Personality assessments and outcomes in medical education and the practice of medicine: AMEE Guide No. 79

Abstract

In a paradigm of physician performance we propose that both “cognitive” and “noncognitive” components contribute to the performance of physicians-in-training and in-practice. Our review of the relevant literature indicates that personality, as an important factor of the “noncognitive” component, plays a significant role in academic and professional performances. We describe findings on 14 selected personality instruments in predicting academic and professional performances. We question the contention that personality can be validly and reliably assessed from admission interviews, letters of recommendation, essays, and personal statements. Based on conceptual relevance and currently available empirical evidence, we propose that personality attributes such as conscientiousness and empathy should be considered among the measures of choice for the assessment of pertinent aspects of personality in academic and professional performance. Further exploration is needed to search for additional personality attributes pertinent to medical education and patient care. Implications for career counseling, assessments of professional development and medical education outcomes, and potential use as supplementary information for admission decisions are discussed.

“In the physician or surgeon no quality takes rank with imperturbability [which] means coolness and presence of mind under all circumstances and the physician who has the misfortune to be without it loses rapidly the confidence of his patient.”

Sir William Osler, 1922, pp. 3–4

Introduction

Considering the notion of professionalism in medicine, it is increasingly acknowledged that at least two major complementary components contribute to the performances of physicians-in-training and in-practice. One component includes a set of “cognitive” abilities, which are often reflected in intellectual capabilities, performances on examinations of recalling factual information and tests of declarative knowledge. The other component, often described under the rubric of “noncognitive” or personal qualities, includes features such as personality attributes, attitudes, interests, values and other personal characteristics (Gonnella et al. 1993, 1998). The “imperturbability” attribute, described in the opening epigraph, reflects a noncognitive personal quality of a physician that, according to Sir William Osler, is associated with patient outcomes.

Despite the acknowledgement of the contribution of “noncognitive” qualities to predicting performance, career interest, and clinical outcomes (Gonnella et al. 1993, 1998), the assessment and development of personal qualities in general, and personality attributes in particular, have not received appropriate attention in medical education research. A recent interest in the development and assessment of “professionalism” in medicine (Stern 2006) has, to a certain extent, shifted the attention of medical education leaders, faculty, and researchers toward assessing and cultivating personality attributes relevant to the notion of professionalism in medicine (Veloski & Hojat 2006).

The “noncognitive” dimension of performance includes vast and diverse arrays of personality, social and cultural variables. A comprehensive treatment of all these “noncognitive” variables is beyond the intended scope of this Guide. Thus, for a more manageable and parsimonious presentation, in this Guide we will attempt to focus on those selected personality attributes that have been empirically studied in medical education research.

Personality in the context of medical education and patient care

In the context of medical education and patient care, we define personality as a configuration of characteristics and behavioral tendencies that comprise an individual's unique features, developed based on a combination of several interacting elements such as constitutional predisposition, rearing environment, quality of early attachment relationships, interpersonal and critical life event experiences, social and cultural environment as well as formal and informal education. The aforementioned conceptualization is consistent with the definition of personality proposed by the American Psychological Association (APA 2007, p. 689).

Table

Practice points

Cognitive (e.g. intellectual abilities) as well as noncognitive (e.g. personality) factors contribute to academic and professional performances. The contribution of personality attributes to performance assessments is greater in the clinical than preclinical phase of medical education and can be explained by the trait-activation theory. There is inadequate evidence supporting the validity and reliability of approaches such as admission interviews, letters of recommendation, essays, and personal statements to assess the personality of applicants to medical school or residency programs. There are plausible explanations (conceptual and methodological) for the modest correlations observed between measures of personality attributes on the one hand and educational and clinical outcomes on the other hand that provide grounds for optimism concerning the use of personality assessments in medical education. Based on the conceptual relevance and currently available empirical evidence, we propose that measures of two personality attributes, “conscientiousness” and “empathy,” be considered in predicting educational and clinical outcomes, in career counseling, in assessments of educational outcomes, as well as potential “tie-breakers” for admission decisions among candidates with similar academic qualifications. The search for additional personality attributes pertinent to medical education and the practice of medicine should be a priority in the research agenda for medical education. Lingering doubts about the role of personality in the performance of physicians-in-training and in-practice results in a futile and never-ending search for additional evidence, which would be counterproductive, because waiting for certainty is waiting for eternity.

A number of researchers in medical education have confirmed that personality measures can contribute valuable information to the selection of medical students (Walton 1987; Helle et al. 2010). It is suggested that the assessment of personality must be given a major role in physician training (Walton 1987). It is reported that intellectual abilities account for about 35% of the variance in performance, but inclusion of personality information increased the common variance to 75% (Walton 1987). In a longitudinal study of internal medicine residents, Girard and Hickman (1991) found that 48% of the variation in clinical ranks and 38% of the variation on American Board of Internal Medicine (ABIM) examinations could be explained by psychological and personality variables.

In our own study with medical students, we noticed that a set of personality measurers (e.g. appraisal of stressful life events, general anxiety and test anxiety, external locus of control, intensity and chronicity of loneliness experiences, extraversion, self-esteem, perceptions of early relationships with parents and peers and measures of over- or under-confidence) could significantly predict performance on medical licensing examinations (Hojat et al. 1988). Also, we found that higher scores on measures of self-esteem and extraversion, lower scores on loneliness, and perceptions of satisfactory relationship with parents in childhood (Hojat et al. 2004a) could predict global ratings of clinical competence in core clerkships in medical school. Furthermore, inclusion of a set of the aforementioned personality measures to the prediction model could substantially increase (from 0.32 to 0.56) the magnitude of correlations between academic attainment predictors already in the statistical model (previous academic grades and scores of the Medical College Admission Test, the MCAT) and the criterion measure (scores on Part 1 of the examinations of the National Board of Medical Examiners) (Hojat et al. 1988).

In another study, we noticed that ratings of interpersonal skill, assigned by residency program directors, were significantly and positively correlated with reports of satisfactory early relationships with mothers and peer prior to medical school, but negatively associated with scores on measures of anxiety, neuroticism, and loneliness (Hojat et al. 1996a). However, research findings, using a variety of personality measures to improve the predictive validity of academic performance have not been consistent (Pollock et al. 1982; Aldrich 1987; Weiss et al. 1988), probably due to conceptual and methodological issues, like variability in the definition and understanding of the constructs of both predictors and criterion measures and other factors as we will describe later.

A number of personality instruments have been used in medical education research such as the NEO Personality Inventory (e.g. Lievens et al. 2002; Ferguson et al. 2003) for measuring the big five factors of personality; the California Psychological Inventory (e.g. Hobfoll et al. 1982; Tutton 1993, 1996); the Eysenck Personality Inventory/Questionnaire (EPI/EPQ; e.g. Roessler et al. 1978; Lipton et al. 1984; Westin et al. 1986); the 16 Personality Factor (16PF) Questionnaire (e.g. Lipton et al. 1984; Green et al. 1993; Peng et al. 1995); the Myers-Briggs Type Indicator (MBTI; e.g. Turner et al. 1974; Tharp 1992); and the Jefferson Scale of Empathy (JSE; e.g. Hojat et al. 2002a, 2002b, 2002c; Hojat 2007), among others. Summarizing the findings on all or even most of the personality instruments that have been used in medical education research is beyond the scope of this Guide. Herein, we will present highlights of findings on selected personality measures that have been more frequently used in medical education research.

Medical education researchers have utilized various personality measures in an attempt to demonstrate that it is not only possible to predict academic performance in medical school (Doherty & Nugent 2011), but also to predict the career interest of medical students (Otis & Weiss 1973; O’Donnell 1982; Walton 1987; Hojat et al. 1998, Hojat & Zuckerman 2008), as well as patient outcomes (Hojat et al. 2011a; Del Canale et al. 2012). The idea that certain personality attributes can predict student performance and career choices is appealing, considering the concern about selecting and training “appropriate” practitioners in different specialties who possess the “right kind” of personality for their specialty of interest (Cob et al. 1977; Di Renzo 1986). Perhaps of equal merit is the prospect of tracing the development or change in certain personality attributes that are important in the care of patients.

A paradigm of physician performance

The importance of the contribution of personality to different aspects of physician behavior can be illustrated in a model of physician performance. We have previously described a paradigm of physician performance (Gonnella et al. 1993, 1998) in which we proposed that a combination of three basic elements of medical knowledge, procedural/technical skills, and personality contributes to the performance of physicians-in-training, and physicians-in-practice. In addition, we suggested that for a more optimal outcome, a physician must be equipped with these three elements to perform simultaneously three essential roles as a clinician, an educator, and a resource manager (Gonnella et al. 1993, 1998).

This multidimensional conceptualization of physician performance (depicted in Figure 1) has important implications not only for the assessment of performance of medical students and physicians, but also for guiding educational programs aimed at professional development for undergraduate, graduate, and continuing medical education.

Figure 1. Taxonomy of physician performance*. *Adapted from Gonnella et al. (1993). Three examples are presented in each cell.

Conventional approaches to obtain personality information in medical education

Some researchers, administrators and faculty in both undergraduate and graduate medical education believe that personality testing is not necessary or appropriate. They hold this opinion because they believe that personality can be assessed from admission interviews, letters of recommendations, personal statements and essays written by the applicants. These approaches, however, have shortcomings when used as a source of personality assessment. There are other researchers who are of the opinion that no personality instruments can validly and reliably predict performances of physicians-in-training and in-practice. In this Guide, we will shed some light on this issue.

Admissions interview

Face-to-face interviews are required as part of the admission process in almost all medical schools and residency programs in the United States and Canada. A great majority of these interviews are unstructured with no uniform questions and no standard assessment procedures. It is believed that interviews provide an opportunity to include the human touch in decision-making and that they help in assessing personal qualities (Albanese et al. 2003). It is claimed that the admissions interviews provide important information in selecting potential students (Puryear & Lewis 1981; Eddins-Folensbee et al. 2012). However, convincing empirical evidence is not yet available to confirm the validity and reliability of admission interviews (Ferguson et al. 2002; Kanter 2012). Compounding this issue is the fact that medical students themselves, without any training, sometimes perform interviews with new applicants in order to supplement the staff and faculty resources needed for interviewing a large number of applicants. Interestingly, no significant difference has been observed between faculty and students interview ratings (Gelmann & Stewart 1975; Elam & Johnson 1992; Eddins-Folensbee et al. 2012).

Reliance on interviews conducted by untrained staff or students can jeopardize the validity of the selection process by giving advantage to those applicants who present themselves well in interview settings. The unstructured interviews by untrained interviewers with no standard scoring guidelines may predict nothing other than an applicant's skills in role playing (Musson 2009). No wonder that the predictive validity of admission interviews has been reported to be disappointingly low (Walton 1987). It is interesting to note that despite all of the aforementioned limitations, in a national survey with residency program directors in the United States, an applicant's interview was considered as the most important selection tool (Wagoner et al. 1986). The use of the interviews in the undergraduate and graduate selection processes provides a unique opportunity to talk with applicants and may be helpful in observing a candidate's reaction to questions, but uncertainties remain open regarding the validity and practical outcomes of admission interviews (Antonovsky et al. 1979; Hobfoll & Benor 1981; Green et al. 1991).

Letters of recommendation

Most medical schools in North America require letters of recommendation to be submitted by those who are fairly familiar with the academic performance and personal qualities of the applicants. Letters of evaluation written by medical school deans also play a great role in the selection of candidates for residency. There is no convincing empirical evidence in support of the predictive validity of letters of recommendation in medical schools. In our own empirical study using a multivariate statistical model, we found that the level of recommendation contained in the letters written by undergraduate pre-medical education advisors did not contribute significantly to the prediction of academic performance in medical school beyond the grades obtained prior to medical school (Zeleznik et al. 1983). Furthermore, letters of recommendation may be biased and flattering with no substantial empirical link to later performance (Walton 1987).

Personal statements, letters of intent and essay

Some medical schools require applicants to write an essay, letter of intent or some personal statements about, for example their interest in medicine, career goals and future plans. There are very few studies on the predictive value of essays or personal statements. In one study, the content of candidates' personal statements was analyzed, and no evidence was found to support its predictive validity (cited in Ferguson et al. 2002). Typically, letters of intent or essays submitted by applicants are evaluated by untrained readers and are assessed on informal criteria (Musson 2009). Even more questionable is whether candidates themselves, without any help, write the statements, essays and letters of intent (Musson 2009).

Because of the aforementioned shortcomings, Haque and Waytz (2012) suggest that one appropriate approach for the assessment of personality of physicians-in-training is to administer psychometrically sound personality instruments.

A benign neglect

It is interesting to note that despite the recent emphasis placed on personal qualities relevant to professionalism in medicine (Stern 2006; Veloski & Hojat 2006), and in spite of the accumulating volume of research by psychologists on the importance of personality in professional development and personal, social and professional behaviors, there seems to be a lack of enthusiasm among medical education leaders, faculty and researchers to take a fresh and serious look at the assessment and cultivation of personal qualities in medical education and in patient outcomes. We speculate that the following factors may contribute to this benign neglect.

First, some have lingering doubts about the role of personality in the performance of medical students and physicians. Proof is needed for supporting the link between personality, academic performance, clinical competence and the quality of patient care. In this Guide, we will attempt to provide some empirical evidence in support of the proposition that the role of personality is important in the professional development of physicians-in-training and in-practice.

Second, a variety of personality instruments have been used over the years in medical education research. Some may argue that from the vast array of personality measures, no specific personality attributes have been identified as the most favorable for predicting performance in medical school and in the practice of medicine, nor in patient outcomes. This ambiguity leads to confusion about choosing specific personality measures with strong associations with medical education and patient outcomes. In this Guide, we will attempt to provide some theoretical and empirical evidence for specific personality measures with the most convincing and consistent support for their validity, reliability and utility in medical education research.

Third, some skeptics may believe that there is no need for independent assessments of personality attributes, because, they maintain that, indicators of intellectual capability, such as academic attainment and professional achievements require specific personal qualities such as achievement motivation, interest, and self-esteem, which are inseparable factors in academic success. According to this view, personality factors are assumed to be embedded in any assessment of academic attainment; therefore, including personality measures in the assessment of performance would be redundant and unnecessary. In this Guide, we will provide support for the notion that personality measures can uniquely and independently contribute to incremental validity in the prediction of educational and clinical outcomes.

Fourth, there are those who believe that personality attributes are not amenable to change. Therefore, there is no point to assess those personal traits or implement programs to enhance those personality features that have already been formed based on genetic predisposition and early life experiences. In this Guide, we will attempt to provide evidence that some personality attributes are amenable to positive and negative changes during the course of medical education, and that targeted educational programs can enhance those personality attributes.

Fifth, some believe that items of personality tests are often transparent, and can thus be “faked” or answered in a way that is recognized as socially desirable. According to this belief, results of self-reported personality tests are not valid because respondents can manipulate their answers to intentionally produce a socially desirable image of themselves. In this Guide we will provide evidence that with proper instructions and control, social desirability response bias cannot substantially distort the findings.

We will attempt to refute the aforementioned speculations and provide empirical evidence in support of the notion that personality is important to be taken into consideration in the admission process, and in the professional development of physicians. In our exploration, we will rely on relevant empirical evidence reported in medical education literature and in our own research findings from the Jefferson Longitudinal Study of Medical Education (Hojat et al. 1996b; Gonnella et al. 2011).

Information about the Jefferson Longitudinal Study and its publications is posted at: http://jdc.jefferson.edu/jlsme.

Purpose

The main purpose of this Guide is to present an overview of research findings on personality assessments and outcomes in medical education research. We will embark on a journey to explore the terrain of personality in medical education research, and to search for specific personality attributes that are associated with educational and clinical outcomes.

