Hojat, Mohammadreza PhD; Louis, Daniel Z. MS; Markham, Fred W. MD; Wender, Richard MD; Rabinowitz, Carol; Gonnella, Joseph S. MD
Empathy, an essential component of the physician–patient relationship, may be linked to positive patient outcomes. Although this notion is consistent with the conceptual view of physician–patient relationships,1–3 empirical data supporting the association between physicians' empathy and tangible clinical outcomes are difficult to find. Several studies generally support the notion that the quality of the physician–patient relationship (as a proxy for empathic engagement in patient care) has a positive influence on patient outcomes.1–6
Published reports also suggest that indicators of empathic engagement in patient care, such as physician–patient communication, verbal interaction (e.g., positive talk), nonverbal cues (e.g., appropriate touch, eye contact, bodily posture, gestures), as well as length of the encounter can lead to increased patient satisfaction7–10 and better compliance.11–13 Relationships have been reported between some measures of empathy and psychotherapeutic effectiveness,14,15 patients' feelings of being important,16 physicians' accuracy of diagnosis,17 and accuracy of prognosis.18 To our knowledge, however, no empirical study has used a psychometrically sound measure of physicians' empathy to examine the relationship between physicians' empathy and laboratory measures of intermediate clinical outcomes.
The purpose of this study therefore was to provide an evidence-based scientific foundation for the study of empathy as a clinically important factor in patient outcomes. We designed this study to test the following hypothesis: Diabetic patients of physicians with high empathy scores would have better clinical outcomes than patients whose physicians had low empathy scores.
Patients who participated in the study were selected from a pool of 7,269 outpatients, treated between July 2006 and June 2009 by 29 attending physicians in the Department of Family and Community Medicine at Thomas Jefferson University. We examined the billing records of these patients and selected 1,154 patients who met four predetermined criteria for eligibility: (1) had the diagnostic code for diabetes on their billing record, (2) were between 18 and 75 years of age at the time of their first visit, (3) had at least two office visits with the physician during the past 36-month time period, and (4) spent at least two-thirds of the total office visits with the attending physician identified as the patients' primary caregiver. Laboratory data were available for these patients through December 2009.
The electronic records of 181 patients did not include the results of either the hemoglobin A1c or the low-density lipoprotein cholesterol (LDL-C) tests. The records of 921 patients included the A1c test results, and the records of 943 patients had the LDL-C test results. The results of both tests were available for 891 patients; these were selected as the final sample of this study. The patients' median age was 56 years, 531 (60%) were women, and the number of office visits per patient during the study period ranged from 2 (32 patients, or 4%) to 50 or more (10 patients, or 1%), with a median of 10 visits.
Measurement of physicians' empathy
Although a few research instruments are available to measure empathy in the general population, none is content-specific to patient care.19(pp63–74) The Jefferson Scale of Empathy (JSE) was developed in 2001 at our medical school as an instrument to measure empathy in the context of medical education and patient care.19–23 Developed following an extensive review of the literature, the instrument relies on the definition of empathy in the context of patient care as a predominantly cognitive attribute that involves an understanding of the patient's experiences, concerns, and perspectives, combined with a capacity to communicate this understanding and an intention to help.24,25 The scale includes 20 items answered on a seven-point Likert-type scale (Strongly Agree = 7, Strongly Disagree = 1).
Evidence in support of the JSE's construct validity,19,20,23 criterion-related validity,20,21 predictive validity,26 internal consistency reliability,20,21,23 and test–retest reliability23 has been reported for physicians. The JSE has received broad acceptance and has been translated into 38 languages to date. (More information about versions of the JSE can be found at www.tju.edu/jmc/crmehc/medu/oempathy.cfm.)
Measures of clinical outcomes
The most recent results from the patients' hemoglobin A1c and LDL-C were used as indicators of the patients' metabolic control.27–29
After receiving approval from the institutional review board of Thomas Jefferson University, we administered the JSE to all 31 attending physicians in the Department of Family and Community Medicine in 2009. All of the physicians practice in the same office located in an urban setting where minority patients constitute more than half of the patients. The chair of the department provided the physicians with a brief explanation of the study's purpose and encouraged them to complete the JSE and return it directly to the Center for Research in Medical Education and Health Care in the provided addressed envelope for processing and statistical analyses. Each physician was identified by a numeric code printed on the JSE scanning form and was told that the code would be used to correlate their JSE scores with his or her diabetic patients' hemoglobin A1c and LDL-C test results.
