Empathy has been identified as a major element of professionalism in medicine.¹ Cultivating empathy is among the goals of medical education and has been endorsed by professional medical organizations.^2,3 The National Board of Osteopathic Medical Examiners lists the cultivation of humanistic behavior and exhibition of compassionate treatment, altruism, and empathy among the required competencies in the practice of osteopathic medicine.^4(pp26-27) 

A European medical education guide⁵ suggests that assessing and enhancing personal qualities such as empathy must be considered a mandate—not just declared desirable—by leaders in medical education and medical professional organizations. Furthermore, it has been proposed that the assessment of medical school applicants’ empathy should be considered a supplementary measure for admissions decisions, particularly in situations in which applicants have comparable academic qualifications.⁶ 

However, empathy is an elusive concept that is difficult to define and hard to quantify. To clarify the conceptual ambiguity associated with empathy, clinical empathy in the context of health care professions education and patient care was defined as a predominantly cognitive (rather than an affective or emotional) attribute that involves an understanding (rather than a feeling) of pain and suffering of the patient combined with a capacity to communicate this understanding and with an intention to help.^7,8 The 4 key terms in this definition are italicized to underscore their significance in the construct of clinical empathy. 

Based on the aforementioned definition of clinical empathy, the JSE was developed for measuring empathy specifically in the context of medical education and patient care.^7-10 Before the development of the JSE, no psychometrically sound instrument was available for measuring empathy in that context. Although a few instruments were available and used for measuring empathy in medical education research,^7(pp57-68) none had face and content validity in the context of patient care.^11,12 Thus, there was a need for a content-specific and context-relevant empathy-measuring instrument. The JSE was developed in response to that need. 

Researchers in the United States and abroad have accumulated an abundance of evidence in support of the psychometrics of the JSE through samples of health care professionals, allopathic medical students, and practitioners.^{7(pp84-128,276-286)} For example, statistically significant associations have been reported between allopathic medical students’ scores on the JSE and ratings of clinical competence given by medical school faculty.¹³ Also, statistically significant associations were observed between students’ JSE scores and ratings of clinical competence given by standardized patients in objective structured clinical examination stations.^14,15 

More importantly, statistically significant associations have been reported between physicians’ scores on the JSE and tangible clinical outcomes in diabetic patients in the United States¹⁶ and Italy.¹⁷ Although researchers at the Cleveland Clinic failed to reproduce significant associations between physician empathy and patient outcomes,¹⁸ this inability to reproduce the findings was attributed to issues related to the internal and external validity of the findings.¹⁹ Another study²⁰ in China showed that physician scores on the JSE had a mediating role in the immunocompetence of patients with advanced prostate cancer. 

Also, statistically significant correlations have been reported between medical students’ JSE scores and conceptually relevant measures, such as indicators of communication skills,²¹ from observations of empathetic communication patterns with standardized patients.²² In a study at the Cleveland Clinic,²³ statistically significant associations were observed between physicians’ JSE scores and standardized measures of patients’ communication experiences with their physicians. Also, studies have found that medical students’ scores on the JSE are associated with personal qualities that are conducive to relationship building; these qualities include empathetic concern and perspective-taking^24,25; warmth, dutifulness, and compassion⁸; sociability²⁶; agreeableness; openness to experience; conscientiousness and extraversion²⁷; emotional intelligence^28-30; cooperativeness³¹; desirable professional behavior³²; positive attitudes toward patient-centered care and orientation toward integrative patient care in osteopathic medical students³³; positive social influence³⁴; humanistic excellence by peer nomination³⁵; personal accomplishment³⁶; and orientation toward teamwork and interprofessional collaboration in allopathic³⁷ and osteopathic medical students.^33,38 

Conversely, JSE scores were negatively correlated with personal qualities that are detrimental to relationship building, including aggression-hostility,^26,39 indicators of burnout such as depersonalization and emotional exhaustion,^36,40,41 and neuroticism.⁴² In a 2018 study⁴³ (using POMEE data), we found strong evidence in support of the measurement properties, underlying components, and latent variable structure of the JSE in a nationwide sample of first-year matriculants to osteopathic medical schools. 

