Participants were enrolled prospectively in the Department of Veterans Affairs Polytrauma Rehabilitation Centers TBI Model Systems (VA PRC TBIMS) national database (a multicenter, longitudinal study of TBI outcomes). Follow-up interview was attempted for all participants. Currently, there are 5 sites across the country enrolling service members and Veterans with a TBI into the VA PRC TBIMS database, which has been in existence since 2010. All VA PRC TBIMS enrollees are age 18 or older and were transferred to a comprehensive rehabilitation program at one of the participating Polytrauma Rehabilitation Centers.⁴¹ All participants provided informed consent directly or by legal proxy. This analysis was conducted with a subset of participants meeting study-based inclusion and exclusion criteria as detailed below.

VA TBIMS database enrollees who enrolled and were discharged between August 2009 and June 2015 were considered for analysis. Further inclusion criteria were 1) completion of inpatient rehabilitation, 2) enrollment within two years of index TBI, and 3) eligibility for and completion of a two-year post-injury follow-up. Individuals were excluded if deceased, refused interview, or withdrawn from the study at the time of follow-up. Individuals who were widowed at time of injury or were missing marital status at follow-up were also excluded from analyses. Veterans or service members are enrolled in the database if they are admitted to inpatient rehabilitation and have a diagnosis of TBI, even if TBI is not the primary reason for admission.

After enrollment, trained research assistants obtained demographic information, including pre-injury relationship status, and premorbid status via interview with the participant or from a proxy for those unable to provide reliable data for themselves.⁴¹ Research staff collected data on injury characteristics, military history, and FIM™ scores via review of medical charts from the acute and rehabilitation hospital stays, consistent with the protocol for the TBIMS database.⁴⁷ Relationship status at 2 years post-injury was collected as part of the 2 year follow-up interview conducted by telephone with the person with injury or a proxy. The window for follow-up data collection was within three months of the injury anniversary date.

Data were analyzed using statistical software R v3.2.2 (R Foundation for Statistical Computing, Vienna, Austria). Descriptive statistics were expressed as 1st quartile; median; 3rd quartile for continuous variables and percentage (count) for categorical variables. Subgroup analysis was performed among those married at enrollment and among those who were single, divorced, or separated at enrollment respectively.

For the subgroup analysis of those married at enrollment, due to a relatively small sample size (i.e. N=33 negative changes at two-year follow-up), a logistic regression model with Firth’s penalized likelihood approach was fitted for predicting negative change (Yes vs. No) at two-year follow-up as a function of each risk factor to evaluate the univariate association between negative change and each risk factor. Redundancy analysis was performed to check the collinearity among all the risk factors to be included in the multivariable model. None of the variables were found to be redundant. A multivariable model with Firth’s penalized likelihood approach was fitted for predicting negative change at two-year follow-up as a function of age at injury, education, injury severity, cause of injury, injury during deployment, FIM cognitive score at discharge, FIM motor score at discharge, mental health utilization prior, and problematic substance use.

For the subgroup analysis of those who were single, divorced, or separated at enrollment, a similar approach was used for predicting positive change (Yes vs. No) at two-year follow-up among the combined subset. When performing redundancy checks among all the risk factors to be included in the multivariable model, cause of injury was found to be redundant, with 77.7% of its variation explained by other covariates. Therefore, cause of injury was dropped from the multivariable model.

The odds ratio (OR) estimate, its’ corresponding 95% confidence interval (CI) and p-value were computed for each risk factor based on the above models. P-values ≤ 0.05 were considered to indicate statistical significance.

A total of 712 participants were enrolled in the VA PRC TBIMS database. As detailed in Figure 1, 101 participants were excluded from analyses due to not having enrolled within two years of TBI, and 158 were not eligible for two year follow-up. Of the remaining sample of 453 eligible participants, five were expired at time of follow-up. At the two-year follow-up, 70 participants were lost, refused, or withdrew (15% lost to follow-up rate), leaving a sample of 378 participants with completed two year follow-up evaluations. Participants were further excluded if missing marital status at two-year follow-up (N=18) or were coded as widowed, leaving a final sample of N=357.

A total of 357 individuals remained in the final analytic sample. The majority of participants sustained a severe TBI and were injured stateside. The majority of injuries were caused by vehicular accidents, with blast injuries the second most frequent. There were an equivalent number of persons single and married at the time of injury, with fewer divorced or separated. Among those married at enrollment, 23% reported mental health utilization during the year prior to injury and 16% reported problematic substance use. Among those single/divorced/separated at enrollment, 21% reported mental health utilization during the year prior to injury, and 37% reported problematic substance use.

