(PS13-47) The Longitudinal Relationship Between Anxiety Sensitivity Cognitive Concerns, Insomnia, and Suicidal Ideation in U.S. Military Personnel

Background: Suicide risk has been steadily rising among U.S. military personnel, prompting efforts to identify causal mechanisms. Research suggests anxiety sensitivity cognitive concerns (ASCC; e.g., “when I have trouble thinking clearly, I worry that there is something wrong with me”) and insomnia (e.g., difficulty initiating or maintaining sleep, daytime dysfunction, etc.) are positively associated with one another and are both risk factors for future suicidal ideation (SI). However, the sequential relationship between ASCC, insomnia, and SI has yet to be tested to our knowledge in a military sample. 

Methods: Using longitudinal data collected through the Military Suicide Research Consortium we assessed the magnitude and direction of these relationships. ASCC was measured using five cognitive items from the Anxiety Sensitivity Index-3, insomnia was measured using the Insomnia Severity Index, and SI was measured using the Depressive Symptom Index-Suicidality Subscale. Two regression models (n₁= 611; n₂=593) were tested: model 1 predicted time 3 (T3) SI from time 1 (T1) ASCC with time 2 (T2) insomnia as a mediator, and model 2 predicted T3 SI from T1 insomnia with T2 ASCC as a mediator. Two structural equation models (SEM) were also analyzed to account for missing data.

Results: Model 1 supported a partial mediation by insomnia of the relationship between ASCC and SI. In Step 1, T1 ASCC was significantly and positively associated with T3 SI (βc = .39, SE = 0.04, p< .001), with a medium effect size (sr² = 0.152). In Step 2, T1 ASCC was significantly and positively associated with T2 insomnia (βa = .45, SE = 0.04, p< .001), with a medium effect size (sr² = 0.198). In Step 3, both T1 ASCC (βc’ = .34, SE = 0.04, p< .001, sr² = 0.090) and T2 insomnia (βb = .13, SE = 0.04, p=.003, sr² = 0.013) were significantly and positively associated with T3 SI.

Model 2 supported a full mediation by ASCC of the relationship between insomnia and SI. In Step 1, T1 insomnia was significantly and positively associated with T3 SI (βc = 0.21, SE = 0.04, p< .001), with a small effect size (sr² = 0.043). In Step 2, T1 insomnia was significantly and positively associated with T2 ASCC (βa = 0.40, SE = 0.04, p< .001), with a medium effect size (sr² = 0.159). In Step 3, T2 ASCC was significantly and positively associated with T3 SI (βb = 0.36, SE = 0.04, p< .001, sr² = 0.110), while the association between T1 insomnia and T3 SI dropped to non-significance (βc’ = 0.06, SE = 0.04, p=0.126, sr² = 0.003).  

SEM supported partial mediation of both models.

Conclusion: Taken together, ACSS and insomnia appear to have a bidirectional relationship. Given that model 2 supported full mediation, ASCC may be a more heavily weighted initial treatment target for reducing future SI in military samples. However, more research is needed to determine if the longitudinal relationship between insomnia, ASCC, and SI is best modeled as a full mediation (as demonstrated by regression) or as a partial mediation (as demonstrated by SEM).