Internal consistency reliability (determined by Cronbach α) in the United States and abroad are mostly reported in magnitudes ranging from 0.70 to 0.89.^7(pp276-286) Stability of scores over time indicated by test-retest reliability (0.60-0.69) has been reported in physicians¹⁰ and in osteopathic medical students who participated in POMEE (Hojat et al, unpublished data, 2019). These studies^24-43 provide strong evidence of the validity and reliability of the JSE. The JSE has been translated into 56 languages and is used in more than 85 countries.⁷ Because of its worldwide use and extensive psychometric support, the JSE has been recognized as the most researched⁴⁴ and widely used instrument in medical education research.⁴⁵ 

Norms are performance data based on large representative samples from well-defined target populations that are used as references to assess, interpret, or place individual performance against the corresponding normative populations. Development of national norms requires pertinent data collected from nationwide and representative samples from target populations. The present study aimed to develop the first national norm tables for the assessment of osteopathic medical students’ scores on the JSE. 

Forty-one of 48 campuses of US colleges of osteopathic medicine participated in this project, representing 85% of all osteopathic college campuses in the country at that time. Student participants were broken into 5 samples: first-year matriculants who responded to the study survey at the beginning of the 2017-2018 academic year (sample 1) and first-year students (sample 2), second-year students (sample 3), third-year students (sample 4), and fourth-year students (sample 5) who completed the study survey at the end of the 2017-2018 academic year. 

The web-based survey consisted of questions about respondents’ demographics (age, gender, race/ethnicity), academic background, career interest, work experience before medical school, and year of medical school education, plus the following 2 scales: 

One of the shortcomings of self-reported personality surveys is the confounding effect of good impression response bias. Respondents can manipulate their answers to produce good impressions. Such attempts to present oneself in a more socially acceptable light is known as the “social desirability response set.” Because most items of the JSE are transparent, they can be answered in a way that seems more socially acceptable. Such attempts to manipulate responses can confound research findings and lead to invalid conclusions. A unique feature of the POMEE study was the statistical control for the confounding effect of good impression response bias on the JSE scores using the Infrequency Scale of the ZKPQ, which allowed us to identify and remove those with exaggerated good impression response bias from data used to develop norm tables. 

The study survey underwent several iterations and 2 pilot testings (alpha and beta). The purpose of the alpha testing was to improve the clarity and comprehensiveness of the study survey; reviewers included medical education researchers and medical students. The purpose of the beta testing was to detect any possible bugs or crashes in its online administration and to explore the limits of the web-based product when using different desktop and mobile devices. Participants in beta testing were mostly student volunteers from 1 allopathic and 1 osteopathic medical school. 

The Jefferson team submitted the institutional review board application for the project to the institutional review board of Thomas Jefferson University. An exemption status was granted. All participating colleges submitted applications to their own institutional review board and received approval related to human subjects research. 

We arranged to have 1 or 2 senior administrators or faculty-level research coordinators from each participating college to serve as a liaison between the college and research teams at AACOM and Jefferson. These liaisons, who also administered the web-based survey at their college campuses, proved to be vital to the project.⁴⁷ 

Reminders to complete the survey were sent 1, 2, and up to 3 weeks after the initial invitation was sent. Surveys were considered usable and were included in the study if respondents provided year of medical school education and responded to at least 16 items of the JSE and all items of the Infrequency scale. 

We used the t test for comparing mean differences in age between each sample and its respective population. We also used the z test for proportions to compare gender composition of samples and their respective population. Additionally, the χ² test was used to determine goodness of fit on race/ethnicity of samples and populations. Statistical significance was set at P<.05. For determining the practical (or clinical) significance of the findings, we considered any effect size estimate of 0.20 or less as practically unimportant, regardless of statistical significance.^48,49 

We received 16,760 usable surveys. The usable sample of first-year matriculants represented 83% (n=6009) of its respective population. Usable samples for the first-, second-, third-, and fourth-year students represented 50% (n=3616), 41% (n=2764), 36% (n=2413), and 40% (n=1958) of their respective populations. 

The mean age was 24.8 years for first-year matriculants, and the mean ages increased progressively by about 1 year for each class in medical school (25.5, 26.3, 27.4, and 28.6 in years 1-4, respectively). About half of the students in each group were men (53%, 51%, 49%, 50%, and 49% in first-year matriculants and in years 1-4, respectively). Only 1% or less of the students in each group self-identified as transgender, gender variant/nonconforming, or declined to answer. 

The majority of respondents were white/Caucasian (60%-68%), followed by Asian (19%-27%), Hispanic/Latino/Spanish (3%-6%), and black/African American (3%-4%) (Table 1). 

We compared age, gender, and race/ethnicity of the usable samples with their respective populations using AACOM data to examine representativeness of the study samples. The study samples closely resembled their respective populations with regard to age, gender, and race/ethnicity. The means of students’ ages in each of the 5 samples and their respective populations were very similar. However, results of t tests comparing sample and population mean ages showed statistically significant differences (P<.01) that can be attributed to the large size of the samples. However, the magnitudes of the effect size estimates of the mean differences were lower than 0.20, indicating a lack of clinically important differences.^48,49 

Gender proportions were also similar between samples and their respective populations. Although the results of the z tests for proportions were statistically significant (P<.01), the magnitude of none of the effect size estimates for proportions (by using the arcsine transformation) reached the 0.20 mark, indicating that the observed differences in gender proportions between samples and their respective populations were clinically unimportant. 

