Abstract
Although the effects of combat exposure on mental health receive a good deal of attention, less attention has been directed to the long-term effects of combat exposure on physical health, apart from combat injuries. Using the 2010 National Survey of Veterans, the author evaluates the long-term effects of combat generally, as well as more specific dimensions of combat experience, including exposure to the dead and wounded. The results indicate that combat exposure increases the likelihood of poor health and disability years later, though exposure to the dead and wounded is even more powerful and accounts for the entire unadjusted relationship between combat service and health. Furthermore, only a small part of the relationship between combat and health is attributable to service-connected disabilities. More of the relationship can be explained by social and behavioral processes subsequent to combat. Relative to veterans who were not exposed to the dead or wounded, veterans with combat exposure are more likely to smoke, less likely to be married, and more likely to report being unprepared for the transition to civilian life. Together these factors explain a large part of the relationship between combat and health, even more than is explained by service-connected disabilities. The effects of feeling unprepared on physical health are as large as those of exposure to combat. The results encourage greater appreciation of combat exposure as a source of stress proliferation, with ongoing implications for health channeled through experiences in civilian life.
Although research has begun to piece together the short- and long-term effects of combat on psychiatric disorders, it has focused less on more general aspects of physical health (Levy and Sidel 2009). To the extent that it has focused on physical health, most investigations have focused on a narrow set of specific issues or obvious service-connected problems. For instance, research has often focused on the direct and enduring consequences of combat injuries, including traumatic brain injury and limb damage (Cross et al. 2011; Glynn 2012). Other research has focused on different long-term effects, though in a similarly specific fashion. Research has focused, for instance, on the effects of exposures to specific toxins, such as agent orange and its relationship to cancer (Chretien and Chretien 2013). Related to this focus is how the U.S. Department of Veterans Affairs (VA) determines service-connected disability, adjudications that necessarily involve understanding the long-term effects of certain exposures and events (see Autor, Duggan, and Lyle 2011 for trends in disability and employment among Vietnam-era veterans).
From the standpoint of understanding the health effects of combat these investigations are illustrative and important, but they are limited in that they precede from a set of potential claimants and certain presumptive conditions backward, rather than trying to more inclusively understand the long-term effects of combat across a range of health outcomes and service experiences. Focusing on the continuation of service-related injuries and exposures might be appropriate from the standpoint of serving veterans or for understanding the epidemiology of combat wounds but is increasingly out of step with research on the long-term and capacious effects of trauma. Recent evidence points to much broader possibilities. In a study of aging veterans, based on a supplement to the Health and Retirement Study, Taylor, Ureña, and Kail (2016) demonstrated the critical role of specific exposures for long-term health among veterans. They found that although veterans generally report good health, those exposed to combat report worse health years later, as assessed through their self-rated health and physical functioning. Furthermore, Taylor et al. found that these long-term effects are not driven by health behaviors, nor are they driven entirely by service-connected disability.
This line of investigation is promising and can be pushed further. The general case for effects of traumatic exposures on physical health is quite strong, and the long-run effects of combat may be underappreciated. Research has linked general psychological trauma, for instance, to impairments in psychoneuroimmunology, which can affect a variety of systems in the long term (Kendall-Tackett 2009). Furthermore, wartime exposures have been linked to elevated daily stressors long after returning to civilian life, setting the stage for a variety of health problems (Miller and Rasmussen 2010). Furthermore, the likelihood that veterans are exposed to some sort of trauma is much higher than of being physically injured themselves: in Iraq, estimates of nonfatal musculoskeletal trauma are around 17 of every 1,000 soldiers deployed to a combat theater in a year (Goodman et al. 2012). The risks of service may be higher than usually thought.
A focus on trauma can also help focus attention on social processes subsequent to service as much as it does specific service-related experiences. Although the trauma of combat is certainly premised on specific experiences, it is also possible that combat matters to health because of person-environment fit: combat produces an especially wide gulf between veterans’ experiences in the armed forces and the demands of civilian life. In this vein, social scientists are increasingly attuned to how total institutions can have adverse long-term effects insofar as the demands of life in such institutions diverge from those on the outside. The long-term effects of incarceration on health, for instance, are partly shaped by the difficulties of reintegration (Schnittker, Massoglia, and Uggen 2012). The same sort of disjuncture may be apparent the military, too, especially among combat veterans. In general, the military provides training and experience relevant for a variety of occupations, but combat experience is less relevant in this regard, and the transition to civilian life may be especially difficult for those with combat-related exposures. In this vein, studies point to the difficulties some veterans experience in their social relationships upon return to civilian life, a finding not at all inconsistent with the experiences of returning prison inmates (Keane et al. 1985). Struggles of this sort may be detrimental to their long-term health.
