Empathy as related to gender, age, race and ethnicity, academic background and career interest: A nationwide study of osteopathic medical students in the United States

Abstract Context Research on associations between medical student empathy and demographics, academic background and career interest is limited, lacks representative samples and suffers from single institutional features. This study was designed to fill the gap by examining associations between empathy in patient care, and gender, age, race and ethnicity, academic background and career interest in nationwide, multi‐institutional samples of medical students in the United States and to provide more definitive answers regarding the aforementioned associations, with more confidence in the internal and external validity of the findings. Methods Four nationwide samples participated in this study (n = 10 751). Samples 1, 2, 3 and 4 included 3616 first‐year, 2764 second‐year, 2413 third‐year and 1958 fourth‐year students who completed a web‐based survey at the end of the 2017‐2018 academic year. The survey included questions on demographics, academic background and career interest, the Jefferson Scale of Empathy, and the Infrequency Scale of the Zuckerman‐Kuhlman Personality Questionnaire to control for the effect of ‘good impression’ response bias. Results Statistically significant and practically important associations were found between empathy scores and gender (in favour of women), race and ethnicity (in favour of African‐American and Hispanic/Latino/Spanish), academic background (in favour of ‘Social and Behavioural Sciences’ and ‘Arts and Humanities’ in Samples 1 and 2) and career interest (in favour of ‘People‐Oriented’ and ‘Psychiatry’ specialties). Conclusions Special features of this study (eg, nationwide representative samples, use of a validated instrument for measuring empathy in patient care, statistical control for the effect of ‘good impression’ response bias, and consistency of findings in different samples from multiple institutions) provide more definitive answers to the issue of correlates of empathy in medical students and increase our confidence in the validity, reliability and generalisability of the results. Findings have implications for career counselling and targeting students who need more guidance to enhance their empathic orientation.


| INTRODUC TI ON
Empathy is the heart of the art of patient care. Empathy has been described as the most frequently mentioned personal quality of the humanistic physician 1 and a major element of professionalism in medicine. 2 Cultivating empathy is listed amongst the goals of medical education, endorsed by professional medical organisations. [3][4][5] Clinical empathy in patient care has been defined as a predominantly cognitive (as opposed to affective) attribute that involves an understanding of patient's experiences, concerns, pain and suffering, combined with a capacity to communicate this understanding and an intention to help. 6(p. 74) Empirical research shows that medical students' empathy is positively associated with how faculty members rate their clinical competence, 7 and physician's empathy has been found to predict positive clinical outcomes. 8,9 Significant correlations have also been reported between medical student's empathy and personality attributes conducive to relationship building (for a review see Hojat et al 10  Although a number of studies have examined associations between empathy and gender amongst medical students, [10][11][12][13][14][15] few have researched empathy and career interest in medical students, and empirical research on medical students' empathy in relation to age, race and ethnicity, and academic background is scarce. Almost all published studies on the aforementioned issues involve single-institution research using small and non-probabilistic accessible sampling designs that limit the internal and external validity of the findings. We designed this study to fill these gaps and shed light on associations between medical students' empathy and gender, age, race and ethnicity, academic background and career interest with nationwide, multi-institutional research, using large representative samples of medical students in the United States (US).

| Participants
Participants included a national sample of 10 751 (out of a total

| Study survey
We used a web-based survey that consisted of questions about students' gender, age, race and ethnicity, academic background and career interest, plus the following two scales:

| The Jefferson Scale of Empathy (S-version)
This 20-item instrument was developed by Hojat and colleagues 15 for measuring clinical empathy in the context of patient care. Items are answered on a 7-point Likert-type scale (1 = strongly disagree; 7 = strongly agree). Ample evidence supports the psychometrics of the Jefferson Scale of Empathy (JSE) in samples of medical and other health professions students in the USA and abroad. 6(pp. 83-128, 275-286) The JSE has been translated into 56 languages, and used in more than 85 countries. 6 Because of its worldwide use and extensive psychometric support, the JSE has been recognised as the most researched instrument in medical education research 16  Also, significant associations were observed between students' JSE scores and ratings given by standardised patients in the objective structured clinical examination (OSCE) stations. 18,19 More importantly, significant associations have been reported between physicians' scores on the JSE and tangible clinical outcomes in diabetic patients in the USA 8 and abroad. 9 Internal consistency reliability, determined by Cronbach's coefficient alpha, is mostly reported in the 0.70s and 0.80s, 6 and stability of scores over time by test-retest reliability (in the 0.60s) has been reported in physicians, 20 allopathic medical students 13 and osteopathic medical students. 10

