Doing the right thing: Researchers retract clinician burnout study after realizing their error

A journal is retracting and replacing a 2016 study which found that nearly two-thirds of clinicians who focus on end-of-life care experienced burnout, after the authors found an error that had dramatically inflated the findings. 

The article, “Prevalence and predictors of burnout among hospice and palliative care clinicians in the U.S.,” appeared in the Journal of Pain and Symptom Management, and has been cited 72 times, according to Clarivate Analytics’ Web of Science. The authors, led by Arif Kamal, of Duke University, included researchers at Mayo Clinic, the American Academy of Hospice and Palliative Medicine, the University of Pittsburgh and other institutions. 

But while working on a subsequent paper, the authors realized that something was amiss with their first article. The two studies revealed strikingly different rates of burnout in the surveys the researchers had conducted, 62% vs. 39%, an unusual finding given that a relatively short time had elapsed between the surveys, they said. A closer look revealed a critical error.

The journal has issued two retraction notices, one abbreviated and one a bit longer.  

The shorter version states:

This article has been retracted at the request of the Editor-in-Chief. A critical error was discovered in how the MBI score was tabulated in the 2016 analysis. This resulted from an inadvertent error in the coding of survey scores. Specifically, the MBI is comprised of 22 questions, with responses recorded on a 7-point Likert scale ranging from 0 to 6. However, the database for the research stored the MBI responses as values between 1 and 7. This necessitated subtracting a value of 1 from all MBI items before scoring the questionnaire. This was not done and resulted in overestimation of the prevalence of burnout. To address this, the authors have submitted a revised manuscript and that revision has been accepted as a replacement for the original paper DOI:10.1016/j.jpainsymman.2019.11.001

Here’s the longer version — a pdf document that somehow gives all of the authors affiliations in Brazil (which Kamal told Retraction Watch he was trying to get the journal to correct): 

We wanted to bring to your attention a notable error in our publication from 2016.1 In this publication, we report the findings of a field-wide survey of AAHPM members regarding burnout and related issues. We reported a burnout rate of 62% among survey participants, calculated using the Maslach Burnout Inventory-22 (MBI), a widely recognized standard for burnout measurement in health services field. In the Fall of 2018, we conducted another burnout survey among several membership organizations in the National Coalition for Hospice and Palliative Care, using several survey components from the earlier study, including the MBI. In the course of performing the more recent analysis, we found a burnout rate much lower than what was reported in 2016.2 Understanding that dramatic shifts in burnout prevalence are not often found, particularly across short time intervals, we re-ran our analysis from the 2016 manuscript. In doing so, we found a critical error in how the MBI score was tabulated in the 2016 analysis. This resulted from an inadvertent error in the coding of the survey scores. Specifically, the MBI comprises 22 questions, with responses recorded on a seven-point Likert scale ranging from 0 to 6. However, the database for our earlier research stored the MBI responses as values between 1 and 7. This necessitated subtracting a value of 1 from all MBI items before scoring the questionnaire. This was not done and resulted in overestimation of the prevalence of burnout. To address this, we have submitted a revised manuscript and that revision has been accepted as a replacement for the original paper. Notably, our revised findings now include the following: 1) the prevalence of burnout in the cohort is lower (38.7%) than what was previously published and 2) there was no longer a significant difference in burnout comparing physicians (41.9%) and other palliative care clinicians/workers (37.1%) (P ¼ 0.17) in the univariable analysis, although clinical role remained a significant predictor of burnout in the multivariable model. We invite the JPSM readership to read the revised manuscript for further details on the methods to address the error and updated findings. We are sincerely regretful of our error and appreciate the ability to update our field on this important topic.

Like Retraction Watch? You can make a tax-deductible contribution to support our work, follow us on Twitter, like us on Facebook, add us to your RSS reader, or subscribe to our daily digest. If you find a retraction that’s not in our database, you can let us know here. For comments or feedback, email us at [email protected].