Because they point to both direct and indirect effects of trauma, these pathways span the range of the potential effects of combat. They suggest, among other things, effects that might emerge only years later and have no direct relationship to injuries received while in service. In the present study I seek to push the literature further in the direction of trauma. I seek to evaluate the long-term effects of combat exposure on physical health across a range of specific experiences, including participation in conflicts from World War II forward, as well as exposures occurring during peacetime. I also seek to understand the pathways behind these effects, including the contributions of service-related injuries and disability, as well as the role of psychiatric disorders. In addition, I explore the relevance of preparation for the civilian transition, social reintegration, and health behavior.
Background
Research on the effects of combat on health has focused largely on the direct consequences of combat. This is both a natural and an obvious focus, though it is one that fails to alert researchers to all the potential long-term effects of combat. Furthermore, it constrains the focus to a particular set of mediating mechanisms and processes that, at least implicitly, circumvent emphasis on the social and psychological determinants of health. A focus on combat wounds, for instance, highlights the difficulties of full recovery among those who suffer from such injuries, but it neglects injuries that occur outside combat theaters, as well as linkages between combat and health that emerge subsequent to service and thus are not “service connected” in any obvious way. By the same token, a focus on posttraumatic stress shines a spotlight on a particularly disabling psychiatric condition for which combat plays an especially prominent role, but it obscures other forms of ongoing stress, which are apparent even among those who do not suffer from posttraumatic stress disorder (PTSD). Understanding the long-term effects of combat on physical health requires greater appreciation of other possibilities and of ongoing strain.
The Direct Effects of Combat on Health
The direct effects of combat on health pertain to combat-related injuries experienced personally by the combatant. In some instances, the long-term effects of combat experience on long-term morbidity and mortality are obvious. Some war veterans, of course, suffer from injuries sustained during combat and have persistent forms of disability as a result. The wars in Iraq and Afghanistan produced about 50,000 wounded American personnel (Schoenfeld and Belmont 2016). In addition, some injuries worsen with time. Although many of the wounded are treated in the combat theater shortly after their injury, some are treated only after they return to civilian life and their injuries grow worse (Goodman et al. 2012). In addition, some veterans suffer from the long-term effects of toxic exposures used in wartime, including agent orange in Vietnam or exposure to oil-well fires and depleted-uranium munitions in the Gulf War (Jones 2006; Jones et al. 2002). Studies of Vietnam-era veterans link both herbicide exposure and combat exposure to later health outcomes (Stellman, Stellman, and Sommer 1988). Although perhaps not routinely associated with combat, the risk of such exposures is quite broad. Surveys of Iraq war personnel, for instance, have found that 98 percent report potential exposure to at least one environmental agent, whereas only 35 percent report receiving incoming hostile fire (Vasterling et al. 2006).
Other direct effects pertain to psychiatric disorders, though in this case, too, the idea is that combat is directly relevant to the onset of a disorder. The effects of combat on psychiatric disorders have received a good deal of attention and, in fact, are a major contributor to PTSD in men (Kessler et al. 1995). Veterans exposed to combat are at a considerably higher risk for developing PTSD relative to other veterans (Solomon and Mikulincer 2006; Solomon, Shklar, and Mikulincer 2005). The risk is especially high within a short window of time following combat, though the disorder can also return 15 to 20 years later (Solomon and Mikulincer 2006). Among Iraq war veterans, the percentage who have met screening criteria for PTSD may be as high as 17 percent (Hoge et al. 2007). The evidence linking combat exposure to psychiatric disorders is perhaps especially strong for PTSD, but it is not limited to that disorder. Evidence from veterans serving in more recent conflicts finds a relationship between combat exposure and mood and anxiety disorders as well (Fiedler et al. 2006). Other evidence points to even more general relationships with emotional well-being. Among veterans returning from Iraq, 19 percent reported mental health problems and 35 percent report using mental health services within a year of returning home (Hoge, Auchterlonie, and Milliken 2006). Of Iraq and Afghanistan veterans seen at VA health care facilities, about 25 percent receive mental health diagnoses of some kind (Seal et al. 2007). In the long term, impaired mental health may undermine physical health as well. Veterans with PTSD, for instance, have unusually high levels of somatic symptoms (Hoge et al. 2007). Of note, though, longitudinal studies exploring the effects of combat exposure on physical health have not found that the effects can be explained by mental health, at least as measured by the symptoms of depression (Taylor et al. 2016). This evidence, too, suggests that there is more to the long-term effects of combat than mental health. In all these cases, though, the effects can still be regarded as direct in that they stem from combat experiences.
