Revisiting the trajectory of medical students’ empathy, and impact of gender, specialty preferences and nationality: a systematic review

Background Empathy allows a physician to understand the patient’s situation and feelings and respond appropriately. Consequently, empathy gives rise to better diagnostics and clinical outcomes. This systematic review investigates the level of empathy among medical students across the number of educational years and how this level relates to gender, specialty preferences, and nationality. Method In accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA), the authors conducted a systematic search of studies published between February 2010 and March 2019 investigating the level of empathy among medical students. The databases PubMed, EMBASE, and PsycINFO were searched. Studies employing quantitative methodologies and published in English or Scandinavian language and examining medical students exclusively were included. Results Thirty studies were included of which 24 had a cross-sectional and 6 a longitudinal study design. In 14 studies, significantly lower levels of empathy were reported by increase in the number of educational years. The remaining 16 studies identified both higher, mixed and unchanged levels. In 18 out of 27 studies it was reported that females had higher empathy scores than males. Only three out of nine studies found an association between empathy scores and specialty preferences. Nine out of 30 studies reported a propensity towards lower mean empathy scores in non-Western compared to Western countries. Conclusion The results revealed equivocal findings concerning how the empathy level among medical students develops among medical students across numbers of educational years and how empathy levels are associated with gender, specialty preferences, and nationality. Future research might benefit from focusing on how students’ empathy is displayed in clinical settings, e.g. in clinical encounters with patients, peers and other health professionals.


Background
Empathy is usually categorised as either affective (emotional), cognitive or a combination of both. The essence of affective empathy is compassion and the ability to enter into other peoples' feelings (Einfühlung). Cognitive empathy is described as "the ability to understand someone's situation without making it one's own" [1]. In the clinical setting and within the context of the patient-physician relationship, it is predominantly the cognitive empathy type that is valued and strived for. Mercer and Reynolds [2] define clinical empathy as the ability to a) understand the patient's situation, perspective and feelings (and their attached meanings); b) to communicate that understanding and check its accuracy and c) to act on that understanding with the patient in a helpful (therapeutic) way. An empathic physician is able to sense the patient's feelings while at the same time sustaining his or her professionalism [3]. Empathy has been shown to contribute substantially to building and maintaining a good patient-physician relationship [4]. Studies on empathy among general practitioners (physicians specialised in general practice) concluded that a general practitioner's display of empathy creates a relationship built on trust, openness, and safety and that a general practitioner's empathic attitude makes the patient feel supported and listened to [5,6]. Consequently, patients are more likely to disclose accurate and important information about themselves resulting in better diagnostics and clinical outcomes [7][8][9]. Steinhausen et al. [8] found that patients who rated their physician as having "high physician empathy" using the Consultation-and-Relational-Empathy (CARE) measure had a 20-fold higher probability of a better self-reported medical treatment outcome compared to patients who rated their physician to have "low physician empathy." Furthermore, studying patients with diabetes, Hojat et al. [9] found a strong correlation between an empathic physician (measured through the Jefferson Scale of Physician Empathy (JSPE)) and lower values of lipoprotein cholesterol (LDL) and glycosylated haemoglobin (HbA1c). Beyond clinical outcomes, empathic communication has been shown to enhance patient satisfaction, compliance and patient empowerment [10][11][12]. Additionally, regarding physicianrelated benefits of empathy, physicians who perceive themselves as being empathic experience empathy as a source of professional satisfaction and meaningfulness protecting against burn-out [5,13,14]. As an offshoot of the large body of research documenting the beneficial effects of physician empathy, empathy development among medical students has become a comprehensive research topic. Moreover, the association between levels of empathy among medical students and variables such as gender, nationality and/or specialty preferences has received an increased focus among researchers. Hojat et al. [15] found that medical students interested in primary care specialties had higher empathy scores than students showing interests in technology and procedure orientated specialties. Female and male physicians are furthermore shown to approach the patient-physician relationship differently [16]. For example, female physicians value psychosocial factors more than male physicians and engage to a larger extent in patient-centred and/or relationship-centred communication [17]. These varying cultural, social and psychological influences on empathy levels are also reflected in the fact that findings from studies conducted in different countries vary to a high degree [18,19]. Several research studies using student self-report measures to measure empathy levels have documented that a significant decline in empathy occurs among medical students as their training progresses [20,21]. Contrary to these finding, however, other studies have shown that empathy levels among medical students increase or that they are maintained [22][23][24]. Neumann et al. [25] published a systematic review on student empathy in 2011, concluding on the basis of 18 studies that empathy levels decline during medical education due to, mainly, an increase in student-patient contact and interaction. Colliver et al. [26] however, conducting a metaanalysis a year earlier, concluded that student empathy levels only decrease to a minimal degree if at all. Since then, more studies on the subject have been published that assumingly reflect all the new educational initiatives taking in relation to the medical curriculum that have empathy cultivation and preservation as a key goal, such as accompanying patients on medical visits making home visits, and reading medically related literature and poetry (narrative medicine) [27,28]. Summarising the above, empathy is an important concept in health care and within educational research. However, as a consequence of many different definitions and understandings of empathy, and of different ways of measuring empathy, research in the area has also led to ambiguous results. There is therefore a need for an updated overview and review of the most recent research evidence regarding empathy among medical students.
The aim of this study was to perform a systematic review in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Guidelines (PRISMA) [29] of the literature published between February 2010 and March 2019. We sought to answer the following questions: 1. What are the empathy levels among medical students across the number of educational years? 2. How do levels of empathy relate to gender, specialty preferences, and nationality?

