American Journal of Epidemiology Advance Access published online on April 22, 2008
American Journal of Epidemiology, doi:10.1093/aje/kwn083
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Invited Commentary: How Healthy is the "Healthy Warrior"?
1 Department of Psychology, Boston University, Boston, MA
2 Massachusetts Mental Health Center Public Academic Psychiatry Division of the Beth Israel Deaconess Medical Center, Department of Psychiatry, Harvard Medical School, Boston, MA
3 Research Service, Boston Veterans Affairs Healthcare System, Brockton, MA
Correspondence to Dr. Rosemary Toomey, Department of Psychology, Boston University, 648 Beacon Street, 6th Floor, Boston, MA 02215 (e-mail: rosemary_toomey{at}hms.harvard.edu).
Received for publication February 19, 2008. Accepted for publication February 29, 2008.
 |
ABSTRACT |
|---|
 TOP ABSTRACT References |
|---|
In this issue of the Journal, Larson et al. (Am J Epidemiol 2008;000:000–00) report incidence rates of 11.8% for any mental disorder and 1.6% for post-traumatic stress disorder among US Marines deployed during Operation Iraqi Freedom and Operation Enduring Freedom. Various methodological reasons can help explain why these rates are lower than those found in previous wars. One primary reason is varying methods of diagnosis. Other reasons include differences in percentages of active-duty personnel, windows of observation, and methods of calculating rates. In addition, comorbidity for some mental disorders and specific risks for developing certain disorders complicate interpretation of rates. Nevertheless, Larson et al. document evidence for the "healthy warrior" effect—namely that deployed Marines have fewer mental disorders than nondeployed Navy and Marine Corps personnel, with the exception of stress disorders. Suggestions are made for directions of future research into this effect.
cohort studies; incidence; mental disorders; military personnel
Abbreviations: CAPS, Clinician-Administered PTSD Scale; OEF, Operation Enduring Freedom; OIF, Operation Iraqi Freedom; PTSD, post-traumatic stress disorder; SCID, Structured Clinical Interview for DSM Disorders
The study by Larson et al. (1), published in this issue of the Journal, examines incidence rates of mental disorders among US Marines deployed in Operation Iraqi Freedom (OIF) and Operation Enduring Freedom (OEF) as compared with contemporary nondeployed US Navy and Marine Corps groups and historical Navy and Marine reference cohorts. This type of study enables us to gain a more thorough understanding of the psychological impact of this war in comparison with the first Gulf war and other wars.
The authors report a 6.4 percent cumulative incidence rate of mental disorders in their OIF/OEF deployed group, excluding persons deployed to combat who were diagnosed with a psychiatric disorder after enlistment (6 percent of the overall sample). Incidence rates of all disorders except stress disorders were significantly lower in combat-deployed personnel than in nondeployed personnel. The incidence of post-traumatic stress disorder (PTSD) was significantly greater in the deployed group (1.5 percent vs. 0.6 percent), while the incidence of acute stress disorder was equivalent in both groups (0.3 percent). When the subgroup with precombat psychiatric diagnoses was included, the cumulative incidence rate rose to 11.8 percent, while the rates of stress disorders rose only slightly (PTSD: 1.6 percent; acute stress disorders: 0.4 percent). Incidence rates of other diagnoses remained significantly lower in the deployed group, except substance abuse diagnoses, for which incidences were 4.7 percent in both groups. Including persons with precombat diagnoses seemed to primarily increase the rates of substance abuse, adjustment, and mood disorders in the group as a whole.