Similarly, composition of race/ethnicity was similar between the study samples and their respective populations, with one exception. There was a substantial 12% difference in the Asian ethnic group in the fourth-year students (20% in the sample and 8% in the population). However, a reverse change was also observed for fourth-year students in the “other/no response” category (6% of the sample and 23% of the population). The effect size estimates of the goodness of fit tests showed values less than 0.20 for all of the study samples, with the exception of the fourth-year students, for whom the effect size was 0.67, most likely due to a discrepancy between proportions of self-reported Asian students in the sample and its respective population. 

To minimize confounding factors in creating the national norm tables, we excluded students (2%-3%) who scored higher than 3 on the Infrequency Scale of the ZKPQ. Also, because of notable gender differences in empathy reported in many studies with medical students,^{7(pp169-187),8,10,13,50} including POMEE, we excluded those who did not specify their gender as male or female (<1%) from the data used in preparing each of the norm tables and those who scored higher than 3 on the Infrequency Scale of the ZKPQ. 

We combined first- and second-year students in a single corresponding national norm table for preclinical students because JSE means (SDs) and score distributions were similar for students in these 2 years (115.0 [12.5] in the first year and 114.7 [12.1] in the second year). Similarly, a single corresponding national norm table for students in the clinical phase of medical school was created because JSE means (SDs) and score distributions were similar for third- and fourth-year students (113.4 [13.0] in the third year and 113.0 [13.2] in the fourth year). Final samples used in the national norm tables included 5818 first-year matriculants (3071 men and 2747 women) (Table 2); 6115 students (3042 men and 3073 women) in the preclinical phase (Table 3); and 4216 students (2099 men and 2117 women) in the clinical phase of medical school (Table 4). 

Because of the notable gender difference on the JSE, we took students’ gender into consideration in the development of norm tables by preparing percentile ranks for men and women separately. If a student's gender is known, we strongly recommend using the corresponding gender-related percentile ranks from the norm tables. If a student's gender is unknown, then the percentile rank on the norm tables for men and women combined can be used as an estimate. 

The following guidelines can be used to determine the percentile rank of an individual's JSE score. First, locate the student's raw empathy score interval in the first column of the appropriate national norm table. Then, move across the row to locate the score's corresponding percentile rank in the column that corresponds to the student's gender. For example, if a score of 125 is obtained by a third-year male student, first find the raw score interval (125-126) by looking in the first column of Table 4, then find the corresponding percentile rank for that score interval by looking across the 125-126 row until reaching the column showing the percentile rank for men, which reads 88%. This percentile rank means that a score of 125 on the JSE for a third-year male student places him in the 88th percentile of all male students in that population. In other words, 88% of the male students in this norm group scored below 125. Using the same guidelines, a third-year female student with a similar score will be placed in the 79th percentile. 

This study, as part of the larger nationwide POMEE with a broader scope (Hojat et al, unpublished data, 2019) is the first of its kind, to our knowledge, to be undertaken in medical education research. It is unique in its scope because it includes a large national sample of medical students at all levels of medical school education, from first-year matriculation before exposure to formal medical education to completion of medical school. We believe the rich nationwide data in the POMEE database will serve to address a number of issues of interest to medical education researchers, beginning with the national norms for osteopathic medical students at all levels of medical school. Future studies are underway to improve our understanding of relationships between empathy scores and demographic variables (age, gender, race/ethnicity), undergraduate major, career interest, and employment experience before medical school. Patterns of changes in empathy scores in different years of osteopathic medical school compared with findings in allopathic medical students will also be explored. 

Among the strengths of this ground-breaking study are that it uses large and representative samples at all levels of medical education; it uses a well-established empathy-measuring instrument specifically developed for administration to medical students with face and content validities and strong psychometric support in allopathic and osteopathic medical students; and it controls for good impression response bias. One limitation of this study is relatively low response rates (<50%) in samples of second-, third- and fourth-year students. However, this limitation is mitigated by support for the representativeness of the samples based on gender, age, and race/ethnicity. 