The Indirect Effects of Combat on Health
Another model for understanding the long-term effects of combat exposure on health focuses on indirect effects (Levy and Sidel 2009). Such pathways receive far less attention, though frameworks for appreciating indirect effects usefully shift the focus: they draw attention to processes that emerge once a veteran returns to civilian life, and furthermore, they foreground civilian experiences as much as combat exposures (Miller and Rasmussen 2010). This shift highlights several things. For one, the indirect approach highlights the potential significance of exposures that do not involve direct physical harm. In fact, they imply that the harms of combat exposure can exist independent of any such harm. In addition, they highlight the relationship between combat and health behavior, as well as ongoing stress in civilian life. In the indirect approach, combat exposure can be regarded as an event that sets off a process of stress proliferation with enduring consequences for chronic disease (Pearlin 1999). Although combat may have set the process in motion, its long-term effects are fomented by other experiences.
A related pathway to poor health operates through the difficulties of reintegration. Military personnel may suffer more to the extent that they feel unprepared to return to civilian life. The process of reintegration can be challenging, and the challenges of returning personnel are psychological as much as social: returning service personnel must understand the role of deployment in their lives and how it fits into their identity once they have returned (Adler, Zamorski, and Britt 2011). For many service members, this transition is relatively easy, involving a return to predeployment social relationships and to jobs for which their service provided good preparation. For others, the transition is more difficult and is likely to be even more difficult for those whose military experience is sharply divergent from what they encounter as civilians. In this regard, combat experiences might be especially destabilizing, a point dating back at least to The American Soldier (Stouffer [1949] 1965). Men who return from combat might, for instance, be especially aggressive, with implications for the social lives. Consistent with this, evidence from veterans from more recent conflicts finds a relationship between combat exposure and interpersonal conflict (Milliken, Auchterlonie, and Hoge 2007), including feeling easily irritated by a spouse (Shay 2002). Combat experiences involving injuries or casualties to one’s unit may be especially potent in this regard, insofar as they can produce long-term guilt, second-guessing, and fatalism (Adler et al. 2011). In addition, veterans exposed to combat are more prone to risk-taking behavior, characterized as a sense of invincibility (Killgore et al. 2008).
In the same vein, veterans exposed to combat also change their health behavior. Smoking among veterans is common. One study found that nearly half of Vietnam War veterans were nicotine dependent, far more than suffered from PTSD (Eisen et al. 2004). Analyses of veterans and their nonveteran siblings provide further evidence for an effect of military service on smoking: service in the military in early adulthood has a strong causal effect on smoking intensity over the life course (London et al. 2016). Smoking among veterans is consequential. Analyses of the long-term effects on the health of World War II and Korean War veterans shows that about 35 percent to 79 percent of the excess mortality among veterans is attributable to military-induced smoking (Bedard and Deschênes 2006). Although the overall prevalence of smoking has gone down over time, it remains high among armed forces personnel. Indeed, the prevalence of smoking among young veterans today is as high as among adults in the late 1960s (Brown 2010). To the extent that smoking is the result of stressful exposures, it may be more common among veterans with combat experience than among other veterans.
Several studies have assessed the indirect effects of combat exposure. One way to consider the long-term effects of combat on physical health is to link combat-related stress to later life outcomes, seemingly far removed from earlier exposures. Evaluating surviving veterans of World War II in 1992 and 2010, one study found that wartime stress was a significant predictor of mortality (Bramsen et al. 2007). Much of this association, though, was more direct than it appeared: it was mediated by psychiatric disorders, and furthermore, the association was particularly strong for wounded soldiers (Bramsen et al. 2007). Nonetheless, evidence bearing on the long-term effects of combat indicates that not all of the long-term effects of combat appear to be related to physical injuries or wounds. A common theme across various studies exploring the issue is the potency of exposure to the dead, dying, and wounded. A study of hospital admissions for shell shock in World War I, for instance, found that admissions followed patterns in casualties, increasing when casualties were high and declining when they fell (Jones, Thomas, and Ironside 2007). Another study explored military and medical records of Civil War veterans (Pizarro, Silver, and Prause 2006). The authors explored various wartime experiences, including being wounded, but also explored the percentage of one’s company who were killed. Although the Civil War was especially violent, the percentage killed was still independently associated with cardiac, gastrointestinal disease, and other comorbid conditions years later. Using an older sample of veterans, most of whom served in Vietnam, Taylor et al. (2016) found that exposure to the dead and wounded had significant effects on later health, even as simple exposure to combat did not.
