Introduction

Recent media outlets report an alarming number of military veterans who are ending their lives each day (Zoroya 2014). Paired with the tragic news of casualties suffered in the Afghanistan and Iraq wars, it may be natural to assume the approximately 22 military veteran suicides a day are young men and women who are returning from these wars. However, approximately 70 % (more than 15 of the 22 military veterans) are at least 50 years of age, largely affecting an older population of military veterans (Kemp and Bossarte 2012). Alongside the tragedy of losing someone to suicide, exposure to suicide (i.e., personally knowing someone who has died by suicide) can have damaging effects for those left behind (Cerel et al. 2014). For example, suicide exposure has been linked to psychological health complications, insomnia, hypervigilance, and dissociative episodes in military populations (Carr 2011). Persisting symptoms can develop into posttraumatic stress disorder (PTSD). Specifically for military populations, the suicide of a veteran can have similarly profound effects on other numerous military personnel, including former unit members, medical personnel who respond, mental health providers working with veterans, and an undeterminable number of other family and community members.

Shneidman (1972) conservatively estimated that each suicide will affect the everyday lives of at least six people. However, more recent reports suggest this number is grossly underestimated (Cerel et al. 2014). Given the increasing rate of suicide among military veterans, the impact of suicide exposure on veterans warrants investigation (Cerel et al. 2015). One avenue that has received little attention is the role of family variables as protective factors against the negative impact of suicide exposure. Previous research has suggested that marriage and having children can be protective in certain situations (Frech and Williams 2007; Cutright and Frenquist 2005); yet, these studies did not examine whether marital status was associated with lower levels of PTSD in veterans who have been exposed to suicide. Therefore, it is important to examine whether marital status is protective against PTSD in suicide-exposed veterans.

PTSD is a mental health disorder that is the direct result of exposure to a traumatic event. The criteria for PTSD are defined as clinical symptoms that occur after an individual is exposed to one or more traumatic events (American Psychiatric Association 2013). Symptoms can occur as intrusions (e.g., recurring and intrusive memories, distressing dreams), avoidance (e.g., avoiding external reminders that could cause distressing memories), negative changes in thought or mood (e.g., inability to remember aspects of the trauma, negative beliefs about the world), and arousal (e.g., irritable behavior, angry outbursts).

An estimated 20 % of Operation Enduring Freedom and Operation Iraqi Freedom veterans have PTSD, 15 % of Gulf War veterans have PTSD, and 30 % of Vietnam veterans suffer from PTSD in any given year (Gradus 2015). As a result, their spouses may experience associated trauma. Research has shown that veterans with PTSD experience family-related problems at a much higher rate than those veterans who do not experience PTSD. For example, divorce rates amongst veterans with PTSD are two times higher than those who do not suffer from the issue (Price and Stevens 2014). Other family-related problems include higher rates of domestic violence, increased sexual problems, and parenting struggles (Price and Stevens 2014). These problems are not limited to veterans of recent wars. In fact, research on Vietnam veterans shows similar results (President’s Commission on Mental Health 1978) suggesting that more needs to be done to understand the variety of factors associated with PTSD and its negative effects on the family unit.

Although determining the cause of PTSD among veterans can be extremely difficult when taking into account the numerous traumatic events experienced during service, suicide exposure is a traumatic experience that is easily identified as one of many stressors that military men and women face both in and away from the battle field (Cerel et al. 2015). Almost half of the adult general population has been exposed to the suicide of someone in their lifetime (Cerel et al. 2016). Cerel et al. (2015) reported that veterans exposed to suicide were almost twice as likely to have diagnosable depression and more than twice as likely to have diagnosable anxiety as individuals who were not exposed to suicide. They also found that individuals exposed to suicide were more likely to report suicide ideation than those not exposed.

Beyond the exposure to another’s suicide, individuals’ perceptions of the exposure’s impact on their lives are an important predictor of mental health outcomes. The examination of one’s perception of the event is a way to determine the amount of stress an individual may be experiencing due to the event. Previous studies found that one’s perception of his or her relationship to the decedent is a large indicator of the effect of the suicide (Cerel et al. 2016, 2015). However, these studies did not account for an individuals’ perception of how the suicide impacted their lives. Some individuals may admire or rely on another individual without feeling emotionally close to him or her; yet, the individuals’ death by suicide could still have a large impact on their lives with lasting effects. Lazarus and Folkman (1984) suggest that people differ in their sensitivity and vulnerability to certain events, as well as their interpretations and reactions. They found a stressful situation to be when an individual recognizes the demands of a given situation exceed their available coping resources (Lazarus and Folkman 1984). Perceived stress or impact of an event can have recognizable effects on both somatic and mental health; therefore the advantage of measuring it is to better understand an individual’s interpretation of their experience and their coping abilities to deal with the situation beyond a specific event in a given period of time (Lavoie and Douglas 2011). Therefore, measuring perceptions about the suicide’s impact is an important step.