The intention is not to undertake a comprehensive review of the literature, nor do we aim to conduct a meta-analytic study. Rather, we limit our exploration by visiting only some selected landmarks that we consider important and typical of the landscape in our journey.

Selected personality instruments frequently used in medical education

Based on our reading of the relevant literature, we have selected 14 instruments used in medical Education research. Five of these instruments have been used more frequently than others in medical education research: The NEO Personality Inventory (NEO PI-R), the 16PF Questionnaire, the California Psychological Inventory (CPI), the MBTI and the JSE.

Selected findings from each of these instruments will be described in more detail than the other nine selected instruments (Table 1). Although our selected instruments are not, by any means, an exhaustive list of all instruments used in medical education, the five aforementioned instruments, based on our literature search, seem to be the most frequently used personality instruments in medical education research.

Table 1.  Selected personality instruments used in medical education research

Most frequently used

1. Measurement of the Five Factors of Personality (NEO PI-R)

2. The 16 Personality Factor Questionnaire (16PF)

3. The California Psychological Inventory (CPI)

4. The Myers Briggs Type Indicator (MBTI)

5. The Jefferson Scale of Empathy (JSE)

Other selected instruments

6. The Eysenck Personality Inventory (EPI)

7. The Minnesota Multiphasic Personality Inventory (MMPI)

8. The Profile of Moods State (POMS)

9. The Temperament and Character Inventory (TCI)

10. The Personal Qualities Assessment (PQA)

11. The Maslach Burnout Inventory (MBI)

12. The Medical Specialty Preference Inventory (MSPI)

13. The Jefferson Scale of Attitudes toward Physician-Nurse Collaboration

14. The Jefferson Scale of Physician Lifelong Learning

In a great majority of our reviewed studies, the associations between measures of personality and indicators of academic performance, ratings of clinical competence and career interests have been examined. However, there is a dearth of empirical research on the link between measures of personality and quantifiable indicators of patient outcomes. Despite this limitation, in this Guide, we will present some of the available empirical data in support of a link between personality attributes and patient outcomes.

1. Measurement of the five factors of personality

A review of the literature on personality and its measurement indicates that a great volume of published research in recent years examined specific personality attributes under a rubric of the five-factor model (FFM) of personality. These factors are often referred to as Openness, Conscientiousness, Extraversion, Agreeableness and Neuroticism (or emotional stability which is the opposite of neuroticism). The acronym OCEAN was used by Hoffman and colleagues (2010) to represent these big-five factors of personality, respectively. The FFM of personality provides a comprehensive framework in describing major personality attributes to determine behavioral, social, professional, and academic differences (Goldberg 1990, 1992, 1993; Deniston & Ramanaiah 1993).

The five factors were originally extracted based on an extensive psycho-lexical analysis of thousands of English words describing personality, supported subsequently by empirical findings resulting from factor analytic research (Goldberg 1990, 1992; Costa & McCrae 1992). The FFM is based not only on theories of personality but is also grounded on a variety of biological, psychological and social perspectives, and an integration of both nature and nurture underpinnings of personality development (McCrae & Costa 1989, 1997; Goldberg 1993; De Raad & Perugini 2002). Evidence suggests that at least some components of the five factors are inherited (e.g. excitability component of the Neuroticism factor) which supports the view on biological roots of some personality attributes (Jang et al. 1996).

The FFM, or some variant of it, currently a popular model of personality among psychologists, has been studied extensively and used by many personality researchers (Musson 2009). Each of the five factors includes a number of facets or components. For example, the Openness factor includes facets such as fantasy, aesthetics, feelings, ideas, actions, imagination, preference for variety, curiosity and intellectual qualities (Costa & McCrae 1992). The Conscientiousness (C) factor includes components such as competence, dutifulness, achievement striving, self-disciplined, deliberation and order. The Extraversion (E) factor includes facets such as sociability, warmth, activity, positive emotions, assertiveness, gregariousness and excitement- seeking. The Agreeableness (A) factor encompasses facets such as trust, compliance, straightforwardness, altruism, tender-mindedness and modesty; and the Neuroticism (N) factor includes components such as anxiety, anger, depression, hostility, self-consciousness, impulsiveness and vulnerability (Costa & McCrae 1992).

The NEO-PI-R, which has been widely used in personality studies and in medical education research, is one of the instruments available for the assessment of the big five factors. This instrument, developed by Costa and McCrae (1992) is the first published instrument designed specifically to measure the big five factors of personality (De Raad & Perugini 2002). The original inventory was developed to measure the three factors of Neuroticism, Extraversion and Openness, hence named NEO Personality Inventory (PI) which was revised (NEO-PI-R) to include two additional factors of the FFM (Agreeableness and Conscientiousness).

The revised self-report form of this instrument consists of 240 items answered on a five-point scale, measuring not only the big five personality factors (48 items per factor) but also six personality facets within each factor (eight items per facet). A shorter version of this instrument (NEO-FFI, 60-item) is also available for measuring the big five factors without detailed measurement of the facets within each factor (Costa & McCrae 1992). Other personality instruments, such as the Zuckerman-Kuhlman Personality Questionnaire (ZKPQ, Zuckerman 2002), were also developed to measure the big five factors or some variant of the FFM.

Performance: The associations between the five personality factors and academic performance have been addressed in a number of studies. For example, in their cross-sectional and longitudinal studies, Lievens and colleagues (2002) administered the Flemish translation of the NEO-PI-R to 785 students in five Flemish universities and found that the chance of success in the pre-clinical years of medical school was better for students who scored high on the Conscientiousness factor. High scores on the Openness factor significantly predicted the final scores in the third year of medical school (Lievens et al. 2002). It was also found that more proactive facets of the Conscientiousness factor such as “self-discipline” and “achievement striving” could predict medical students’ academic achievement better than more regulatory facets of the Conscientiousness factor such as “order,” “deliberation” and “dutifulness.” (Lievens et al. 2002).

The Extraversion factor was the only factor that negatively correlated with examination results in the first year of medical school, suggesting a restricting effect of this factor at the beginning of the academic career (Lievens et al. 2002). These investigators concluded that significant variation exists among medical students in terms of personality, reflected in the FFM, which is linked to academic success. In particular, they placed an emphasis on the findings that scores on the Conscientiousness factor could strongly predict students’ success in preclinical years of medical school (Lievens et al. 2002).

In a study by Helle and colleagues (2010), the five factors of personality inventory and a test of visual perceptual skills, designed to assess an individual's visual perception (Martin 2006) was administered to 150 second-year medical students at the University of Turku in Finland. Results showed that the Conscientiousness factor and one element of visual perceptual ability (spatial relationship awareness) predicted performance on the diagnostic classification in microscopic observation in an undergraduate course in pathology at the beginning of the course. In a study of 176 students attending the Nottingham Medical School in the UK, Ferguson and colleagues (2003) found that the Conscientiousness factor was the best predictor of academic performance in the pre-clinical phase of medical education. In a meta-analytic review, the Conscientiousness factor was found to be a significant predictor of job performance in other occupations as well (Tett et al. 1991).

In a longitudinal study, Lievens and colleagues (2009) followed up on students who participated in their original study (Lievens et al. 2002). It was found that grade point averages in the first year rather than personality factors were the most important predictors of attrition in preclinical years. However, as the students progressed through medical school, the Openness, Conscientiousness and Extraversion factors became increasingly important predictors of academic success in the clinical phase of medical education. Consistent with these findings, McManus and his colleagues (2004) in a 12-year longitudinal study of medical students who attended five medical schools in the UK reported that perception of stress and burnout were predicted by scoring high on the Neuroticism, low on the Extraversion factor, and low on the Conscientiousness factor.

The Conscientiousness factor has long been recognized as a crucial predictor of job performance in medicine as well as in other professions (Barrick & Mount 1991; Behling 1998; Hurtz & Donovan 2000). Also, research findings suggest that sociability, a prominent feature of the Extraversion factor, is an important mediating variable in the clinical performance of medical students (Ferguson et al. 2003; McManus et al. 2004; Hojat et al. 2004a; Knights & Kennedy 2007; Tyssen et al. 2007; Lievens et al. 2009). However, their results on the Extraversion factor are less consistent in the preclinical than clinical phases of medical education (Piedmont et al. 1991; Lievens et al. 2002).

Lievens and colleagues (2009) used the expressions of “getting along” as a reflection of the Extraversion and Openness factors, and “getting ahead” as a reflection of the Conscientiousness factor. Extraversion and Openness to experiences are two personality attributes that facilitate physician-patient interpersonal relationships, and thus can contribute to optimal clinical outcomes. The importance of the Extraversion and Openness factors in clinical performance has also been confirmed in a study by Piedmont and colleagues (1991).

In a study by Haight and colleagues (2012), the relationships between personality measures and medical student preclinical and clinical performances in 175 students at the Saint Louis University School of Medicine were examined. It was found that scores of the MCAT correlated with academic examinations, whereas scores on the Conscientiousness and Extraversion factors correlated with indicators of clinical performance and humanism nominations. More specifically, the Conscientiousness factor could predict clinical skills, but the Extraversion factor was a significant predictor of indicators of clinical skills that relied heavily on interpersonal interactions. In a review article, Doherty and Nugent (2011) examined the relationships between personality and academic performance in medical school. They concluded that the Conscientiousness factor can predict long-term success in medical education, and the Extraversion factor is an important mediating factor in clinical performance.

In a longitudinal study to examine changes in the validity coefficients of personality in predicting academic performance of an entire 1997 cohort of medical students in six Flemish universities in Belgium, Lievens and colleagues (2009) reported that the Openness, Conscientiousness and Extraversion factors (and most of their facets) showed an increase in the magnitude of the their validity coefficients as students progressed through medical school. For example, the validity coefficient for the Extraversion factors shifted from a negative correlation of −0.11 in the first year of medical school to a positive correlation of 0.31 in the last year of medical school. The changes in the validity coefficients from the first to the last year of medical school were from 0.18 to 0.45 for the Conscientiousness, 0.02 to 0.30 for the Openness, −0.01 to 0.17 for the Agreeableness and 0.03 to −0.07 for the Neuroticism factors (Lievens et al. 2009).

The significant change in the validity coefficient of the Openness factor from the preclinical to the clinical years of medical school deserves some explanations. Openness has been linked to academic ability and divergent thinking (Goff & Ackerman 1992; McCrae 1996). However, the magnitude of its validity coefficients in predicting academic achievement has not been impressive (Hough 1992; Barrick et al. 2001). Openness is a personality factor that facilitates acceptance and adequate adjustment to the wide variation of changes (LePine et al. 2000) that is encountered during the clinical phase of medical education. Lievens and colleagues (2009) suggested that although there may be no advantages to being open to new experiences in the preclinical years of medical school, this personality attribute increasingly becomes relevant in clinical education and in applied settings. Therefore, openness to experiences seems to be more beneficial in the clinical phase of medical education, which requires interpersonal interaction with patients.

With regard to the findings on variation in the magnitude and direction of validity coefficient for the Extraversion factor, Lievens and colleagues (2009) speculated that while it might not be beneficial for medical students to be extraverted during the preclinical years, this quality becomes important later in the clinical years when human interaction is required for achieving optimal outcomes. Consistent with this notion, Rolfhus and Ackerman (1999) found that extraverts, compared to introverts, obtained lower scores on knowledge tests. Thus, the negative correlation between extraversion and performance in the preclinical phase of medical education could be due to the fact that extraverted students are likely to spend more time socializing (Chamorro-Premuzic & Furnham 2003) and be involved in nonacademic activities such as sports and social events. Instead, introverted students may spend more time reading the high volume of course materials in the preclinical years. In addition, it has been reported that extraverts are more likely than introverts to be easily distracted, while introverts are more likely to focus on cognitively demanding tasks (Entwistle & Entwistle 1970), which help them to obtain better grades on tests of acquisition of factual information; a key feature of examinations in the preclinical years (Sanchez et al. 2001).

Conversely, it has been reported that extraverts are more likely to obtain better assessment marks in the activities performed in group settings, clerkships, practicums or seminar classes (Furnham & Medhurst 1995). Thus, extraverts seem to have the personality attributes needed to perform better in educational environments that require interpersonal interactions (Ferguson et al. 2000). Therefore, one can expect that those who score high on the Extraversion factor would not perform as well as their introverted counterparts in the preclinical phase of medical education. This can explain the change of the validity coefficient of the Extraversion factor from negative in preclinical to positive in the clinical years of medical school.

The lack of predictive validity for the Neuroticism factor in medical school found in the Lievens and colleagues study (2009) is not surprising. Those high in Neuroticism are prone to anxiety; thus, less likely to perform well on academic tests that require concentration and recall of factual information. In addition, high scorers on Neuroticism are more vulnerable to test-taking anxiety and evaluation apprehension during examinations of recalling factual information early in medical school. Some studies on emotional stability, the opposite of neuroticism, have shown positive relations between emotional stability and performance in college students (e.g. Cattell & Kline 1977; Lathey 1991; Sanchez et al. 2001) and in medical students (Barratt & White 1969).

The facets of the Agreeableness factor such as trust, altruism, modesty and tender-mindedness (Costa & McCrae 1992) can facilitate physician-patient relationships, thus it is expected that scores on the Agreeableness factor predict the clinical competence of medical students. The facets of the Agreeableness factor are positively associated with clinical performance in medical students (Gough et al. 1991; Shen & Comrey 1997). However, findings on relationships between this factor and performance in medical school are not consistent.

Among the big five factors, the Conscientiousness factor has been found most consistently to predict academic achievement in both preclinical and clinical phases of medical education (Costa & McCrae 1992; Goff & Ackerman 1992; Blickle 1996; De Raad 1996; De Raad & Schouwenburg 1996; Busato et al. 2000) and performance in the work environment (Barrick & Mount 1991; Salgado 1998; Dudley et al. 2006; Burch & Anderson 2008). Obviously, facets of this factor such as achievement striving, competence, dutifulness, self-discipline, order and deliberation (Costa & McCrae 1992) can provide a plausible explanation as to why the Conscientiousness factor is the best predictor of academic success in undergraduate college students (Wolfe & Johnson 1995), as well as graduate college students (Wiggins et al. 1969), and in medical school (Lievens et al. 2009).

Tyssen and colleagues (2007), in a six-year longitudinal study of 421 students who were accepted into four medical schools in Norway reported that low levels of Conscientiousness combined with high levels of Neuroticism and low levels of Extraversion could increase susceptibility to stress in medical school, thus negatively affecting academic performance. In a study by Merlo and Matveeskii (2009), the hypothesis that personality attributes such as conscientiousness, social skills, and mental well-being, are associated with success in clinical anesthesia was confirmed, which was in agreement with findings from previous research (Gough, et al. 1991; McDonald et al. 1994; Reich et al. 1999).

Career interest: Personality attributes contribute to an individual's behavior, preferences and interests including career choices. Empirical studies provide support for the notion that personality is linked to specialty interests in medical students and physicians. For example, Borges and Savickas (2002) found that scores on extraversion and openness to new experiences could distinguish surgeons from other physicians. It was also reported (Borges & Savickas 2002) that physicians in nonsurgical specialties were less adaptive to change (e.g. low on the Openness factor). Myers and Davis (1976) found that pathologists were less extraverted, and experienced more negative effects in their career due to lower levels of sociability and less dominating personalities. Psychiatrists were described as being imaginative, curious, looking for variety and experiencing deep feelings which are among features of the Openness and Agreeable factors (Borges & Savickas 2002).

