1. Introduction
Nowadays, risk-taking in military populations is often a substantial public health concern and a leading cause of preventable deaths among military personnel [
1]. Reports estimate that 31% of these deaths are caused by accidents linked to risk-taking post-deployment [
2]. This data mirrors studies demonstrating that veterans involved in lethal road traffic accidents were more likely to drive above the speed limit, not wear a seatbelt, smoke and misuse alcohol [
3]. Violence and fighting have also been highlighted as a major concern in military populations, with prevalence estimates ranging from 13% to 58% [
4]. The impact of risk-taking is further compounded by the significant associated health care costs, with alcohol and smoking attributed to cost the US military over
$1.2 annually [
5].
Despite consistent reports on the impact of risk-taking among military personnel, few studies have investigated help-seeking populations. Help-seeking veterans are outlined as a specific high-risk group for risk-taking due to their elevated health difficulties [
6]. A longitudinal study found that help seekers who suffer from physical and mental health problems often engage in various risk-taking behavior long after leaving the military [
7], further exacerbating their difficulties [
8]. Worryingly, reports illustrate that 62.4% of veteran deaths following treatment were linked to behaviors that could be prevented [
1].
Strong theoretical frameworks for elevated risk-taking in help-seeking veterans, however, remain ambiguous. By nature, military personnel are exposed to situations of uncertainty and violence, with shooting, fighting, and inter-unit competitions forming core components of basic training [
9]. Some frameworks illustrate that the normalization of these behaviors is linked to automatic behaviors in civilian life [
10]. Moreover, the elevated prevalence of mental health difficulties among help-seekers, including traumatic brain injuries and Post Traumatic Stress Disorder (PTSD), have been associated with biological changes in the brain, affecting attention and decision making [
11,
12,
13,
14]. Research conducted on US veterans shows that PTSD symptoms consistently correlate to increased substance misuse, suicide attempts, thrill-seeking and aggression [
15]. Other theories propose that individuals who join the military are naturally more inclined to risk-taking [
16]. A report published by the Department of Defence on 28,546 serving personnel highlighted that 78% classified as sensation-seekers [
17].
Whilst these initial findings provide relevance to inform prevention for US personnel, corresponding research into risk-taking among UK veterans remains scarce. In comparison to the US, UK personnel often differ in terms of military service, socio-demographics, and health difficulties, and the availability of post-deployment care shadows the US [
18]. Insights into the predictors of risk-taking among UK help-seeking veterans are urgently needed to advance risk assessments and interventions across healthcare providers. This is the first study to examine the potential predictors of risk-taking behavior in a nationally representative sample of UK help-seeking veterans.
2. Materials and Methods
2.1. Ethics
This study received approval from the Combat Stress Ethics Committee. Participants gave informed consent and all data were anonymized to ensure no identifiable participant information was reported.
2.2. Design
The current study employed a cross-sectional design to explore the predictors of risk-taking in help-seeking veterans. The primary outcome variable was a composite variable of risk-taking, aligning with previous epidemiological methods [
19]. Potential predictors included demographics and health presentations.
2.3. Participants
Participants were required to be UK veterans, defined as having served a minimum of one day in the UK military and have contacted Combat Stress (CS) between January 2015 and January 2016. CS is a national UK charity providing clinical mental health services for veterans with PTSD and is commissioned by the National Health Service. To create a representative sample of UK help-seeking veterans, no exclusion criteria were applied. The total number of veterans meeting our criteria was 3335. From this, a 20% (n = 667) randomly selected sample was obtained via random number generation. 403 participants completed the study and 51 opted out. After removing any participants who were deceased (n = 4) or had invalid contact details (n = 63), the final adjusted response rate was 67.2% (n = 403).
2.4. Measures
2.4.1. Demographics.
A self-report questionnaire collected demographics, including; sex (male/female), age, relationship status (single/in a relationship), employment status (employed/unemployed), whether the participant had left the Armed Forces early (<5 years; yes/no) and the length of time taken to contact Combat Stress (≥5 years; yes/no). Age was grouped into four categories: less than 35, 35 to 44, 45 to 54 and 55 years and over to aid clinicians in identifying the age bracket most at risk.
2.4.2. Health Presentations.
PTSD was assessed using the Posttraumatic Checklist for the DSM-5 (PCL-5; [
20]), which has shown high validity (sensitivity 0.88; specificity 0.69) and reliability in treatment-seeking veterans [
21]. The PCL-5 comprises of 20 items with a Likert scale ranging from 0 (not at all) to 4 (extremely) asking how much in the past month individuals have been affected by symptoms of PTSD. A total score was calculated and participants who scored 38 or more met criteria for probable PTSD [
20]. Individual scores for the symptom clusters were summed to investigate the association between intrusive thoughts (items 1 to 5), avoidance (items 6 to 7), negative alterations (questions 8 to 14), hyperarousal (items 15 to 20) and risk-taking.
Common mental health difficulties were assessed using the General Health Questionnaire (GHQ-12) due to the high validity (sensitivity 0.76; specificity 0.83) in clinical populations [
22]. The GHQ-12 uses 12 items with a Likert scale ranging from 0 to 3. The last six item scores were reverse scored to match the same scale as the first six. A score above four marked a probable diagnosis of anxiety and depression [
23].
Traumatic Brain Injury (TBI) was assessed using the Brain Injury Screening Index (BISI; [
24]). The BISI has been used in previous veteran studies [
8]. Participants reported their most serious head injury and endorsement of any one symptom indicated a TBI.