All attending physicians agreed to participate in the study. Two physicians with minimal patient responsibilities (they provide primary care for fewer than 45 patients) were excluded from the statistical analyses. Because the patients' electronic records were linked to their physicians' empathy levels by the numeric codes, the patients remained anonymous. For the purpose of statistical analyses, we classified the physicians into three groups according to the distribution of their JSE scores: high- (top third), moderate- (middle third), and low-scoring physicians (bottom third).
The results of the A1c tests were classified into three categories according to the standards of National Quality Measures Clearing House (www.qualitymeasures.ahrq.gov): good control (<7.0%), poor control (>9.0%), and moderate control (≥7.0% and ≤9.0%). A1c levels were used to measure the adequacy of blood glucose control according to national standards developed on the basis of numerous studies showing a strong relationship between A1c levels and the development of complications from diabetes such as microvascular disease and neuropathic problems. The levels of LDL-C were also classified into three categories: good control (<100 mg/dL,), poor control (>130 mg/dL), and moderate control (≥100 and ≤130 mg/dL).27–29 The proportions of patients in each of the aforementioned categories were calculated.
The chi-square test was used to examine the significance of associations between physicians' levels of empathy and levels of the hemoglobin A1c and LDL-C test results. In addition, pairwise differences in the proportions of patients with good and poor control test results for physicians receiving high or low empathy scores were examined by using the z test for proportions. The possible confounding effects of physicians' and patients' gender and age, and patients' type of health insurance in the relationships between empathy scores and medical test results, were examined using logistic regression analyses. Statistical analyses were performed using SAS version 9.1 for Windows.
Descriptive statistics on empathy scores for the three groups of high-, moderate-, and low-scoring physicians are reported in Table 1. As shown in the table, the differences for the mean empathy scores for the three groups of physicians were statistically significant by analysis of variance and Duncan multiple comparison test (F(2, 27) = 77.0, P < .001, high scorers > moderate scorers > low scorers).
The proportions of patients in the good, moderate, and poor control A1c and LDL-C categories were compared for the high-, moderate-, and low-empathy-scoring physicians. The frequency and percent distributions of A1c test results by levels of physicians' empathy are reported in Table 2.
The association between the three levels of the hemoglobin A1c test outcomes and the three levels of physicians' empathy was highly significant (χ2(4) = 22.04, P < .001). The likelihood of good control (A1c < 7.0%) was significantly greater in the patients of physicians with high empathy scores than in the patients of physicians with low scores (56% and 40%, respectively; z = 4.0, P < .01). Conversely, the likelihood of poor control (A1c > 9) was significantly lower in the patients of physicians with high empathy scores than it was in the patients of physicians in the low-scoring group (15% and 26%, respectively; z = −3.7, P < .01).
Frequency and percent distributions of the LDL-C test results by levels of physicians' empathy and summary results of statistical analysis are also reported in Table 2. Similar to the previous findings, the association between the three levels of LDL-C test outcomes and the three levels of physicians' empathy was highly significant (χ2(4) = 15.55, P < .001), both of which provide support for our research hypothesis. The results of LDL-C test outcomes showed that the likelihood of good control (LDL-C < 100) was significantly higher for the patients of physicians with high empathy scores than for the patients of physicians with low scores (59% and 44%, respectively; z = 3.8, P < .01). Conversely, the likelihood of poor control (LDL-C ≥ 100) was lower for the patients of high-empathy-scoring physicians than for the patients of low scorers (14% and 24%, respectively; z = −3.3, P < .01).