Healthy Soldiers and Healthy Warriors
The aim of this study is to evaluate the long-term effects of combat exposure on health. Studies exploring the long-term effects of military service, however, must contend with the nonrandom selection of personnel for military service. At least in a volunteer army, members of the armed forces are likely to be healthier on average than nonmembers, referred to as the healthy soldier effect (Bollinger et al. 2015). Although this effect may have deteriorated since 9/11, it still implies that comparisons between veterans and nonveterans are biased by selection: veterans will appear healthier by virtue of having been selected to service on the basis of their health. In the present study, however, I am interested in understanding the effects of combat among veterans, for which selection processes are likely to produce conservative estimates. In this vein, some evidence points to the so-called healthy warrior effect (Larson, Highfill-McRoy, and Booth-Kewley 2008). This implies a different kind of selection, as revealed in some previous studies. Exploring the incidence of psychiatric disorders among Marines, for example, one study found that psychiatric disorders (except PTSD) had a lower incidence among combat-deployed personnel than among military personnel not deployed to a combat zone (Larson et al. 2008). In accounting for this pattern, the authors argued that poor psychological fitness is often detected early in recruit training or combat operations, resulting in the discharge of psychologically at-risk personnel prior to combat or at least early in deployment. Given this, any analysis of the putative effects of combat among veterans should be sensitive to the amount of time a veteran served. Short service times are like to be at least partly correlated with early psychological problems (leading to “washing out” of training). These issues, among others, can be addressed in the data used in the present study.
Analytic Strategy
I evaluate the long-term effects of combat exposure on health over three empirical stages. First, I characterize the different specific exposures associated with military service during specific periods and their effects on health. Second, I evaluate the sensitivity of such effects by, among other things, limiting the sample to those who served during a draft era (to minimize the influence of selection into the armed forces), to those who served longer than two years (to minimize the influence of selection into combat), and to those with no service-connected disability (to focus on indirect effects rather than the continuation of direct combat injuries). Finally I evaluate potential mediating pathways, all situated on experiences subsequent to the end of military service, including social integration and self-reported preparation for the return to civilian life.
Data and Methods
Data for this study are drawn from the 2010 National Survey of Veterans (NSV), the sixth in a series of nationwide surveys of veterans (Westat 2010). The NSV was designed for administrative purposes, particularly to help the VA better meet the needs of veterans. To this end, the survey mostly contains information on the benefits veterans are entitled to. However, the 2010 survey also contains useful information for evaluating the long-term effects of wartime service. In particular it contains information on health, as well as information on the timing and nature of military service.
The survey was conducted using a self-administered mailed survey, supplemented with a Web-based instrument. Respondents were identified using address-based sampling (using a sample drawn from the full U.S. Postal Service Delivery Sequence File, combined with information on likely veteran households, based on lists from the VA and the U.S. Department of Defense) (Westat 2010). Data were collected in two stages. In the first, potential respondents were sent a prenotification letter and a short screening questionnaire. From this, veterans were identified and provided with the full survey, as well as the option to participate in the same survey via the Web. In total, 8,710 surveys were received. The response rate for the screening was low (32.3 percent), though the response rate for the veteran survey itself was much higher (66.7 percent). The estimated percentage of veterans who responded to the survey was 38.8 percent, though the sampling frame of the survey did not include homeless or institutionalized veterans, or veterans living outside the United States. After eliminating missing cases on the health variables, the final sample size was 8,265.
Variables
Dependent Variables
The key dependent variables are threefold. Respondents were asked to rate their general health status, from excellent to poor. Self-rated poor health represents respondents who reported their health as either “poor” or “fair.” In addition, respondents were asked if they needed assistance for 11 activities: bathing, eating, moving from a bed, using a toilet, walking around the house, dressing, preparing meals, managing money, performing household chores, using a telephone, and taking medications. Because the presence of disability is relatively low, the variable used here represents those who reported needing assistance on any of the activities, whether “some” assistance or “a lot.” The final health variable represents the likely presence of a psychiatric disorder. Respondents were asked whether they were treated for a psychiatric disorder in the preceding six months, whether as an outpatient or inpatient. Ascertained in this fashion, this variable likely reflects the treatment of severe psychiatric disorders, though because veterans have robust access to health care, the variable is unlikely to be biased by an inability to use services, as it often is among those with less robust coverage.
Military Experiences and Combat Exposure
Respondents were asked a variety of questions about their military experiences. They were asked when they served and in what branch. From these responses, several variables were created. Branch of service was coded as Marines, Air Force, Navy, or Coast Guard (relative to Army). Because veterans could have served in multiple branches, these variables are not exclusive. In addition, variables were created corresponding to service during war or an active conflict (though not necessarily overseas deployment in that conflict). These included World War II (December 1941 to December 1946), Korea (July 1950 to January 1955), Vietnam (August 1964 to April 1975), and Operations Enduring Freedom, Iraqi Freedom, and New Dawn (September 2001 and later). Those who served in the Persian Gulf war were not uniquely identified, though that operation was short. All these variables are relative to service during nonconflict periods and, as before, are not exclusive in that some personnel served over multiple conflicts (or during both conflict and peacetime).