Finally, physical health problems were measured using the self-reported question “do you have any problems relating to the following physical health areas? Tick all that apply”. A list of 13 problems was taken from an NHS screening tool used within a South East Hospital in England. These included; diabetes, cardiovascular problems, blood pressure, respiratory problems, liver or kidney problems, amputation, neurological problems, digestive problems, chronic pain, poor mobility, hearing impairment, sight impairment, communication problems. The total number of problems was summed and scores were then evenly split into either a ‘high-risk’ or ‘low-risk’ group for physical health problems, as informed by the sample median post-hoc [
25].
2.4.3. Risk-Taking Behaviors
The primary outcome ‘risk-taking behavior’ was a composite variable based on previously established military risk-taking behaviors, including; heavy smoking [
26], alcohol misuse [
27], risky driving [
28], and fighting [
3]. Cases of risk-taking behaviors across variables were summed to create a continuous mean risk score [
19].
Survey items regarding smoking asked participants whether or not they smoked and if so, how many cigarettes they smoked a day. If participants smoked ≥20 cigarettes a day, they were classed as risky smokers [
28].
Risky alcohol consumption was measured using the Alcohol Use Disorders Identification Test (AUDIT) to identify high drinking levels due to the high reliability and validity (sensitivity 0.86; specificity 0.89) showed in clinical and military populations [
29,
30]. The AUDIT comprises of 10 items rating quantity and frequency of alcohol consumption over the past month on a scale of 0 to 4, where 0 represented ‘never’ and 4 was ‘daily or almost daily’. A score of 16 or above was classed as risky drinking [
31].
Risky driving was assessed using a measure adapted from a King’s College study on risk-taking in military personnel [
28]. Participants were asked the following questions: how fast they drive in a built-up area using forced-choice options: (1) within 5 mph of the limit (2) 6–10 mph above the limit and (3) more than 10 mph over. Participants were then asked how fast they drive on the motorway: (1) within 10 mph of the limit (2) 11–20 mph above the limit and (3) more than 20 mph over. If participants scored in the highest category in either question (3 points), this indicated risky driving [
28].
Fighting was examined through the Walter Reed Aggression Scale (WRA-4). This measure was chosen due to it being specifically designed for military populations and used in a range of previous veteran studies [
3,
32,
33]. The measure comprised of four items rating both verbally and physically aggressive behavior on a 5-point Likert scale, where 1 represented ‘never’ and 5 represented ‘5 or more times’. Veterans were asked how often they exhibited the fighting behaviors in the past month: “got angry at someone and yelled or shouted at them,” “got angry with someone and kicked or smashed something, slammed the door, punched the wall”, “got into a fight with someone and hit the person,” or “threatened someone with physical violence.” Answering ‘yes’ to any of these behaviors within the past month indicated fighting behavior [
33].
2.5. Procedure
The data collection for this study was conducted at Combat Stress in Leatherhead from January to September 2016. The questionnaire was first mailed to participants three times and then remaining non-responders were phoned by the researcher to minimize response and selection bias. After three calls, no additional contact attempts were made.
2.6. Statistical Analysis
The data was cleaned using frequency lists to check for invalid characters, numbers outside of the valid range and missing values, to minimize human errors. A multiple imputation protocol was put in place where a mean score would only be calculated to replace a missing value when less than 20% of the questionnaire data were missing; otherwise, data were excluded [
34].
A risk variable was created from the total scores of each measure (smoking, alcohol misuse, risky driving, and fighting) which were coded according to their clinical cutoff scores. Scores above the clinical cut-off point were coded as ‘high-risk’ as they met criteria for caseness indicating risky behavior, and scores below the cut-off were coded as ‘low-risk’. Once binary variables for each variable were created, the risk-taking variables were combined to create a continuous composite primary outcome variable of ‘risk-taking behavior’ as guided by the additive model by Pickett et al. (2002) [
19].
First, unadjusted linear regressions were employed to explore the relationship between each demographic variables (age, sex, relationship status, employment status, early service leaver status and time to contact CS) and the composite risk-taking behavior. The same analyses were repeated adjusting for these demographics.
Second, separate linear regression models were fitted to each health outcome (PTSD, common mental health difficulties, TBI and physical problems) to explore predictors of risk-taking. A multivariate regression was conducted adjusting for each of the other health outcomes. For example, if PTSD was examined, common mental health difficulties, TBI, and physical health were adjusted for. In addition, all analyses adjusted for age and relationship status.
Finally, a linear regression model was independently fitted to each PTSD symptom cluster, with risk-taking as the primary outcome. To minimize the influence of common confounders, subsequent analyses adjusted for the other symptom clusters, age and relationship status. All analyses were conducted in 2017 using STATA Version 13.0 (StataCorp LLC; College Station, TX, USA).
5. Conclusions
This was the first study to explore demographics and health as predictors of risk-taking in a sample of UK help-seeking veterans. 87% of our sample had engaged in risk-taking in the past month, with fighting (77.9%) and heavy drinking (40.2%) most common. Veterans who were younger and in a relationship were more likely to engage in greater risk-taking, alongside those who met case criteria for probable common mental health difficulties, PTSD and TBI. Finally, a direct association was found between higher hyperarousal, more negative alterations in mood and cognition and increased risk-taking. The development of additional screening and focused interventions for risk-taking is vital to lessen morbidity rates and preventable deaths among veterans.