Military Suicide Research Consortium common data elements: Bifactor analysis and longitudinal predictive ability of...

To enhance and standardize the assessment of suicidal self-directed violence (SDV) in military populations, the Military Suicide Research Consortium developed the Common Data Elements (CDEs). Previous research supported the CDEs as assessing a higher-order factor of suicidal SDV in military populations. The present study had two aims: 1) confirm the bifactor structure of the CDEs in a high-risk sample, and 2) assess the ability of the factorially derived suicidal SDV factor to predict suicide attempts and return to care for suicidal ideation over 3-month follow-up. Utilizing a sample of service members referred for a psychiatric evaluation (N = 1,044), the CDE structure was assessed with confirmatory bifactor modeling. Logistic regressions and receiver operating characteristic (ROC) analyses were used to assess the suicidal SDV risk factor's prediction of suicide attempts and return to care for suicidal ideation during follow-up (n = 758). Bifactor modeling suggested adequate fit for the overarching suicidal SDV risk factor. Logistic regressions supported the overarching suicidal SDV risk factor as a predictor of suicide attempts (OR = 4.07, p < .001) and return to care for suicidal ideation (OR = 2.81, p < .001) over follow-up. However, ROC analyses suggested that the model including the suicidal SDV risk factor was only significantly better at classifying suicide attempts over follow-up (not return to care for suicidal ideation) than the model that did not include it (AUC difference = 0.15, p < .001). Findings suggest that the shared variance assessed across CDEs better predicts future suicide attempts beyond any individual suicide-related constructs. (PsycInfo Database Record (c) 2020 APA, all rights reserved).