Given that return from deployment may be as stressful as preparing to deploy, continued support from family and community is needed and may be a factor aiding in the prevention of PTSD. Kline et al. (2014) found that marriage was a protective factor in preventing suicide in a sample of National Guard service members, while Griffith (2012b) also found that social support is linked to fewer PTSD symptoms in a National Guard sample. Similarly, individuals reporting higher levels of social support tend to have lower rates of psychological distress and psychiatric disorders (Cohen and Syme 1985; Cohen and Williamson 1991; Sherbourne and Hays 1990). A spouse is perhaps the most important support figure post-deployment as the service-member processes combat experiences and readjusts to civilian life. Problems may arise, however, when the support system of the spouse becomes disorganized due to trauma either partner has experienced (Basham 2008). Studies of trauma survivors have also shown that those who perceive that they are cared for by others cope better than those who perceive that they have few resources (Krause 2004; Norris et al. 2002).

When examining whether family resources have a protective nature, Hill’s (1958) ABC-X model provides an excellent resource for conceptualizing the relationships between variables (McCubbin and Patterson 1983). In its most basic form, the ABC-X model proposes that the impact of a stressor (A) on an outcome variable (X) is affected by the available resources (B) and the way in which the stressor is perceived (C). Figure 1 displays how these components could play out among veterans who have been exposed to suicide. For veterans, knowing someone who has died by suicide would be a substantial stressor. However, if one has protective resources and perceives the suicide exposure as having a low impact on his or her life, a negative outcome (i.e., having clinically-significant PTSD symptoms) could be inhibited. Based on previous research and the ideas proposed by the ABC-X model, the following hypotheses should be true: (1) The proportion of veterans who have clinically-significant PTSD symptoms will be different among non-married and married veterans depending on whether they perceive the suicide exposure to have low or high impact on their lives; (2) Suicide exposure variables predict likelihood of having clinically-significant PTSD symptoms independent from other demographic and service-related variables for both non-married and married veterans; and (3) The perception of suicide exposure’s impact on veterans will have a larger effect on non-married veterans’ likelihood of having clinically-significant PTSD symptoms about the suicide than on married veterans’ likelihood of having clinically-significant PTSD symptoms.

Fig. 1
figure 1

Conceptual model of ABC-X model applied to suicide exposure. The stressor suicide exposure is filtered through resources available through marital status as well the perceived impact of the exposure to determine whether the veteran is likely to develop PTSD

Full size image

Method

Participants

Respondents were primarily male (89.8 %) and Caucasian (92.4 %), with ages ranging from 21–94 years old (M = 60.1, SD = 14.7). Additional ethnicities included African-American (4.6 %), Native American (1.2 %), Hispanic (1.2 %), Asian American (.3 %), and individuals who identified as Mixed or of multiple ethnicities (.3 %). Just over half (55.8 %) of respondents served in the Army, while 16.7 % served in the Navy, 15.3 % served in the Air Force, 8.3 % served in the Marines, .2 % served in the Coast Guard, and 3.7 % identified as serving in multiple branches. Number of deployments ranged from 0–30 (M = .93, SD = 2.17), and time spent in the armed forces ranged from 1–35 years (M = 8.20, SD = 7.98).

Procedures

Participants in this study were taken from a larger sample (N = 1736) that was recruited as part of a dual-frame, random digit-dial survey on suicide exposure in the Commonwealth of Kentucky from July 2012 to June 2013. For further information regarding the RDD methodology, see Cerel et al. (2016). Participants were included in this secondary analysis if they identified as a military veteran who had been exposed to suicide (n = 434). An a priori power analysis using G*Power (Faul et al. 2007) based on a two-tailed test with an alpha (α) value of .05, a beta (β) value of .20, and a medium effect size (odds ratio) of 2.5 yielded a recommended sample size of 54. Additionally, a sensitivity power analysis was calculated based on a two-tailed alpha (α) value of .05, a beta (β) value of .20, and our sample size of 423, which indicated sufficient power to detect a small effect size (odds ratio) of 1.4.

Measures

Marital Status

Participants were asked “What is your marital status?” and given five response options: single, married, divorced, widowed, and separated. For the purposes of this manuscript, these options were dichotomized (due to the small numbers of veterans in the divorced, widowed, and separated categories), with single, divorced, separated, and widowed individuals assigned to not married (0), and married individuals assigned to the married (1).

Perceived Impact of Suicide

The perceived impact of the participant’s exposure to suicide was measured by the question “Thinking about the effect of the person’s suicide on your life, what response is closest to your experience?,” which included 5 response options anchored by the death had little effect on my life (1) to the death had a significant or devastating effect on me that I still feel (5). The five response options were then dichotomized by combining the response options: the first three indicators into the low impact category, and the last two indicators of the scale into the high impact category. Items were dummy coded low impact (0) and high impact (1).

