Evaluation & the Health Professions / September 2004
Kim et al. / EFFECTS OF PHYSICIAN EMP ATHY
THE EFFECTS OF
PHYSICIAN EMPATHY ON
SUNG SOO KIM
Michigan State University
MARK V. JOHNSTON
Kessler Medical Rehabilitation Research
and Education Corporation
EVALUATION & THE HEALTH PROFESSIONS, Vol. 27 No. 3, September 2004 237-251
© 2004 Sage Publications
The present study attempted to develop new
scales of patient-perceived, empathy-related
constructs and to test a model of the relation-
ships of physician empathy and related con-
Five hundred fifty outpatients at a large uni-
versity hospital in Korea were interviewed
with the questionnaire. The data were ana-
lyzed using structural equation modeling.
Patient-perceived physician empathy signifi-
cantly influenced patient satisfaction and
mation exchange, perceived expertise, inter-
personal trust, and partnership. Improving
physician empathic communication skills
should increase patient satisfaction and com-
pliance. Health providers who wish to
improve patient satisfaction and compliance
should first identify components of their
empathic communication needing improve-
Keywords: patient; physician; relationship;
partnership; trust; empathy
AUTHORS’ NOTE: Requests for reprints
should be sent to Sung Soo Kim, KMRREC,
1199 Pleasant Valley Way, West Orange, NJ
For example, such outcomes of health care as patient’s overall satis-
faction with health care services and compliance with medical regi-
the patient and the provider and are particularly related to the physi-
cian’s empathic communicative behaviors (DiMatteo & Hays, 1980;
DiMatteo, Taranta, Friedman, & Prince, 1980; Hojat et al., 2002;
Linn & Wilson, 1980; Olson, 1995; Pelz, 1982). In recent years, a
ical education along with diagnostic expertise and technical skills
recommended that physicians learn to become compassionate and
(Association of American Medical Colleges, 1998; Meryn, 1998).
However, the widely recognized importance of physician empathy
does not seem to be well reflected in modern medicine yet (Mercer,
Watt, & Reilly, 2001). Furthermore, the current literature on physi-
cians’ empathic communication has several limitations. First, little
empirical research has been done to develop and test a measure of
patient-perceived physician empathy and its effects on patient out-
comes. A measure of physician empathy that reflects actual feelings
of patients is needed to provide direct information on effective empa-
thy. A majority of the existing measures of empathy tend to view
empathy as either an affective or cognitive construct, and they tend to
employ subject ratings of their own empathic concerns. For example,
Hogan’s (1969) Empathy Scale took empathy to be a cognitive pro-
primarily affective process.
A few other empathy instruments have also been developed. For
empathy, and Davis’(1983) Interpersonal Reactivity Index measures
a person’s empathy. However, none of these instruments were in-
Hojat et al.’s (2001) Jefferson Scale of Physician Empathy, was in-
troduced to specifically measure physician empathy. But, like other
any investigators have provided ample empirical evidence for
238Evaluation & the Health Professions / September 2004
likely to affect patient outcomes. Empathy must be perceived and felt
by patients to be effective (Squier, 1990). One study (Free, Green,
Grace, Chernus, & Whitman, 1985) compared the effects of empathy
as rated by three different groups (patients, therapists, and clinical
supervisors) and found that only patient-perceived empathy was sig-
nificantly related with patient outcomes. Similarly, another study
(Kurtz, 1972) found that client-perceived empathy, but not therapist
self-reported empathy, was highly associated with therapy outcomes.
Therefore, it would appear best to use a patient-perceived empathy
scale to measure physician empathy.
and affective) view of physician empathy. Historically, a controversy
exists over whether empathy is an affective or cognitive construct or
both. The most prevailing current view among empathy theorists and
researchers holds that empathy entails both affective and cognitive
elements and that only the clear recognition of both the affective and
cognitive aspects can improve our understanding of empathy (Davis,
Third, whereas empirical research on physician empathy has been
sparse in Western countries, research on patient-physician communi-
One definition of the patient’s perception of physician empathy is
This understanding and acceptance consists in turn of two compo-
nents: cognitive and affective. The cognitive aspect of physician em-
pathy isdefined asthe physician’sability to accuratelyapprehend the
mental state of his or her patients (the ability to take another person’s
the patients. The affective aspect of physician empathy is defined as
the physician’s ability to respond to and improve his or her patients’
mation exchange, perception of physician expertise, interpersonal
Kim et al. / EFFECTS OF PHYSICIAN EMPATHY239
trust, and partnership with respect to patient satisfaction and compli-
ance based on data collected in Korea.