Larson et al. (1) contrast their lower rate of mental disorders with rates from the first Gulf war and the Vietnam War. They correctly point out some reasons why their PTSD rate of 1.6 percent in the entire group is lower than that in other reports, such as the impact of false-positive findings generated by surveys. One comparison they make is with the PTSD rate of 6.2 percent in the first Gulf war that I and my colleagues reported (2). The authors state that our rate was determined by "PTSD screening criteria." Actually, our rate of PTSD based on screening (a score of 
50 on the PTSD symptom checklist) was 9.8 percent of the deployed group. In contrast, our rate of 6.2 percent was generated by means of a structured interview, the Clinician-Administered PTSD Scale (CAPS). Structured interviews are considered the gold standard for diagnosing mental disorders, and this interview's specific focus on PTSD makes it more detailed than other structured interviews. Larson et al. (1) obtained their information using databases on inpatient and outpatient psychiatric diagnoses from military treatment facilities and government-reimbursed private clinics. Chart diagnoses often underestimate the true prevalence of a psychiatric diagnosis. For example, in an attempt to validate a PTSD screening instrument in 97 veterans seeking care at Department of Veterans Affairs substance abuse clinics, Kimerling et al. (3) used the CAPS as the criterion measure. The CAPS identified 32 patients with PTSD, but only eight of these patients had a PTSD chart diagnosis. Larson et al.'s rates were also lowered by their not having records from privately reimbursed clinics or diagnoses made in the war zone with no postcombat follow-up, although they are pursuing the latter data. It is critically important to consider the method of diagnosis when comparing rates across studies.
The authors mention that history can be an inexact guide for estimating the future psychological costs of war (1), and they cite the recently revised estimates of PTSD rates from the Vietnam War (4). Again, different diagnostic methods accounted for most of the difference between the original and revised diagnoses. In the National Vietnam Veterans Readjustment Survey, Kulka et al. (5) originally estimated a 30.9 percent lifetime prevalence of PTSD based on an algorithm using the Mississippi Scale for Combat-Related PTSD and the Diagnostic Interview Schedule. The algorithm was created in part using Structured Clinical Interview for DSM Disorders (SCID) interviews in a subsample of 260 veterans. Dohrenwend et al. (4) reexamined those study data. A reduction to 22.5 percent was generated by reevaluating and extrapolating from the original SCID diagnoses and correcting problems related to 1) distinguishing PTSD diagnoses caused by war-related traumatic events versus non-war-related traumatic events and 2) linking symptoms specifically to traumatic events. The latter problem is what accounts for many false-positive incidence rates when survey data are used. The final revised rate of 18.7 percent resulted from adjustment for objective documentation of exposure to a traumatic event and the additional diagnostic criterion of functional impairment specified in the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (as compared with the Diagnostic and Statistical Manual of Mental Disorders, Third Edition, Revised). Thus, the primary difference of 8.4 percent (30.9 percent vs. 22.5 percent) was attributed to more careful adherence to the diagnostic criteria at the time of the study and assurance that the PTSD was war-related, while a lesser difference of 3.8 percent (22.5 percent vs. 18.7 percent) was attributed to the other adjustments. In a further revision, McNally (6) used a slightly revised impairment threshold judged to be more clinically significant, which resulted in even lower estimates (National Vietnam Veterans Readjustment Survey current PTSD: 15.2 percent; Dohrenwend: 9.1 percent; McNally: 5.4 percent).
Having a variety of choices about diagnostic methods is reasonable. Frequently, constraints of time, personnel, and finances lead study investigators away from lengthy interviews such as the CAPS. It is unlikely that investigators would be able to get the resources needed to apply the CAPS to the 140,322 contemporary troops in Larson et al.'s study (1). Even when using structured interviews, there can be challenging diagnostic dilemmas and judgment calls. As I noted above, one of the reasons for Dohrenwend's revised PTSD estimates included the elimination of PTSD cases in which the traumatic event was unrelated to combat. However, there may be traumatic events experienced in a war zone that are not combat-related. The perfect example would be rape by a fellow soldier. Such a trauma would not have occurred if the victim were not an enlisted soldier. One can easily argue that such a victim should have the right to the same kinds of services as persons with combat-related PTSD, yet it is also easy to see that such a case is unique. The word of caution is that when comparing rates across studies and, perhaps more importantly, when making policy decisions based on study findings, such details of diagnostic methodology are critical.
Additional methodological reasons can explain different study rates of psychiatric diagnoses. For the purposes of illustrating some of these reasons, our study (2) can again be used. As Larson et al. (1) stated, their sample did not contain Reservists, which served to lower the risk for stress disorders. In contrast, our study sample contained 36.3 percent Reservists, 28.6 percent National Guard personnel, and 35.2 percent active-duty personnel; such a mixture would be expected to generate higher PTSD rates. The observation window for Larson et al. was 4 years and 2 months, while our observation window was 2.5 years. They calculated incidence rates, while we calculated prevalence rates during the period of deployment. Depending on the duration of illness, we would expect these different types of rates to vary. Investigators in other studies have reported lifetime, 12-month, or current prevalence rates. Estimates based on restricted observation time windows cannot be compared with lifetime prevalence rates. "Current" rates from the National Vietnam Veterans Readjustment Survey were assessed 10–12 years postservice, while our study examinations were conducted approximately 10 years after the first Gulf war. Rates from the Larson et al. study were assessed within 4 years and 2 months of enlistment; thus, additional PTSD cases with delayed onset may continue to appear in their sample as time proceeds.