| Measuring attempts to make 'good impression' responses
Respondents to self-reported personality tests can manipulate their answers to produce good impressions. Such attempts to present a more socially acceptable version of ourselves are known as the 'social desirability response set' 21 and can confound research findings, leading to invalid conclusions.
Most of the JSE items are transparent; thus, respondents can produce 'good impression' answers. We used the 'Infrequency' Scale of the Zuckerman-Kuhlman Personality Questionnaire (ZKPQ) 22 to control for 'good impression' response bias. This 10-item scale (true or false responses) was developed to identify subjects with invalid records (a sample item: 'I never met a person that I didn't like').
According to the author of this scale, scores higher than 3 on this scale indicate questionable validity of the respondent's record. 22 This scale has previously been used with medical students. 6(p. 127),23

| Procedure
The web-based survey for this study evolved through several iterations and two pilot studies. The study was approved by the

| Statistical analyses
We used Pearson correlations to examine associations between JSE scores and age. Also, we used analysis of covariance in which group classification was the independent variable, the JSE score was the dependent variable and the score on the 'Infrequency' scale of the ZKPQ 22 served as a covariate. We calculated effect sizes (mean differences in terms of standard deviation [SD] unit) for the statistically significant differences to determine the practical (clinical) importance of the statistically significant findings. 24,25 Effect sizes of 0.20 or less were considered negligible, thus practically unimportant. 24,25 Mean scores of the JSE were adjusted using the score of the 'Infrequency' scale of the ZKPQ as covariate, to control for the effect of 'good impression' response bias (less than 3% of respondents in each study sample scored above the cut-off point of 3 on this scale).

| Gender
Women constituted 49% (n = 5271) of the total participants. Table 1 shows their composition in each sample: Sample 1, 48% (n = 1738); Sample 2, 50% (n = 1383); Sample 3, 49% (n = 1180); and Sample 4, 50% (n = 970). Mean scores and SDs on the JSE for each sample and total participants by gender, and summary results of statistical analyses are also presented in Table 1. In all four samples, women consistently obtained higher JSE mean scores than men. For the total participants, the mean score for men was 111.82 (SD = 13.30) and for women it was 116.96 (SD = 10.82); the difference was statistically significant (F (1,10 619) = 472.68, P < .01) and practically important (effect size = 0.42). A similar pattern of findings was observed in each of the study samples.

| Age
The mean age in Sample 1 was 25.

| Race and Ethnicity
The majority of respondents in each of the study samples were White/Caucasian (65% of the total participants, n = 7022), followed by Asian (21% of the total participants, n = 2304), Hispanic/ Latino/Spanish (5% of the total participants, n = 485) and African-American (3% of the total participants, n = 324). (Table 2). In each of the samples, 1% or fewer identified as American Indian/Alaskan or Hawaiian/Pacific Islander. The JSE mean scores, SDs and summary results of statistical analyses are reported in Table 2.
The highest JSE mean score for the total participants was  The majority of students majored in 'Biological Sciences' (60% of the total participants, n = 6423) and the fewest majored in 'Arts and Humanities' (4% of the total participants, n = 401). The pattern of distribution of undergraduate majors was similar in the four study samples (Table 3). Means and SDs of the JSE scores by academic background and summary results of statistical analyses are presented in Table 3.

| Academic background
The highest mean scores for empathy of the total participants were obtained by students who majored in 'Social and Behavioural

| Career interest
A list of the 23 specialties most frequently pursued by graduates of colleges of osteopathic medicine was included in the survey, as well as options of 'Other' and 'Undecided.' Respondents were asked to indicate the specialty they planned to pursue after medical school.
We classified the specialties into the following broad categories:   (1) 33 (1) 35 (1) 22 (1) 129 (1) Adjusted F-ratio Respondents who did not report their gender or who were included in the 'other' category were excluded from the statistical analysis.   (5) 155 (6) 156 (6) 128 (6) 616 (6) Adjusted F-ratio Respondents who did not report their ethnicity or who were included in the 'Other' category were excluded from statistical analysis.   (15) 367 (13) 308 (12) 278 (14) 1484 (14) Adjusted F-ratio Respondents who did not report their undergraduate major or who reported majors in the 'other' category were excluded from statistical analysis. Radiology, and Surgery), and 'Other' specialties (including specialties chosen by fewer than 20 students). We retained 'Psychiatry' in its own category because in previous research, psychiatrists obtained the highest scores on the JSE 20 and we were interested to ascertain if that was also the case with osteopathic medical students.