Anesthesiologists, surgeons and psychiatrists, compared to obstetricians/gynecologists, showed a common feature by sharing a higher mean score on the Openness factor. Family physicians were found to be mixed in this factor (Borges & Savickas 2002). Lower scores on the Extraversion factor were shared by anesthesiologists and surgeons, but family physicians and psychiatrists were more Agreeable than obstetricians/gynecologists and surgeons (Borges & Savickas 2002). Family physicians, who were characterized as sympathetic, trusting, cooperative and altruistic, showed higher scores on the Agreeableness and Conscientiousness factors, but varied regarding the Openness factor (Borges & Savickas 2002).

Hoffman and colleagues (2010) studied a group of 204 residents (in surgery, medicine, pediatrics and anesthesiology), and another group of 207 medical students, and compared their scores on the big five factors with norms for the general population. They found that surgery residents scored higher on the Conscientiousness, and Extraversion factors but lower on Openness. Medical students scored on average high on Extraversion which placed them in the same group as students who studied law, economics, psychology, education, and political and social sciences. Conversely, medical students' extraversion scores were significantly higher than students of other academic majors such as sciences and applied sciences (Lievens et al 2002). In a study by Magee and Hojat (1998), using the NEO PI-R, it was found that male and female physicians who were nominated as positive role models in medicine, compared to the general population, scored significantly higher on the Conscientiousness Factor, and on personality facets such as achievement striving, activity, competence, dutifulness, trust, assertiveness and altruism. They scores lower than the general population on the vulnerability facet of personality. In another study, internal medicine residents, compared to the general population, scored higher on the Openness Factor, and on the idea, achievement striving, excitement seeking, fantasy, feelings and deliberation facets (Hojat et al. 1999c).

Chibnall et al. (2009) compared 133 third-year medical students at Saint Louis University School of Medicine with 163 police officer recruits. Discriminant function analysis showed that the factors of Conscientiousness, Neuroticism and Openness could accurately classify 77% of medical students and police recruits. Medical students scored higher on Openness and Neuroticism but lower on the Conscientiousness factor than police recruits. This pattern of findings, according to study investigators can be explained by the fact that Openness and Neuroticism, in contrast to Conscientiousness, do not seem to be virtues for police and security personnel. A high level of Conscientiousness in police recruits can be expected, considering facets of this personality factor such as order, dutifulness and self-discipline which are desirable characteristic for police recruits (Chibnall et al. 2009).

Barrick and Mount (1991) examined the relationship between the five personality factors and indicators of job performance in five occupational groups which included professionals (physician were in this occupational group), police, managers, sales persons and skilled/semi-skilled workers). Results showed that the Conscientiousness factor consistently predicted performance in all occupational groups. Extraversion predicted performance in two occupational groups that required social interaction, such as managers and sales persons and training proficiency in all occupations was linked to the Openness and Extraversion factors (Barrick & Mount 1991). Medical students, compared to students in philosophy, languages and history scored significantly higher on the Consciousness and Extraversion factors (Lievens et al. 2002). Medical students compared to humanities students scored lower on the Extraversion and Openness factors (Bunevicius et al. 2008).

The FFM of personality has received attention among personality researchers, and is recognized as the most parsimonious and comprehensive model of normal adult personality (Costa & McRae 1992; Yamagata et al. 2006). Although its use in medical education research is not yet widespread, its potential for providing useful information for personality research in medical education is worthy of consideration (Chibnall et al. 2009).

Overall, the results of the FFM in medical education research generally suggest that among all big five factors, the Conscientiousness factor seems to be a more consistent predictor of academic performance in medical school, and the Extraversion factor appears to be linked to preferences in some “people-oriented” specialties that require more intense patient-physician interaction.

• The NEO-PI-R is a widely used instrument in personality research for the assessment of the big five personality factors: Openness, Conscientiousness, Extraversion, Agreeableness and Neuroticism.

• The Conscientiousness factor and its facets (achievement striving, deliberation, dutifulness, order, and self-discipline) are conceptually more relevant to performance of physicians-in-training and in-practice.

• Empirical data support the link between scores on the Conscientiousness factor and performance measures in the preclinical and clinical phases of medical education.

1. The 16 Personality Factor Questionnaire

The 16 Personality Factor Questionnaire (16PF) is a well-known instrument developed by Cattell (1943, 1946, 1947, 1948), Cattel & Kline (1977) and Cattell et al. (1993). It is one of the oldest personality instruments, first published in 1949, revised several times with the most recent version released in 1993. It contains 185 items which provide scores for the 16 primary personality factors.

The 16 personality factors were determined and based on an extensive factor analytic study of a large number of personality attributes derived from a psycho-lexical hypothesis suggested by Allport and Odbert (1936), based on the assumption that if a word exists for a personality attribute then that attribute must be real. The primary 16 factors are Warmth, Reasoning, Emotional Stability, Dominance, Liveliness, Rule-Consciousness, Social Boldness, Sensitivity, Vigilance, Abstractedness, Privateness, Apprehension, Openness to Change, Self-reliance, Perfectionism and Tension. Higher order factor analysis of the primary 16 personality factors resulted in the five global personality traits which resemble the FFM of personality. These five global personality traits are: Openness-Tough Mindedness (analogous to the Openness factor in the FFM), Self-Control (analogous to the Conscientiousness factor in the FFM), Extraversion (similar to the Extraversion factor in the FFM), Independence-Accommodation (analogous to the Agreeableness factor in the FFM), and Anxiety (analogous to the Neuroticism factor in the FFM) (Conn & Rieke 1994; Hofer & Eber 2002). The 16PF Questionnaire enjoys strong psychometric support (Cattell et al. 1970). It is one of the most frequently used instruments in a large volume of personality studies, and has also been used in medical education research.

Performance: In a study by Manuel et al. (2005), 206 medical students at the University of Cincinnati, School of Medicine completed the 16PF Questionnaire. The scores on the Warmth factor were positively correlated with indicators of clinical data gathering skills. However, scores on the Abstractedness and Privateness factors were negatively correlated with the assessment of clinical skills. In addition, measures of communication skills correlated positively with Warmth, Emotional Stability and Perfectionism factors and negatively with the Privateness factor. The investigators concluded that some personality factors from the 16PF questionnaire can predict medical students’ clinical skills (Manuel et al. 2005).

In explaining their findings, Manuel and colleagues (2005) speculated that high scorers on the Warmth factor are likely to have the following features: attentive to others, easy-going and likeable; high scores on the Emotional Stability factor are likely to be adaptive, mature and in control; and high scorers on the Perfectionism factor are likely to be self-disciplined, socially precise and organized. All of the aforementioned personality attributes contribute positively to clinical skill assessments. In addition, those investigators suggest that high scorers on the Abstractness factor of the 16PF Questionnaire are likely to be impractical, and high scorers on the Privateness factor are likely to be discreet and shrewd. Thus, these personality attributes were expected to be negatively related to assessments of clinical competence (Manuel et al. 2005).

In another study by Green et al. (1991), the 16PF Questionnaire was administered to 129 medical students at the University of Wales, College of Medicine. No relationship was found between personality scores, performance in medical school, and subsequent academic success. It was concluded that the 16PF questionnaire would be unlikely to help in the assessment of applicants and medical students (Green et al. 1991). Similar findings were obtained in another study of 146 students at the University of Wales, College of Medicine by the same investigators (Green et al. 1993).

Contrary to findings reported by Green and colleagues (1991), a study in Malaysia by Peng and colleagues (1995) showed that the 16PF Questionnaire could make a distinction between students who were at risk of academic failure and their counterparts who were not. In their study, 101 students completed the Bahasa Malaysian translation of the 16PF Questionnaire at the beginning of medical school. The relationships between personality scores and academic success at the end of the second year of preclinical training in medical school were examined. Personality attributes such as being enthusiastic (a feature of the Liveliness factor), venturesome (a feature of the Social Boldness factor), imaginative (a feature of the Abstractedness factor) and experimenting (a feature of the Openness to Change factor) correlated positively with indicators of success. A personality attribute such as being self-assured (a feature of the Apprehension factor) was negatively correlated with performance measures (Peng et al. 1995). Students who were academically in trouble were more likely to be reserved, less emotionally stable, and more apprehensive than others. The authors concluded that the 16PF Questionnaire is a useful instrument for identifying the personality profile of students who are likely to have academic problems (Peng et al. 1995).

Huxham et al. (1985) administered the 16PF questionnaire and the EPI (Eysenck & Eysenck 1964, 1975) to a cohort of 142 medical students in Australia in the second and sixth years of medical school. These investigators were interested in examining changes in personality during medical school. They concluded that during the study period, medical students became brighter, more mature, more venturesome, more tough-minded, more trusting, more self-assured, more self-controlled and more extraverted (Huxham et al. 1985).

Career interest: The 16PF Questionnaire has also been used to examine specialty differences. For example, Reeve (1980) used the 16PF to compare anesthesiologists and general practitioners. It was found that the former group was more likely to be self-sufficient (a feature of the Self-Reliant factor), dominant (a feature of the Dominance factor), tense (a feature of the Tension factor) and introverted. In another study, Borges and Osmon (2001) used the 16PF questionnaire to investigate personality differences among anesthesiologists compared to family physicians and general surgeons. Anesthesiologists seemed to have a different level of suspiciousness and skepticism (features of the Vigilance factor) than the other two groups of physicians. Family physicians differed significantly from general surgeons and anesthesiologists with regard to Rule-Consciousness and Abstractedness factors which indicate that family physicians were more rule bound and imaginative, which is somewhat consistent with Taylor (1993), and the Taylor et al. (1990) description of family practitioners.

By using the 16PF questionnaire, Chowdhury and colleagues (1987) showed that internal medicine residents had a tendency to be skeptical and aloof. Residents in psychiatry were characterized by greater tolerance for frustration, emotional maturity, and stability. Psychiatry residents were also found to be more tender-minded, compared to internal medicine residents who were more realistic and practical. Psychiatry residents showed a high-level capacity for abstract thinking, faster learning and a quicker grasp of ideas (Borges & Savickas 2002).

Findings on the 16PF Questionnaire generally suggest that the instrument has limited success in predicting academic performance in medical school or in predicting specialty interest of medical students.

• The 16PF Questionnaire which provides scores for 16 primary personality factors is one of the oldest personality instruments used in medical education research.

• Although this instrument generally enjoys strong psychometric support in personality research, its success in predicting specialty interest and performance of physicians-in-training and in-practice is limited.

1. The California Psychological Inventory

The California Psychological Inventory (CPI) is a frequently used self-report personality instrument, originally developed by Harrison Gough, which has been revised three times. It is a lengthy instrument, and its current form includes 434 items measuring 20 folk scales: Dominance, Capacity for Status, Sociability, Social Presence, Self-acceptance, Independence, Empathy, Responsibility, Socialization, Self-control, Good Impression, Communality, Well-being, Tolerance, Achievement via Conformance, Achievement via Independence, Intellectual Efficiency, Psychological Mindedness, Flexibility and Femininity-Masculinity (Gough 1987). The scales were not developed based on factor analytic research; however, a factor analytic study of the CPI suggests that the big five factors of personality can also be measured by the CPI (Soto & John 2009).

The CPI has been used in a number of medical education studies in predicting academic performance and specialty interest. In a review article, Ferguson et al. (2002) claimed that the CPI was the most commonly used personality instrument in medical education.

Performance: Gough and colleagues reported several studies using the CPI with medical students and residents to predict their performances (Gough et al. 1963, 1964, 1991). Gough and Hall (1967) reported that the CPI could differentiate cognitive performance of successful from unsuccessful students in medical school. Reich et al. (1999) demonstrated that certain scales of the CPI were associated with poor clinical performance among residents.

In a study with all applicants to the School of Medicine at Ben-Gurion University, statistically significant correlations (albeit low in magnitude), were found between interview ratings and scores of the following scales of the CPI: Dominance, Self-acceptance, Well-being, Tolerance, Responsibility and Achievement via Conformance (Hobfoll et al. 1982). Interview ratings were based on a global judgment of overall assessment of attributes such as empathy, responsibility, personal integrity, intellectual flexibility and tolerance of ambiguity.

In a discriminant analyses, it was also found that scores of the CPI scales of Achievement via Independence, Self-acceptance, Dominance and Achievement via Conformance were the best overall predictors of cognitive performance and teaching staff ratings (Hobfoll et al. 1982). Teaching staff ratings were based on the judgment of five teaching staff on whether a student fits an idealized model with regard to personality attributes such as self-initiative, interpersonal sensitivity and intellectual flexibility. No significant correlation was found between scores of the CPI and ratings of clinical competence (Hobfoll et al. 1982).

Ferguson and colleagues (2002) in their review article reported that the following eight scales of the CPI had more consistently emerged as significant predictors of success in medical education: Dominance, Tolerance, Sociability, Self-acceptance, Well-being, Responsibility, Achievement via Conformance and Achievement via Independence. Their summarized review findings indicate that scores on the Dominance scale correlated negatively with undergraduate multiple choice examination grades (r = −0.26); Tolerance correlated negatively with the ability to use numerical information (r = −0.25); and Well-being and Achievement via Conformance correlated positively with success in oral examinations (0.22 and 0.32, respectively) (Ferguson, et al. 2002).

In a study by Hodgson and colleagues (2007), it was found that physicians who demonstrated unprofessional behavior during medical school, compared to those who did not, scored significantly lower on four CPI scales. Results were in agreement with findings in which indicators of unprofessional behavior of medical students, extracted from the excerpts of negative comments in medical students’ academic records, could be grouped into domains of irresponsibility, lack of self-improvement and poor initiative (Papadakis et al. 2005).

The level of professionalism in medicine was found to be significantly associated with scores of the CPI scales of Responsibility (r = 0.53), Communality (r = 0.50) and Well-being (r = 0.46) (Hodgson et al. 2007). Significant differences were observed on scores on the CPI scales of Responsibility, Sociability, Self-control, Communality and Well-being between those who had a record of unprofessional behavior and their counterparts without such a behavior (Hodgson et al. 2007). These findings suggest that the CPI, administered at matriculation to medical school, could predict unprofessional behavior during medical school (Hodgson et al. 2007).

Career interest: Gough et al. (1991) administered the CPI to first-year anesthesiology residents and reported that they were self-confident, had superior interpersonal skills, and were goal seeking as indicated by their high scores on the CPI scales of Dominance, Social Presence and Achievement via Independence, respectively. Coombs et al. (1993) compared surgical and nonsurgical specialists who graduated from the University of California, School of Medicine. They administered the CPI and other personality scales at the beginning and at the end of medical school, and found no pronounced difference between the two groups on any of the CPI scale scores (Coombs et al. 1993).

Overall, regarding the above-mentioned findings on the use of the CPI in medical education research, we agree with the concluding remarks by Hobfoll and colleagues (1982) that this personality instrument has a limited value in predicting students’ performance in medical school and the specialty interest of physicians in training.

• The California Psychological Inventory (CPI) is a lengthy instrument and one of the most commonly used personality inventory in medical education research.

• Despite a large volume of research, the CPI seems to have a limited value in predicting specialty interest and performance of medical students and physicians.

1. The Myers-Briggs Type Indicator

The Myers-Briggs Type Indicator (MBTI) is a widely used personality instrument developed in the 1950s by Isabel Briggs Myers and her mother Katherine Cook Briggs based on Carl G. Jung's psychological typology (Jung 1933, 1971; Myers 1962; Myers & Caully 1985). The test includes 144 forced-choice items (in a longer Form Q and 93 items in a shorter Form M) designed to measure four bipolar personality types and their combinations: Introversion-Extraversion (I or E type), Sensing-Intuition (S or N type), Thinking-Feeling (T or F type) and Judging-Perceiving (J or P type). Based on the scores on the aforementioned personality types, the test taker can be further classified into one of the 16 combined personality types. For example, higher scores on Introversion (as opposed to Extraersion), Intuition (as opposed to Sensing), Thinking (as opposed to Feeling), Judging (as opposed to Perceiving) will classify individual's combined personality type in the Introversion-Intuition-Thinking-Judging category, or the INTJ type.

