Introduction: Veterans Health Administration (VHA) electronic health records (EHR) are a rich source of health information, but Veterans who use VHA are systematically different than those who do not. EHR data provides active surveillance on utilizers within that system but may be subject to collider bias if selection is conditioned on the exposure-response function. VHA studies of military-related exposures may be especially prone to collider bias because VHA use is related to both military exposures and health status. This collider bias may be especially inflated in the context of mental health outcomes because veterans with mental health symptoms are more likely to engage in care-seeking behavior at VHA. This project will evaluate the magnitude of collider bias in associations between deployment and post-traumatic stress disorder when limiting the sample to Veterans who receive care from VHA.

Methods: This study used prospective cohort survey data from the Millenium Cohort Study (MCS), a study with up to 24 years of follow up that was designed to evaluate the impact of military service on health outcomes. Active servicemembers are sampled and are prospectively followed through the military separation period, providing self-reported health information among a sample of veterans that includes VHA users and VHA non-users. We estimated relative risk separately in veteran participants and VHA users.

Results: In a cohort of 146,688 veterans, 71,017(48%) used VHA. In unadjusted models, VHA users who experienced deployment with combat  were at an increased risk for PTSD (RR 1.4; 95% CI 1.3 to 1.5) compared to non-deployers, and this association persisted in the target population of veterans  (RR 1.8; 95% CI 1.7 to 1.9).

Conclusion: Results suggest that magnitude of associations differ between VHA users and veterans. In population-based epidemiologic research of veterans, the utility of VHA EHR may be limited by collider bias.