Grouping specialties into broad categories of 'People-Oriented'
and 'Technology and Procedure-Oriented' has been used in medical education research. 6,12,20,26 Specialties that require frequent and continuous encounters with patients and preventive care consultations are grouped in the 'People-Oriented" specialties and specialties that require more technical and procedural skills are grouped in the 'Technical and Procedure-Oriented' specialties.
In this study, 'People-Oriented' was the most frequently chosen category in each study sample and was selected by 45% (n = 4867) of the total participants. The proportion expressing an interest in pursuing a 'People-Oriented' specialty (mostly a primary care specialty) increased as students progressed through medical school (from 35% in Sample 1, n = 1257, to 63% in Sample 4, n = 1231) ( Table 4).

| D ISCUSS I ON
Our results confirm some of the previous findings about associations between empathy, gender and career interest and provide new insights into associations between empathy, race and ethnicity and academic background. Our finding of higher empathy scores amongst women aligns with most of those reported in allopathic 12,13 and osteopathic medical students. 14,15,27 The gender difference in empathy has often been attributed to social learning and cultural factors. 6 However, evidence regarding gender-specific behaviours observed in infants and toddlers (eg, infant's reactive crying) 28 suggests that women's empathic inclination may have hard-wired roots, in addition to reflecting social learning and cultural factors.
Empirical research on empathy and age is scarce in medical students. Our findings of no substantial correlation between JSE scores and age agree with a few studies in health professions students. 6(p159) It may be speculated that the negligible correlation between JSE and age in health professions students could be an artifact of the 'restriction of range' phenomenon in students' ages, which does not allow the corresponding correlation to capture the full range of the relationship. More research is needed to confirm this speculation.
Empirical research on empathy and race and ethnicity in medical students is scarce. One reason is that such studies undertaken in a single institution often lack a sufficient number of available students in the under-represented race and ethnic groups to allow meaningful statistical analyses. Our findings that African-American as well as Hispanic/Latino/Spanish students obtained the highest mean empathy scores are interesting and call for further research to explore underlying reasons. A study with nursing students 29 found no significant association between race and ethnicity and scores on the JSE. However, consistent with our findings, in a multi-institutional study with allopathic medical students, African-American students obtained significantly higher JSE scores than White/Caucasian and Asian/Pacific Islander students. 19 The higher JSE scores in the under-represented African-American and Hispanic/Latino/Spanish minority groups may be explained by the notion of the "wounded healer effect" 6(p140), 30 , which describes that those who have experienced suffering can better understand the suffering of others by sharing common experiences.
This effect suggests that those who have experienced discrimination and social injustice may be more sensitive to the suffering of others and develop more empathic understanding of others who are in need of help.
Empirical research on empathy and students' academic backgrounds is also scarce. This is the first large-scale study to examine associations between empathy and undergraduate majors in medical students. Our findings do not agree with those reported in nursing students 29,31 in which no significant association was observed be- More empirical research is needed to confirm this speculation.
Consistent with our findings, several studies have reported significant differences in empathy scores in allopathic medical students who expressed an interest in a 'People-Oriented' specialty and those who expressed an interest in a 'Technology and Procedure-Oriented' specialty. 12 This pattern of findings has also been reported amongst practising physicians. 20 In her doctoral dissertation, Bailey 33 reported that medical students who planned to pursue a career in specialties requiring extensive and prolonged encounters with patients received significantly higher empathy scores than their counterparts who planned to pursue procedure-oriented specialties.
Previous studies on specialty interest and empathy in osteopathy medical students reported mixed results. One study reported that students who were planning to pursue 'People-Oriented' specialties scored higher on the JSE than their peers who were planning to pursue 'Technology and Procedure-Oriented' specialties. 34   102 (5) 116.21 (11.11) 396 (4) 116.93 (12.28) Other specialties f 740 (20) 501 (18) 441 (18) 342 (17) 2024 (19) Undecided 842 (23) 646 (23) 66 (3) 10 (<1) 1564 (15) Adjusted F-ratio Respondents who did not report their specialty plan, those who were undecided or those who reported specialties in the 'other' category were excluded from statistical analysis. f Includes those who did not report their specialty plan, those who selected 'Other' specialty or those who selected specialties that were chosen by <20 students. **P < .01.
another study with osteopathy medical students did not find such a relationship. 27 There are some differences between allopathic and osteopathic medical education philosophies. For example, in osteopathic medical education, a greater emphasis is placed on provision of holistic care, hands-on approaches to diagnosis and treatment, and integrative patient-centred care. 35,36 Thus, it is important to examine similarities and differences in research findings on empathy between allopathic and osteopathic medical students. Findings of this study regarding associations between empathy, gender and specialty interest are generally consistent with those in allopathic medical schools. We need comparable data on ethnicity and academic background to explore similarities and differences.

| Limitations and strengths
A limitation of this study is the lower than 50% response rates in

| CON CLUS IONS
This nationwide study of empathy in osteopathic medical students offers the most definitive insights to date into associations between empathic orientation in patient care and gender, race and ethnicity, academic background and career interest amongst osteopathic medical students. Our results have implications for medical students' career counselling and can also help medical schools monitor and target those who need more guidance to improve and sustain their empathic orientation towards patient care.

CO N FLI C T S O F I NTE R E S T
The authors declare that they do not have a conflict of interest.

E TH I C A L A PPROVA L
This study was approved by the Institutional Review Boards of Thomas Jefferson University and all participating colleges.