For the present study, the most important variables pertain to specific combat exposures. Four variables were coded, with increasing specificity across them: a variable was coded for any conflict-period service, any combat experience (asked as any service in “combat or a war zone”), any exposure to the dead and wounded (was the respondent “ever exposed to the dead, dying, or wounded”), and any exposure to toxins (was the respondent “ever exposed to environmental hazards”). The latter was limited to “probable” exposure, on the basis of respondents’ saying that they were either “definitely” or “probably” exposed to toxins. The other variables in this set were definite, with categorical yes/no answers. As will be elaborated on later, these variables, too, are not exclusive. About 46 percent of those serving during wartime participated in combat, relative to 13 percent at other times. Furthermore, of those serving in combat, about 70 percent were exposed to the dead or wounded, relative to about 16 percent of those not serving in combat. Coding the variables in this fashion—where each variable does not require a positive response to the one preceding it—expands the potential for traumatic exposures beyond combat alone, an important feature of the analysis.
Sample Restriction Characteristics
For purposes of testing the effects of combat exposure, variables were created to identify useful sample restrictions, in service of addressing selection. These include service during the draft era (prior to 1974), when exposure to combat was more likely to be more random relative to later periods (though certainly not perfectly so). In addition, a variable for short service was created, corresponding to respondents who served less than two years, potentially reflecting a lack of readiness for service or combat.
An especially critical sample restriction is the elimination of those with service-related disabilities. This study is interested in the long-term and especially indirect injuries of war, some of which will be hidden to the extent that research focuses on service-connected disability. For purposes of this analysis, service-connected disability was defined broadly. Respondents with service-connected disabilities were identified with a positive response to any one of several indicators: the receipt of any service-connected disability compensation, reporting that any service-connected disability kept them from getting or holding a job, or whether the VA determined that the respondent had a service-connected disability rating greater than zero. The latter is particularly consequential in that it will eliminate from consideration veterans with even relatively mild forms of disability.
Mediating Mechanisms
Three sets of variables were used as potential mediating mechanisms. Respondents were asked about their smoking status, including whether they were currently or were ever smokers. Respondents were also asked if they were currently married or not. In addition, respondents were asked how well prepared they were “leaving the military for the civilian job market.” A variable indicating feeling unprepared was coded on the basis of respondents’ reporting they were either “poorly” or “very poorly” prepared. Such reports were especially common among Vietnam-era veterans, though the feeling was apparent in a variety of veterans.
Control Variables
The models also include controls for sex, age (including an age-squared term in all the models), race/ethnicity (non-Hispanic white, black, Hispanic, Asian, and American Indian), and education (less than high school, high school, some college or associate’s degree, and bachelor’s degree or higher).
Results
Table 1 presents descriptive statistics for the key variables, arrayed over period of service. All major conflicts are presented, as well as periods with no active conflict. Because some veterans served over multiple conflicts, the sum of the individual samples, represented in columns, exceeds the total sample size. Several things are notable. For one, exposure to the dead or wounded is not limited to those who served during conflict periods. Furthermore, of all the conflicts contained in the sample, reports of exposure to death were most frequent among those serving in Operations Enduring Freedom, Iraqi Freedom, and New Dawn. Although veterans of this conflict also report the best physical health, they report the most use of psychiatric services.
Summary Statistics, National Survey of Veterans 2010 (n = 8,265).
Note: OEF/OIF/OND = Operations Enduring Freedom, Iraqi Freedom, and New Dawn.
Table 2 presents a basic model for each of the three health variables. In each case the model is logit regression. All the control variables are presented, though in later tables they are not. And in contrast to later tables, exposure to combat is presented here only as service during a specific conflict period (relative to peacetime service). The results reveal several patterns. In general, service during a conflict period is more detrimental to health, though Vietnam and World War II are most consequential. In addition, branch of service is also significant, with those branches least involved in combat operations, the Coast Guard and, to a lesser extent, the Navy, demonstrating the best health. Veterans of branches with the highest risk for poor health are from the Army and Marines.
Logit Regression of Health on Select Independent Variables, National Survey of Veterans 2010 (n = 8,265).
Note: Values in parentheses are standard errors. OEF/OIF/OND = Operations Enduring Freedom, Iraqi Freedom, and New Dawn.
p < .05. **p < .01. ***p < .001.