In a cross-sectional study of a single institution, Rizkalla and Henderson⁵¹ found that osteopathic medical students in preclinical years of training (years 1 and 2) had similar empathy scores and that students in clinical years of training (years 3 and 4) also had similar empathy scores. Similar findings in POMEE led us to group first- and second-year osteopathic medical students and third- and fourth-year osteopathic medical students into preclinical and clinical norm tables, respectively. However, we must emphasize that the national norm tables developed in this study can only be used for osteopathic medical students. Different sets of norms should be created for allopathic and other health professions students. 

The national norms developed in this study have important implications for assessing osteopathic medical students’ individual and group scores against national norms. Percentile ranks in norm tables corresponding to students’ raw scores can provide supplementary measures for admissions to medical school or to postgraduate medical education programs. Medical schools and postgraduate medical education programs are accountable to select qualified applicants with the best potential to become caring physicians.⁶ Traditional admission measures (eg, Medical College Admission Test, undergraduate grade point averages) may help in selecting those who can successfully pass examinations of recalling factual knowledge in the early years of medical school, but are not predictive of caring for the patient. The contention that admission interviews, letters of recommendation, essays, and personal statements can reliably assess personal qualities such as caring attitudes and empathetic orientation awaits empirical verification.⁵ We believe that the assessment of empathy from using the norm tables reported in the present study can be helpful to fill this gap. Concern about the possibility of a respondent's attempt to fake responses in self-reported tests such as the JSE can be addressed by using approaches to detect the social desirability response set, or the good impression response bias measured by using the Infrequency Scale of the ZKPQ in the present study. 

Looking forward, we believe national norms could serve as a tie-breaker for applicants with similar academic qualifications. Moreover, the national norm tables could help to identify the degree of students’ need for remedial education to improve their empathy. In addition, national norms tables could potentially be used to assess the outcomes of educational programs, interventions, or remedies implemented in different years of osteopathic medical school training. Thus, not only can these norm tables be used to assess individuals’ scores against national data, but they can also help to evaluate group changes in percentile ranks over time to assess outcomes of targeted educational programs. Further research is needed to investigate these and other uses of the national norm tables for osteopathic medical students. 

Interest in recognizing, assessing, and enhancing empathy among health care professionals is increasing, particularly as studies uncover improved educational and patient outcomes associated with increased empathy in health professions students and practitioners. The national norm tables for osteopathic medical students reported in this nationwide study provide a compass for colleges of osteopathic medicine as they review admissions criteria and student performance. Further research in the POMEE's ongoing national longitudinal study will evaluate changes in empathy scores over time and will elucidate specific educational and institutional factors that are associated with positive and negative changes as students progress through medical school. 

We thank the many people who contributed to this project, especially those who made a significant contribution to this project’s success: members of the AACOM team, especially Thomas Levitan, MEd, who was instrumental at the onset of this project in developing the survey instrument; Jefferson research team colleagues Shira Carroll, BS; Daniel Z. Louis, MS; J. Jon Veloski, MS; and Vittorio Maio, PharmD, MS, MSPH; Tracy Treadway; and Luke Kedziora, BS, at the Jefferson Department of Information Services and Technology, who helped us in preparing the web-based research survey for online administration and correcting computer glitches during and after pilot testing; deans of participating Colleges of Osteopathic Medicine and each of the research coordinators from participating college campuses (listed here in alphabetic order): Michael Becker, DO; Joe Bianco, PhD; Linda Boyd, DO; Joseph Brewer, PhD; Lisa Carroll, MD; Mark Clark, PhD; Karen Clayton, PhD; Glenn E. Davis, PhD; Robyn Dreibelbis, DO; Kyle Henderson, PhD; Ana Maria Homs, PsyD; Justina Hyfantis, PhD; Britt Johnson, PhD; Gretchen Lovett, PhD; Susan Mackintosh, DO; Elizabeth McClain, PhD; Edward Magalhaes, PhD; Patience Mason, MEd; Terrence Miller, PhD; Malcolm Modrzakowski, PhD; David Mokler, PhD; Bruce Newton, PhD; Lorree Ratto, PhD; Sean Reeder, DO; Mireille Rizkalla, PhD; Raquel Malina Romanick, JD; Miko Rose, DO; Katherine Ruger, EdD; Amy Schlueter, MPA; Trish Sexton, DHEd, MS; Dana Shaffer, DO; Robert Sorrells, PhD; Vivian Stevens, PhD; Mary Ann Taylor, PhD; Clinton Whitson, MS; and Rynn Ziller, EdD. We are also thankful to Dorissa Bolinski, BA; and Pamela Walter, MFA, for editorial polishing of this article. Last, but not least, we give special thanks to all of the thousands of osteopathic medical students in the academic year of 2017-2018 who made history in medical education research by voluntarily and willingly completing and submitting the online survey of this project.