Posttraumatic Stress Disorder

The 7-item Short-Screening Scale for PTSD was used to measure self-reported PTSD symptoms specifically about the impact of the suicide exposure. Participants are asked whether they have experienced a list of PTSD symptoms (e.g., “Do you become jumpy or easily startled by ordinary noise or movements?”) with two response options, no (0) or yes (1). Responses are summed (ranging from 0–7), and scores greater than or equal to four indicate that the individual meets the probable criteria for a PTSD diagnosis. The sensitivity and specificity of the scale are 80 and 97 %, respectively (Breslau et al. 1999). Internal reliability of this scale yielded a Cronbach’s α of .87. Test for normality indicated that responses for this item were negatively skewed, even after applying a log transformation. Therefore, responses were dichotomized using the standardized cut-off scores. Scores equal to four or higher indicate clinically-significant PTSD symptoms.

Data Analyses

All statistical analyses were performed using SPSS 22 for Windows. Descriptive information was calculated for all predictor and outcome variables. To test H1, a three-way loglinear analysis was conducted using dichotomized variables of (Non-married = 0 and Married = 1; Low Perceived Impact = 0 and High Perceived Impact = 1) to determine whether the proportion of non-married and married veterans who have clinically-significant PTSD symptoms is different based on their perceived impact of suicide exposure. Finally, to test H2 and H3, hierarchical binary logistic regression analyses were used to determine the ability of suicide exposure variables to predict likelihood of PTSD diagnosis, after controlling for other contextual variables (i.e., demographic and military-related variables). For this analysis, the data was also stratified by marital status to determine whether effect sizes differed for non-married veterans compared to married.

Results

The majority of respondents were married (74.7 %); roughly 11.2 % were divorced, 7.6 % were single and had never married, 6.0 % were widowed, and .5 % were separated. The number of individuals who died by suicide known by each participant ranged from 1–50 (M = 3.03, SD = 4.14), and time since most recent death by suicide ranged from 0–73 years (M = 20.06, SD = 17.10). Of the 426 suicide exposed veterans who reported the amount of impact of the event, 7.9 % of participants reported the death had a significant or devastating effect on their lives, 9.5 % reported the death disrupted their lives in a significant way, 20.5 % reported the death disrupted their lives for a short time, 41.9 % reported the death had somewhat of an effect on their lives but did not disrupt them, and 20.3 % reported the death had little effect on their lives.

Table 1 includes descriptive information. The three way loglinear analysis produced a final model that retained the married × perceived impact of suicide exposure and perceived impact of suicide exposure × clinically-significant PTSD symptoms interactions. The likelihood ratio of this model was χ 2 (2, N = 414) = 3.33, p = .189. The married × suicide impact interaction indicated that veterans were less likely to have high-impact suicide exposure if they were married, χ 2 (1) = 8.17, p = .006, ϕ = −.14. The odds of a married veteran reporting high-impact suicide exposure were 2.19 times lower than the odds of a single veteran reporting high-impact, OR = .46, 95 % CI [.26, .79]. The suicide impact × clinically-significant PTSD symptoms interaction indicated that veterans with high-impact suicide exposure were more likely to have clinically-significant PTSD symptoms than veterans with low-impact suicide exposure, χ 2 (1) = 68.88, p < .001, ϕ = .41. The odds of a veteran with high-impact suicide exposure having clinically-significant PTSD symptoms was 10.04 times higher than veterans with low-impact suicide exposure, OR = 10.04, 95 % CI [5.42, 18.56].

Table 1 Sample demographics (N = 434)
Full size table

For both non-married and married participants, hierarchical binary logistic regression models including suicide exposure variables statistically enhanced the ability to predict the likelihood of having clinically-significant PTSD symptoms (see Table 2). The Hosmer and Lemeshow tests for significance were not statistically significant which specifies the models were a good fit. The full model for non-married veterans accounted for 60 % of the variance in likelihood of having clinically-significant PTSD symptoms. After controlling for demographic and service-related variables, which together explained 23 % of the variance, suicide exposure variables explained an additional 37 % of the variance. Female, non-married veterans were 4.68 times more likely to have clinically-significant PTSD symptoms (95 % CI [1.07, 20.54]), and non-married veterans were 4.58 times more likely to have clinically-significant PTSD symptoms for each level increase in perceived impact of the suicide exposure (95 % CI [2.12, 9.90]).