The MBTI has been widely used in educational counseling, human resource management and in medical education research. In an early large-scale study, Myers and Davis (1965) used data from the MBTI collected in the 1950s from 45 medical schools on 5355 students. It was found that there were approximately equal numbers of medical students in all of the personality types; thus, it was concluded that medicine is a diverse field that can benefit from a variety of personality types; each can match a desirable personality constellation for a particular specialty. In another large-scale study, a total of 7190 medical students completed the MBTI and it was reported that there were more Intuitive, Feeling and Judging types among medical students compared to the general population (McCaulley 1977, 1981).

Performance: Some studies using the MBTI reported a link between personality types and academic performance in medical students. For example, in one study medical students who were classified as the Sensing-Thinking (ST) type obtained the highest scores in a neurochemistry course (Wild & Skipper 1991). In another study with 114 students at the University of New Mexico, School of Medicine, the failure rate was highest in the medical licensing examination (National Board of Medical Examiners, Part 1) among those who were classified as the Intuitive-Feeling (NF) type (O’Donnell 1982). In a study by Tharp (2009), the highest grades in an undergraduate physiology course were achieved by students with a Sensing preference. Kim (1999) reported that medical students with a Thinking preference performed better in medical school than their counterparts with a Feeling preference.

The MBTI was administered to 263 osteopathic medical students at Midwestern University/Chicago College of Osteopathic Medicine to examine the relationship between personality types and performance on the MCAT (Sefcik et al. 2009). No significant correlation was obtained between personality types and performance on the MCAT. However, the NF personality type students were more likely to score lower on the Comprehensive Osteopathic Medical Licensing Examination-USA (COMLEX-USA, Level 1) (Sefcik et al. 2009).

In a study by Ornstein and colleagues (1987), the association between personality types, and residents’ laboratory test ordering behavior was examined. Participants included 39 family medicine residents at the University of South Carolina Medical Center in Charleston who treated 1326 hypertensive patients in 14 006 visits, and ordered 7361 laboratory tests. Results indicate that the Introvert and Intuitive types were likely to order more tests than the Extravert and Sensing types; however, the findings did not reach the traditional level of statistical significance (Ornstein et al. 1987). McNulty and colleagues (2006) examined the relationships between personality types and learning style. Findings showed that although the use of computer-aided instruction was positively correlated with the Sensing rather than Intuitive personality types, higher use of discussion forums (as opposed to lecture and tutorial) was associated with Perceiving-Judging type. From their study with 137 medical students, Wild and Skipper (1991) concluded that the relationships between personality types and academic performance may be more complicated than had been discussed in the literature.

Career interest: The MBTI has been used more frequently in research on career interest and in understanding the relationship between personality types and specialty choices. It has been argued that because of the variety of diverse specialties, medicine is a discipline that can attract students with a variety of personality types (Myers & Davis 1965). In a large scale study to examine personality types and specialty choice, all U.S. medical schools were surveyed to identify those schools in which the MBTI had ever been administered to their students.

The schools were also asked whether they would be willing to share their MBTI data if available (Stilwell et al. 2000). Twelve schools expressed their willingness to share data. Findings on the available data linking personality types with specialty choice showed that the Thinking type students were likely to choose primary and non-primary care specialties at about the same rate; however, the Feeling type students were significantly more likely to select primary care specialties. Similarly, the Extravert type students chose primary and non-primary care specialties at about the same rates.

Those who pursued family medicine were more likely to be the Feeling rather than Thinking type (Stilwell et al. 2000). Also, the Feeling types chose surgical specialties at a significantly lower rate than did the Thinking types. Finally, the Introvert types pursued surgical specialties at a significantly lower rate than did the Extravert types (Stilwell et al. 2000). These investigators further divided those who chose non-primary care specialties into two groups of surgical and nonsurgical specialties. Their findings suggest that gender, E–I and T–F types could predict interest in surgical specialties (e.g. being male, extraverted and thinking types) (Stilwell et al. 2000).

The personality types of applicants to an otolaryngology residency program were compared to those of the general population and physicians in other medical specialties (Zardouz et al. 2011). It was found that otolaryngology applicants were likely to have an Extravert-ST-Judging personality profile. These investigators also reported that Thinking (T) and Judging (J) types were more common than Feeling (F) and Perceiving (P) types among the otolaryngology residency applicants (Zardouz et al. 2011).

A study by McCaulley (1978) showed that those who were attracted to ophthalmology and otolaryngology had similar personality types. However, those choosing ophthalmology were more people-oriented and those choosing otolaryngology were more technology-oriented. The proportion of Extravert-ST-Perceiving type was significantly higher in otolaryngology residency applicants (8%) than the general population (3%) (Zardouz et al. 2011). In different studies, personality types of physiatrists (Sliwa & Shade-Zeldow 1994), pediatric residents (Lacorte & Risucci 1993) and emergency department staff have been compared (Boyd & Brown 2005), and no pronounced differences in their personality types have emerged.

Harris and Ebbert (1985) used the MBTI to examine differences in personality types between first-year family medicine residents and rural primary care physicians. Results showed that the residents were significantly more Intuitive (as opposed to Sensing) and more Feeling (as opposed to Thinking) types. The authors concluded that family medicine residents differed from rural primary care physicians in how they gather information. Family medicine physicians also tend to rely on their intuition (N) rather than sensing (S) perception when gathering information (Myers & Davis 1976; Harris & Ebbert 1985; Friedman & Slatt 1988; Taylor et al. 1990).

Obstetricians and gynecologists have been described by Myers and Davis (1976) as being more likely to be Extravert and Sensing types. Findings of a longitudinal study using the MBTI (McCaulley 1978) showed that the obstetrics-gynecology specialty attracted individuals with a Sensing type, whereas Friedman and Slatt (1988) found that medical students who entered obstetrics-gynecology tended to score high on ST-Judging dimensions. Myers and Davis (1976) reported that pediatricians showed a large proportion of the Extraversion-Sensing-Feeling-Judging type, as well as Introverted-Sensing-Feeling-Judging types. However, Friedman and Slatt (1988) found that medical students interested in pediatrics yielded less distinctive MBTI profiles. They also found that medical students who were interested in psychiatry were more likely to display an Introverted-Feeling-Perceiving personality type (Friedman & Slatt 1988).

Myers and Davis (1976) found that surgeons were more likely to display the Extraverted and Sensing (S) type, whereas Friedman and Slatt (1988) found that students interested in surgery yielded less distinctive MBTI types. Findings of a longitudinal study (McCaulley 1978) showed that the surgical subspecialties of general, orthopedic and obstetrics/gynecology, which deal with straightforward problems requiring technical skill, attracted individuals with a Sensing (S)-type personality.

The Sensing type has often been reported to be common among obstetricians (Myers & Davis 1976; McCaulley 1978), general surgeons and orthopedic surgeons (McCaulley 1978). Neurological, plastic and thoracic surgeons (McCaulley 1978) often score high on the Intuitive dimension and thus could be characterized as imaginative, curious and having a need for variety (Borges & Savickas 2002). Although hospital-based and support specialties, such as pathology and radiology have not received as much attention in MBTI studies, some researchers have addressed personality types in these specialties. For example, Myers and Davis (1976) reported that pathologists tended to be the Introvert, Intuitive and Thinking type; and Friedman and Slatt (1988) reported that students interested in pathology did not display a distinct personality type in the MBTI. Using the MBTI, Myers and Davis (1976) showed that anesthesiologists were characterized as both Introverted-ST-Perceiving and Introverted-Sensing-Feeling-Perceiving types.

It has been reported that compared with data from the 1950s, the type distribution of physicians has remained relatively unchanged, with the exception of a trend toward more Judging types. It is also reported that women in medicine, today as compared to those in the 1950s when medicine was more male-dominated, are more representative of the general population in the Feeling personality type (Stilwell et al. 2000). From the published studies, it seems that Feeling type students and women were more likely to choose primary care specialties.

Although research findings on the link between personality types from the MBTI and specialty interest do not provide a consistent and clear picture, a more frequently reported conclusion that can be drawn from the MBTI studies is that surgeons are more likely to be the E type (extraverted) (Myers & Davis 1976; McCaulley 1978; Stilwell et al. 2000), suggesting that they tend to be sociable and active. Another frequently reported finding from MBTI studies of medical specialists is that family physicians are likely to have a Feeling personality type (Harris & Ebbert 1985; Friedman & Slatt 1988; Taylor et al. 1990; Stilwell et al. 2000; Borges & Savickas 2002), which can be helpful at least in medical students’ career counseling.

Despite the large volume of medical education research in which the MBTI has been used, one cannot determine with confidence which personality type performs better in medical school, and which personality type predicts interest in a specific specialty and subspecialty. In addition, some of the findings on personality types and specialty choice seem counter-intuitive such as family physician's Introvert and surgeons Extravert personality types because intuitively, family physicians require more social skills than surgeons to maintain long-term relationships with their patients.

Overall, the MBTI does not seem to be a useful instrument in predicting academic performance in medical school. Although the MBTI has been widely used in medical education research and in career counseling, this instrument has little credibility among research psychologists (APA 2007, pp. 604–605).

• The MBTI probably is the most widely used personality instrument in medical education research on career counseling and specialty choices.

• Despite the large volume of research, this instrument does not have high credibility among psychologists and personality researchers.

1. The Jefferson Scale of Empathy

The Jefferson Scale of Empathy (JSE) (20 items) was specifically developed for measuring empathy in the context of medical education and patient care, relying on the conceptualization of empathy as a predominantly cognitive (as opposed to affective or emotional) attribute that involves an understanding (as opposed to feeling) of patients’ pain, experiences, concerns and perspectives, combined with a capacity to communicate this understanding, and an intention to help (Hojat 2007, 2009; Hojat et al. 2009). This conceptualization makes a distinction between empathy (predominantly a cognitive attribute) and sympathy (predominantly an effective attribute). The two concepts have different consequences in patient care (Hojat 2007; Hojat et al. 2011b). For example, empathy in abundance is always beneficial in patient care, while sympathy in excess can be detrimental, causing emotional dependency in patients and leading to emotional exhaustion, burnout and compassion fatigue in physicians (Hojat 2007, Hojat et al. 2011b). This distinction was recognized by Nightingale et al. (1991) in their empirical study in which they observed that physicians’ empathy had a different measurable effect than sympathy on their clinical decision making behavior.

Three versions of the JSE are available: One for administration to medical students (S-Version), one for administration to physicians and other health professionals (HP-Version) and one for administration to students in any health profession fields other than medicine (HPS-Version). These versions are similar in content with slight changes in wording to reflect students’ orientation toward empathy in medical education (S-Version), other health profession education (HPS-Version), and behavioral tendencies toward empathic engagement in patient care (HP-Version) in physicians and other health professionals. For example, an item in the S-Version that reads “It is difficult for a physician to view things from patients’ perspectives” reads as “It is difficult for me to view things from my patients’ perspective” in the HP-Version, and reads as “It is difficult for a health care provider to view things from patients’ perspectives” in the HPS-Version.

In exploratory factor analytic studies, three factors of “perspective taking,” “compassionate care” and “walking in patients’ shoes” have emerged in samples of medical students and physicians in the United States (Hojat et al. 2001a, 2002c) and abroad (Alcorta-Gaza et al. 2005; Di Lillo et al 2009; Kataoka et al. 2009; Rahimi-Madiseh et al. 2010; Roh et al. 2010; Shariat et al. 2010). The three-factor model was also reproduced in confirmatory factor analytic studies with medical students in England (Tavakol et al. 2011) and in medical students in Iran (Shariat & Habibi 2013). Ample evidence has been reported in support of the validity and reliability of the JSE in medical and other health profession students, physicians, and other practicing health professionals. The JSE has enjoyed broad international attention by medical education researchers, has been translated into 43 languages thus far, and used in over 60 countries. It has been described as one of the most researched and widely used instruments in medical education (Colliver et al. 2010). Information about the JSE is posted at: www.tju.edu/jmc/crmehc/medu/oempathy.cfm).

To our knowledge, before the development of the JSE, no psychometrically sound instrument was available to measure empathy specifically among medical students, residents and physicians. There was a need for such an instrument, and in response the JSE was developed to measure empathy in the context of patient care. A few instruments exist for measuring empathy in the general population (for a review, see Hojat 2007, pp. 63–74). However, none of those instruments is content-specific and context-relevant to medical education and patient care.

The following four of these instruments have been frequently used in medical education research. The Interpersonal Reactivity Index (IRI, Davis 1983) includes 28 items tapping both cognitive and emotional empathy, and contains four scales: perspective-taking, empathic concern, fantasy and personal distress. A sample item (from the perspective-taking scale) is “I sometimes try to understand my friends better by imagining how things look from their perspective.” Another instrument is the Empathy Scale (Hogan 1969) which includes 64 items. A sample item is “I have seen some things so sad that I almost felt crying.”

The third instrument is the Emotional Empathy Scale (Mehrabian & Epstein 1972) which includes 33 items intended to measure “emotional empathy” (synonymous to sympathy). A sample item is “It makes me sad to see a lonely stranger in a group.” There is another instrument, the Balanced Emotional Empathy Scale (BEES, Mehrabian 1996), which includes 30 items, and according to its author intended to measure “vicarious emotional empathy.” A sample item is “Unhappy movie endings haunt me for hours afterward.” As indicated before, and reflected in the content of the sample items, none of these instruments seem to have “face” and “content” validity specific to medical education and patient care. With the exception of the BEES, extensive psychometric data have been published for the other three instruments in the general population (Hojat 2007, pp. 66–69, 72–73). Thus, the JSE is the only instrument featuring “face” and “content” validities in the context of medical education and patient care.

Performance: A significant association has been reported between medical students’ scores on the JSE and medical school faculty's global ratings of students’ clinical competence in core clinical clerkships in the third year of medical school (Hojat et al. 2002a). This association can be explained by the fact that the ability to communicate with patients and understand their concerns (key features in the conceptualization of empathy) is often taken into consideration in the assessments of students’ global clinical competence. No significant association was observed between scores of the JSE and grades on objective (multiple-choice) examinations of medical knowledge (Hojat et al. 2002a), which was consistent with findings reported by other researchers (Hornblow et al. 1977; Kupfer et al. 1978; Diseker & Michielutte 1981; Austin et al. 2005). In a group of Mexican medical students, significant associations were found between scores of the JSE and academic performance in medical school (Alcorta-Garza et al. 2005).

Significant associations have been found between JSE scores on the one hand, and simulated patients’ evaluations of students’ empathic engagement in objective structured clinical exam stations (OSCE, Berg et al. 2011a, 2011b), peer nominations on professionalism attributes (Pohl et al. 2011), and scores of attitudes toward interprofessional collaboration (Hojat et al., 2012c; Ward et al. 2009), on the other hand.

Career interest: Scores of the JSE have been associated with specialty choice. For example, several studies reported that physicians in “people-oriented” specialties (e.g. general internal medicine, family medicine, pediatrics and psychiatry) scored higher on the JSE than others who were practicing “technology-oriented” or “procedure-oriented” specialties (e.g. pathology, radiology anesthesiology, surgery) (Hojat et al. 2002b, 2002c).

