Reintegration Problems and Treatment Interests Among Iraq and Afghanistan Combat Veterans Receiving VA Medical Care

Our sampling frame included the 181,611 Iraq-Afghanistan combat veterans who made at least one visit to a VA facility within the continental United States between October 2003 and July 2007. This represents 60% of Iraq-Afghanistan veterans who used VA services within this time frame. The remaining 40% were not classified as combat veterans.

We used stratified random sampling without replacement. We selected this sampling strategy to increase the precision of our estimates and models. Because sampling from demographically more homogeneous groups produces estimates with smaller error variance than a simple random sample ( 27 ), we sampled from strata that included veterans of the same race-ethnicity and gender who were living in geographically similar areas. To this end, we divided the United States into six regions (Northeast, Southeast, Upper Midwest, Southern Midwest, Northwest, and Pacific Coast). Each regional stratum was divided into four gender (male or female) and race (white or nonwhite) combinations. From each of the resulting 24 strata we randomly selected 55 Iraq-Afghanistan combat veterans for recruitment (N=1,320). Because one-fifth of Iraq-Afghanistan veterans had missing race-ethnicity data, we randomly selected an additional 15 men and 15 women with missing race-ethnicity information from each of the six regions (N=180). Later, we used veterans' self-report to reclassify race and ethnicity and to verify deployment. We then constructed estimates using Horvitz-Thompson type estimators, with weights equal to the inverse of sample inclusion probabilities.

Of the 1,500 veterans originally identified for survey, 274 were excluded for the following reasons: deceased (N=8), veteran of other war eras (N=89), could not be located through U.S. Postal Service after three attempts (N=167), or currently redeployed to Iraq or Afghanistan (N=10). Of the 1,226 Iraq-Afghanistan combat veterans who remained eligible, 754 (62%) returned surveys by July 14, 2008.

Veterans received a prenotification letter describing the study, followed two weeks later by a cover letter, 12-page questionnaire, and $5 incentive. The cover letter reiterated the study's goals and described the risks, benefits, and voluntary nature of participation. Return of the survey signified veterans' consent to participate in the study. Nonresponders received a reminder letter and two more mailings of the questionnaire. Data were collected between April and July 2008.

We used VA administrative databases to obtain the following sociodemographic information for both responders and nonresponders: age, gender, race-ethnicity, military component (active duty versus National Guard or reserve), receipt of service-related disability benefits (any benefits and benefits specifically for PTSD), use of VA mental health services within the past year (any versus none), and distance in miles to nearest VA and community-based outpatient clinics. We also extracted the past two years of the ICD-9-CM codes for PTSD, anxiety disorders other than PTSD, depression, substance use disorders (excluding nicotine dependence), psychoses, and traumatic brain injury.

The study questionnaire assessed veteran characteristics, physical and mental health, perceived community reintegration problems, and treatment interests and preferences regarding intervention service delivery (in person or over the Internet). The research team, which included clinicians with expertise in deployment-related readjustment problems and in measure development, developed the initial version of the questionnaire based on literature reviews, early findings from ongoing studies, and clinical experience. A focus group of one veteran and three active-duty service members provided the investigators with anonymous feedback on the content and format of the initial version of the survey. After incorporating this feedback, we pilot-tested the survey with a sample of 87 Iraq-Afghanistan combat veterans. We finalized the survey after analyzing response patterns and reviewing participant comments in the comment section.

One item assessed overall difficulty in readjusting to civilian life over the past 30 days on a 5-point scale ranging from 1, no difficulty, to 5, extreme difficulty. Sixteen items assessed specific problems over the past 30 days in the following functional domains: social relations, eight items; productivity, three items; community participation, two items; perceived meaning in life, one item; and self-care and leisure activities, two items. Responses ranged from 1, no difficulty, to 5, extreme difficulty. Items were modified for relevance to this population from the social relations, life activities, and self-care domains of the World Health Organization Disability Assessment Schedule II ( 28 ), supplemented with content from the Community Integration Questionnaire and Community Integration Measure, which were developed for traumatic brain injury research ( 29, 30 ). Internal consistency of these 16 items, as measured by Cronbach's coefficient alpha, was .95. Nine yes-no items were used to assess problem experiences since returning home from Iraq or Afghanistan, including four items on potentially harmful behaviors and one item each on divorce or separation, legal problems, job loss, problems accessing health care, and loss of spirituality or religious life.

To assess interest in services for readjustment problems, participants checked individual items on a list of 12 possible services. They also indicated how they would want to receive information and services for community reintegration problems. In addition, because use of the Internet has become increasingly common for health care service delivery ( 31, 32 ), we assessed access to the Internet and frequency of Internet use.

Overall physical and mental health status was assessed with the 12-item Short-Form Health Survey (SF-12v2), with mental and physical health component summary scores normalized such that they could be compared with values obtained in the U.S. population, which has a mean score of 50 and a standard deviation of 10 ( 33 ). To assess probable PTSD, we used the Primary Care PTSD Screen ( 34 ), which is used by the VA and Department of Defense (DoD). A cutoff score of 3 yielded .76 sensitivity and .92 specificity for clinical PTSD in a sample of active-duty soldiers who returned from combat in Iraq ( 35 ). We screened for alcohol and drug problems using the Two-Item Conjoint Screen ( 36, 37 ). This screen is also included in the DoD Postdeployment Health Reassessments ( 3 ). A cutoff score of 1 had .80 sensitivity and specificity among 18- to 59-year-old primary care patients ( 37 ).

