Historically, overall tobacco use has been consistently higher among U.S. military personnel compared to civilian populations [1–7]. Tobacco has potentially far-reaching consequences for the U.S. military, including tobacco-related adverse effects on combat readiness, risks to the health and welfare of military beneficiaries, and avoidable costs incurred from added health care services as well as lost productivity. According to the Institute of Medicine, smoking affects military readiness by impairing physical performance and endurance, reducing vigilance and cognitive function, increasing the risk of motor vehicle crashes and other unintentional injuries, as well as resulting in work absenteeism [8]. The health effects of tobacco on all users, including military beneficiaries, include cardiovascular disease, respiratory disease, and cancer, as well as short-term adverse effects, such as acute respiratory illnesses, impaired wound healing, periodontal disease, and peptic ulcer disease [9]. Sadly, tobacco use accounts for one-sixth (nearly 17%) of the deaths among Department of Defense (DoD) beneficiaries, primarily among military retirees [10]. In fact, several investigators have estimated that the cost to the DoD resulting from tobacco use could exceed $1 billion per year [10, 11].

Recognizing the negative impact of smoking on military readiness, the DoD implemented several tobacco prevention/intervention programs, such as tobacco cessation counseling, direct access to specific tobacco cessation products, and instituting new workplace tobacco control policies [12, 13]. Institutional tobacco control policies included creating tobacco-free areas and buildings, restrictions on workplace product promotion and advertisements, and increasing the price for tobacco products within the military commissary system. Although these intervention programs and policies may have contributed to a 21% decline in smoking prevalence between 1980 and 1998, the prevalence of smoking has subsequently increased by 2.3% from 1998 to 2005 [14]. Unfortunately, overseas military deployments were shown to be associated with smoking initiation or resumption [15]. Thus, some of the increases in military smoking trends in the early 2000s may have been reinforced by prolonged deployments to Iraq and Afghanistan.

With efforts to reduce smoking in U.S. military Service members, recent data indicated that smoking prevalence in U.S. military and civilian populations were close to equivalent by 2015, 14% and 15%, respectively [16, 17]. However, smoking prevalence among veterans or those who separated from the military remained as high as 30% from 2010 to 2015 [18]. It is important to note that approximately one-third of veterans who enrolled in the Department of Veterans Affairs health care system never smoked. In contrast, the comparable figure in the U.S. general population was 57% [19]. Military and veteran populations differ from the general U.S. population with respect to age, sex, health status, and socioeconomic status. Hence, tobacco prevention, control and cessation programs used in civilian populations may need to be modified to increase effectiveness among military and veteran communities. While some prospective studies have been conducted to identify smoking behaviors among military and civilian populations, to our knowledge, no studies have examined smoking behavior at the time that active duty military Service members transition to veteran status. Given that this transition period could be stressful to Service members, an analysis of cigarette smoking patterns at this critical time of transition may prove beneficial in identifying more robust tobacco prevention and control measures among both military Service members and veterans. Therefore, the objective of our study was to examine cigarette smoking among military Service members during and after the time of transition in order to better understand and guide targeted smoking cessation, as well as prevention efforts.

Overall, 151,567 Service members volunteered, consented, completed a baseline survey, and were enrolled in panels 1–3 of the Millennium Cohort. Among these 151,567 Millennium Cohort Study participants, 81,564 Service members met the eligibility criteria (in military service at baseline with completion of the first follow-up survey) and were included in our study. Service members who had missing covariate data (n = 14,535) were then excluded from our analyses, resulting in a final study population of 67,029 participants, among whom 25% (n = 16,433) had separated from the military and were classified as veterans. Given that approximately 18% of eligible participants were excluded from our study due to missing covariate data, we used multiple imputation techniques to ensure that bias was not introduced by using only participants with complete data. These additional analyses resulted in similar findings.