Statistical control for gender, age, and type of insurance
Logistic regression was used to examine the unique contribution of levels of physicians' empathy in predicting optimal clinical outcomes after controlling for physicians' and patients' gender and age, and patients' health insurance. In the first logistic model, the outcomes of the hemoglobin A1c test were dichotomized according to whether they had achieved good control (<7.0%, n = 452). The independent variables included physicians' and patients' gender; physicians' age (>50 years and ≤50 years); and patients' age (median split, >56 years, n = 443; and ≤56 years, n = 448). Patients were grouped into three categories based on their type of health insurance: private insurance (n = 470), Medicare (n = 312), and Medicaid (n = 108). The sole uninsured patient was deleted from the logistic regression analyses. Summary results of the logistic regression analysis are reported in Table 3.
The results indicate that the physicians' empathy scores were associated significantly and uniquely with the prediction of good A1c outcomes. The odds ratio of 1.8 obtained for physicians' empathy (high versus low) indicated that for an increase from the low- to the high-scoring category, the odds of good control of hemoglobin A1c (<7.0) increased by 80%. Also, for an increase from the low-scoring to the moderate-scoring category of physicians' empathy, the odds of good control of A1c increased by 50%. Physicians' gender (being male was associated with good control of patients' A1c outcome), physicians' age (younger age was associated with good control of patients' A1c), and patients' type of insurance (Medicare was associated with good control) also contributed significantly to the model. Patients' gender and age did not contribute. The Hosmer–Lemeshow goodness-of-fit test showed that the model was mathematically sound (χ2(8) = 7.03, P = .53). These findings indicated that the physicians' degree of empathy was a unique and significant contributor to the prediction of good control of hemoglobin A1c for diabetic patients, beyond the contributions of gender and age of the physicians and patients, and type of patients' health insurance.
In another logistic regression model, we classified the results of the LDL-C test into two categories in which an LDL-C test result of less than 100 was regarded as good control. The same predictors used in the previous model were included as the independent variables. The summary results of this analysis are reported in Table 3.
The odds ratios for physicians' empathy reported in the table indicated that for an increase from a low- to a high-scoring category of physicians' empathy, the odds of good control of LDL-C (<100) increased by 80%. Also, for an increase from a low- to a moderate-scoring category of physicians' empathy, the odds of good control of patients' LDL-C increased by 40%. Patients' gender also contributed significantly to the prediction model (being male was associated with good control of the LDL-C). Neither physicians' gender and age, nor patients' age or type of health insurance, predicted the medical outcome. The Hosmer–Lemeshow goodness-of-fit test indicated a mathematically sound model (χ2(8) = 2.94, P = .94). These findings indicated that physicians' empathy contributed uniquely and significantly to the prediction of good control of LDL-C test outcomes for diabetic patients beyond the gender and age of physicians and patients, and patients' health insurance. These findings provide additional support for our research hypothesis.
Confirmation of the hypothesis that physicians' empathy is associated with positive clinical outcomes suggests that empathy should be viewed as an integral component of physician competence. The methods used did not permit us to delineate the cause-and-effect mechanism that would explain our findings. A number of explanations are plausible and worthy of further investigation. One possible explanation is that greater empathy in the physician–patient relationship enhances mutual understanding and trust between physician and patient, which in turn promotes sharing without concealment, leading to better alignment between patients' needs and treatment plans and thus more accurate diagnosis and greater adherence.
Research provides support for the aforementioned explanations. For example, it has been reported that physicians' understanding of their patients' perspective—a key feature in the definition of physician empathy19—enhances patients' perceptions of being helped,30 improves emotional management and patients' empowerment,1 and increases patients' perceptions of social support networks.1,19 In a factor analytic study, 52% of the variance in patients' ratings of satisfaction with their medical care was accounted for by the physicians' level of interpersonal warmth and respect,31 which are among the features of physician empathy.19 In another study with diabetic patients, dietitians' empathic engagement proved to be predictive of patient satisfaction and successful consultations.32 In yet another study, the researchers found that empathy was the most important quality for being a “good physician.”33 Further research is needed to explore the underlying mechanisms of the link between physicians' empathy and patient outcomes.
We noticed that the highly empathic physicians saw a smaller number of patients than the other groups of physicians. One may speculate that this could be due to highly empathic physicians spending more time with their patients, thus leading to fewer patients being seen. These physicians may also be involved in additional academic activities. These and other speculations need to be empirically tested in future research.