Based on the literature review (Kim, 2000), we hypothesized two
separate causal chains: one from (a) cognitive empathy to (b) cogni-
the other from (a) affective empathy to (b) partnership to (c) interper-
sonal trust. In addition, we hypothesized that cognitive information
exchange, perception of physician expertise, partnership, and inter-
personal trust would all directly influence both patient satisfaction
hypothesized that highly empathic physicians (both highly cognitive
and affective) would have more satisfied and compliant patients. We
further predicted that cognitive information exchange and percep-
tion of physician expertise should have a stronger effect on compli-
ance than on satisfaction, whereas partnership and interpersonal trust
should have a stronger effect on satisfaction than on compliance.
eted university hospital in Pusan, Korea, in 1999. The patients at this
site represent a wide variety of illnesses, physician-patient relation-
ships, and types and lengths of medical examination.
The six interviewers recruited the subjects in the pharmacy lobby,
where people waited to pick up their prescriptions after seeing the
doctors. The interviewers promised that the patient would remain
anonymous and that physicians would not see individual responses.
After 2 weeks of the survey, 550 questionnaires were collected out of
approximately 800 outpatients.
240 Evaluation & the Health Professions / September 2004
Eight latent variables were measured. Six of them addressed the
patient’s perception of physician communication skills: (a) cognitive
empathy, (b) affective empathy, (c) cognitive information exchange,
(d) partnership, (e) physician expertise, and (f) interpersonal trust.
The final two variables involved patient satisfaction and patient com-
pliance. Each latent variable had between two and seven observed
indicators, each of which was measured via a five-point Likert scale
ranging from strongly agree (1) to strongly disagree (5).
selected and modified several items from Barrett-Lennard’s (1981)
Relationship Inventory for perceived-physician empathy, Comstock
et al.’s (1982) Satisfaction Questionnaire for patient satisfaction,
DiMatteo etal.’s(1993) GeneralAdherence Scalefor patient compli-
ance, Roter, Hall, and Katz’s (1987) satisfaction questionnaire for
Cognitive Information Exchange and Partnership, and Anderson and
expertise and interpersonal trust. In addition, the questionnaire asks
the patient’s age, sex, education, and income. Reliability was esti-
mated using Cronbach’s alpha. As Table 1 shows, the coefficients for
all the scales, except for cognitive empathy (0.68), were well above
We tested our model via Structural Equation Modeling (SEM) us-
ing the EQS version 5.7b for Windows. SEM also prefers confirma-
tory factor analysis (CFA).
We used a comprehensive two-step modeling approach (Anderson
& Gerbing, 1988) to develop, test, and estimate our model. First, we
Kim et al. / EFFECTS OF PHYSICIAN EMPATHY241
The results easily met the preliminary fit criteria suggested by
Bagozzi and Yi (1988), showing the absence of both negative error
variances and correlations greater than one. Although χ2(467) = 911,
p < .001, the χ2statistic is highly sensitive to sample size. Therefore,
other fit indexes, which are not sensitive to sample size, were exam-
ined (Bagozzi & Yi, 1988; Bentler, 1990; Fan, Thompson, & Wang,
1999). As Table 2 shows, both the comparative fit index (CFI) and
CFI-Robust were 0.948 and 0.957, respectively, which indicates an
acceptable model fit to the data (Bagozzi & Yi, 1988). RMSEA was
0.043, which also indicates a very good fit because values less than
0.05 show a very good fit (Steiger, 1989). All of these indicate a very
good fit to the data.