This pattern of finding was observed not only among practicing physicians (Hojat et al. 2002b, 2002c), but also among first year medical students who completed the JSE on the first day of medical school (orientation day) before being exposed to medical training (Hojat et al. 2005). In addition to completing the JSE, these students were asked about the specialty they were planning to pursue after graduation from medical school. Students planning a “people-oriented specialty (e.g. family medicine, general internal medicine, general pediatrics, psychiatry) scored higher on the JSE than their peers who chose “technology/procedure-oriented” specialties (e.g. pathology, anesthesiology, radiology, surgery) (Hojat et al. 2005).

The significant differences in the mean scores of the JSE observed among physicians in “people-oriented” and “technology/procedure-oriented” specialties can be partially explained by the fact that physicians with different degrees of interpersonal skills are naturally inclined to pursue specialties that demand certain degrees of interpersonal skills (Harsch 1989). The differences might also be a function of medical training by the amount of emphasis that is placed on interpersonal skills training in different specialties. Obviously, the “people-oriented” specialties, such as general internal medicine, require a higher degree of interpersonal skills than “technology/procedure-oriented” specialties, such as pathology, diagnostic radiology or anesthesiology. However, our findings that entering medical students with higher scores on the JSE, before being exposed to formal medical education, are interested in pursuing “people-oriented” specialties suggest that interpersonal skills training in medical school may not be the only factor that prompts students to pursue specialties that require such skills.

In addition, among consistent findings was the gender difference in mean scores of the JSE in the favor of female medical students (Hojat et al. 2001a, 2002a, 2002b) and in practicing physicians (Hojat et al. 2002c) in the United States and abroad (Hsiao et al. in press; Alcorta-Garza et al. 2005; Kataoka et al. 2009; Shariat et al. 2010; Suh et al. 2012; Zenasni et al. 2012). This pattern of finding for women's higher empathy is consistent with those reported in the general population. Several explanations can be offered for a gender difference in empathy. For example, it has been suggested that women are more receptive to emotional signals (Trivers 1972), a quality that can contribute to a better understanding and hence to a greater empathic engagement. Also, on the basis of the evolutionary theory of parental investment (Trivers 1972), women are inclined to invest more than men in the caring for their children and develop more caring attitudes toward their offspring which is also reflected in their social relationships. The findings on gender differences in empathy are also in agreement with the reports on the practice style of female physicians who are likely to spend more time with their patients (Bertakis et al. 1995), and render more preventive and patient-oriented care (Maheux et al. 1990; Hojat et al. 1995a). All of these factors can lead to forming an empathic engagement in patient care.

Other correlates: Significant associations have been reported between scores of the JSE and some personality measures. For example, in a study with medical students (Hojat et al. 2005b), we found that the scores of the JSE were significantly and positively correlated with Sociability scores (measured by the Zuckerman-Kuhlman Personality Questionnaire [ZKPQ], Zuckerman 2002). In addition, a significant but negative correlation was observed between the JSE and Aggressive-Hostility scores from the ZKPQ. Furthermore, higher scores on the JSE were significantly associated with higher levels of students’ self-reported satisfaction with their early relationships with their mothers (Hojat et al. 2005b), which provides support for the notion that empathy is nurtured by the quality of the early mother-child attachment relationship (Hojat 1998, 2007).

A statistically significant correlation has been observed between scores of the JSE and a measure of patients’ perceptions of physician empathy among physicians in a family medicine residency program (Glaser et al. 2007). However, the relationship between scores of the JSE and patients’ assessments of physician empathy did not reach the conventional level of statistical significance (p < 0.05) in a study with internal medicine residents (Kane et al. 2007) or in another study with a small group of attending psychiatrists (Aggarwal 2007).

In a study of pharmacists, Reisetter (2003) reported significant correlations between JSE scores and those of the Physician Belief Scale (PBS; Ashworth et al. 1984; McLellan et al. 1999). For example, a positive correlation was found between scores of the JSE (compassionate care factor) and the PBS scores on clinicians’ concern about their ability to address their client's psychosocial problems. Conversely, a negative correlation was found between JSE scores (walking in the patient's shoes factor) and scores on one of the PBS scales that measures the difficulties perceived by the clinician to examine the client's psychosocial problems. In a study with medical students, no significant association was observed between scores of the JSE and psychostimulant drug abuse; however, psychostimulant drug abuse was significantly associated with scores on the “aggressive-hostility” factor of the ZKPQ in medical students (Bucher et al. 2013).

Clinical outcomes: Clinical outcome is a complex notion because it depends not only on physician performance, but also non-physician factors, such as insurance regulations, governmental policies, patients’ social-cultural background and beliefs, environmental, technical and human resources that contribute to the quality of patient outcomes (Gonnella et al. 1993). Perhaps because of this complexity, there is a scarcity of empirical evidence on the clinical outcomes of personality measures in medical education research. However, there are studies that report some indicators of empathy in the context of patient care to be associated with patient outcomes. For example, it was shown that specific features of empathic engagement in patient care, marked by understanding, communication, positive language, appropriate touching, eye contact and bodily posture, can lead to patient satisfaction (Hall et al. 1988; DiMatteo et al. 1993; Zachariae et al. 2003; Kim et al. 2004), greater compliance (DiMatteo et al. 1986; Falvo & Tippy 1988; Squier 1990), patients’ feelings of being important (Colliver et al. 1998), accuracy of diagnosis (Barsky 1981), accuracy of prognosis, (Dubnicki 1977) and lower rates of malpractice litigation (Beckman et al. 1994; Levinson et al. 1997).

It has been reported that physicians’ understanding of their patients’ perspective, a key feature in the conceptualization of physician empathy (Hojat 2007), enhances patients’ perceptions of being helped (Eisenthal et al. 1979), improves patients’ empowerment (Street et al. 2009), and increases patients’ perception of a social support network (Eisenthal et al. 1979; Hojat 2007; Street et al. 2009). In a study with diabetic patients, dietitians’ empathy was found to be predictive of patient satisfaction and successful consultations (Goodchild et al. 2005). Physicians’ understanding of their diabetic patients’ beliefs about their illness was associated with better self-care outcomes such as improved diet and increased blood glucose self-testing (Sultan et al. 2011). In a study with internal medicine residents, a lower level of empathy was associated with a higher rate of incidents of medical errors (West et al. 2009).

To our knowledge, there are only two empirical studies in which a direct link between scores of a validated measure of empathy developed in the context of patient care (JSE) and tangible clinical outcomes has been reported. In one study with 29 family medicine physicians and their 891 patients with diagnoses of diabetes mellitus, it was found that physicians’ scores on the JSE were predictive of optimal clinical outcomes in the patients (indicated by medical test results of hemoglobin A1c <7.0% and LDL-C<100) (Hojat et al. 2011a). In another large scale study with 242 primary care physicians and their 20 961 patients diagnosed with diabetes mellitus in Italy, it was found that physicians’ higher scores on the JSE were significantly associated with lower rates of metabolic complications (coma, diabetic ketoacidosis, hyperosmolar state) that required hospitalization of their patients (Del Canale et al. 2012).

In a recent editorial, we indicated that empathic engagement in patient care revolves around reciprocity and mutual understanding that evokes “psycho-socio-bio-neurological” responses in both physicians and patients (Hojat et al. 2013). These mechanisms provide plausible explanations for the observed associations between physician empathy and clinical outcomes. For example, at the psychosocial level, empathic engagement lays the foundation for a trusting relationship. Constraints in communication will diminish when a trusting relationship is formed. In the secure space of a trusting relationship, the patient begins to tell the tale of his/her illness without concealment. This in turn leads to a more accurate diagnosis and greater compliance, which ultimately will result in a better quality of care.

At the bio-neurological level, empathic engagement is analogous to a synchronized dance between involved parties, which is orchestrated by bio-neurological markers. For example, the interpersonal attunement in empathic engagement can activate some pro-social endogenous neuropeptides or hormonal changes (e.g. oxytocin, vasopressin) (Heinrichs & Domes 2008). In addition, a set of neurons, known as the mirror neuron system (MNS) is discharged when observing another person performing a goal-directed act, as if the observer is performing the act (Rizzolatti et al. 1996; Gallese 2001). In other words, the same set of neuron cells that is discharged in the acting person, will be implicated in the person who observes the act, without actually performing it. The MNS is believed to play an important role in understanding the experiences of others, which is the key ingredient of empathic communication. Of course, more research will further clarify the associations between physician empathy and clinical outcomes in a variety of diseases and settings and the underlying mechanisms.

Overall, findings of studies in which the JSE was used showed that empathy scores were significantly associated with indicators of clinical competence and were predictors of tangible patient outcomes. Furthermore, it was found that scores on the JSE were associated with career interest and specialty choices. Also, research findings confirmed that empathy can be enhanced and sustained by targeted educational programs (Hojat et al. 2012a; Van Winkle et al. 2012). The psychometric support and empirical findings suggest that the JSE is a promising instrument for measuring a personality attribute that is conceptually relevant to patient care, and empirically linked to clinical performance in medical school, career interest and patient outcomes.

• The JSE was specifically developed to measure empathy in the context of medical education and patient care.

• The JSE is supported by strong evidence in support of its validity and reliability in medical students, physicians and other health professions students and practitioners.

• Empirical data support the associations between scores of the JSE and indicators of clinical performance in medical school, and interest in broad areas of “people-oriented” and “technology/procedure-oriented” specialties.

• Empirical evidence is available in support of a link between physicians’ scores on the JSE and tangible clinical outcomes.

Other personality measures: In addition to the aforementioned personality instruments, a large number of other personality measures have been used with medical students, residents, and practicing physicians to predict their performance, adjustment and specialty interests. We selected a few of them for inclusion in this Guide.

1. The Eysenck Personality Inventory

The Eysenck Personality Inventory (EPI) (Eysenck & Eysenck 1964) and its successor The Eysenck Personality Questionnaire (EPQ) (Eysenck & Eysenck 1975) have been used in a number of medical education studies. The EPQ includes three scales of Extraversion, Neuroticism, and Psychoticism. It also contains a “Lie” scale to detect a “faking good” tendency. In a study with students at Jefferson Medical College (Fenderson et al. 1999), it was found that students in the top 20% of the class who declined an invitation to participate in an honors program in pathology scored higher on the Neuroticism scale of the EPQ. In another study, we found that medical students who received lower marks on clinical competence were more likely to score lower on the Extraversion scale of the EPQ (Hojat et al. 2004a). It was also found that those who were in the top half of their class in clinical competence evaluations scored significantly lower on the Neuroticism scale of the EPQ (Hojat et al. 1996a).

Westin and colleagues (1986) studied final year medical students at the University of Trondheim in Norway. They administered the EPI and calculated a measure of the students’ tendency to make assertive versus non-assertive decisions using three selected patient management problems. The relationships between personality measures (from the EPI) and an index of decision assertiveness (decisiveness), indicators of medical knowledge (grades and educational rank) and time used in making decisions were examined. A high degree of decisiveness in the clinical simulations was significantly correlated with Extraversion scores of the EPI. Decisiveness scores were negatively correlated with time spent on the examination of case histories. No relationship was found between decisiveness and grades in medical school, educational rank, Neuroticism scores of the EPI, or the students’ gender or age (Westin et al. 1986).

In a study by Ashton and Kamali (1995), second year medical students at the University of Newcastle in the UK completed the EPQ and a questionnaire about their alcohol, tobacco, cannabis and other illicit drug consumption, and their physical fitness. Compared to a previous study conducted about a decade earlier, no significant change was observed in students’ personality, prevalence of cigarette smoking, levels of caffeine consumption and participation in sports. However, students’ use of cannabis and other illicit drugs increased two-fold (Ashton & Kamali 1995). Golding and colleagues (1983) and Golding and Cornish (1987) reported significant correlations between personality factors and drug abuse in students. Specifically, tobacco and alcohol consumption and experience with cannabis and illicit drugs, which is detrimental to academic attainment, correlated with scores of the Psychoticism scale of the EPQ.

1. The Minnesota Multiphasic Personality Inventory

The Minnesota Multiphasic Personality Inventory (MMPI) is a widely used personality instrument, primarily for the assessment of mental health (Tellegen & Ben-Porath 2008). It has also been used in medical education research. For example, John et al. (1976) reported that poor academic performance in medical students was predicted by MMPI scores. In their study of medical students, Schonfield and Donner (1972) observed a link between higher scores of the masculine pole of the masculinity-femininity scale of the MMPI and interest in the technology-oriented specialties. In their comparisons of medical and law students, Solkoff and Markowitz (1967) used the MMPI and found that medical students were more introspective and idealistic and more sensitive to the needs of others, whereas law students were more likely to be extroverted and masculine oriented. It has also been reported that scores on the MMPI could predict physician burnout (McCranie & Brandsma 1988).

The MMPI was used in a study comparing accelerated and traditional students at three points in time: matriculation, after ten weeks, and after 62 weeks of medical school (Nathan et al. 1989). No significant difference was observed between the two groups of students. These investigators concluded that concerns about the relative immaturity of younger accelerated students and a corresponding inability to cope with the stressful environment of medical school might be unfounded (Nathan et al. 1989).

1. The Profile of Mood States

The Profile of Mood States (POMS; McNair et al. 1981) measures six mood-related dimensions: “Tension-Anxiety,” “Depression-Dejection,” “Anger-Hostility,” “Vigor-Activity,” “Fatigue-Inertia” and “Confusion-Bewilderment.” The POMS also assesses an overall personality attribute of mood disturbance by adding the scores of the six mood-related scales. A consistent pattern of findings that emotions vary throughout the school years was observed in eight studies (Mitchell et al. 2005). In two studies (Ford & Wentz 1984; Uliana et al. 1984), it was found that “Anger-Hostility” scores rose during the first year of residency training. Another study reported that scores on an additional scale such as Fatigue-Inertia worsened throughout the year (Gordon et al. 1986). In yet another study, Bellini and colleagues (2002) found that scores on Anger-Hostility, Fatigue-Inertia and Depression-Dejection all rose by the fifth month of internship.

These attributes are amenable to change by targeted programs. For example, in one of our studies, we noticed that a course in “mindfulness-based stress reduction” could reduce any psychological stress of students reflected in their significantly lower mean posttest scores on Tension-Anxiety and Confusion-Bewilderment, and higher mean scores on Vigor-Activity (Rosenzweig et al. 2003). In a similar study with primary care physicians who participated in a mindfulness meditation and self-awareness course, improvements in the POMS scores were observed (Krasner et al. 2009).

1. The Temperament and Character Inventory

The Temperament and Character Inventory (TCI), developed by Cloninger (1986, 1987) is a self-report instrument that measures four temperament and three character dimensions of Cloninger's personality model (Cloninger 1986, 1987; Cloninger et al. 1991, 1993). The four independent temperament dimensions are “Novelty Seeking,” “Harm Avoidance,” “Reward Dependence” and “Persistence.” The three character dimensions are “Self-Directedness,” “Cooperativeness” and “Self-Transcendence.”

In one study in Japan, the TCI was administered to 119 second year medical students at Osaka City University Graduate School of Medicine (Tanaka et al. 2009). It was found that scores on Persistence, Self-Directedness, Cooperativeness and Self-Transcendence were positively associated with a measure of intrinsic academic motivation. In a multiple regression analysis when adjustments were made for age and gender, it was found that scores on Persistence, Self-Directedness and Self-Transcendence were positively associated with intrinsic academic motivation that can lead to better academic performance in medical school (Tanaka et al. 2009).

