|
Abstract
|
An erratum is being issued to the article published in 2021 titled Suicide among veterans with amyotrophic lateral sclerosis.1 In this research study, investigators used data from national Veterans Health Administration electronic health records (EHR) that are captured in the Department of Veterans Affairs (VA) Corporate Data Warehouse (CDW). Study-specific databases for research projects within the VA, including this one, are provisioned by the VA Informatics and Computing Infrastructure (VINCI) group such that the study team is given access only to EHR records that meet study criteria. Study investigators were notified after the publication of the article that the database for this project was impacted by a larger data provisioning issue within the VA, and as a result, the data provisioned for this project were incomplete. The VINCI team corrected the provisioning issues and provided updated, complete data for the cohort. Additionally, we have verified that the new data were provisioned correctly and have rerun all our analyses with the new data to ensure accuracy. It should also be noted that the CDW EHR data are live data, meaning that the underlying data can change from day to day. Thus, we expected and found some small variation in data, regardless of the provisioning issues. Overall, findings were not meaningfully impacted.
The following corrections and clarifications apply to this publication.
In the Abstract, Results section, it should now read: The hazard of suicide was 3.96 times higher for those with ALS than for those without (95% confidence interval [CI], 2.63-5.97; p < .0001). After adjusting for covariates, those with ALS remained at increased risk (hazard ratio, 3.48; 95% CI, 2.31-5.25; p < .0001).
In the Analysis section, it should now read: The proportional hazards assumption was tested for the fixed variables in the models, age, and gender, via interactions between time and each variable. The interaction with age was significant in both adjusted models and therefore remained in the models to account for nonproportionality. The interaction with gender was nonsignificant and was removed from the models as there is no evidence of a proportional hazards violation for that variable.15
In the Results section, it should now read: Data were obtained for 7429 individuals with ALS and 1 580 757 without ALS.
In the Results section, it should now read: In the unadjusted analysis, those with an ALS diagnosis, relative to those without an ALS diagnosis, were found to be 3.96 times more likely to die by suicide, a medium to large effect.
In the Results section, it should now read: For those who died by suicide, we assessed the method of suicide. Of the 23 veteran suicide decedents in the ALS group, 21 (91.3%) used firearms as their method of suicide. In the no ALS group, 73.7% (N = 3422) of suicide decedents used firearms as their method of suicide. This results in a statistically nonsignificant, yet medium to large odds ratio of 3.75 (95% confidence interval [CI], 0.88-16.0; p = .06).
In Table 1, demographic values have slightly changed, with no changes to significance, and can be viewed in Updated Table 1.
In Table 2, hazard ratios and 95% CIs were slightly different, though no differences in p values were noted (see Updated Table 2).
Language: en |