242 Evaluation & the Health Professions / September 2004
Description of Empathy Scale and Outcome Measures
Cognitive empathy scale
1. Interested in knowing what my experience means to me.
2. Still understands me when I am not clear.
3. Always knows exactly what I mean.
1. Responds to me mechanically.
2. Tries to keep me from worrying.
3. Respects my feelings.
4. Shows interest in me.
5. Shows caring about my psychological well-being.
6. Shows great concern for my well-being.
7. Cares about me.
1. Followed my doctor’s suggestions exactly.
2. Exactly complied with this doctor’s drug regimens.
1. Overall, I am discontented with this doctor.
2. I have confidence in this doctor.
3. Overall, I am satisfied with this doctor.
4. Compared to the other doctors I have been to, this
doctor is very good.
NOTE: All standardized coefficients were significant at α = 0.05.
We next tested the full structural model. The CFI and CFI-Robust
were 0.931 and 0.940, respectively, and the RMSEA, which was
0.049, indicated a very good fit of the model (see Table 3).
The latent predictor variables modeled intercorrelated highly (r =
.68- .91, p < .001). Independent variables also correlated highly with
patient satisfaction (r = .74- .88, p < .001) but modestly with patient
compliance (r = .25-.40, p < .01). The existence of these high
intercorrelations among factors raised an issue of discriminant valid-
ity (whether each factor is a distinct construct). We used rather strin-
gent criteria suggested by Anderson and Gerbing (1988) to test
discriminant validity. A chi-square difference test was conducted on
the values obtained for the constrained models (with correlations
between constructs or latent variables fixed at 1.0) and on the uncon-
strained models using two constructs whose correlation is estimated
(Jöreskog, 1971). Each chi-square test examined a null hypothesis
that the two constructs were perfectly correlated. If the null hypothe-
Kim et al. / EFFECTS OF PHYSICIAN EMPATHY 243
Goodness of Fit Results for Confirmatory Factor Analysis
Comparative fit index (CFI)
90% confidence interval
Goodness of Fit Results for Full Structural Model
Comparative fit index (CFI)
90% confidence interval
construct measures were distinctly developed, measuring different
concepts. Thus, even though some of the constructs were highly cor-
related, the structural models were capable of separating out their
effects on other variables. The empathy model was tested by estimat-
ing the full structural model shown in Figure 1.
Direct effects are regression coefficients that are calculated while
controlling for all the variables that affect a given endogenous
The cognitive component of physician empathy led to better exchange of
cognitive information, and the affective aspect of physician empathy
led to partnership.
Abetterexchange ofcognitiveinformation ledtoincreasedperception of
Partnership led to increased interpersonal trust.
Partnership had a higher association with patient’s satisfaction than did
cognitive information exchange.
Patient’s perception of physician expertise had virtually the same effects
on both compliance (0.503) and satisfaction (0.507).
total effects(indirect effect+direct effect)on patient compliance and
patient satisfaction. Table 5 shows that partnership and perceived
affective empathy had the greatest (and near equal) effects on both
compliance and satisfaction; perceived physician expertise had the
third greatest effect on both compliance and satisfaction; interper-
faction but negative effects on compliance.
244 Evaluation & the Health Professions / September 2004
Test of Discriminant Validity:
2Test Between the Constrained and Unconstrained Models
Cognitive Information Exchange
Perception of Physician Expertise
NOTE: All χd(1)
2were significant at α level of 0.05. Each test had one degree of freedom.
munication skills significantly and substantially influenced patient
satisfaction and patient compliance. Moreover, the results were con-
246 Evaluation & the Health Professions / September 2004
Standardized Total Effects (Indirect + Direct)
of Each Predictor on Compliance and Satisfaction
Perception of physician expertise
Cognitive information exchange
Figure 1: Physician Empathy Model
(< .50) between the 2 variables.
*p < .05. **p < .01. ***p < .001.
sistent with the specific causal model proposed, which assumes the
mediating factors such as partnership and perception of physician
expertise. For Korean patients, emotional aspects (e.g., partner-
iors played the most important roles in increased satisfaction and
Unexpectedly, perception of physician expertise was one of the
best predictors of both patient satisfaction and compliance, although
Korean patients weremoreconcerned with their physician’saffective
aspects of communicative skills.