Search strategy
The review was conducted according to the PRISMA guidelines [29]. AJ and FA conducted a systematic search in March 2019 informed by the research questions. Three databases were searched: PubMed, EMBASE, and Psy-cINFO. The following search words were used: 'empathy' AND 'medical student' AND ('decrease' OR 'increase'). Additionally, synonyms, the National Library of Medicine's Medical Subject Heading terms (MeSH) and subject headings were identified and applied (see Additional file 1). During the full-text screening, we also performed a manual search of reference sections to identify studies not found through the database searches.

Inclusion and exclusion criteria
Inclusion criteria were the following:

Selection of data
Titles and abstracts of the studies were screened. In the case of uncertainty, full texts were read. Disagreement between reviewers (AJ and FA) regarding inclusion of the studies was settled through discussion until concordance was reached. Afterwards, AJ and FA read the full texts of the eligible studies. Together, the authors summarised and analysed the methods, results, and discussion sections of the studies. Independently, we applied methodological quality assessment tools on the different studies according to study design. Crombie's items [30] were applied to cross-sectional studies (n = 24) and consist of seven items rated as "yes" (1 point), "unclear" (0.5 points) and "no" (0 points), with a maximum of 7 points. The quality of the longitudinal studies (n = 6) was assessed by employing a structured 33-point checklist from Tooth et al. (see Additional file 2) [31]. Possible disagreements were discussed and settled and there was inter-rater reliability.

Included studies
The search resulted in 1501 studies, of which 347 were duplicates (see Fig. 1). A total of 1154 studies were screened by title and abstract. Among these, 41 studies were selected for full-text reading since they fulfilled the inclusion criteria. During full-text reading, reference sections were also screened, which revealed another 12 eligible studies. A total of 53 studies were full-text screened. We excluded 23 of the 53 studies since they did not apply to our aim (n = 20) or were in a language other than English or Scandinavian (n = 3). Altogether 30 studies were included in the review.

Study characteristics Study design and sample sizes
Of the 30 studies included in the review, 24 studies were cross-sectional and 6 studies longitudinal (see study characteristics and main findings in Table 1). Sample sizes of the cross-sectional studies varied from 129 [28] to 5521 [48] participants. In the longitudinal studies, sample size varied from 72 [52] to 1653 [55] participants.

Scales
All cross-sectional studies employed the Jefferson Scale of Physician Empathy student version (JSPE-S), except for four studies using the following scales: the Basic Empathy Scale [40], Measure of Patient-Centered Communication (MPCC) [28], Reading the Mind in the Eyes (RMET) and Balanced Emotional Empathy Scale (BEES) [37], and Empathic Skill Scale Form B and Conflict Tendency Scale [23]. All longitudinal studies used JSPE-S except for one that applied the Interpersonal Reactivity Index scale (IRI) [55]. One longitudinal study applied both an observational Objective Structured Clinical Examinations (OSCE) evaluation and JSPE-S [38]. Likewise, a crosssectional study used the Measure of Patient-Centered Communication (MPCC), which is also an observational scale that measures empathy [28].

Country
The studies were conducted in 20 different countries.

Quality assessment and risk of bias in the included studies
The quality assessment tools were used to identify the risk of bias. All included studies employed self-reporting questionnaires. Consequently, reporting bias was present which may have influenced the results. Three studies used small sample sizes, including respectively 129 [28], 77 [32], and 122 [56] study participants Hence, the findings of those studies may not be representative of the student population measured and it might over-and/or underestimate the outcome measures.
One obvious limitation of the cross-sectional studies design was their inability to report changes over time. On the contrary, longitudinal studies could describe changes over time. Only one study used a control group of non-medical students, increasing its quality since it enabled comparison.
All studies, except for one [37], employed validated scales to examine the level of empathy. One study [53] employed the English validated JPSE-S on students who did not have English as their native language.