Larson et al. excluded persons with postenlistment precombat diagnoses from some of their analyses. Exclusion of this subgroup presumably served to isolate the specific psychiatric effects of combat itself, although not definitively. Rates from the entire group would better estimate the true impact of deployment, because those persons with precombat diagnoses still went to combat. Indeed, such persons are probably more vulnerable to the development of additional mental disorders, and it would be useful to explore this empirically, especially in light of the high comorbidity associated with PTSD (7).
In calculating their incidence rates, Larson et al. eliminated subjects from the incidence density denominator once they were diagnosed and no longer considered at risk. This assumes that there are common risk factors for the development of mental disorders. While there is shared risk for the development of mental disorders in general, evidence suggests there are also risks unique to particular disorders. Using a twin study design, Chantarujikapong (8) examined the magnitude of genetic and environmental contributions to the lifetime co-occurrence of symptoms of generalized anxiety disorder, panic disorder, and PTSD. Panic disorder symptoms and PTSD shared additive genetic influences with symptoms of generalized anxiety disorder but also had disorder-specific additive genetic contributions. The remaining variance for all three disorders was due to unique environmental factors—that is, environmental factors experienced by one twin (e.g., trauma) rather than shared by both twins. There were significant contributions from unique environmental factors that overlapped between the disorders as well as those that were specific to each disorder. Because combat trauma is a risk factor for developing PTSD and other mood and anxiety disorders (9), the transition from precombat to combat changes the risk profile for individuals. Based on what appeared to be higher percentages of substance abuse, adjustment disorders, and mood disorders diagnosed prior to combat, these persons were eliminated from the incidence density denominator before the unique risk factor of combat was present for development of PTSD. The possible impact of these types of issues on the incidence rates reported is unclear.
While there are many methodological factors that may have influenced the PTSD incidence noted by Larson et al. (1), these factors were held in common in their deployed and reference groups. Thus, the authors' finding that most diagnoses were significantly less prevalent in the deployed group as compared with the nondeployed group provides evidence for the "healthy warrior effect." Evidence for this effect is strengthened by their data documenting that their deployed group was significantly less likely to have been hospitalized for a psychiatric disorder in their initial 6 months of service and less likely to have separated early from the military (1). It is not clear how to reconcile these data with evidence from previous wars showing that deployed groups have greater rates of mental disorders than nondeployed groups. If Larson et al. had lifetime rates for their sample and they were lower than rates from other large-scale epidemiologic studies, that would provide further evidence of the healthy warrior effect, particularly if the comparison groups had lifetime rates more similar to those of the large-scale epidemiologic studies. Haley (10), who coined the phrase "healthy warrior effect," recommended correcting this bias by individually matching units from the nondeployed reference populations to deployed units. In the case of the available data points in the Larson et al. study, this could perhaps be accomplished by excluding persons from the reference group in order to achieve comparable rates of early psychiatric hospitalization and early separation, while maintaining comparable demographic matching between groups. The groups could then be compared again with regard to rates of psychiatric diagnoses.
The "healthy warrior effect" is parallel to the "healthy worker effect," on which there is more literature. The effect of this bias is the potential underestimation of risks to workers or warriors. There are initial and continuing selection biases. For example, Le Moual (11) distinguished between the "hire effect" and the "survivor effect." As applied to our veterans, the "hire effect" would be the selection bias of who enlists in the military in the first place. Presumably, the effect would be enhanced in times of easy access to recruits, when the military has the luxury of setting higher entrance criteria. The "survivor effect" refers to the selection bias of who remains in a job, or in the military, and who gets deployed. One recommendation for minimizing this bias from the occupational literature is to include in the worker group those workers who were initially hired but eventually left the job. As applied to the Larson et al. data (1), this would mean including persons who separated early from the military in the deployed group. Given that the risk factor of interest is combat, only those who separated from the military once deployed would be included.