Satisfaction is a distinct construct, which is different from any of
the other variables in the model, but it correlated highly with the oth-
concepts might not have been as clear to them as to patients in the
United States. Thus, they may not have been able to clearly distin-
guish them. However, discriminant validity showed that satisfaction
was not perfectly correlated with any of the other variables. Future
research should distinguish them even more sharply than we have
here, and intervention studies can test whether improving the cogni-
tive and affective empathic behaviors of physicians leads to better
compliance and satisfaction.
Interestingly, the fact that partnership had the strongest effect on
both patient compliance and satisfaction mayreflectKorean patients’
processes in Korea.
The negative relationship between information exchange and out-
comes, especially compliance, was surprising and deserves
able is highly correlated). Note however that the bivariate correlation
between information exchange and compliance was positive (r = .246).
reflects effects of unmeasured factors, such as an unexpected symp-
tom or problem in illness management (e.g., complexity of treatment
regimens, side effects, and complications). When these are encoun-
tered in patient care, considerably greater information would need to
Kim et al. / EFFECTS OF PHYSICIAN EMPATHY247
be exchanged, regardless of empathy, expertise, or other factors. In
such a situation, compliance would be more difficult, and a negative
coefficient would result.
Another possible explanation is that because of the institutional-
ized passive patient roles, Korean patients may not often expect to
receive a lot of information or explanation from their physicians. On
the other hand, the Korean physicians, because of the social norms
assigned to them (dominant roles), might try to maintain their
expected authority and dominance in the physician-patient relation-
ship by limiting the flow of information to patients.
Consequently, as the Korean patients encounter physicians who
behave otherwise, they may begin to question the physician’s author-
ity. Highly expressive physician behaviors might connote insincerity
sequently, the patients might be less likely to believe in and comply
with their physician’s medical recommendations because the physi-
cian’s behavior would not fit the typical image of physicians in the
hierarchical authority structure of the Korean medical setting. In the
future, cross-cultural research is needed to investigate whether this
holds true across different cultures.
PRACTITIONERS AND EDUCATORS
The effective use of empathic communicative skills may be one of
the best ways to improve patient satisfaction and patient compliance.
By increasing patient compliance, the physician is likely to improve
patient health. By increasing patient satisfaction, the physicians can
ing physicians and by malpractice suits. Thus, as one management
they are more satisfying to patients.
248Evaluation & the Health Professions / September 2004
LIMITATION AND FUTURE RESEARCH
There are several limitations to this study. The measure of com-
pliance did not include an item about the difficulty of following the
medical regimen, which is found to influence compliance (Turk &
Meichenbaum, 1991). The sampling method used in this study may
have limited the generalizability of the results because the subjects
wereasked to voluntarily participate in the study. Those patients who
volunteered may have been different from those who did not. Also,
the study was done solely on outpatients, not on inpatients, and the
patients were recruited in one regional hospital in Korea, which may
not be nationally representative. Further research with larger popula-
tions is needed to examine possible differences between inpatients
and outpatients as well as between large-hospital patients and small-
clinic patients. Patients’ communication behaviors might also influ-
ence physicians’ empathic communication (Beisecker & Beisecker,
patient expectations about physician communication styles. Some
patients might prefer authoritarian physicians, whereas others might
not (Street, 1990). Because data are cross-sectional and correlational
in nature, the direction of causality is not proven. Experimental re-
effects on patient outcomes.
Last, there is evidence that empathetic communication styles may
substantially influence important aspects of patient outcomes, and
these differences may, with some variation, hold across a number of
oftheseeffectsacrosscountries andtodevelopinterventions thatwill
demonstrably improve patient-physician communication.
Anderson, J. C., & Gerbing, D. W. (1988). Structural equation modeling in practice: A review
and recommend two-step approach. Psychological Bulletin, 103(3), 411-423.
to assess interpersonal trust in patient-physician relationship. Psychological Reports, 67,
Kim et al. / EFFECTS OF PHYSICIAN EMPATHY249
Association of American Medical Colleges. (1998). Medical school objectives project.
Retrieved January 20, 2004, from http://www.aamc.org/meded/msop
Bagozzi, R. P., & Yi, Y. (1988). On the evaluation of structural equation models. Journal of the
Academy of Marketing Science, 16(1), 74-94.
