Citation

Maple M, Sanford R, Pirkis J, Reavley N, Nicholas A. J. Affect. Disord. 2020; 274: 1228-1231.

Copyright

(Copyright © 2020, Elsevier Publishing)

DOI

10.1016/j.jad.2020.06.012

PMID

32663965

Abstract

The authors regret that the use of SPSS 24 for the analysis undertaken in this manuscript did not correctly calculate odds ratios. Due to the weighting of the data, STATA v15 should have been used to perform these analyses. When STATA is used, several of the marginally significant findings become non-significant. In summary, the following variables of interest related to exposure are no longer significant in the final binary logistic regression model: reported closeness, time since death, and number of suicide exposures. Similarly, the following control variables are no longer significant in the final model: gender, education level, and geographic region. However, knowing a friend or family member at risk, marital status, and age remain significant at the p≤0.001 level. While perceived impact and Aboriginal or Torres Strait Islander status are only significant at p≤0.01.

All revised tables and associated text is located below which provides full details of all amendments following the analytic error. Significant was set at p≤0.001. Note that in each instance of a change detailed below, the amended text replaces the original text that precedes it.

The authors would like to apologise for any inconvenience caused.

Language: en