Notice of Retraction due to a programming error

To the EditorOn behalf of our coauthors, we write to report a programming error and other errors that affected the results in our article, “Effect of a Program Combining Transitional Care and Long-term Self-management Support on Outcomes of Hospitalized Patients With Chronic Obstructive Pulmonary Disease: A Randomized Clinical Trial” published in the December 11, 2018, issue of JAMA .We write to explain what happened and to request retraction of this article.

In this study, we had tested a 3-month hospital-initiated program that combined transition and long-term management support for patients hospitalized for chronic obstructive pulmonary disease (COPD). We had originally reported that the intervention, compared with usual care, resulted in a fewer number of mean COPD-related hospitalizations and emergency department visits at 6 months per participant (0.72 [95% CI, 0.45-0.97] vs 1.40 [95% CI, 1.01-1.79]) and an adjusted difference in the 100-point St George’s Respiratory Questionnaire (SGRQ) score between groups of −6.69 (95% CI, −12.97 to −0.40; P  = .04). The correct results reverse the main finding to more COPD-related hospitalizations and emergency department visits in the intervention group vs the usual care group (1.40 [95% CI, 1.10-1.79] vs 0.72 [95% CI, 0.46-0.97]) and an adjusted difference in the SGRQ score that is no longer statistically different (5.18 [95% CI, −2.15 to 12.51]; P  = .11). Here we describe the detected errors and the additional analyses that have been conducted.

The identified programming error was in a file used for preparation of the analytic data sets for statistical analysis and occurred while the variable referring to the study “arm” (ie, group) assignment was recoded. The purpose of the recoding was to change the randomization assignment variable format of “1, 2” to a binary format of “0, 1.” However, the assignment was made incorrectly and resulted in a reversed coding of the study groups. Even though the data analyst created and conducted some test analysis programs, they were of the type that did not show any labeling of the arm categories, only the “arm” variable in a regression, for example. After detecting this error, we promptly reported it to our institutional review board and appropriate offices within our university, alerted JAMA , and proceeded to confirm whether the error had affected the analytic data sets, which we found to be the case. We therefore started a complete data reanalysis, with 2 biostatisticians performing double programming and an independent analysis of study primary outcomes to ensure the validity of the reported results. As noted here, this reanalysis showed reversed study findings, with a higher number of hospitalizations and emergency department visits in the intervention compared with the usual care group.

All study data were subsequently reanalyzed to detect any other errors. This included data review, repeating all the data preparation and programming, and full reanalysis. Over the course of this reanalysis, we detected an error in imputing missing values for the SGRQ, whereby the worst possible score (100) was incorrectly imputed for missing values of participants who had died beyond the 6-month study period. The correct approach would have been to classify those values as missing because those participants had not died by the 6 months after discharge study end point. As noted here, after correction of this error, we found no significant difference in the co–primary outcome of change in health-related quality of life as measured by the SGRQ between the intervention and usual care groups. We also detected an error in summarizing the baseline medication classes in Table 1, and 2 hospitalizations that were not counted in the initial analysis.

To reduce the occurrence of future similar programming errors, the Johns Hopkins Biostatistics Center has instituted a new standard operating procedure for checking randomization assignment to be followed in all trial analyses. To ensure that the group assignment used in any of the trial analyses is correct, a verification process will be included at the beginning and end of each analysis program. This process is intended to confirm that the group assignment separately provided by the trial team matches the group assignment used in the analysis program. The matching confirmation is reviewed by a second biostatistician/analyst before its use in the results.

With this notice of retraction, we are publishing a new article with complete corrected findings.Given the corrected finding of a paradoxical increase in acute care use in the intervention group, we conducted a post hoc exploratory analysis to evaluate the difference in treatment effect between the study groups in subgroups of patients whose characteristics might affect the primary acute care visits outcome. The results of this additional analysis are included in the new article.

We apologize to the readers and editors of JAMA for any confusion caused by these errors and the erroneous original report of the trial findings and this subsequent retraction of the article. We appreciate the opportunity to publish a new article with the correct findings and an additional analysis. All of our coauthors agree with this retraction notice.

Article Information

Corresponding Author:Hanan Aboumatar, MD, MPH, Johns Hopkins School of Medicine, 750 E Pratt St, Fifteenth Floor, Baltimore, MD 21202 ( [email protected] ).

Conflict of Interest Disclosures:Dr Aboumatar reported receiving grants from the Patient-Centered Outcomes Research Institute. Dr Wise reported receiving grants and personal fees from AstraZeneca/Medimmune, Boehringer Ingelheim, and GlaxoSmithKline; personal fees from AbbVie, Contrafect, Novartis, Pulmonx, Roche, Spiration, Sunovion, Merck, Circassia, Kiniksa, Pneuma, Propeller Health, Syneos, Verona, Bonti, Denali, and Aradigm; and grants from Pearl Therapeutics and Sanofi.