Barrett-Lennard, G. T. (1981). Empathy cycle: Refinement of a nuclear concept. Journal of
Counselling Psychology, 28(2), 91-99.
Beisecker, A. E., & Beisecker, T. D. (1990). Patient information-seeking behaviors when com-
municating with doctors. Medical Care, 28, 19-28.
that correlate with patient satisfaction. Journal of Medical Education, 57, 105-112.
Davis, M. H. (1983).Measuringindividual differences in empathy: Evidencefor a multidimen-
sional approach. Journal of Personality and Social Psychology, 44(1), 113-126.
duct: A study of patient satisfaction in a family practice center. Journal of Community
Health, 6, 18-34.
DiMatteo,M. R., Hays, R. D., Gritz, E. R., Bastani, R., Crane, L., Elashoff, R. M., et al. (1993).
Patient adherence to cancer control regimes: Scale development and initial validation. Psy-
chological Assessment, 5, 102-112.
DiMatteo, M. R., Taranta, A., Friedman,H. S., & Prince, L. M. (1980).Predicting patient satis-
faction from physicians’nonverbal communication skills. Medical Care, 18, 376-387.
Fan, X., Thompson, B., & Wang, L. (1999). Effects of sample size, estimation methods, and
ing, 6(1), 56-83.
Hogan, R. (1969). Development of an empathy scale. Journal of Consulting and Clinical Psy-
chology, 33, 307-316.
cian empathy: Definition, components, measurement, and relationship to gender and spe-
cialty. The American Journal of Psychiatry, 159, 1563-1569.
Hojat, M., Mangione, S., Nasca, T. J., Cohen, M. J. M., Gonnella, J. S., Erdmann, J. B., et al.
(2001). The Jefferson Scale of Physician Empathy: Development and preliminary
psychometric data. Educational and Psychological Measurement, 61(2), 349-366.
Jöreskog, K. G. (1971). Statistical analysis of sets of congeneric tests. Psychometrika, 36, 109-
compliance) in Korea. Unpublished doctoral dissertation, Michigan State University, East
Kurtz, R. R. (1972). Different approaches to the measurement of therapist empathy and their
relationship to therapy outcomes. Journal of Consulting and Clinical Psychology, 39(1),
Linn, L. S., & Wilson, R. M. (1980). Factors related to a communication style among medical
house staff. Medical Care, 18, 1013-1019.
Mehrabian,A., & Epstein, N. (1972).A measure of emotional empathy. Journal of Personality,
Medical Journal, 322, 865.
250Evaluation & the Health Professions / September 2004
Meryn, S. (1998). Improving doctor-patient communication. British Medical Journal, 316,
Olson, J. K. (1995). Relationships between nurse-expressed empathy, patient-perceived empa-
thy and patient distress. Image: the Journal of Nursing Scholarship, 27(4), 317-322.
Pelz, S. L. (1982). Toward a theory of patient satisfaction. Social Science & Medicine, 16, 577-
Roter, D., Hall, J., & Katz, N. (1987). Relations between physicians’ behaviors and analogue
patients’satisfaction, recall, and impressions. Medical Care, 25(5), 437-451.
Squier, R. (1990). A model of empathic understanding and adherence to treatment regimens in
practitioner-patient relationships. Social Science & Medicine, 30(3), 325-339.
Steiger, J. H. (1989).EzPATH: A supplementarymodulefor SYSTAT andSYGRAPH. Evanston,
IL: Systat Software.
Street, R. L. (1990). Communication in medical consultation: A review essay. The Quarterly
Journal of Speech, 76, 315-357.
Street, R. L., Jr., & Wiemann, J. M. (1987). Patient satisfaction with physician’s interpersonal
involvement, expressiveness, and dominance. Communication Yearbook, 10, 591-612.
Turk, D. C., & Meichenbaum, D. (1991). Adherence to self-care regimens: The patient’s per-
spective. In R. H. Rozensky, J. J. Sweet, & S. M. Tovian (Eds.), Handbook of clinical psy-
chology in medical settings (pp. 249-266). New York: Kluwer Academic Publishers.
Kim et al. / EFFECTS OF PHYSICIAN EMPATHY251