We thank the authors of the three letters to the editor for their constructive comments regarding our review. The possibility of empathy decline during medical education is a serious issue that needs to be carefully considered and further discussed. First, all three letters recommend the use of statistical effect-size measures, given the “different types of Likert scales.” We agree that this is a potential problem, and in our review we specifically reported results in Table 1 for each of the 11 studies separately, so readers could see that empathy decline was quite small in every study—regardless of scale. We also reported in the text the mean decline for each of the three empathy instruments separately (−0.2 for the Interpersonal Reactivity Index–Empathy Concern subscale with a five-point scale, −0.2 for the Jefferson Scale of Physician Empathy [JSPE] with a seven-point scale, and −0.3 for the Balanced Emotional Empathy Scale with a nine-point scale) to further emphasize that the declines were nearly identical and quite small—regardless of instrument. We thought that, for these studies, the results reported in terms of the original measurement scale were more direct and more meaningful (actual magnitude of decline in empathy scores in terms of the rating scale anchors) than results on a standardized effect-size scale (reported in standard deviation units).
We agree that the JSPE is based on an extensive research base and has a solid psychometric foundation—possibly the most researched and widely used instrument in medical education. Our question is whether physicians' self-reports of empathy are indicators of patients' perceptions of physician empathy—two recent studies showed modest (r = 0.48) and nonsignificant (r = 0.24) correlations—and in particular whether the very small/weak declines in physician self-reported empathy in the 11 studies warrant concerns about diminished patient perceptions and impact on care. We also agree that a 100% response rate may be “unobtainable” in most studies, and we certainly do not have an “expectation of a 100% response rate.” However, we stress that the self-report empathy declines reported were small and that response bias is an especially plausible threat to the validity of small effects. In brief, the 11 studies we reviewed are exemplary quasi-experimental studies, but as with all quasi-experimental studies, plausible threats to validity necessarily qualify research conclusions (such as empathy decline), as must the meaning of the measures (self-reports) used to assess empathy.1
Jerry A. Colliver, PhD
Professor, Department of Medical Education, Southern Illinois University School of Medicine, Springfield, Illinois; email@example.com.
Melinda J. Conlee
Editorial writer, Department of Medical Education, and managing editor, Teaching and Learning in Medicine, Southern Illinois University School of Medicine, Springfield, Illinois.
Steven J. Verhulst, PhD
Professor, Department of Medical Education, and director, Statistics and Research Consulting, Southern Illinois School of Medicine, Springfield, Illinois.
J. Kevin Dorsey, MD, PhD
Professor, dean, and provost, Southern Illinois University School of Medicine, Springfield, Illinois.
1 Colliver JA, McGaghie WC. Quasi-experimentation and unresolved threats to validity: The reputation of medical education research. Teach Learn Med. 2008;20:101–105.