Return to search

Predicting and explaining incident and ongoing depression in U.S. Army National Guard members: a lifecourse perspective

The National Guard is a unique, part-time subset of the U.S. military that has been increasingly deployed during recent conflicts, often has a different set of life circumstances compared to full-time Active Duty servicemembers, and is studied much less frequently than are Active Duty populations. Depression, one of the most common mental disorders among both civilian and military populations in the United States (US), is associated with a range of comorbid mental and physical health conditions. The associations between stressful life events throughout the civilian lifecourse—including during childhood—and adult-onset and persistent depression have been documented in some demographic groups, but have not yet been studied in a National Guard population. Stressful civilian life events may be particularly important in this population, due to frequent transitions between military and civilian employment and engagement among Guard members. We used data from the Ohio Army National Guard Mental Health Initiative to investigate the relationship between two domains of civilian life experiences from across the lifecourse and adult depression: (1) early-life adverse experiences, such as being mistreated during childhood, and (2) more recent stressful experiences outside of deployment, such as financial problems or divorce.

First, we estimated the relationships between each of these two domains of exposure and the rates of incident depression across four years using Cox proportional hazards models. We found that male servicemembers who reported at least one out of four traumatic childhood events assessed had a 77% higher rate of incident depression during follow-up compared to those who reported no traumatic childhood events, after adjusting for race and age group (95% CI (confidence interval): 1.33, 2.49). When further adjusting for posttraumatic stress disorder (PTSD) in the time between childhood events and depression, this relationship only slightly attenuated (aHR (adjusted hazard ratio) = 1.71, 95% CI: 1.24, 2.35), suggesting that the relationship between traumatic childhood events and adult depression is not driven by PTSD. Furthermore, when stratified by income level, the association between traumatic childhood events and depression was stronger among men making $40,000 per year or less, with an adjusted hazard ratio of 2.06 (95% CI: 1.22, 3.49), compared to men making more than $40,000 per year (aHR = 1.63, 95% CI: 1.09, 2.45).

We also found that men who reported at least one out of nine stressful events assessed in the prior year (a time-varying exposure updated over time) had twice the rate of incident of depression overall compared to men who reported no past-year stressful events (95% CI: 1.52, 2.72), adjusting for race, age group, and past-year PTSD. We observed imprecise associations between these exposures and incident depression among women that should be interpreted with caution due to the small sample size (aHR for one or more childhood events: 0.74, 95% CI: 0.35, 1.61 and aHR for past-year stressors: 1.07, 95% CI: 0.57, 2.01).

Our second study departed from traditional epidemiologic null hypothesis testing methods, taking instead a prediction approach to studying incident depression. Supervised machine learning—including methods such as tree classification and random forests—have the flexibility to identify predictors that are not pre-specified, and can be used for hypothesis generation. Among both male and female soldiers, we found that reporting verbal abuse by a parent or guardian during childhood, being of mid-level rank status in the military, recently deploying to a non-conflict area, having been robbed, and having been mistreated were all important predictors of incident depression across five years of follow-up. PTSD and traumatic events in adulthood (including combat-related experiences) as well as having children appeared more important for prediction among men compared to women, while military characteristics (e.g., years of service) as well as hearing about traumatic events happening to others (e.g., learning that a family member was in a serious car accident) appeared more predictive of depression for women compared to men. We also identified subgroups of individuals with certain combinations of predictors who were at high risk of depression onset, such as men with both past-year PTSD and a casualty in their unit during their most recent deployment. Overall, prediction accuracies of our algorithms were moderate to good when cross-validated.

Our third study returned the specific focus to our two main exposure domains of interest, childhood traumas and adult civilian stressors, but took a different approach for understanding depression as an outcome. While our first two studies assessed depression as a binary construct, our third study identified latent sub-groups of depression symptom patterns—or trajectories—across follow-up using latent class growth analysis, and estimated the associations between life stressors and membership into these different trajectory groups. For both men and women, a four-group depression model was identified, including a stable, symptom free group (showing essentially no depression symptoms at any point during follow-up) that included about 62% of the overall sample, an increasing depression symptom group including 13% of the sample, a decreasing depression symptom group with 16% of the sample, and a “chronic” depression symptom group representing 9% of the sample (staying essentially steady around 4-5 symptoms throughout follow-up).

After controlling for sex, race, and age group, soldiers who reported one or more traumatic childhood events had 3.57 times the odds (95% CI: 2.53, 5.05) of belonging to the chronic depression symptom group compared to the symptom free group. Reporting childhood events was also associated with being in the decreasing and increasing depression symptom trajectory groups compared to the symptom free group (aOR (adjusted odds ratio): 2.33, 95% CI: 1.75, 3.11 for the decreasing symptom group and OR: 1.78, 95% CI: 1.29, 2.45 for the increasing symptom group). When controlling for sex, race, age group, and past-year PTSD, time-varying adult stressors had the largest effect on depression symptoms for the increasing depression symptom group compared to other groups, particularly in the last two years of follow-up (where there was an adjusted difference of 1.02 symptoms at each year, for stressors compared to no stressors). The decreasing depression symptom and symptom free groups saw a negligible difference in symptoms when comparing one or more stressors to no stressors, while about a half of a symptom difference was seen for the chronic depression symptom group, unchanging across the follow-up time.

All three studies in this dissertation indicated the importance of considering stressful life events that occur outside of deployment when studying the mental health of National Guard servicemembers. These findings may be particularly relevant given the frequent switch between military and civilian engagement in the National Guard, and the relative neglect of this group within military research. Furthermore, our novel machine learning findings helped to bridge the gap between population-level and individual-level prediction of depression among National Guard members. Although replication studies are needed, the results of this dissertation may help inform potential intervention strategies for depression in order to reduce the overall disease burden of the U.S. Army National Guard.

Identiferoai:union.ndltd.org:bu.edu/oai:open.bu.edu:2144/40706
Date07 May 2020
CreatorsSampson, Laura
ContributorsGalea, Sandro
Source SetsBoston University
Languageen_US
Detected LanguageEnglish
TypeThesis/Dissertation

Page generated in 0.0029 seconds