing and responding to opportunities for compassion; (4) non-verbal communication of caring (e.g. facing the patient, eye contact); and (5) verbal statements of acknowledgement, validation, and support. A possible common denominator among these interventions is assuring the patient of true physician presence and focus, and that they are not going through their current medical condition alone, but that they have the full attention and support of the physician. This report found heterogeneity in the curricula studied, as well as outcome measures used to test the effectiveness of the curricula. Patient perspective of physician compassion has previously been demonstrated to be associated with improved clinical outcomes. [91][92][93] However, in this systematic review we only identified eight studies measuring physician empathy/compassion from the patient perspective. A commonality among these studies was a focus on taking time to be fully present, listening, and/or having awareness of the patient's emotional state. Learning and incorporating such skills shifts the focus from narrow biomedical inquiry to knowing the patient as a whole person, which has been demonstrated leads to increased patient-reported trust in their provider. [70] Such trust has been demonstrated to improve compliance with prescribed therapies and has been suggested to improve clinical outcomes. [6] We also found that teaching providers specific skills and behaviors increased patient assessment of physician compassion. Thus, we propose the design of future curricula should include the training of nonverbal behaviors such as sitting, body position towards the patient, calm tone of voice, and eye contact. These results are consistent with previous evidence that physicians who sit during consultation are considered to be more compassionate compared to those that stand. [94] Similarly, Sherer et al found that observers rated psychology therapists who sit in close proximity to the patient (91 cm), in addition to provide consistent eye contact (90% of the time), as having more empathy, warmth, and genuineness compared to those that sat further (213 cm) and provided minimal eye contact (10% of the time). [95] Future curricula should also focus on educating providers on the importance of listening and identifying empathy and compassion opportunities, as well as provide guidance/examples on how to respond to these opportunities with statements of support, acknowledgement, and validation.
[48] Finally, we propose that testing of these curricula should incorporate validated outcome measures, which measure actual patient perception of physician empathy and compassion, as opposed to standardized patients or third-party reported measures. [96] Given that the patient experience of compassion (or lack thereof) is likely what drives the association between physician compassion and clinical outcomes, future research should employ patient assessment of physician compassion for testing the effects of training curricula on clinical outcomes. Importantly, burnout among physicians has a major economic toll on health care, as well as a major toll on the health of patients. [97] Burnout has been identified as a major public health issue, with recent reports identifying that approximately 50% of physicians are experiencing burnout. [98] There is now evidence that compassionate patient care may be beneficial for physicians. Specifically, compassionate patient care may enhance physician resilience and resistance to burnout. [99][100][101] Therefore, empathy and compassion training curricula may be an effective therapy to reduce physician burnout. Thus, in addition to measuring patient perspective of physician compassion, future curricula should also incorporate specific skills and behaviors demonstrated to improve physician self-assessment of empathy and compassion (Table 1).
This systematic review also identified methods of training that are most effective. The preponderance of evidence to date suggests that in addition to didactic lectures, incorporating a curriculum in which physicians can practice learned skills might be the best for enhancing physician empathy and compassion. Specifically, similar to medical training in which clinical skills are practiced in real time with actual patients and oversight by practicing physicians, incorporating similar methods to "practice" compassion in the clinical setting appears to be beneficial. Additionally, in certain clinical settings, and with patient consent, videotaping interactions with patients as a method to provide feedback on verbal and non-verbal behaviors appears to enhance compassionate behaviors. Of similar importance we identified curriculum and training methods that were not effective. We found studies that only incorporated didactic and/or small group sessions were the least likely to be effective (13/22). None of the four studies that focused on using art were found to improve self-reported empathy scores. [44,54,62,65] In addition, only one of the five studies that focused on Balint group training had a positive effect on the reported outcome measure. [43,57,61,74,79] We recognize there are important limitations of this systematic review to consider. First, all 52 included studies had some risk of bias according to the Cochrane Collaboration's tool for assessing the risk of bias in clinical trials. Therefore, the results of the included studies must be interpreted with some caution. Second, there were varying educational scenarios, curricula studied, and outcomes measurements used, resulting in a high degree of heterogeneity. Therefore, we were not able to perform a quantitative meta-analysis to determine the effects of specific curricula on any particular clinical outcome. Third, there was a paucity of studies evaluating long-term outcomes. Thus, we are not able to determine if the effects of the tested curricula are sustained over time. However, one study found that teaching specific verbal and non-verbal behaviors resulted in increased supportive behaviors at 12 months after the training. [84] Fourth, there is significant overlap between the constructs of empathy and compassion, [5] and to date there is no agreed upon instrument to measure empathy and compassion in health care. [102] While the majority of the studies reported an outcome measure of empathy (48/52), it is possible that a component of compassion was also being measured. For example, the Consultation and Relational Empathy (CARE) measure is stated to measure empathy; however, one of the items of this measure specifically asks, "how was the doctor at showing care and compassion?" Thus, given the complex nature of the empathy/compassion relationship it is unlikely that the intervention curricula affected, or the outcome measures assessed, either empathy or compassion alone. Therefore, we believe it is not possible to precisely differentiate the two constructs in this report, and future research is required to further delineate and define the different effects of empathy versus compassion training. Fifth, none of the studies assessed direct clinical outcomes of patients, and as such it is not clear whether the observed changes in the indices of empathy or compassion training have meaningful implications for patient health outcomes. Sixth, although we searched the databases generally considered to be the most important sources to search, [33] and we did not exclude studies based on publication type, it remains possible that pertinent studies were conducted and either not published or not captured by our search strategy.

Conclusion
In summary, current evidence suggests that training can enhance physician empathy and compassion. This report has collated the medical education literature on skills and behaviors that enhance physician empathy and compassion, and provides a framework from which researchers and educators can develop evidence-based curricula.