In another study by Jiang and colleagues (2003), associations between the TCI scores, anxiety and fatigue were examined in 162 first-year and 89 fifth-year students from Saga Medical School in Japan. Significant and positive correlations were found in the TCI scores on the Harm Avoidance and scores on measures of anxiety and fatigue (general fatigue, psychological fatigue and physical fatigue). In addition, scores on Self-Directedness were negatively correlated with scores on trait anxiety and fatigue (Jiang et al. 2003). The TCI scores on Harm Avoidance and Self-Directedness, as predictors for fatigue-related disorders in medical students (Jiang et al. 2003), can influence academic performance in medical school. The associations between TCI scores and anxiety and depression have been addressed in other studies (Cloninger 1986; Crowley et al. 1993; Joffe et al. 1993; Tanaka et al. 1997, 1998; Hansenne et al. 1999).

1. The Personal Qualities Assessment

The Personal qualities Assessment (PQA) was developed in Australia, designed to assess personal qualities considered important for the study and practice of medicine and other health professions. The PQA questions are grouped into three scales. The first is a measure of individual differences in cognitive skills; the second is a measure of being involved or detached (empathy, self-confidence, narcissism and aloofness); and the third is a measure of ethical or moral orientation (Munro et al. 2005; Powis et al. 2005; James et al. 2009). In their study with Scottish medical students, Lumsden and colleagues (2005) found that students’ cognitive ability was similar in men and women, but women were more empathic and had better communication orientation.

Also, no significant differences were observed on any of the PQA measures between those who attended the state-funded or independent schools. Those with a deprived family background tended to score lower on the scale of cognitive skills. The study investigators concluded that fairness in the medical school admissions process might be improved by using personality instruments to objectively identify desirable qualities in future doctors (Lumsden et al. 2005).

1. The Maslach Burnout Inventory

The Maslach Burnout Inventory (MBI) is a 22-item self-administered instrument that measures three components of burnout: “Emotional Exhaustion,” “Depersonalization” and “Personal Accomplishment” (Maslach et al. 1996). Three studies used the same data collected from family medicine residents to analyze different aspects of burnout (Rafferty et al. 1986; Purdy et al. 1987; Lemkau et al. 1988). Findings indicate that family medicine residents, regardless of gender, exhibited moderate to high levels of burnout, especially on Emotional Exhaustion and Depersonalization scales. In another study, it was reported that 76% of internal medicine residents met criteria for burnout (Shanafelt et al. 2002). These studies suggest that burnout in different specialties is an important factor that must be taken into consideration to improve patient outcomes. It has been reported that burnout not only negatively can influence performance in medical school, but also can exert adverse impact on professionalism in medicine (Bellini et al. 2002; Shanafelt et al. 2002; Thomas 2004).

This instrument has been used for the assessment of educational programs to reduce psychological distress. For example, in a study with primary care physicians, it was found that a course in mindful meditation and self-awareness could cause a significant decrease in Emotional Exhaustion and Depersonalization, and an increase in the Personal Accomplishment component (Krasner et al. 2009).

1. The Medical Specialty Preference Inventory

The Medical Specialty Preference Inventory (MSPI): Career preference in medical specialty and career indecision have been studied in medical education by using the MSPI (Zimny 1979, 1980, Sodano & Richard 2009). The revised version of the MSPI (150 items, revised in 2002) calculates interest scores for six major specialties (family medicine, internal medicine, obstetrics/gynecology, pediatrics, psychiatry and surgery). Borges and colleagues (2005) reported that physicians whose specialty interest (measured by the MSPI) was congruent with their actual area of practice were more satisfied with their job than those with incongruent match between specialty of interest and of practice.

Career indecision can also be determined by the pattern of scores on this inventory (Richard 2005). For example, Walters (1982) examined the relationship between career indecision and academic performance. It was found that students who were classified as “low-interest undecided” obtained significantly lower medical school grades compared to “decided” students; whereas “high-interest undecided” students did not differ from the “decided” students. Another study examined the association between career indecision and personality. Students classified as “low-interest undecided” showed less personal integration compared with “decided” students (Walters 1982). Furthermore, as suggested by Walters (1982), a lack of commitment to a career in medicine (reflected in the low interest category) could negatively influence academic performance in medical school. Some reported that vocationally undecided students were more likely to perform poorly in medical school than their vocationally decided counterparts (Rose & Elton 1971; Lunneborg 1975, 1976).

Students in the “low-interest undecided” group obtained significantly higher scores on measures of the impulse expression scale of the Omnibus Personality Inventory (OPI, Heist & Yonge 1968) and obtained significantly lower scores on measures of personal integration, personal bias and altruism, measured by the OPI compared to students in the “high-interest undecided” students (Walters 1982). In a longitudinal study of predictive validity, Glavin et al. (2009) reported that the MSPI scorers could correctly predict medical students’ future specialty choice 58.1% of the time.

1. The Jefferson Scale of Attitudes Toward Physician-Nurse Collaboration

This 15-item scale was developed to measure orientation toward collaboration and teamwork between physicians and nurses (Hojat & Herman 1985; Hojat et al. 1997a, 1999a). The scale was developed in response to a need for a validated instrument to measure an important aspect of professionalism in medicine, namely teamwork and interprofessional collaboration (Veloski & Hojat 2006).

Psychometric evidence in support of this scale has been reported among American (Hojat et al. 1997a, 1999a; Ward et al. 2008), Mexican (Hojat et al. 2001b), Italian and Israeli samples (Hojat et al. 2003b). This scale has been translated into several languages (e.g. Spanish, Hebrew, Persian/Farsi, Turkish, Japanese, and Chinese) and used by medical and nursing education researchers in different countries (Yildirim et al. 2005; Ardahan et al. 2010; Hansson et al. 2010; El Sayed & Sleem 2011; Onishi et al. 2012). In a review article, this scale was listed among the recommended instruments for measuring physician-nurse collaborative relationships (Daugherty & Larson 2005). Three underlying factors of “shared education and teamwork,” “caring as opposed to curing” and “physician authority” emerged in factor analytic studies of this scale (Hojat et al. 1999a). A significant correlation has been found between scores on this scale and the JSE (Ward et al. 2009). Also, scores of this scale was significantly correlated with scores of a validated measure of attitudes toward physician-pharmacist collaboration (Hojat et al. 2012c).

1. The Jefferson Scale of Physician Lifelong Learning

This is a 16-item instrument developed to measure another element of professionalism in medicine, namely, lifelong learning (Veloski & Hojat 2006). Data are available in support of the psychometrics of this instrument (Hojat et al. 2009, 2010, 2012b). Factor analytic studies show three reliable factors in this instrument: “learning beliefs and motivation,” “attention to learning opportunities” and “skills in seeking information.” (Hojat et al. 2006, 2009, 2010, 2012b). These factors correspond to the key features of lifelong learning often described in the literature, were empirically supported in a study with medical students (Brahmi 2007).

In a large-scale study of 3195 physicians who graduated from Jefferson Medical College, we collected survey data from physicians who were classified into three groups: Full-time clinicians (n = 1127), academic clinicians (n = 1612) and others (n = 456). The reliability coefficients (coefficient alpha and test-retest) of the instrument ranged from 0.72 to 0.86 in these groups of physicians. We found that the academic clinicians scored significantly higher on the lifelong learning scale than the full-time clinicians (Hojat et al. 2009, 2010, 2012b). A Significant association was observed between scores of the lifelong learning scale and medical school class rank in both groups of academic clinicians and full-time clinicians (Hojat et al. 2009, 2010, 2012b). Also, significant correlations were found between scores on this instrument and the criterion measures of reported commitment to lifelong learning, learning motivation, information seeking skills, professional accomplishments, career satisfaction and academic performance in both full-time clinicians and academic clinicians (Hojat et al. 2009, 2010, 2012b).

Professional accomplishments such as publishing in a professional journal, research presentation at national professional meetings, and receiving professional awards and honors were significantly associated with scores of the Jefferson Scale of Physician Lifelong Learning in both groups of physicians (Hojat et al. 2009, 2010, 2012b). No significant gender difference was observed on the scores of lifelong learning. However, we noticed that physicians in internal medicine scored higher than others, and those with combined MD-PhD degrees had higher scores on this scale (Hojat et al. 2009, 2010, 2012b). The Jefferson Scale of Physician Lifelong Learning has also been adapted for administration to medical students with satisfactory psychometric support (Wetzel et al. 2010).

• Although the EPI, MMPI, POMS, TCI, PQA, MBI, MSPI, Physician-Nurse Collaboration and Physician Lifelong Learning are all useful exploratory instruments in medical education research, the last four are more specific than others for physicians-in-training and in-practice.

• The last two instruments (Physician-Nurse Collaboration and Physician Lifelong Learning), plus the JSE which was previously described, are particularly important as measures of oft mentioned elements of professionalism in medicine.

Discussion

Understanding of the link between personality, performance and patient outcomes is important because of its potential implications for admission decisions, career counseling, enhancing personal qualities pertinent to patient care and for predicting academic and clinical success. In this Guide, we provided evidence in support of relationships between personality and academic and professional performance. However, we noticed that the literature on the link between personality and specialty interest is somewhat sketchy with no consistent results. A loose relationship has been reported among a few personality attributes and some specialties. In addition, more variation is reported in personality attributes within specialties than between them (Borges & Savickas 2002). Because common personality attributes are found in physicians in different specialties, it seems that no specific personality attribute uniquely fits any specific specialty (Borges & Savickas 2002). Empathy though may be an exception when broader specialty areas are taken into consideration (e.g. “people-oriented” and “technology/procedure-oriented”). Thus, we will focus on the role of personality in performance of medical students and physicians rather than career interest.

Validity concerns

One noticeable finding on the link between personality and performance is that the reported predictive validity coefficients are often modest in magnitude. Perhaps this is one of the reasons that some have questioned the utility of personality measures in medical education. The modest validity of personality measures in medical education research, though, should not be surprising, given the conceptual and methodological issues involved in studying the relationships between personality measures on the one hand, and criterion measures on the other hand.

There are several plausible explanations for the modest validity of personality measures in medical education research. Among them are the followings.

1. Multidimensionality of personality

Personality is not unidimensional. Different personality researchers have devised different sets of personality constructs, as by its very nature the field requires. Thus, personality is judged through the lens of the researcher and includes all of the biases of the test authors. Accordingly, as we described before, a variety of self-report personality instruments have been used in medical education research for measuring different aspects of personality. The selection of the specific personality attributes from a variety of instruments to be used in medical education research is of paramount importance since those selected attributes must be conceptually relevant to performance in the context of medical education and patient care; and the link must be supported by empirical evidence.

1. Construct dissimilarity

Construct similarities and dissimilarities between personality attributes and criterion measures can contribute to the magnitude of correlations among them. Obviously, a correlation of a larger magnitude is expected between two conceptually relevant variables, such as scores on empathy and ratings of interpersonal skills, than between two conceptually less relevant variables, such as scores on empathy and grades on a multiple choice examination of recalling factual information (Hojat et al. 2002a).

1. Changes in predictor-criterion matching

Poor predictor-criterion matching in medical education research (Hough et al. 1990; Hough 1992) can contribute to the underestimation of validity of personality measures (Lievens et al. 2009). An important issue related to the observed variation in the predictive validity of personality measures during the course of medical education is that the nature of the criterion measures (performance indicators) changes from preclinical to clinical phases of medical education.

The conventional medical school curriculum has been divided into preclinical and clinical phases. Early in medical school, during the preclinical phase, students take courses related to the sciences that are basic to medicine (e.g. anatomy, physiology, biochemistry). These courses are typically assessed by examinations of recalling factual information and declarative knowledge. Later in medical school, the curriculum shifts to the clinical phase, and medical students rotate across various clerkships that often require patient contact. Students’ performance is usually assessed by faculty's ratings of clinical competence, or by standardized or simulated patients in OSCE stations, oral examinations or other methods.

Different sets of ability or skills are often involved in the performance of medical students during preclinical and clinical phases of medical education. For example, the ability to recall, compartmentalize and organize factual information, as well as test-taking skills, under the rubric of “cognitive” abilities, often contribute to success in the preclinical phase. However, communication and interpersonal skills, bedside manner, attitudes, personal qualities or characteristic (referred to as “noncognitive” attributes), often contribute to the assessments of competence in the clinical phase (Haight et al. 2012).

In our own research, we noticed that measures of cognitive abilities contributed more than those of the noncognitive attributes to the prediction of performance in the preclinical phase of medical education. However, a shift toward a higher validity coefficient was observed when personality measures were included to predict clinical competence in the clinical phase of medical education (Hojat et al. 1993).

The oft-reported findings of the increase in the predictive validity of personality measures from the preclinical to clinical phase of medical education can be explained by the trait-activation theory (Lievens et al. 2009). In other words, personality traits that are important for clinical performance manifest themselves during the clinical phase of medical training. This notion is consistent with the view in organizational psychology about different components required for performance in various jobs (Borman & Motowidlo 1993). The theory of trait-activation provides a plausible explanation as to why measures of academic abilities prior to medical school (e.g. grades on examinations of declarative knowledge, scores on entrance examinations such as the MCAT) have shown a declining predictive validity as students progress from preclinical to clinical phases in medical school; while the predictive validity of personality measures increases in the clinical phase of medical school training (Humphreys & Taber 1973; Lin & Humphreys 1977; Lievens et al. 2009).

1. Proximal and distal criterion measures

Based on the aforementioned discussion, one can reasonably expect that personality measures are more likely to predict the “distal” performance (in clinical phase) rather than “proximal” criterion measures (in the preclinical phase) in medical school. Accordingly, the predictive validity and utility of personality measures would be underestimated when using the “proximal” criterion measures, which leads us to another issue; the time interval between recording of predictors and criterion measures.

Personality measures are often administered early in medical school sometimes during the admission process. Measures of performance in the clinical phase of medical education in North America are recorded usually after completion of the second year of medical school. This is a relatively long time interval to examine predictive validity. Specific experiences or events occur during this time period that can confound the predictive validity. The time interval between administering the personality test and recording criterion measures in the personality research reported in the psychological literature is usually a few months and rarely exceeds a year or two (Lievens et al. 2009). However, in this particular situation, distal performance (measures of clinical competence) is more relevant to personality attributes than proximal performance (grades on sciences basic to medicine). Therefore, the confounding effects of the time interval between testing and the criterion measure, subsequent to gaining new experiences, could suppress the true relationships between personality measures and distal performance, adding to the complexity of validity research on personality testing in medical education.

1. Restriction of range

Another reason for the modest validity coefficient of personality measures in medical school is that a correlation coefficient is highly dependent upon the range and variability of the measures. Restriction of range, due to selection and attrition, can shrink validity coefficients. Therefore, inferences drawn from correlation coefficients may be misleading; because all things being equal, the more restricted the range of scores, the lower the validity coefficient. The true relationships between correlated measures cannot be captured when only those who successfully completed their medical training are included in the final statistical analyses; thus, eliminating those in the bottom tail of the score distribution who could not successfully completer medical school. The resulting “ceiling effect” would lead to a lower validity coefficient (Gough et al. 1963).

1. Nonlinear relationships

A nonlinear relationship between some measures of personality and some indicators of academic attainment can lead to a decrease in predictive validity. When the nature of a relationship is nonlinear, the magnitude of the Pearson correlation coefficient will become negligible. A curvilinear relationship (i.e. inverted U shape) has been reported between anxiety and performance in medical school (Shen & Comrey 1997; Ferguson et al. 2002), which is consistent with the arousal theory (Yerkes & Dodson 1908) suggesting that individuals perform better at their optimal arousal level, below and above which performance is likely to fall. In almost all validity studies on personality measures in medical education research, the linearity assumption has not been tested (Shen & Comrey 1997).