The majority of our study population, shown in Table 1, were men (69.6%), aged 17–34 (66.5%), non-Hispanic whites (76.1%), married (57.3%), and possessed less than a bachelor’s degree from post-secondary education (68.3%). At baseline, compared to our Service member group, our veteran group had higher proportions of younger individuals and were skewed towards being female, not married, less educated, in enlisted ranks, in the Marine Corps, and previously part of the active component. Notably, smoking prevalence at baseline was higher in veterans (20.1%) than in Service members (17.4%). Both Service members and veterans showed a statistically significant decline in smoking prevalence over time (p = 0.0005 and 0.0009, respectively) (Fig 2).

When examining factors associated with baseline smoking, TRS was a significant risk factor in the unadjusted model, with all levels of TRS having increased odds of smoking compared to TRS of 3 years or above. We also observed unadjusted dose-response relationships between smoking and age, deployment, and alcohol consumption, as well as number of life stressors (Table 2). Additionally, unadjusted associations between smoking and sex, race/ethnicity, education, marital status, BMI, panel, military rank, Service component, Service branch, military occupation, PTSD, depression, and anxiety were statistically significant at alpha = 0.05 level. In the adjusted analysis, all variables remained statistically significant except for TRS, which was forced into the adjusted model as the primary exposure of interest. The only factor that remained strongly significant in the adjusted model was alcohol consumption (infrequent drinker: adjusted odds ratio [AOR]: 2.29, 95% confidence interval [CI]: 2.04–2.58; light drinker: AOR: 2.90, 95% CI: 2.57–3.26; moderate drinker: AOR: 4.06, 95% CI: 3.61–4.58; heavy drinker: AOR: 6.63, 95% CI: 5.86–7.51). Other factors which retained a modest, but still statistically significant effect in the adjusted model, included life stressors, PTSD, depression, anxiety, and certain demographic and military-related variables (see Table 2).

The results of our longitudinal analyses, shown in Table 3, include both unadjusted and adjusted odds ratios for smoking by separation status, as well as selected covariates of interest. Separation was not a risk factor in the unadjusted model. However, when adjusted for baseline smoking and other covariates, separation became a protective factor, and veterans were 20% less likely to be a smoker than Service members. Other variables significantly associated with current smoking status in the unadjusted analysis included smoking at baseline (a 35-fold increase), alcohol consumption, life stressors, PTSD, depression, anxiety, and certain demographic and military-related variables. Notably, deployment without combat exposure appeared to be a protective factor, while deployment with combat exposure was a risk factor. In the adjusted analysis, separation became a statistically significant protective factor for smoking, and the 17–24 age group changed from being a significant risk factor to a protective factor. Baseline smoking remained the most influential risk factor for current smoking status (AOR: 27.32, 95% CI: 25.94–28.78).

Furthermore, the adjusted analysis revealed a strong dose-response relationship between smoking and alcohol consumption (infrequent drinker: AOR: 2.70, 95% CI: 2.23–3.27; light drinker: AOR: 2.78, 95% CI: 2.29–3.37; moderate drinker: AOR: 3.25, 95% CI: 2.68–3.95; heavy drinker: AOR: 4.25, 95% CI: 3.47–5.19), and a more modest dose-response association was found between smoking and life stressors. Smoking was also associated with being non-Hispanic white, Army, on active duty, and screening positive for PTSD or depression. Additionally, the occupational group of electrical/mechanical equipment repair was also associated with smoking. Conversely, participants at reduced risk of current smoking were women, 17–24 years old, who were married, overweight or obese, officers, combat specialists, in panel 3, and college-educated (earned at least a bachelor’s degree).

Within our study population, 20% of those who would soon become veterans and 17% of those who would continue as Service members were identified as current smokers at baseline. Although smoking prevalence declined consistently in both veteran and active duty Service member populations over the 15-year interval of our study, smoking prevalence remained higher among veterans at all time points, confirming the 2009 report by the Institute of Medicine [8]. However, when explicitly examining any association between military separation and smoking, the findings of our study did not fully agree with the hypothesis that the time of transition from the military to civilian life is a risk factor for smoking. We identified, in the unadjusted model, that those who would soon leave active military Service were more likely to smoke at baseline than their counterparts who remained in the military throughout the study time period. However, after adjusting for other covariates, time remaining in Service was no longer a significant risk factor for smoking. In the longitudinal analysis, after adjusting for smoking at baseline and other characteristics, military separation in fact became a protective factor for smoking. We might speculate that serving in the military contributed to a culture of smoking (camaraderie, social norms, peer-pressure) as compared with different experiences at re-entry into civilian life, such as workplace smoking prohibition. It is also possible that Service members lost the financial resources in support of their smoking behavior while on active duty and, therefore, stopped smoking after military separation.