The levels of empathy across number of educational years
Significantly lower levels of empathy by increase in number of educational years were found in 14 out of 30 studies. Of these, 12 were cross-sectional studies [21, 33, 34, 39-41, 45-48, 50, 52, 54] and two were longitudinal [52,56]. All except one [55] of the cross-sectional studies used JSPE-S. Four cross-sectional studies [23,27,28,44] reported a higher level of empathy among medical students at a higher year of medical school. Five crosssectional studies [24,42,43,49,51] and one longitudinal study [32] found no statistically significant difference in empathy scores across the different years of medical education. Hasan et al. [53] reported higher empathy scores with higher educational years up until the fourth year, where a decreasing trend was observed. A crosssectional study [37] differentiated between emotional and cognitive empathy and found a higher cognitive empathy level in final year students compared to first-year students. On the contrary, a longitudinal study [55] found no change in cognitive empathy.
Chen et al. [38] conducted a longitudinal study, applying both self-administered empathy measures and observed empathy in an OSCE. It showed higher self-administered        cross-sectional study [27] year and senior students, between genders, and between specialty preferences than first-year medical students -Females had higher empathy scores than males -   -Singleinstitution survey empathy scores among second-year students compared to third-year students and the opposite for the observed empathy scores. In another longitudinal study by Chen et al. [35] higher levels of empathy were found up to the thirdyear of education, followed by a persistent decline. Smith et al. [56] conducted a longitudinal study applying both JSPE-S and the Questionnaire of Cognitive and Affective Empathy (QCAE). The two scales revealed incongruent results: the QCAE score increased over time while JSPE-S measured a decrease over time.

Gender
Female students were reported to have higher empathy scores compared to male students in 16 cross-sectional and 2 longitudinal studies [21, 24, 27, 34-37, 39, 40, 42-44, 47-50, 53, 56]. One longitudinal study by Quince et al. [55] found a lower level of emotional empathy among men compared to women who did not show any change. No gender differences were found in relation to cognitive empathy and no differences between genders were detected in seven cross-sectional [23, 28, 33, 41, 45,   46, 54]. Three studies did not investigate the differences in empathy across genders [32,38,51].

Specialty preferences
Nine cross-sectional studies investigated a possible relation between empathy scores and specialty preferences of the students [27,28,33,34,39,43,45,46,53]. Three studies detected higher levels of empathy among students who preferred a "people-orientated" specialty [28,43,45]. No association between specialty preferences and empathy scores was found in the remaining six studies. None of the longitudinal studies examined specialty preferences.

Main findings
This systematic review aimed to investigate the level of empathy among medical students across the educational years and how the measured empathy levels relate to gender, specialty preferences, and nationality. In reviewing studies from 20 different countries, variations were found in the level of empathy among medical students across the number of educational years. Nearly half of the included studies [21, 33, 34, 39-41, 45-48, 50, 52, 54, 56], of which only two [52,56] were longitudinal, reported lower empathy scores with higher educational years. The remaining 17 studies [23, 24, 27, 28, 32, 35-38, 42-44, 49, 51, 53, 55, 56] identified both higher, mixed or unchanged levels of empathy throughout the medical education.
Most studies [21, 24, 27, 34-37, 39, 40, 42-44, 47-50, 53, 55, 56] found a tendency towards higher levels of empathy among female students as compared to male students. Out of nine cross-sectional studies, only three [28,43,45] reported an association between empathy and specialty preferences. Furthermore, studies from non-Western countries reported a lower level of mean empathy scores as compared to Western countries. These findings thus differed from the previous review by Neumann et al. [25] which concluded that empathy decreases by an increase in the educational years particularly among those preferring "non-people-orientated" specialities. While different results might be explained by differences in study populations, study design (longitudinal vs cross-section), the instrument used, local culture, etc., this review tells us that we cannot make the often quoted statement that "empathy declines with level of training".