1. Multicollinearity

The genuine relationship between predictors (personality measures) and criterion measures (performance indicators) cannot be captured when predictors are themselves highly correlated. This phenomenon, known as multicollinearity, contributes to underestimating the predictive validity of personality attributes. The modest contribution of some personality measures in multiple regression analyses could be an artifact of multicollinearity, which must be taken into consideration when assessing the validity of personality measures.

1. Volunteer bias

Volunteer bias in research can also confound validity coefficients. Medical students’ willingness to voluntarily participate in medical education research varies by gender, ethnicity and academic achievement (Callahan et al. 2007). We have shown that research volunteers in medical school, on average, perform better during and after medical school, compared to their unwilling classmates (Callahan et al. 2007). This finding suggests that volunteer participants in medical education research cannot fairly represent the entire population of medical students. This leads to the self-section bias and raises question about the validity of research in medical education when participation is voluntary. This issue is exacerbated by the requirement of voluntary participation in human subject research for granting approval by most universities’ research ethics committees (e.g. the institutional review board, IRB). Nevertheless, high participation rates and evidence of the representativeness of the volunteer sample in relation to the population being studied can provide support for the validity.

1. Variation in methods of assessment

Variation in methods of assessment is another factor that can contribute to the modest validity coefficients between personality and criterion measures in medical school. Self-report personality measures rely mostly on Likert-type scales. Criterion measures of cognitive performance in medical school are often assessed by multiple choice or true-false formats (in the preclinical phase) and by observational methods, ratings of clinical competence by the faculty or assessments by simulated patients in the clinical phase of medical education. In their discussion of the concept of a multitrait-multimethod matrix, and convergent and discriminant validation, Campbell and Fiske (1959) reported that similarity and dissimilarity in the methods of assessment can contribute to the magnitude of the relationships among the measures. Thus, variation due to the methods of assessment can obscure the outcomes.

1. Gender effects

Gender is another variable that can confound predictive validity assessments (Hojat et al. 1999b). Gender was not an important factor in early studies of medical education when medical students and physicians were predominantly male (Zeldow & Daugherty 1991). The influx of women to medicine in the later part of the past century and gender differences observed in personality, performance, career motivation, and specialty preference suggest that gender must be considered as a contributing variable in validity studies in medical education research.

It has been reported that women on average fall behind their male counterparts during the preclinical phase of medical education, but they usually catch up to or sometimes surpass men on some measures of clinical competence during the clinical phase of medical education (Hojat et al. 1997b; Halpern et al. 1998; Ferguson et al. 2002). In addition, female physicians are rated higher on personal qualities such as helpfulness, human relationships, expressiveness, intrinsic career motivation, family responsibility and job security; while men obtained higher marks on personality features such as independence, decisiveness, self-confidence, extrinsic career motivation and orientation toward income and prestige (Buddedberg-Fischer et al. 2003).

Gender differences have also been observed in career choices (Hojat et al. 1999b). For example, historically women have been more likely to choose “people-oriented” specialties that require intensive patient contact, while men have been more likely to prefer “technology-oriented” specialties that require performing complicated procedures (Buddedberg-Fischer et al. 2003; Hojat 2007). A detailed discussion of whether the underlying reasons for gender differences are the results of social learning (Bandura 1986), or hard-wired gender specific inclination (Halpern 1992, 1997; Valian 1999) is beyond the intended scope of this Guide. Regardless of the reasons for gender differences, it is important to examine and control gender effects for a fair assessment of predictive validity of personality measures in medical education research.

1. Race and ethnicity effects

In addition to gender, race and ethnicity can contribute to the validity of personality measures. Given the changing demographic and ethnic composition of medical students and physicians, particularly in the United States, and the emphasis placed on ethnic diversity in the medical workforce (AAMC 2004; Nickens et al. 1994), it is important to control for ethnic status as a possible intervening variable in the validity studies. Our research findings suggest that ethnicity contributes significantly to the assessment of cognitive (Rosenfeld et al. 1992; Veloski et al. 2000) and noncognitive measures (Berg et al. 2011a) in medical education research.

A number of studies also confirm the role of ethnicity in medical school admissions and academic attainment (Rosenfeld et al. 1992; Esmail et al. 1995; McManus et al. 1995; Ready 1995; Crump et al. 1999; Tekian 1997; Girotti 1999; Hardy 1999; Lumb & Vail 2000; Giordani et al. 2001; Ferguson et al. 2002). These studies suggest that it is important to consider ethnic status as a possible confounding variable for a fair assessment of the predictive validity of personality measures in medical education.

Reasons for optimism

Despite all of the aforementioned conceptual and methodological limitations, the findings of the modest predictive validity of personality measures in medical education and practice are still encouraging for the following reasons:

First, a meta-analysis of more than 33 000 studies revealed that the average effect size in personality research is only 0.21, a modest effect size by any standard (Richard et al. 2003). In other meta-analytic studies of the validity of personality tests, the overall validity coefficient has been reported to be in the 0.20 s (Ghiselli & Barthol 1953; Schmitt et al. 1984).

Second, it has been reported that the average predictive validity coefficient in undergraduate medical education is 0.30 (Ferguson et al. 2002), which means that only 9% (r²= 0.30²= 0.09) of the variance of the criterion measure could be accounted for by the predictors. Third, the average predictive validity coefficient in graduate medical education is reported to be only 0.14 (Ferguson et al. 2002), meaning that only 2% of the variance of the criterion measure could be accounted for by the predictors.

Fourth, in our analysis of correlations among performance variables before and after medical school by using data collected from five medical schools, we found modest correlations in the magnitude of 0.30–0.40, which are typical in medical education research (Gonnella et al. 1993). These findings provide some grounds for optimism regarding the utility of personality assessments in medical education.

Considering the above-mentioned reasons, the modest predictive validity coefficients of personality measures in relation to performance in undergraduate and graduate medical education should not be pessimistically considered as disappointing.

Social desirability response bias

In addition to the issue of modest validity, another reason for hesitation to use personality instruments in the assessment of physicians-in-training and in-practice is the issue of social desirability response bias that can also be relevant to the validity of personality tests. Constructing socially neutral items that measure personality is difficult and raises questions about the face and content validity of such items. The degree to which socially desirable responses have a confounding effect on test scores can be a function of the test taker's perception of the purpose of personality testing. For example, when testing is used for research purposes (non-penalizing situation) respondents are less inclined to depict themselves as socially desirable than when the test is used to screen applicants for employment or admission.

There are very few studies on the effects of “faking” in personality test outcomes (Hough et al. 1990). We conducted an empirical study to examine the possible effect of socially desirable responses (Hojat et al. 2005b) in which we administered the JSE and other personality tests, including the ZKPQ to 422 first-year medical students. The ZKPQ includes an “Infrequency” subscale that was developed to detect intentionally false responses by identifying respondents with an invalid pattern of responses (Zuckerman 2002). Scores on this subscale can be regarded as indicators of social desirability response bias. Attempts to give socially desirable responses were determined by a cutoff score of 3, which the test's authors suggested would identify respondents whose patterns of responses were of questionable validity. An examination of the distribution of scores on this subscale indicated that 4.9% of the respondents attempted to give false “good impression responses” or to respond carelessly without regard for the truth (Hojat et al. 2005b).

In an unpublished study with medical students, we used two approaches to examine the possible effects of social desirability response bias on the outcomes of our research on the JSE. First, analyses of data regarding the relationship between scores on the JSE and on personality measures clearly demonstrated that the general research outcomes remain virtually unchanged whether or not respondents who responded carelessly to the instrument were included (determined by scores on the “Infrequency” subscale that were above the cutoff point of 3). This finding was expected because of the small proportion of respondents who scored above the cutoff point.

Second, we used the analysis of covariance (ANCOVA) method to control the effect of giving false responses on the research outcomes by using the “infrequency” score as a covariate. Again, we noted no substantial change in the general pattern of results. These findings suggest that social desirability response bias did not distort the validity of the JSP score.

Our findings were consistent with the results of an earlier study by Matthews and colleagues (1981), who reported that their derived index of empathy was not affected by social desirability response bias or by scores on a “good impression” scale. Two other studies reported no significant correlations between empathy scores obtained on the Emotional Empathy Scale and social desirability response bias (Mehrabian & Epstein 1972; Mehrabian & O’Reilly 1980).

Despite these findings, the confounding effects of giving false responses and attempting to present a socially acceptable image in penalizing testing situations (e.g. in admission or employment) need to be addressed further in future studies. One approach that may minimize the effect of social desirability response bias is reminding the respondents to reply truthfully, since their intentionally false responses can be detected by a scale embedded in the test which will invalidate the test results. One of the available measures (e.g. Infrequency subscale from the ZKPQ) could be used for that purpose. For example, pattern of endorsement of items such as “I never met a person I did not like” or “I have always told the truth” (from ZKPQ) can give a clue as to whether a respondent is honest in completing the test.

Are personality attributes amenable to change?

The lingering doubts that personality attributes are not amenable to change may contribute to a hesitation to assess personality and implement programs to enhance personality attributes in physicians-in-training. The question of whether personality attributes are amenable to change has roots in the old nature versus nurture debate on whether personality is inherited and inborn or learned over the course of social and emotional development. Proponents of nature over nurture place great emphasis on the notion that genetic predisposition has an undeniable role in the development of human behavior. Some developments in the Human Genome Project have provided more fuel in support for that argument (Collins 1999). However, proponents of nurture over nature use Watsonian classical conditioning (Watson 1924), Skinnerian operant conditioning (Skinner 1938) and Bandura's (1986) social learning theory as evidence that personality can be molded by principles of behavior modification, personal experiences, social learning and educational interventions; thus, they conclude that environment and learning could have a prominent role in the development of personality.

However, most scholars today are of the opinion that it is the interaction of nature and nurture that contributes to the development of personality. Human beings are born with some potential for “engageability,” which is triggered and developed to a certain degree by environmental, social, experiential, and educational factors (Neubauer & Neubauer 1990). Abundant research evidence has been accumulated in support of the proposition that social and educational environments play an important role in the development of personality including the shaping of interpersonal skills and caring attitudes (Hojat 2007). There are empirical studies showing that some personality attributes can be changed as a result of positive or negative educational experiences in medical school. This notion is supported by the findings on the erosion and enhancement of empathy during medical school.

Erosion of empathy during medical education

A number of studies have shown that during the course of health professions education, a person's capacity for empathy can undergo positive, negative, or no change (see Hojat 2007, pp. 181–184 for a review). Some studies have reported a significant decline in the scores of the JSE during the clinical phase of medical education (Hojat et al. 2004b, 2009, Chen et al. 2007; Hojat 2007, Newton et al. 2008). In our more recent longitudinal study of four classes of medical students at Jefferson Medical College (Hojat et al. 2009), a significant decline in scores of the JSE was observed in third-year medical students when the curriculum shifts to clinical training and patient care, and the decline did not rebound during the rest of medical school training.

Such a decline on the scores of the JSE was also noticed in another study with internal medicine residents as they progressed through residency training (Mangione et al. 2002). However, the decline in empathy in this study did not reach the conventional level of statistical significance. The findings of erosion of empathy during medical education are consistent with those reported by Whittemore and colleagues (1985), Bellini and colleagues (2002) and Bellini and Shea (2005). A similar decline in empathy scores also was observed among nursing students who had more exposure to patient care than others (Ward et al. 2012).

Consistent with the above-mentioned findings, an early study by Becker and Geer (1958) reported that medical students become somewhat cynical during the course of medical education. By the third year of medical school, according to Becker and Geer (1958), the students realized that they were no longer motivated by an idealized view of medicine, leading to a hedonistic shift, shown also by Whittemore and colleagues (1985), and by Feudtner and colleagues (1994). In a study by Zeldow and colleagues (1987), a modest but “unmistakable” shift (according to the study authors) toward hedonism between the freshman and junior year of medical school was observed in two cohorts of students. According to the investigators, these changes perhaps reflect a less idealized view of the self and a less sentimental view of the medical profession (Zeldow et al. 1987).

In explaining changes in empathy, medical students reported a lack of positive role models, lack of time to form an empathic relationship with patients, excessive workloads, disrespectful and overly demanding patients, over-reliance on computer-based diagnostic and therapeutic technology, and a market-driven health care system as factors that contribute to erosion of empathy (Hojat et al. 2009) and the escalation of cynicism (Hojat 2007).

Despite the overwhelming evidence of the erosion of empathy during medical education, skeptics have raised concern about the significance of these findings in undergraduate and graduate medical education (Colliver et al. 2010), but such critics have not been left unchallenged by our team (Hojat et al. 2010) and other empathy researchers in medical and dental education (Newton 2010; Sherman & Cramer 2010).

Findings of erosion of empathy in undergraduate (Hojat et al. 2004b, 2009; Chen et al. 2007) and graduate medical education (Bellini et al. 2002; Bellini & Shea 2005; Mangione et al. 2002) suggest that if a personality attribute, such as empathy, can decline by negative educational experiences, it can also be enhanced by positive educational experiences and targeted interventions.

Enhancement of empathy in medical education

The link between empathy, clinical competence and patient outcomes (Hojat et al. 2011a; Del Canale et al. 2012) makes it critical that we nurture empathy in physicians-in-training and physicians-in-practice. The cultivation of empathy in undergraduate medical education has been listed among learning objectives endorsed by the Association of American Medical Colleges (AAMC 2008). Also, the ABIM recommended that humanistic qualities such as empathy be instilled and assessed as an essential part of graduate medical education (ABIM 1983).

The reported decline in empathy during undergraduate and graduate medical education coupled with the findings that empathy should be viewed as a component of physician competence that has implications for patient outcomes, beg for the development of targeted educational programs to sustain and enhance empathy among physicians-in-training and physicians-in-practice. Research has shown that empathy must be considered as an important component of a health care provider's overall competence and is a significant factor in optimal patient outcomes (Hojat et al. 2009; Del Canale et al. 2012).

These findings suggest that leaders and faculty at all levels of health profession education (e.g. undergraduate, graduate and continuing education) must implement targeted educational remedies to enhance and sustain empathy in all students and trainees, and assess the educational outcomes. Research shows that empathy can be enhanced with targeted educational programs. For example, the following 10 approaches have been described (Hojat 2009) for improving empathy among health professions students and practitioners: Improving interpersonal skills; analyzing audio or video taped encounters with patients; being exposed to role models; role-playing (e.g. aging games); shadowing a patient (e.g. patient navigator); experiencing hospitalization (e.g. getting admitted to a hospital by presenting fabricated symptoms); studying literature and the arts; improving narrative skills; watching theatrical performances and engaging in small group discussion about difficult patients, e.g. Balint (1957) method.

In a study with pharmacy students, Chen et al. (2008) reported enhancement in the JSE scores among students who participated in an empathy training program. Also, Fernandez-Olano and colleagues (2008) reported a significant increase in the JSE among Spanish medical students and residents who participated in a communication skills training program. However, Cataldo and colleagues (2005) found no significant increase in the JSE scores as a result of Balint training among residents in a family medicine residency program. In a qualitative and quantitative study with 40 staff physicians at the Cleveland Clinic, it was found that a faculty development program using guided narrative writing could influence, to a limited extent, the empathy of practicing physicians (Misra-Herbert et al. 2012).

In one study, it was found that watching a short theatrical play (depicting problems facing elderly patients) could significantly increase scores of the JSE in medical and pharmacy students (Van Winkle et al. 2012). The increase in empathy scores, however, did not last for a long time. In another study, it was reported that shadowing patients by first-year emergency medicine residents in the emergency room for a short period of time prevented decline in empathy. The empathy scores of the control group who followed their routine training schedules declined during the study period (Forstater et al. 2011). In a study with primary care physicians, it was found that participation in a course of mindful meditation and self-awareness could significantly increase scores of the JSE (Krasner et al. 2009).

