Correction: Assessing the effect of empathy-enhancing interventions in health education and training: a systematic review of randomised controlled trials

Winter R, Issa E, Roberts N, et al. Assessing the effect of empathy-enhancing interventions in health education and training: a systematic review of randomised controlled trials. BMJ Open 2020;10:e036471.

The authors want to alert readers to the following corrections made to the published version.

1. The author Riess’ name is misspelt and appears as Reiss in Figures 2 and 3.

2. On page 4 the line “was the only study that used objective measures.” is corrected to “Most studies (18) used only self-report measures ^{23,28,29,32,33,35-39,41-43,45,47,48-50,52}. Four used objective measures ^31,34,46,48 and four used a combination of self- and objective-report tools to measure empathy. ^30,40,44,51”.

3. On page 4 the line 12 studies had 100 or more participants. Has been corrected to 13 studies and include reference 44.

Below is the clarification for the errors that have been brought to the author’s attention that there was a difference between the results that the author has obtained from the meta-analysis and the results published in one of the individual trials.⁴⁴

The author does not believe that they have made an error there. They have obtained the opinion of statistical and methodological colleagues who were not involved in the conduct of the review and believe that their analysis is correct: the difference is not statistically significant, according to standard statistical methods.

Here are the data provided on page 1282 of the trial report (J Gen Intern Med 2012;27:1280–86):

	Training group	Control group
N	54	45
Pre-post change score (mean difference)	0.7	−1.5
SD	7.9	6.0

Although the trial report claims that this difference was significant, we are unable to replicate this using the analysis method that author applied across our review:

	What the paper reports	What RevMan says	What we calculate
Cohen’s d	0.31	0.31	0.31
T-test	n/a	1.5348	1.5348
P value	0.04	0.1281	0.1281

The difference seems to arise from the fact that the trial authors used the Wilcoxon—Mann—Whitney test, whereas RevMan uses a parametric test. The justification in the trial report (page 1283) for using non-parametric tests was “because they are robust to violations of the assumptions of parametric tests”. However, no data were provided in the trial report to show, for example, that the data were not distributed normally, and the other similar trials in our review seem to have normally distributed data and used parametric tests.

In the absence of evidence that the data were not distributed normally, standard statistical textbooks recommend parametric tests.^1–3