Limitations and Concluding Remarks
Achieving the goals of clinical care is a complex endeavor involving multiple factors that include physicians, patients, culture, race, ethnicity, severity of disease, environment of care, and health care regulations, among others. Obviously, controlling for all these factors in clinical research is difficult. In addition to these factors, our study was limited in other ways. For example, we did not control for factors that may have had an impact on the intermediate outcomes we measured, such as severity of disease. Also, the study was conducted in a single institution in an academic setting, and intermediate outcomes were measured for only one complex chronic disease.
We attempted to minimize the impact of some of the limitations in several ways. Although we did not control for severity of disease, we conducted the research with faculty in a large practice setting, which accepts all patients from all insurers and does not consider severity when assigning them. We chose to study diabetes because of its high prevalence, the availability of well-accepted evidence regarding the value of achieving intermediate optimal outcomes, and readily available and accepted quantitative measures for assessing the control of the disease. In addition, because of the chronic nature of the disease, diabetic patients need sustained contact and follow-up with their physicians, making empathic engagement more important. To improve the validity of the findings in future research, it would be desirable to include other factors involved in patient outcomes, such as hospitalization for complications caused by diabetic acidosis, coma, or septicemia.
Despite the study's limitations, however, our findings support the recommendations of such professional organizations as the Association of American Medical Colleges34 and the American Board of Internal Medicine35 to assess and enhance empathic skills in undergraduate and graduate medical education.
Finally, our findings regarding the substantial associations between physicians' empathy and patient outcomes, combined with research findings documenting the erosion of empathy during undergraduate24,36,37 and graduate medical education,38,39 reinforce the need for the assessment and enhancement of empathic skills in both physicians-in-training and practicing physicians. Although questions have been raised about the validity of findings regarding the decline in empathy during medical education,40 such critics have not been left unchallenged.41
Replication of this study in multiple institutions, across different cultures, and with a variety of disease conditions (e.g., hypertension, asthma, infectious diseases, cancer, etc.) can establish physicians' empathy as an important component of their overall competence and as a significant factor in positive patient outcomes, thus placing empathy in the domain of evidence-based medicine.
The authors would like to thank all physicians who participated in this study by completing the Jefferson Scale of Empathy. Also, thanks are due to James J. Diamond, PhD, for his assistance in logistic regression, Kaye Maxwell for her contribution to the management of physicians' data, Mike Devenny for extracting data from patients' electronic records, and Dorissa Bolinski for her editorial assistance.
This study was approved by the institutional review board of Thomas Jefferson University.
1Street RL Jr, Makoul G, Arora NK, Epstein RM. How does communication heal? Pathways linking clinician–patient communication to health outcomes. Patient Educ Couns. 2009;74:295–301.
2Neuwirth ZE. Physician empathy—Should we care? Lancet. 1997;350:606.
3Stewart MA. Effective physician–patient communication and health outcomes: A review. Can Med Assoc J. 1995;152:1423–1433.
4Sanson-Fisher R, Maguire P. Should skills in communicating with patients be taught in medical schools? Lancet. 1980;316:523–526.
5Roter DL, Hall JA, Merisca R, Nordstrom B, Cretin D, Svarstad B. Effectiveness of interventions to improve patient compliance: A meta-analysis. Med Care. 1998;36:1138–1161.
6Staudenmayer H, Lefkowitz MS. Physician–patient psychosocial characteristics influencing medical decision-making. Soc Sci Med. 1981;15:77–81.
7Hall JA, Roter DL, Katz NR. Meta-analysis of correlates of provider behavior in medical encounters. Med Care. 1988;26:657–675.
8Kim SS, Kaplowitz S, Johnston MV. The effects of physician empathy on patient satisfaction and compliance. Eval Health Prof. 2004;27:237–251.
9Zachariae R, Pedersen CG, Jensen AB, Ehrnrooth E, Rossen PB, Von der Maase H. Association of perceived physician communication style with patient satisfaction, distress, cancer-related self-efficacy, and perceived control over the disease. Br J Cancer. 2003;88:658–665.
10DiMatteo MR, Sherbourne CD, Hays RD, et al. Physicians' characteristics influence patients' adherence to medical treatment: Results from the medical outcomes study. Health Psych. 1993;12:93–102.