In our latest study to examine if enhanced empathy can be sustained, we showed medical students video clips of patient encounters selected from three movies. Students were encouraged to present their views on positive and negative episodes of the encounters in the video clips, and discussed the feedback. Enhanced empathy was observed among all students who watched and discussed the video clips of patient encounters (as compared to a control group who did not). A few weeks later, those who watched and discussed the video clips were divided into two groups. One group participated in a lecture and discussion session on the importance of empathy in medical education and patient care. The other group watched a documentary movie. Enhanced empathy could be sustained in the first group, but not in the second group. It was concluded that the enhanced empathy could be sustainable when reinforced by additional intervention, but will dissipate without such reinforcement (Hojat et al. in press). We are exploring approaches not only to enhance but also sustain empathy by additional reinforcement during the course of medical education.

The aforementioned findings suggest that targeted educational programs can significantly improve empathy in the context of medical education and patient care. In all of our experiments, we noticed significant variation in the magnitude of changes among participants. Not everyone could equally be influenced by negative experiences (Hojat et al. 2003a, 2009) or equally benefit from the educational programs (Van Winke et al. 2012), which may suggest that constitutional factors provide a window of opportunity for changes, but the size of that window varies among participants. Personality attributes are indeed unevenly distributed in the population. Some people possess some of the personality attributes (positive or negative) in abundance; some in meager amounts, depending upon many factors including genetic predisposition, early relationships experiences, family and social environment, and of course learning and educational factors. Some people seem to have a larger window than others for personality changes; not all seeds sowed in a garden bed grow at the same rate.

Based on the aforementioned findings, we propose that some personality attributes are amenable to change by targeted educational programs, but the degree of change depends on constitutional factors, early attachment experiences, exposure to positive role models and social and educational factors.

Conclusions

Now, that we have come this far in our journey to explore the link between personality and medical education outcomes, physician performance and patient outcomes, the logical questions would be: What have we learned? What guidance can we provide to identify desirable personality attributes? What are the implications?

Medical training is a long and expensive endeavor. The emotional and financial costs are high for those who are admitted to medical school but fail to graduate, and for those who later recognize that their personality is no match for the medical profession (Green et al. 1991). For those who embark on a journey to become a physician but never arrive at their destination, an incomplete medical education cannot appropriately equip them to pursue other ventures (Green et al. 1991). To avoid frustration, and to contain human and financial costs, it is important to select the right candidate with appropriate intellectual abilities and personality for medical training and practice of medicine. Because of the aforementioned issues, we maintain that selection of qualified candidates for medical training is a social responsibility of academic medical centers (Gonnella & Hojat 2001; 2012).

The assessment of personality for selection of medical students and career counseling was recommended by pioneers in personality assessment (e.g. Gough 1967; Gough & Hall 1967). Despite this recommendation, there seems to be a hesitation among researchers and leaders in medical education to use personality instruments for selection purposes. Perhaps one of the reasons for this hesitation is a lack of consensus among personality scholars in medical education about what personality measures to use, and why to use them!

Conceptual relevancy and empirical evidence

The crucial question raised by many is how can we identify the pertinent personality attributes for medical training and practice? We believe that at least two factors could be considered. First, the selected attribute must be conceptually relevant to components of physician performance, and theoretically linked to optimal patient outcomes. A lack of clear conceptual relevancy between a selected personality attribute and a defined outcome measure will undermine the potential value of personality measures in medical education and make it totally unacceptable to society. Second, in addition to conceptual or theoretical relevancy, selection of desirable personality attributes must be evidence-based, meaning that convincing empirical support must be available to back the significant associations between selected personality measures and indicators of performance.

For example, it seems more reasonable to conceptualize that a measure of empathy in patient care would be theoretically relevant to physician clinical competence (Hojat et al. 2002a), and to patient outcomes (Hojat et al. 2011a, Del Canale et al. 2012) than a measure of androgyny (Bem 1974, Yarnold 1993). Therefore, our first task is to choose a manageable number of personality measures that meet the conceptual relevancy, and empirical support requirements. Ease of administration, time needed to complete, and cost-benefit factors can also be taken into consideration.

Selected personality measures

Based on our reading of the relevant theoretical and empirical literature, and on our own research findings from the Jefferson Longitudinal Study of Medical Education Outcomes (Hojat et al. 1996b; Gonnella et al. 2011), at the present time we believe there are two personality measures that can satisfy the two aforementioned conditions more consistently than others.

The conscientiousness factor

At the conceptual level, it seems reasonable to concur that personal qualities such as responsibility, competence, dutifulness, achievement striving, self-discipline, deliberation and order are relevant to a physician performing his or her roles as a clinician, educator and manager (Figure 1). These are all among the facets of the Conscientiousness factor of the big FFM of personality (Costa & McCrae 1992). In addition to the conceptual relevancy, there is ample empirical evidence in support of the validity of this personality attribute in the performance of physicians-in-training and in-practice.

In addition to the findings, we previously reported in describing the FFM of personality, the Conscientiousness factor was found to be a significant predictor of professional success not only in medicine, but also in a variety of other occupational settings (Tett et al. 1991). The Conscientiousness factor is not only a positive predictor of competence in the clinical phase of medical training, but also a significant predictor of performance in the preclinical phase of medical education, even when statistical control was made for previous academic performance (Ferguson et al. 2002). In their meta-analytic research, Barrick and Mount (1991) concluded that conscientiousness is a universal predictor of job performance. The universality of the Conscientiousness factor in academic and professional success has been confirmed in a variety of disciplines and in different academic settings (Hurtz & Donovan 2000; Noftle & Robins 2007; Poropat 2009; Haight et al. 2012), and in predicting healthy behavior (Bogg & Roberts 2004). Thus, conscientiousness is the first personality attribute we selected as being relevant to medical education and the practice of medicine. The scores of the Conscientiousness factor of the NEO PR-I (Costa & McCrae 1992) can well serve as a psychometrically sound measure of this personality attribute.

Empathy in patient care

There is another set of personal qualities such as communication skills, understanding, ethnic and cultural sensitivity, perspective taking ability, teamwork, collaboration and personal and professional ethics that seem desirable for the practice of medicine and in improving clinical outcomes. These are all ingredients of empathy as we conceptualize it (Hojat 2007, 2009; Hojat et al. 2009). We have shown that medical students’ empathy is significantly linked to global ratings of clinical competence (Hojat et al. 2002a). We also showed that medical students’ self-report empathy scores (measured by the JSE) were predictive of ratings of empathic behavior and interpersonal skills (given by the directors of postgraduate medical education programs) about three years later (Hojat et al. 2005a).

Furthermore, and more importantly, we have shown that scores on physician empathy (measured by the JSE) were significantly associated with tangible clinical outcomes in diabetic patients (Hojat et al. 2011a; Del Canale et al. 2012). These findings provide convincing evidence to confirm that empathy is an important component of overall competence for medical students and physicians, and a significant factor in optimal patient outcomes, suggesting that empathy must be placed in the realm of evidence-based medicine. Therefore, we selected empathy as the second personality attribute relevant to the clinical performance of medical students and physicians as well as optimal patient outcomes. The JSE can serve as a psychometrically sound instrument for measuring this attribute. These two personality attributes (conscientiousness and empathy) are among features described by Truax and Carkuff (1967) as the “helping personality” characteristics.

Other personality attributes

It can be argued that it does not seem justifiable to choose only two attributes to represent a complex personality profile for predicting academic and career success for physicians-in-training and in-practice. We concur with this. Being a “good enough” physician, requires more than two personality attributes.

We are not fully satisfied with the two-attribute personality profile. There may be other pertinent personal characteristics that deserve more attention. As an example, in our own research, we noticed that students’ retrospective report of their perception of early relationships with their parents, especially the mother, was a significant predictor of ratings of clinical competence and interpersonal skills given by the directors of postgraduate medical education programs to physician residents (Hojat et al. 1996a). We also found that such positive perceptions of the early relationships with the mother were associated with a positive personality profile (e.g. lower loneliness, lower depression, lower anxiety, higher self-esteem) (Hojat 1998) and also with more positive appraisals of stressful life events, as well as success in medical school (Hojat et al. 2003a). Medical students’ reports of maternal unavailability in childhood were associated with higher scores on the intensity and chronicity of loneliness experiences, more depression, lower self-esteem and more negative appraisal of stressful life events (Hojat 1998; Hojat et al. 2005b).

These findings are in agreement with some human social-emotional development theories, including John Bowlby's attachment theory (Bowlby 1969). Of course more empirical evidence (preferably in longitudinal studies) to show that the quality of early relationships with a primary caregiver can significantly predict clinical performance of physicians-in-training and in-practice would add to our confidence to consider the early relationship information as an important attribute in the personality assessments. Considering the available evidence, however, at the present time, we suggest that the two selected personality attributes be used as potential indicators of success in medical education and the practice of medicine until further research suggests additional measures.

There are other personal qualities that seem conceptually relevant to performance in the context of medical education and patient care. For example, indicators of emotional intelligence, tolerance of ambiguity and emotional regulations seem desirable for optimal clinical performance and patient care, but more convincing empirical evidence is needed to connect these features of personality directly to measures of success in medical school and in the practice of medicine. Empirical confirmation of these links should be placed on the agenda of future research.

It is also interesting to contemplate the idea that similar to the “g” factor in intellectual abilities, there might be a general, or a “g” factor in the personality of competent medical students and physicians. It would be a break-through in personality research in medical education if such a “g” factor was discovered for predicting success among physicians-in-training and in-practice. The possibility of the existence of such a general factor should be examined in future medical education research.

Implications

Obviously, the assessment of pertinent personality attributes has implications for selection, career counseling and for evaluation of professional development and medical education outcomes. The aim of medical school admissions is to select applicants who will perform well in undergraduate and graduate medical education, and who become good physicians. This implies excluding those who might bring the profession into disgrace (Lumsden et al. 2005). Identifying applicants who are likely to become competent physicians is a crucial responsibility of academic medical centers (Haslam 2007; Gonnella & Hojat 2012). The assessment of personality is one step toward achieving this goal. Based on our discussion of the reported findings in this Guide, we suggest that attention be given to applicants’ scores on measures of conscientiousness and empathy at undergraduate and graduate medical education levels to identify those applicants with a more “suitable” personality profile for medical practice. At the least, these measures could be used as potential “tie breakers” in the admission decisions for those applicants with relatively similar profiles on other admission requirements.

The selection of candidates with suitable personalities cannot be appropriately accomplished by admission interviews, letters of recommendation, personal statements, letters of intent and personal essays alone. We concur with Haque and Waytz (2012) that one appropriate approach for the assessment of the personality of physicians-in-training is to administer psychometrically sound instruments, which we suggest must be conceptually relevant and empirically linked to educational and clinical outcomes.

However, in the admission process, serious legal and socio-political concern exists as to whether society is prepared to accept the use of personality assessments for excluding an applicant from medical education and denying the opportunity to become a physician. It seems doubtful that medical school deans and admissions authorities would be willing to accept the legalities of using personality tools for admission decisions purposes. Society, it is argued, may buy into the use of personality assessments for training purposes and for monitoring personal development, but not for admission purposes. Society sometimes may speak louder than empirical evidence, and a bold action is needed to break free from deep-rooted ideas, unless better alternatives become available. Scientific ideas are independent from popular perceptions and democratic choices. Lingering doubts and hesitation to take bold action in utilizing personality assessments in the selection and professional development of trainees in medicine, result in a futile and never-ending search for additional evidence which would be counterproductive for medical education and the practice of medicine; because, waiting to certainty is waiting for eternity.

At a minimum, consideration should be given to the use of personality tests in the development, implementation, and assessment of targeted educational programs at the undergraduate, graduate and continuing medical education levels to enhance those personality attributes that are pertinent to success in medical education and the practice of medicine.

Final remarks

Now, it is time to reflect on what we have learned from our journey to the terrain of personality in medical education research. From what we have seen, we should be able to provide responses to the five questions raised at the outset of our journey.

First, in response to the importance of personality in the process and outcomes of medical education, we have shown in this Guide that personality plays a significant role in the performance of physicians-in-training, and in-practice. While we may not be as certain about the role of personality in specialty choice and selection, a large volume of empirical studies provides convincing evidence, which adds to our confidence, on the importance of personality attributes in predicting performance in medical school and the practice of medicine.

Second, in response to identifying a manageable number of personality attributes most relevant to medical education outcomes, after our review of the literature we have selected the two personality attributes of “conscientiousness” and “empathy” because of their conceptual relevance to physician competence as well as support from a number of empirical studies. Of course, there might be as well additional personality attributes that can serve a similar purpose. Further research is needed to provide convincing and consistent evidence about the validity and utility of such additional personality measures.

Third, in response to the notion of redundancy or overlapping cognitive and noncognitive aspects of performance, we have shown that the two constructs of academic aptitudes and personality are separate entities (Hojat et al. 1988) that uniquely contribute to prediction of performance in undergraduate (Hojat et al. 1988) and graduate (Hojat et al. 1996a) medical education outcomes. In other words, they are complementary, not redundant.

Fourth, in response to the idea of the amenability of personality attributes to change, as an example we have described approaches that enhance empathy in undergraduate and graduate medical education. We reported that even short workshops can influence empathy of trainees in undergraduate (Hojat et al. 2012a) and graduate medical education levels (Forstater et al. 2011), and that the enhanced empathy can be sustained by additional educational reinforcements (Hojat et al. 2012a).

Fifth, in response to the issue of the possibility of “faking” in personality testing, we reported some studies that suggest social desirability response bias may not substantially distort the results when the test is administered in a “non-penalizing” situation. However, the production of an intentional “good impression” in responses is always a possibility; proper instructions and examinations of response pattern on specific items to detect socially desirable responses can be helpful in minimizing “faking” and to identify those with invalid responses.

To recapitulate, in the admission of applicants to medical school, and in the assessment of candidates for postgraduate medical education, as well as in the education of physicians-in-training, and professional development of physicians-in-practice, we should not focus only on academic abilities and intellectual capacities. Because of the contribution of personality to all aspects – personal and professional – of human behavior, we strongly believe that medical education and medicine can profoundly benefit from seriously considering the potential of pertinent personality attributes in the selection and education of intellectually qualified applicants to undergraduate and graduate medical education as well as in professional development of physicians to better perform their roles as clinicians, educators and resource managers.

Medicine which was considered by the public as one of the most highly respected professions of all, is losing ground (Thomas 1985) partly because of the failure of some physicians to preserve their altruistic image (Schlesinger 2002). At the turn of 20th century, George Bernard Shaw equated the image of the medical profession to the faith in God by declaring that “We have not lost faith, but we have transferred if from God to the medical profession.” However, in the past few decades, profound changes in medical education and the health care services, an imbalance in teaching the science and the art of medicine, unduly monetary considerations to contain cost, increasing commercialization of medical care, health insurance policies formulated by nonmedical administrators, the emergence of “defensive” medicine, and loss of the human presence in caring for the patients by its replacement with computerized diagnostic and therapeutic technology have transformed the image of physicians, and eroded the public's trust in medicine (Schlesinger 2002). Perhaps medicine can regain some of its well-deserved reputation, and physicians can reclaim their altruistic image by greater attention to the role of personality in the selection, education, practice and professional development of physicians.

Acknowledgements

We would like to thank Clara Callahan, MD, for her critical review of this Guide and her constructive comments; Jon Veloski, MS, for his review and valuable feedback; Dorissa Bolinski for her editorial assistance and David Evans for the management of the references.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the article.