Table 3 presents regression models for each of the three health indicators. The remaining tables present semi-elasticities, representing average marginal effects and interpretable as a proportional change in the outcome for a unit change in the independent variable (see Williams 2012). For each indicator, three models are presented in an unfolding sequence: the first model for serving during any war period, the second model adding combat exposure, and the third model adding exposure to the dead and wounded, as well as exposure to toxins. The control variables remain the same as in Table 2, though without controls for the specific conflict period. Over the three models, the results reveal the increasing relevance of specific exposures, consistent with previous studies that have employed similar covariates (Taylor et al. 2016). In general, those who serve during years of active combat operations report worse health than those who do not, though the next model reveals that this relationship is driven by those with combat experience, which in turn is driven entirely by those exposed to the dead and the dying, as revealed in the next model. The coefficients for exposure to the dead and wounded range from .327 to .459. Moreover, exposure to the dead and wounded is more consequential than simple exposure to combat, at least for physical health (if not for any psychiatric services), suggesting estimates of the effects of combat per se may understate the significance of combat for health.
Semi-elasticities for Combat Exposure Variables from Health Regression Models, National Survey of Veterans 2010 (n = 8,265).
Note: Values in parentheses are standard errors. All models also include controls for age, age squared, sex, education, race/ethnicity, and branch of service (elasticities not shown).
p < .001.
The remaining tables explore this relationship further, both by testing its vulnerability to alternative specifications (in Table 4) and exploring mediating pathways and, thereby, indirect pathways (in Tables 5 and 6). Table 4 turns to the process producing these associations in more detail. The table again presents results for all three health outcomes, though in this case imposing strategic sample restrictions, including potential direct effects. The table begins by estimating the effects of exposure to the dead, while limiting the sample to those not reporting serving in actual combat. If the effects of exposure to the dead and wounded reflect the trauma of such exposures rather than risks that are correlated with such exposures, especially while serving in a combat role, the effects should be very different once the sample is limited to those who report not serving in combat. For all three outcomes, the results suggest that this is not the case: the effect for exposure to the dead and wounded remains significant and approximately the same magnitude. The next model estimates the same relationships, though limiting the sample to those who served during the draft era. Again, the elasticity for exposure to the dead and wounded remains significant and is little changed. The next model reduces the influence of a different form of selection, this time eliminating those who served less than two years. The elasticities remain similar to what they were before. The final model zeroes in on long-term effects that cannot be related directly to service activities. The model eliminates from consideration veterans with any service-connected disability or a disability that can be directly related to their service experiences. The coefficient for exposure to the dead or wounded, while reduced, remains significant, pointing to effects on health that only emerged later.
Semi-elasticities for Combat Exposure Variables from Health Regression Models, National Survey of Veterans 2010 (n = 8,265).
Note: Values in parentheses are standard errors. All models also include controls for age, age squared, sex, education, race/ethnicity, and branch of service (elasticities not shown).
p < .05. **p < .01. ***p < .001.
Semi-elasticities for Combat Exposure Variables from Reintegration Variables Models, National Survey of Veterans 2010 (n = 8,265).
Note: Values in parentheses are standard errors. All models also include controls for age, age squared, sex, education, race/ethnicity, and branch of service (semi-elasticities not shown).
p < .01. ***p < .001.
Semi-elasticities for Combat Exposure Variables from Health Regression Models, National Survey of Veterans 2010 (n = 6,471).
Note: Values in parentheses are standard errors. All models also include controls for age, age squared, sex, education, race/ethnicity, and branch of service (elasticities not shown).
p < .05. **p < .01. ***p < .001.
Table 5 estimates the same model presented in Table 3, but exploring potential pathways as the outcome. These pathways include smoking behavior (represented as whether the respondent is currently a smoker or not), not feeling prepared for the transition to civilian life (yes or no), and current marital status (represented as married or not). In each case, exposure to the dead and wounded increases risk by increasing the likelihood of smoking (with an elasticity of .212), by increasing the likelihood of reporting being unprepared for civilian life (elasticity of .244), and by decreasing the likelihood of being married, though by a relatively small amount (with an elasticity of –.057).