Table 2 Hierarchical binary logistic regression analysis for predicting clinically-significant PTSD symptoms (N = 423)
Full size table

The full model for married veterans accounted for 34 % of the variance in the likelihood of having clinically-significant PTSD symptoms. Suicide exposure variables explained an additional 20 % of the variance after controlling for demographic and service-related variables, which together explained only 14 % of the variance. For each year increase in age, married veterans were 4 % less likely to have clinically-significant PTSD symptoms (OR = .96, 95 % CI = .94, .99]). Years of combat deployment was also a significant factor for predicting clinically-significant PTSD symptoms in married veterans: For each year deployed in combat, veterans were 3.7 times more likely to have clinically-significant PTSD symptoms (95 % CI [1.56, 8.79]). Compared to the effect size for non-married veterans (OR = 4.58), married veterans were only 2.64 times more likely to have clinically-significant PTSD symptoms for each unit increase in perceived impact (95 % CI [1.82, 3.83]).

Discussion

This study examined the effects of marital status and perceived stress among military veterans in hopes of identifying protective factors for preventing PTSD among those exposed to suicide. Findings show that the perceived impact of suicide has a higher effect on the likelihood of having clinically-significant PTSD symptoms in non-married veterans compared to married veterans. This analysis is unique in its focus on veterans and whether marital status is a protective factor in preventing clinically-significant PTSD symptoms. These results strongly suggest that marriage and perceptions of impact of an event are predictor variables that need to be considered when examining variation in the levels of PTSD.

Although this study highlights an important protective factor for married veterans, it is unclear why or how this benefit occurs. One explanation might be the companionship and accountability offered by the partner in the marital union. The partner shares a genuine interest for the wellbeing of their spouse, and therefore acts as a thermostat monitoring the degree to which their spouse is responding or behaving “normally”. Given that the relationship is a healthy and loving union, the spouse also serves as a compassionate listener offering social and moral support. Another notable benefit of marriage is the ability of a spouse to encourage a hurting partner to express their emotions in a safe environment while reinforcing that they are not alone. Finally, a spouse is also best able to monitor high risk behavior and provide additional resources given the close proximity and daily interactions of a married couple. However, in order to determine whether this explanation holds true, additional research is needed that assesses not only relationship status but also relationship quality.

This study adds to prior research by supporting the findings of Lavoie and Douglas (2011) that suggest perceived stress or impact of an event can have recognizable effects on one’s somatic and mental health status. These findings also confirm that individuals with more social support have lower rates of psychological distress and psychiatric disorders (Cohen and Williamson 1991; Koeske and Koeske 1991; Sherbourne and Hays 1990). This analysis can also help mental health providers, practitioners, and legislative officials understand the importance of healthy marital relationships. Policies that support military marriages in light of lengthy and frequent military deployments may have a positive effect on marital relationships and produce healthier service members. Debriefing efforts after traumatic events occur may also be strengthened by the inclusion of questions that ascertain perception of the said event, and by allowing additional services to be offered to those veterans that identify the event as highly impactful. Service providers may also want to consider promoting community resources that build and strengthen social support networks.

Although this study revealed important findings some notable limitations remain. Due to the majority of respondents’ age being 60–79 years old, this sample is not fully representative of all current military veterans and therefore; may have more limited generalizability to recent returning veterans. Possible cohort effects and other lifespan developmental trajectories, such as chronic PTSD, may also exist among a prominently older veteran population and are worthy of further exploration. The ability to define relational support was also limited to marital status, as the quality of marriage and other measures of social support were not captured in the data. We also could not ascertain temporal sequence of PTSD diagnosis and marriage status, which could give further insight, particularly for those who were divorced, possibly due to relationship issues stemming from untreated PTSD symptoms. Of note, the PTSD symptoms were queries specifically in reference to the suicide exposure itself. Other data that would have indicated support, such as education level and financial income, were not captured in this dataset and could not be included in this analysis. Furthermore, PTSD about other traumatic events was not captured in this study along with other known correlates with PTSD, such as anxiety, depression, and moral injury. We recommend use of these indicators in future studies. Future studies may also be undertaken to ascertain sex differences in the odds of PTSD outcomes.

The study findings support the hypothesis that military veterans who indicated high perceived impact of suicide exposure will have increased odds of having clinically-significant PTSD symptoms. The results also indicated that being married buffered the odds of military veterans having clinically-significant PTSD symptoms in support of our second hypothesis. It is important to note that although each group was only compared to the control group (married/low impact), one may deduct that military veterans who were married and indicated high perceived impact of suicide exposure were half as likely to have clinically-significant PTSD symptoms about the suicide compared to military veterans who were not married and indicated high perceived impact of suicide exposure. The third hypothesis was also supported revealing military veterans who are not married and indicate high perceived impact of suicide exposure have the greatest odds of having clinically-significant PTSD symptoms. Overall, the results give us a greater understanding of the effects of marriage and military veteran’s perception of the impact of suicide exposure on PTSD outcomes.