Gender Differences in VA Disability Status for PTSD Over Time

The Minneapolis VA Medical Center Institutional Review Board approved this study. Data were drawn from our original cohort of nationally representative veterans who first applied between 1994 and 1998 for VA disability status for PTSD and who responded to mailed surveys at two time points: between 1998 and 2000 (time 1) and between 2004 and 2006 (time 2). On average, almost ten years (mean±SD=9.8±1.4) lapsed from time 1 to time 2. We originated the cohort to explore disparities in PTSD disability benefit awards after an original claim, including possible gender disparities (15). At study inception, 100,750 male veterans and 3,866 female veterans were eligible for participation. Although male veterans represented 96% of all applicants for VA PTSD disability benefits, random sampling was stratified by gender to achieve a 1:1 gender ratio.

We used monetary incentives, repeat mailings, and telephone prompts to maximize survey responses. The effective time 1 response rate was 68%, and nonresponse bias was minimal. Specifically, time 1 survey responders did not differ from ineligible veterans and nonresponders on any characteristic, except that survey responders served on average six months more in the Armed Forces than ineligible veterans and nonresponders and were slightly more likely to have recently worked for pay (15). Of the 3,337 veterans who completed time 1 study material, 2,998 (90%) were alive at time 2. Of these, 2,551 (85%) returned useable time 2 surveys. On the basis of their time 1 data, time 2 responders did not differ from nonresponders on clinical characteristics, employment status, or income. However, time 2 nonresponders were approximately two years younger, 5% more likely to be nonwhite, and 4% less likely to have attended college than time 2 responders. Elsewhere we report on changes in symptoms and functioning from time 1 to time 2 (12). Here we focus on changes in PTSD service connection.

Our primary outcomes were gain and loss of PTSD disability status over a ten-year period. We assessed service connection for PTSD using claim information extracted from VBA databases.

To facilitate comparison with our original study (15), we group covariates into the following three categories in this report: clinical variables (PTSD, functioning, and postservice life stressors), demographic characteristics, and military trauma type.

PTSD symptom severity was measured with the Penn Inventory for Posttraumatic Stress Disorder (Penn Inventory) (17), which has excellent internal consistency and test-retest reliability of .87–.93. Among veterans, scores of at least 35 have .90–.98 sensitivity and .94–1.00 specificity for PTSD diagnosis (18,19). In keeping with our prior work (15), we used Penn Inventory scores as a continuous measure.

Functioning was assessed with measures of social functioning, physical functioning, and employment status. We used the Social Adjustment Scale (SAS) (20) to evaluate functional impairment. The SAS measures psychosocial adjustment and functioning across the following domains: major work role; social interactions; marital, parental, and extended-family interactions; and economic self-sufficiency. We used the SAS summary measure, which has scores ranging from 1, best functioning, to 5, worst functioning. Internal consistency was .85. We assessed physical functioning using the RAND Revised Physical Functioning Battery (21). Scores range from 12 (worst impairment) to 36 (no impairment). Its internal consistency was .92. Employment was assessed with a single item. We used medical conditions listed in VBA databases to compute Charlson Comorbidity Index scores (22). The Charlson Comorbidity Index assigns severity-weighted scores to 18 specific medical illnesses and predicts long-term survival.

To assess postservice trauma and hardships, which may affect symptoms and functioning, we used the revised Life Stressor Checklist (23). This scale assesses exposure to natural and human-made disasters, accidental injuries, and other serious traumas and stressors. Scores range from 0 to 11. Internal consistency was .72.

Demographic covariates included time 1 age, race, education, marital status, military branch, military service era, region in which veterans filed their claim, and years since claim initiation.

In-service sexual trauma was assessed at time 1 with the Sexual Harassment Inventory (24).Veterans were considered to have been sexually assaulted if they were forced to have sex against their will or if someone attempted to force them to have sex against their will. Three items from the criminal sexual misconduct subscale of the Sexual Harassment Inventory, plus a fourth question asking about in-service sexual assault unrelated to work, determined veterans’ in-service sexual assault status, which was then dichotomized as a yes-no variable.

Combat exposure was measured at time 1 with a modified 22-item version of the Combat Exposure Index (25). Scores ranged from 0, indicating no combat, to 22, indicating the heaviest combat (Cronbach’s α=.90).

We used chi square and t tests to assess possible gender imbalances in study measures. To address our first two objectives mentioned above, we estimated within each gender (stratum) the rate of gaining and losing PTSD service connection and then the corresponding odds ratios (ORs) and their 95% confidence intervals (CIs).

To address our third objective, we used stratified logistic regression. Because we used a stratified sampling design with oversampling of women, we based all the estimates on weighted stratified analyses, with weights proportional to the inverse of sample selection probabilities. Gaining was defined as the probability of having PTSD service connection at time 2 given the absence of PTSD service connection at time 1. Losing was defined as the probability of not having PTSD service connection at time 2 given the presence of PTSD service connection at time 1.