The final table explores the process that might produce such a relationship in greater detail, in this case focusing on the two nonpsychiatric outcomes. The first model provides a baseline. It eliminates from consideration those who received psychiatric services, as well as those with service-connected disabilities. The model thus represents a conservative estimate of the effect of exposure to the dead or wounded on health, eliminating direct pathways. The second model controls for the three potential mediating pathways, in an attempt to reduce the effect further. The results suggest that part of the relationship between exposure to the dead and wounded and health is channeled through psychiatric disorders, though the models also point to the significance of sociobehavioral pathways. The final model yields elasticities for exposure to the dead that are still statistically significant, though smaller than the same elasticity presented in any of the other models. It also yields elasticities for all three mediating pathways that are significant: not feeling prepared for the transition to civilian life, for instance, has an especially strong effect on health. It is useful to compare the reductions in the coefficient over the tables. Comparing Table 3 (model 3) with Table 4 (model 4) permits an evaluation of the reduction in the elasticity once service-connected disability is eliminated, while comparing Table 4 (model 4) with Table 6 (model 2) permits an evaluation of the reduction in the elasticity once smoking, marital status, and transition preparation are controlled for. In the case of poor health, the elasticity for exposure to the dead and wounded drops from .327 to .276 in the former (16 percent) and from .276 to .193 (30 percent) in the latter. In the case of functional limitations, the elasticity for exposure to the dead and wounded drops from .423 to .402 (5 percent) and from .402 to .343 (15 percent). In both cases, the mediating pathways matter more than eliminating service-connected disability. Altogether the models point to strong indirect effects, premised in part on reintegration difficulties. Of note, reporting not being prepared for the transition has a very strong relationship with poor health, exceeding that for being married if not for being a current smoker. Furthermore, reporting not being prepared for the transition is itself as detrimental to health as being exposed to the dead or wounded.
Discussion
The aim of this study was to estimate the long-term consequences of combat exposure on physical health. To date much of the literature has focused on the effects on combat exposure on mental health, though the results presented here suggest the effects on physical health may be just as consequential. The results indicate that the effects of combat may be underappreciated in at least two ways. For one, the results point to more specific experiences that matter, consistent with some previous research (Taylor et al. 2016). The models progressively unfold with greater specificity, revealing the most potent aspects of combat exposure. Combat exposure alone is a crude indicator. In Table 3, for instance, each consecutive model yields a larger association: any combat exposure matters more than any wartime service, and any exposure to the dead or wounded matters more than any combat exposure. The one exception to this pattern simply reinforces the importance of considering physical health as much as mental health: combat exposure has a larger association than exposure to the dead and wounded only in the case of psychiatric treatment. It is likely that even more granular indicators of combat experiences would yield even larger associations, though presumably affecting a smaller segment of veterans.
This pattern also implies that some of the trauma of military experience will be entirely independent of combat and thus that research may be short sighted in considering only combat per se. To be sure, combat and exposure to the dead and wounded are highly correlated (with a tetrachoric correlation in excess of .7). Nonetheless, about 16 percent of those who report no combat experience report exposure to the dead or wounded, a group that likely includes medical personnel or service personnel witnessing accidents. Although the death rate from noncombat violent causes such as suicides and accidents is lower among troops in Iraq than in the United States generally, such accidents do occur (Buzzell and Preston 2007). The relevance of such exposures is further enhanced by trends in the nature of combat. The lethality of war wounds among U.S. soldiers has declined in recent conflicts, because of advances in body armor and frontline medical responses. This has occurred despite advances in the lethality of firepower. In World War II, about 30 percent of wounded soldiers died from their injuries, whereas in Iraq and Afghanistan about 10 percent died (Gawande 2004). Of course, many injured soldiers will still suffer lifelong injuries to limbs and organs, though one implication of the present study is that as more of the wounded are returned from the battlefield and treated, other service personnel are more likely to be exposed to the traumas of war. The summary statistics presented in Table 1 indicate that in no other conflict was the percent of personnel exposed to the dead or wounded higher than during Operations Enduring Freedom, Iraqi Freedom, and New Dawn.
As much as they highlight the significance of specific experiences in war, the results also point to the relevance of civilian life for understanding the effects of combat. The final table reveals relationships that are partially mediated by smoking behavior, social relationships, and readiness for civilian life. Exposure to the dead and wounded is associated with all three factors, in much the same fashion as for health itself. This pattern is significant for several reasons but especially because it suggests not all the effects of war-related exposures are produced by the immediate effects of trauma. The literature on war exposure and mental health has recognized this possibility, with those advocating for indirect effects emphasizing a variety of psychosocial challenges that emerge only after the trauma and produce, in turn, worse mental health (see Miller and Rasmussen 2010 on this point). The models estimated here point to the relevance of both approaches, though they also encourage more research on other psychosocial factors that exposure to the dead and wounded is likely associated with. At the same time, they encourage research to consider the importance of transition readiness in its own right. Among those who served in combat roles, 24 percent felt unprepared for the transition, but even among those who did not serve in combat 19 percent felt unprepared. Statistically the risks of feeling unprepared for later health are as large as the risks associated with exposure to the dead or wounded.