11DiMatteo MR, Hays RD, Prince LM. Relationship of physicians' nonverbal communication skills to patient satisfaction, appointment noncompliance and physician workload. Health Psychol. 1986;5:581–594.
12Falvo D, Tippy P. Communicating information to patients: Patient satisfaction and adherence as associated with resident skill. J Fam Pract. 1988;26:643–647.
13Squier RW. A model of empathetic understanding and adherence to treatment regimens in practitioner–patient relationships. Soc Sci Med. 1990;30:325–339.
14Burns DD, Nolen-Hoeksema S. Therapeutic empathy and recovery from depression in cognitive-behavioral therapy: A structural equation model. J Consult Clin Psychol. 1992;60:441–449.
15Greenberg LS, Watson JC, Elliot R, Bohart AC. Empathy. Psychotherapy. 2001;38:380–384.
16Colliver JA, Willis MS, Robbs RS, Cohen DS, Swartz MH. Assessment of empathy in a standardized-patient examination. Teach Learn Med. 1998;10:8–10.
17Barsky AJ. Hidden reasons some patients visit doctors. Ann Intern Med. 1981;94:492–498.
18Dubnicki C. Relationships among therapist empathy and authoritarianism and a therapist's prognosis. J Couns Clin Psychol. 1977;45:958–959.
19Hojat M. Empathy in Patient Care: Antecedents, Development, Measurement, and Outcomes. New York, NY: Springer; 2007.
20Hojat M, Mangione S, Nasca TJ, et al. The Jefferson Scale of Physician Empathy: Development and preliminary psychometric data. Educ Psychol Meas. 2001;61:349–365.
21Hojat M, Gonnella JS, Mangione S, et al. Empathy in medical students as related to clinical competence, gender, and academic performance. Med Educ. 2002;36:522–527.
23Hojat M, Gonnella JS, Nasca TJ, Mangione S, Vergare M, Magee M. Physician empathy: Definition, components, measurement and relationship to gender and specialty. Am J Psychiatry. 2002;159:1563–1569.
25Hojat M. Ten approaches for enhancing empathy in health and human services cultures. J Health Hum Serv Admin. 2009;31:412–450.
26Hojat M, Mangione S, Nasca TJ, Gonnella JS. Empathy scores in medical school and ratings of empathic behavior 3 years later. J Soc Psychol. 2005;14:663–672.
27American Diabetes Association. Standards of medical care in diabetes—2009. Diabetes Care. 2009;32(suppl 1):S13–S61.
28Executive summary: Standards of medical care in diabetes—2010. Diabetes Care. 2010;33(suppl 1):S4–S10.
29Nathan DM, Kuenen J, Borg R, Zheng H, Schoenfeld D, Heine RJ. Translating the A1C assay into estimated average glucose values. Diabetes Care. 2008;147:1473–1478.
30Eisenthal SE, Emery R, Lazare A, Udin H. “Adherence” and the negotiated approach to patienthood. Arch Gen Psychiatry. 1979;36:393–398.
31Kenny DT. Determinants of patient satisfaction with the medical consultation. Psychol Health. 1995;10:427–437.
32Goodchild CE, Skinner TC, Parkin T. The value of empathy in dietetic consultation: A pilot study to investigate its effect on satisfaction, autonomy and agreement. J Hum Nutr Diet. 2005;18:181–185.
33Carmel S, Glick SM. Compassionate-empathic physicians: Personality traits and social-organizational factors that enhance or inhibit this behavior pattern. Soc Sci Med. 1996;43:1253–1261.
35American Board of Internal Medicine. Evaluation of humanistic qualities in the internists. Ann Intern Med. 1983;99:720–724.
36Hojat M, Mangione S, Nasca TJ, et al. An empirical study of decline in empathy in medical school. Med Educ. 2004;38:934–941.
39Bellini LM, Baime M, Shea JA. Variation of mood and empathy during internship. JAMA. 2002;287:3143–3146.
41Hojat M, Gonnella JS, Veloski J. Rebuttal to critics of studies of the decline of empathy. Acad Med. 2010;85:1812.