Marital status at enrollment and Year 2 follow-up is summarized in Table 1. Among 357 individuals, 151 (42%) of them were married at enrollment, 134 (38%) were single at enrollment, 56 (16%) were divorced at enrollment, and 16 (4%) were separated at enrollment. Among 151 married at enrollment, 118 (78%) of them remained married at Year 2, and 33 (22%) underwent negative change at Year 2 (i.e. single/divorced/separated). In the combined subset of 206 single/divorced/separated at enrollment, 179 (87%) remained unchanged at Year 2, and 27 (13%) underwent positive change at Year 2 (i.e. married).

Table 2 summarizes study variables for those participants who were married at enrollment, those who remained married at Year 2, and those who underwent negative change at Year 2. Persons who remained unchanged were older than those who underwent negative change (median age: 34.5 years vs. 26 years). There was a higher percentage of persons with more than a high school education in the unchanged relationship status group compared to the negative change group (59% vs. 39%). There were also more people with severe TBIs in the negative change compared to the unchanged group (84% vs. 60%).

The results from univariate and multivariable models representing negative change (i.e., going from married at enrollment to single, divorced, or separated) are shown in Table 3. Based on univariate analysis, age at injury was found to be significantly associated with negative change at Year 2. When we evaluated the impact of all risk factors simultaneously by running the multivariable model, age at injury, education, and prior mental health utilization were found to be significantly associated with negative change. Assuming all other predictors constant, as age at injury increased by 1 year, the odds of negative relationship change decreased by 9% (95% CI: 4% to 15%, p<0.001). The odds of negative relationship change were 65% greater (95% CI: 5% to 88%, p=0.040) for persons with less than a high school education compared to those with greater than high school education. The odds of negative relationship change among persons with prior mental health utilization was 4.80 times greater than those without prior mental health utilization (95% CI: 1.21 to 22.35, p=0.025).

The same analysis was replicated among the subset of persons who were single, divorced or separated at enrollment. Table 4 summarizes study variables for all persons who were single, divorced or separated at enrollment, persons who remained unchanged at Year 2, and those who underwent positive change at Year 2. Compared to those who were married at enrollment, this subgroup had less mental health utilization and more problematic substance use. More persons in the positive change group had been injured during deployment compared to unchanged persons (67% vs. 30%).

The results from univariate and multivariable models representing positive change (i.e., going from single, divorced, or separated at enrollment to married) are shown in Table 5. Based on univariate analysis, age at injury and injury during deployment were found to be significantly associated with positive change at Year 2. When we evaluated the impact of all risk factors simultaneously by running the multivariable model, injury during deployment was found to be positively associated with positive change at Year 2. Assuming all other predictors constant, the odds of positive change among persons injured during deployment was 5.36 times the odds among those injured stateside (95% CI: 1.68 to 20.08, p=0.004).

This study assessed relationship stability among individuals who received inpatient rehabilitation at a VA Polytrauma Center following primarily moderate-to-severe TBI. Findings may differ among individuals with mild TBI, civilians with TBI, those with moderate-to-severe TBI who receive inpatient rehabilitation in a civilian setting, or those admitted to acute care but who do not receive inpatient rehabilitation. The sample size is modest and the proportion maintaining relationship stability high; therefore, the samples in which relationship change was being examined were small. The self-report nature of the marital status variable also resulted in some inconsistencies in responses (e.g., some participants who went from single to divorced) and the possibility exists that changes occurred within the two year timeframe that were not captured. The predictor variables studied were limited to those variables available in the TBIMS database. This database does not include certain variables that might contribute to marital stability, including age at marriage, years of marriage, whether the participant is married to the same person at each time point, partner demographics, religious beliefs, presence of children from the marriage, relationship satisfaction at time of injury, and behavioral dysfunction. Lastly, this study looks at the important question of marriage stability in the first couple of years following TBI, but does not consider the equally important question of marital satisfaction or relationship quality. Future research should endeavor to replicate these analyses with larger samples and include additional variables known to impact marital stability. Additionally, future research should further explore predictors of positive relationship change and explore strengths associated with persons showing positive change in an effort to facilitate them.

In spite of limitations, the current findings have clinical implications for working with injured military service members and Veterans and their spouses/partners. For example, couples/marital counseling could be included as part of Polytrauma rehabilitation, particularly for those patients who have a premorbid mental health history, are younger, and have less education. Identifying at-risk couples early on and getting them into appropriate treatment may better prepare them for navigating marital problems post-TBI. Such counseling may also help injured individuals who may be more susceptible to mental health problems. Practitioners should also be familiar with the unique stressors related to deployment and/or military training in general,⁵² which may require individualized treatment in addition to marital therapy. Moreover, spouses may require their own separate training for serving as a caregiver in an effort to reduce caregiver burden, which has been shown to compromise caregiver health in caregivers of persons with TBI.⁵³ Given the growing research on divorce and marital strain related to military service in general,^54,55 a family systems-based approach⁵⁶ following a TBI among military service members and Veterans should be emphasized in routine rehabilitation efforts.