The results also point to the dangers of insisting on a sharp distinction between mental and physical trauma. The models reveal that even when eliminating those who use psychiatric services, veterans exposed to the dead and wounded suffer from worse health and higher disability. This finding does not mean that the effects of combat trauma on physical health do not reflect some underlying mental health problem, though they do suggest a focus on severe psychiatric disorders could be misleading. Some survey data are relevant to this issue. Among Iraq and Afghanistan veterans seeking services in the VA, for instance, about 41 percent screened positive for PTSD, though much higher percentages reported some kind of psychosocial difficulty (Sayer et al. 2010). More than half reported, for instance, having difficulty sharing personal feelings, 44 percent reported difficulty making new friends, and 57 percent reported problems controlling anger (see Table 3 in Sayer et al. 2010). Although these difficulties are especially common among veterans with PTSD, they are not uncommon among veterans not suffering from the disorder (see Table 5 in Sayer et al. 2010). Furthermore, virtually all veterans in the survey expressed an interest in services to help them readjust to civilian life. Insofar as some of the long-term effects of combat exposure are channeled through social processes following release, the sweep of combat is greater than is perhaps appreciated. Some research has explicitly highlighted this issue, arguing that the effects of war exposures are underestimated to the extent that they are envisioned only as a consequence of direct exposures on mental health and not in ways that show how show war exposures can elevate stressful social and material conditions after deployment (Miller and Rasmussen 2010). The present study adds empirical evidence to this argument.
Limitations
This study has several limitations. First, the analysis rests on cross-sectional data. I used several methods to identify the effects of combat on health, including eliminating subpopulations for whom selection is likely to play a significant role. Most of these adjustments did not affect the results. In addition, the estimates rest on the logical case that exposure to combat is positively selective for health. Nonetheless, these are ad hoc methods for evaluating long-term effects in a cross-sectional observational study. Longitudinal data, along with richer information on childhood background conditions, could further substantiate evidence for the effects of combat. Taylor et al. (2016) were able to use longitudinal data, though the data did not span the period prior to military service, and the average veteran in the study was in his 70s.
In addition, in this study I used general indicators of health, though it would be informative to have more precise indicators of disease. In general, veterans suffering from poor health suffer from a variety of conditions. Indeed, the VA routinely uses the idea of polytrauma to characterize the kinds of comorbidities they find among veterans, a category that includes the presence of traumatic brain injury, chronic pain, and PTSD (Belanger, Uomoto, and Vanderploeg 2009; Gironda et al. 2009). Polytrauma patients are unlikely to account entirely for the patterns revealed here, especially because the analysis eliminates from consideration service-connected disability. Nonetheless, it is likely that self-reported poor health among combat veterans reflects patterns of comorbidity cutting across various systems and organs. This is especially likely if the pathways linking combat to poor health reflect health behaviors, stress, and social risk factors.
In addition, this study, though revealing the value of including specific indicators of combat exposure, encourages the use of even more specific indicators. Some comprehensive scales are available, including the Combat Exposure Scale (Keane et al. 1989). The Combat Exposure Scale also focuses on exposure to the dead and wounded, in addition to the personal risk of death or injury in combat. A more comprehensive assessment would move beyond such exposures, though, to include the context of the exposure (e.g., combatant vs. noncombatant roles) and the depth of exposure (e.g., frequency of exposure to death). At least one scale, the War Events Scale, attempts to assess war-related exposures distinct from combat in a comprehensive fashion (Unger, Gould, and Babich 1998).
This study also did not consider some important individual differences in the response to combat. Research on the relationship between combat exposure and mental health is instructive in this regard (e.g., Britvic, Radelic, and Urlic 2006). At least with respect to posttraumatic stress, the apparent severity of combat exposure is occasionally unrelated to the risk for onset (Koren et al. 2005). The key to understanding PTSD is the perceived threat of an event, though veterans seeing the same actual threat are likely to interpret it in different ways (Koren et al. 2005). Furthermore, even though combat exposure increases the risk for PTSD, combat personnel who immediately receive stress-related treatment have a lower risk even decades after the war (Solomon et al. 2005). To the extent that the long-term effects of combat exposure on health operate through processes of event construal and ongoing psychological stress, the effects may be attenuated by a host of individual and situational differences.
Conclusions
This study highlights several things. It highlights how the effects of combat on health likely involve a variety of psychological and social processes. It encourages a greater focus on physical health, in tandem with the long-standing focus on mental health. And it points to the importance of considering postdeployment factors, including health behaviors and reintegration preparation. In this regard, this study draws parallels between the experiences of combat veterans and other people transitioning from total institutions or out of significant social roles. Although combat is certainly unusual in many critical respects, appreciating its long-term effects on health requires considering, in a more general fashion, the psychological and social aspects of trauma and displacement.
Footnotes
Acknowledgements
I would like to thank Thomas McBride and David Vinge for helpful comments.