Although we only observed a moderate relationship between TRS and baseline smoking in the unadjusted model, this finding may still suggest that Service members may be smoking more prior to military separation. This ambiguous time period before separation (TRS) is defined by DoD as the time interval starting as early as 2 years before the separation date [30]. Our study findings could prove useful for developing and testing effective smoking interventions to support Service members before and after military separation. Our data did not inform differences among persistent smoking, smoking initiation, and/or smoking relapse, and therefore, we were not able to further explore differences between these groups. Future research focusing on identifying any differences in these groups may prove useful when developing and validating effective smoking prevention/cessation programs tailored to a transitioning population.

We also identified other influential factors associated with smoking in our study. Not surprisingly, smoking at baseline had the strongest association with continued smoking at follow-up. Alcohol use, depression, PTSD, and multiple life stressors also had moderate to strong associations with smoking. The majority of factors significantly associated with baseline smoking and current smoking status in our study were consistent with the existing literature [1, 15, 31, 32]. Baseline and current smokers were more likely to be men, non-Hispanic white, not married, not overweight/obese, in enlisted ranks, and possessing less than a bachelor’s degree from any post-secondary education. Service members in the active component are reported to smoke more than National Guard/Reserve personnel [15], also consistent with our findings.

The finding of deployment being a significant risk factor in our baseline model was aligned with the 2008 study by Smith et al. reporting that Service members smoke to relieve stress during deployment [15]. Interestingly, the effect of deployment on smoking moved towards the null in our longitudal analysis, with deployment without combat experience becoming a significant protective factor. It is possible that since deployment was assessed at baseline, the deployment effect diminished over time. The “healthy warrior effect” could also be a reason for this observation [33, 34]. Since good health status is a prerequisite for military deployment, it is possible that some deployed Service members stopped the smoking behaviors that they developed during deployments and gradually returned back to their healthier lifestyle after deployment.

Our study did identify several risk factors that differed from previous research. When comparing smoking prevalence across military Service branches, we found the highest smoking prevalence in the Army, which differed from DoD reports indicating that the Marine Corps had the highest smoking prevalence, followed by the Army [1, 3, 16]. It is possible that the lower smoking prevalence we observed in Marine Corps members was due to volunteer bias [35] since the Millennium Cohort participation rate for Marines was only 13%, compared to 26% in all other Services combined. It is also possible that health-oriented Marines may have been more willing to respond to a health-focused survey, thereby influencing the smoking rates found in our study.

In summary, military separation was not a risk factor for smoking, as we had hypothesized. Even though we found that Service members who were close to separation from the military smoked more than those whose separation was further into the future, this association did not retain statistical significance after adjusting for confounding factors. Our study did show that the prevalence of smoking declined among both veterans and currently serving military members over our defined follow-up period. Not surprisingly, baseline smoking was a strong predictor of smoking status at follow-up, suggesting that senior leaders should continue to actively support and strengthen smoking prevention and cessation efforts across the continuum of military service, starting at accession. Additionally, our study indicates that Service members transitioning out of the military may be at increased risk of smoking prior to their actual separation date. We also observed a strong dose-response relationship between smoking and alcohol consumption, as well as a modest association between smoking and life stressors, supporting the use of smoking cessation and prevention programs targeting specific groups. While confounding factors may have played a role in precisely quantifying these associations, increasing the emphasis on smoking cessation and prevention programs in any pre-separation counseling sessions may still reduce the number of smokers among these higher prevalence populations, and that is a worthy public health goal.