In addition to the traditional risks of combat, the first Gulf war highlighted potential risks from exposure to toxic substances. Given that these effects may be subtle and limited to subgroups of deployed personnel, eliminating any healthy warrior bias is important to delineating the true risk of such exposures. In examining the "healthy warrior effect," it will be important to investigate whether this effect presents more of a bias in active-duty personnel compared with Reservists or in Marines compared with persons in other branches of service.
|
ACKNOWLEDGMENTS |
|---|
Conflict of interest: none declared.
|
References |
|---|
|
TOPABSTRACTReferences |
|---|
- Larson GE, Highfill-McRoy RM, Booth-Kewley S. Psychiatric diagnoses in historic and contemporary military cohorts: combat deployment and the healthy warrior effect. Am J Epidemiol (2008) 000:000–00.
- Toomey R, Kang HK, Karlinsky J, et al. Mental health of US Gulf War veterans 10 years after the war. Br J Psychiatry (2007) 190:385–93.
[Abstract/Free Full Text] - Kimerling R, Trafton JA, Nguyen B. Validation of a brief screen for post-traumatic stress disorder with substance use disorder patients. Addict Behav (2006) 31:2074–9.[CrossRef][Web of Science][Medline]
- Dohrenwend BP, Turner JB, Turse NA, et al. The psychological risks of Vietnam for U.S. veterans: a revisit with new data and methods. Science (2006) 313:379–982.
- Kulka RA, Schlenger WE, Fairbank JA, et al. Trauma and the Vietnam War generation: report of findings from the National Vietnam Veterans Readjustment Study. (1990) New York, NY: Brunner/Mazel.
- McNally RJ. Revisiting Dohrenwend et al.'s revisit of the National Vietnam Veterans Readjustment Study. J Trauma Stress (2007) 20:481–6.[CrossRef][Web of Science][Medline]
- Kessler RC, Sonnega A, Bromet E, et al. Posttraumatic stress disorder in the National Comorbidity Survey. Arch Gen Psychiatry (1995) 52:1048–60.
[Abstract/Free Full Text] - Chantarujikapong SI, Scherrer JF, Xian H, et al. A twin study of generalized anxiety disorder symptoms, panic disorder symptoms and post-traumatic stress disorder in men. Psychiatry Res (2001) 103:133–45.[CrossRef][Web of Science][Medline]
- Koenen KC, Lyons MJ, Goldberg J, et al. A high risk twin study of combat-related PTSD comorbidity. Twin Res (2003) 6:218–26.[CrossRef][Web of Science][Medline]
- Haley RW. Point: bias from the "healthy-warrior effect" and unequal follow-up in three government studies of health effects of the Gulf War. Am J Epidemiol (1998) 148:315–23.
[Free Full Text] - Le Moual N, Kauffmann F, Eisen EA, et al. The healthy worker effect in asthma: work may cause asthma, but asthma may also influence work. Am J Respir Crit Care Med (2008) 177:4–10.
[Abstract/Free Full Text]
CiteULike 
Connotea 
Del.icio.us What's this?
This article has been cited by other articles:
|
|
 |
|
  J. Wilson, M. Jones, N. T. Fear, L. Hull, M. Hotopf, S. Wessely, and R. J. Rona Is Previous Psychological Health Associated With the Likelihood of Iraq War Deployment? An Investigation of the "Healthy Warrior Effect" Am. J. Epidemiol., June 1, 2009; 169(11): 1362 - 1369. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
C. W. Hoge Re: "Psychiatric Diagnoses in Historic and Contemporary Military Cohorts: Combat Deployment and the Healthy Warrior Effect Am. J. Epidemiol., November 1, 2008; 168(9): 1095 - 1096. [Full Text] [PDF]
|
|
|
|
|
|
B. P. Dohrenwend, D. M. Sloan, B. P. Marx, D. Kaloupek, and T. M. Keane Re: "Psychiatric Diagnoses in Historic and Contemporary Military Cohorts: Combat Deployment and the Healthy Warrior Effect" Am. J. Epidemiol., November 1, 2008; 168(9): 1093 - 1094. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||