Prevalence and proportions were weighted to represent the population of Iraq-Afghanistan combat veterans who used VA medical services. We used stratified estimates weighted by the inverse of sample inclusion probabilities to calculate population parameter estimates and their standard errors ( 27 ). Less than 3% of community reintegration items had missing values. Assuming that this degree of missing data depended only on the observed covariates, we imputed missing values using logistic regression methods for multiple imputations ( 38 ). We used stratified logistic regression models to construct odds ratios and 95% confidence intervals ( 39 ) for community reintegration problems by probable PTSD status.Odds ratios were adjusted for demographic characteristics that preceded deployment, including age, gender, race-ethnicity, and military component. When comparing community reintegration problems among those with and without probable PTSD, we used a p value of <.002 as the threshold for statistical significance. Stratified Poisson regression was used to determine whether probable PTSD was associated with the number of community reintegration problems and the number of services of interest for community reintegration problems. All analyses were performed for responders and then adjusted for potential nonresponse bias based on administrative data available for both responders and nonresponders; adjustment was achieved by constructing response propensities and producing a weighted combination of within-propensity class estimates ( 40 ). SAS, version 9.2, and R coding were used to perform the calculations.

Responders did not differ from nonresponders in diagnoses extracted from VA medical records, receipt of mental health services, distance to a VA or community-based outpatient clinic, service connection for any condition, service connection for PTSD, or race-ethnicity. Compared with nonresponders, responders were older, were more likely to be female, and were more likely to have been activated to Iraq or Afghanistan from the National Guard or reserves than from active duty. The decision to adjust for demographic characteristics that preceded deployment was made a priori to increase the precision of our estimates.

The median time between participants' return from their most recent Iraq or Afghanistan deployment and completion of the survey was 42 months (interquartile range 31–54 months). Roughly 22% reported more than one deployment to Iraq or Afghanistan war theaters. One-fifth of the sample had fought in prior U.S. wars. Participant and estimated population sociodemographic characteristics are listed in Table 1 . As expected, given our stratification scheme, survey respondents were roughly balanced in terms of gender and race (white or nonwhite). However, when weighted back to the overall population of Iraq-Afghanistan combat veterans who receive VA medical care, we estimated that 87% of the population was male and 63% was white.

As shown in Table 2, approximately a quarter of the population of Iraq-Afghanistan combat veterans who had received VA medical care had a diagnosis of PTSD and 7% had a substance use disorder diagnosis documented in VA administrative records. However, on the basis of the screening measures included in our survey, we estimated that 41% and 38% of the population may have had PTSD or a drug or alcohol use problem, respectively. A total of 149 (51%) of the 291 study participants who screened positive and 46 (10%) of the 463 who screened negative for PTSD carried a PTSD diagnosis in VA medical records; 33 (13%) of the 262 who screened positive and seven (1%) of the 492 who screened negative for probable drug or alcohol use problems carried a substance use disorder diagnosis in VA medical records. The SF-12v2 scores indicated that Iraq-Afghanistan combat veterans had substantially poorer overall general medical and mental health than the U.S. general population. The population majority (58%) was receiving disability benefits for a service-related condition, with PTSD being most common.

An estimated 40% (95% confidence interval [CI]=35%–45%) of Iraq-Afghanistan combat veterans who used VA medical services perceived some to extreme overall difficulty in readjusting to civilian life within the past 30 days. The estimated average number of specific areas of some to extreme difficulty was 6.33±.27 out of 16. As shown in Table 3, we estimated that at least 25% of these veterans were having some to extreme difficulty in each of the domains assessed. Difficulty in social relations (such as confiding in others and getting along with their spouses, children, and friends) were particularly common. Some to extreme productivity problems (including problems keeping a job and completing the tasks needed for home, work, or school) were reported by 25%–41% of Iraq-Afghanistan veterans. The estimated mean number of dichotomously scored problem experiences since coming home from Iraq or Afghanistan was 3.03±.12 out of nine. Potentially harmful behaviors since coming home from Iraq or Afghanistan were common, with 31% reporting more alcohol and drug use and 57% reporting more anger control problems ( Table 3 ).

An estimated 96% (CI=93%–99%) of Iraq-Afghanistan combat veterans who used VA medical care reported interest in services for community reintegration problems. The estimated average number of services that veterans would consider using for community reintegration problems was 6.84±.17 out of the 12 options presented. As Table 4 shows, veterans most frequently reported interest in obtaining information about VA benefits and about schooling, employment, or job training. The three most popular ways of receiving community reintegration services or information were at a VA medical center, over the Internet, and through the postal mail. Almost all Iraq-Afghanistan combat veterans had access to the Internet and used it regularly.

Probable PTSD was associated with problem drinking or drug use and with worse SF-12v2 mental and physical component summary scores (p<.001). The odds of reporting some to extreme difficulty was significantly higher among those with probable PTSD in each functional area assessed, with odds ratios ranging from 3.10 to 13.78 ( Table 5 ). Similarly, the odds of reporting problems experienced since homecoming were higher among those with probable PTSD, with odds ratios ranging from 2.21 to 8.89 ( Table 5 ). However, functional problems and postdeployment problems were also present among those without probable PTSD. Among those with and without probable PTSD, the most commonly reported problem since homecoming was controlling anger. Finally, veterans with probable PTSD expressed interest in more types of services for community reintegration problems than those without probable PTSD (p<.001). A majority of those with probable PTSD were interested in traditional mental health services for their reintegration problems, such as face-to-face individual therapy (83%; CI=77%–89%) and medications (71%; CI=64%–79%) (data not shown).