Possible explanations for lower and higher levels of empathy
In the literature, several explanations for a decline in empathy have been discussed without demonstrating a clear causal relationship. Some scholars point to the phenomenon of burnout among medical students and refer to the association found in the literature between high burnout level among medical students and low empathy score [65][66][67]. Related, stress among medical students [68][69][70] has also been shown to correlate negatively with empathy [69]. Another explanation put forward in the literature for empathy decline is increased patient contact during clinical training [34,35,38,45,46,50,52]. Chen et al. [38] explained the development towards lower levels of empathy during clinical training as a result of an acculturation process in which superiors and mentors try to protect their students against psychological distress by cultivating a climate of cynicism, emotional distance and detachment among medical students in their contact with patients and at the same time try to safeguard "professionalism" in the clinical setting. Moreover, Li et al. [45] stated that clinical training might encompass intense patient-physician relationships, long working hours and sleep deprivation, leading to lower levels of empathy after clinical training. Furthermore, in the literature, the so-called "hidden curriculum," lack of role models, fear and anxiety in the meeting with the patients, and increased workload are also pointed out as possible reasons for a decline in empathy [46,71,72]. Another explanation mentioned in the literature is that the medical curriculum focuses more on diagnosis and treatment than humanistic values [73]. Shapiro et al. [71] also stated that the biomedical discourse has diverted the students' focus from empathy leading him/her to adopt a mechanistic view on illness that might reduce the patients to a disease or an object.
Discussing the increases in empathy levels that were documented in some of the reviewed studies, Magalhaes et al. [27] pointed out that the medical curriculum has an increased focus on the development of empathy as the educational years progress and that students have increasingly reached an acknowledgment of the importance of empathy in the patientphysician relationship. This point of view was put forward as a possible explanation for the documented increases in empathy. Furthermore, training and competence acquirement through clinical training of communication skills have also been proposed as an explanation for the tendency towards higher levels of empathy in senior year medical students [27,28]. In relation to these explanations it should also be kept in mind that the medical curriculum varies across countries and medical schools.

Gender differences
In the literature, varying explanations for gender differences are suggested. Bertakis et al. [16] found that females are more receptive to emotional signals than males. Furthermore, they are said to show more interest in the patient's family and social life, thereby being able to achieve a better understanding of the patient and reach a more empathic relation. Shashikumar et al. [39] stated that females through evolutionary gender differences are more caring and loving.

Nationalities
Nine studies in our systematic review reported a propensity towards lower empathy scores in non-Western compared to Western countries. All of these studies applied the JSPE-S. Shariat et al. [47] stated that awareness of the cultural differences should be kept in mind when applying the JSPE-S in cultures that differ from the USA, where the JSPE-S was developed. A Japanese psychometric study of the JSPE pointed out that Japanese patients preferred their physician to be calm and unemotional, emphasising that cultural differences could indeed explain the differences in empathy scores between countries and cultures [74].

Specialty preferences
A possible association between levels of empathy and specialty preferences was investigated in nine of the included studies [27,28,33,34,39,43,45,46,53]. Only three studies [28,43,45] reported an association between higher level of empathy among people preferring "people-orientated" specialities. Engaging in an empathic understanding of the patients' feelings and life circumstances is important in all medical specialities since showing an empathic attitude towards the patient has been shown to lead to positive effects on patients' health outcome [8,9]. It can be argued, however, that a focus on empathy is relevant mostly within people-orientated specialities since physicians who work in these specialities are both in need of help regarding empathy preservation (helping patients) and administration (helping themselves so as to decrease the risk of stress and burnout) [3].

Strengths and limitations
A strength of the present systematic review is that the literature search was conducted in three databases. Furthermore, the screening of literature and selection of studies was performed by two reviewers. Moreover, we consider the implementation of a quality assessment of all included studies as a strength. This review has several limitations. Since our search words included words that presume a change e.g. "decrease" and "increase," our search may be too narrow, and there is a risk that relevant studies have been overlooked. Additionally, possible relevant studies in languages other than English and Scandinavian were not included. Another limitation is that only quantitative studies were included. This excluded qualitative aspects that could have contributed to a more varied and profound understanding of the quantitative findings.

Future research
Most of the included studies applied the selfadministered JSPE-S and therefore did not explore the display of empathy that takes place between the patient and the medical student. Sulzer et al. [75] stated that the JSPE-S scale focuses on thoughts and not actions. Furthermore, research has shown that self-reported empathy has only a vague association with the patientphysician relationship in the clinical setting [75]. To improve the knowledge of empathy among medical students, research that includes both cognition, action, and feelings is recommended [75]. Incorporation of non-medical students as control groups is also required in order to gain more insight into whether medical students' levels of empathy differ compared to other university students. Furthermore, future investigators should employ a variety of research designs to investigate the important role of empathy in medical education, such as mixed methods research, observational research, and qualitative research. These studies could focus -not on self-reportingbut rather on patient perceptions of empathic student/physician behaviour. Qualitative research conducted with students could also contribute to new perspectives and insights about student-perceived factors influencing the development of empathy and its expression in clinical care. Lastly, a meta-analysis is desirable since it enables the calculation of statistical significance and heterogeneity.

Conclusions
This systematic review including thirty studies, revealed varied and inconsistent findings on the levels of empathy among medical students. Statistically lower empathy scores by an increase in educational years were found in 14 studies. The remaining studies reported higher [4] and unchanged [6] scores in empathy. In most studies, females were reported to have higher levels of empathy than males. Study participants from non-Western countries reported a tendency towards lower mean empathy scores as compared to those from Western countries. Only a few studies reported a correlation between "people-oriented" specialty preferences and empathy scores. Future research should focus on examining relational empathy in studentpatient interaction using observational scales and qualitative methodologies.