We conducted these analyses hierarchically. Model 1 included PTSD service connection at time 1, gender, and the PTSD service connection at time 1 × gender interaction. We refer to this as the unadjusted model. For model 2, we added clinical variables as a block, including measures of symptoms and functioning and their change from time 1 to time 2. This block tested whether persisting gender differences in PTSD service connection could be attributed to appropriate sources of variation, such as PTSD symptoms or functional impairment. For model 3, we added demographic characteristics other than gender as a block. In model 4, we added in-service sexual assault. In model 5, we substituted combat exposure for in-service sexual assault. It was at this last step that the gender difference in initial PTSD service connection claim awards became nonsignificant in our prior work (15). Using within-gender means for each scale, we imputed missing time 2 values for covariates. The findings were highly similar when we reran analyses without imputed values. These stratified logistic regression models were used to obtain ORs comparing women’s with men’s odds of gaining and losing VA disability status for PTSD.

The sample included 1,394 men and 1,604 women. Seventy-three percent (N=2,181) of the sample identified as white, 16% (N=488) identified as African American, 3% (N=78) identified as American Indian, 1% (N=35) identified as belonging in other racial-ethnic groups, and 7% (N=195) did not report their race. In addition, 5% (N=164) of veterans in the sample indicated that they were Hispanic. As can be seen in Table 1, men and women differed on most characteristics. The rate of service connection for PTSD for women was 18 percentage points lower than that for men at time 1 and 14 percentage points lower than that for men at time 2.

The rate of gain and loss of PTSD service connection differed by gender. Specifically, 314 of the 756 women (42%) and 201 of the 400 men (50%) who had not been awarded PTSD disability status by time 1 had gained it by time 2. Thus the rate of gain was 42% (CI=38%−45%) for women and 50% (CI=45%−55%) for men. Seventy-one of the 848 women (8%) and 49 of the 994 men (5%) with PTSD service connection at time 1 lost it by time 2. Thus the rate of loss of disability benefits for PTSD was 8% (CI=7%−10%) for women and 5% (CI=4%−6%) for men.

Table 2 presents characteristics of women compared with men who gained and lost PTSD service connection over this ten-year period. Table 3 shows that women had lower unadjusted odds of gaining service connection for PTSD than men. However, adjusting for clinical factors eliminated the gender difference. Further adjustment for demographic characteristics and military trauma type did not meaningfully alter the ORs. Table 4 shows that the odds of losing service-connected benefits for PTSD were higher for women than for men. Controlling for clinical variables reduced but did not eliminate the gender difference. However, the gender difference in loss of PTSD service connection became nonsignificant when we added the demographic block. The ORs remained nonsignificant when we added in-service sexual assault and combat to the models.

We found that women were less likely than men to gain a service connection for PTSD after their initial claim was denied. The gender discrepancy in initial PTSD claim awards that we observed previously (15) was not eliminated over time. However, our findings indicated that the gender difference in gaining PTSD service connection after a claim denial was explained by differences in clinical factors associated with gender. Thus, although differences in combat exposure accounted for the gender difference in initial claim awards in this cohort (15), appropriate sources of variation accounted for the gender difference in gain of PTSD service connection among those whose original PTSD claim was denied.

The finding that 42% of women and 50% of men who were originally denied service connection for PTSD eventually obtained it is also of interest. Assuming that these veterans deserved service connection for PTSD, one must wonder why their initial claim was denied. It may be that they did not sufficiently describe their problems at the time of their original claim, or perhaps these veterans and their advocates became more adept at establishing evidence linking the veterans’ PTSD to military service. Particularly complicated cases might have required more time and effort to sufficiently develop. Alternatively, it may be that the initial evaluation was flawed, leading to a decision reversal once the veteran obtained a more accurate evaluation. Taken together, our findings underscore the need for the VA to implement methods to ensure high-quality and equitable disability evaluations for PTSD at the time of initial claim. A recent study indicates that implementation of standardized disability evaluations using evidence-based assessment measures would increase the quality and reduce the variability in PTSD disability evaluations (26).

The rate of loss of PTSD service connection was low for both men and women. Nevertheless, we observed that women were more likely than men to lose VA disability status for PTSD. Although results are not broken down by gender, the rate of exit from the Social Security disability programs is similarly low (27). Adjusting for clinical factors alone reduced but did not eliminate this gender difference. Adjusting for both clinical factors and demographic characteristics, on the other hand, rendered this gender difference statistically nonsignificant. However, we are mindful to not overinterpret our findings in light of the small number of veterans whose original PTSD claim awards were reversed. Indeed, our findings indicate that more research is warranted to understand loss of PTSD service connection and to determine whether gender or other disparities exist.

To our knowledge, this is the only longitudinal study of former VA disability applicants. The study purposefully oversampled women and thus was well prepared to examine gender differences. Important as this cohort is to answer some questions about changes in disability status, these data were collected before the wars in Iraq and Afghanistan; consequently, we do not know whether the findings generalize to these newer veterans. Women now comprise approximately 12% of forces deployed to these wars. Although technically barred until recently from serving in combat, a growing and unprecedented number of female service members are deployed to combat areas in combat support roles (28).

Furthermore, there have been significant changes in VA policy since the inception of this cohort. First, the VA liberalized the evidentiary requirement for PTSD claims such that it does not require corroboration of a stressor related to fear of hostile military or terrorist activity if a VA doctor confirms that the stressful experience recalled by a veteran adequately supports a diagnosis of PTSD and the veteran’s symptoms are related to the claimed stressor. Second, the VA has made significant advances in PTSD care, including roll-outs of evidence-based psychotherapies for PTSD (29,30), which may allow newer veterans to avoid PTSD’s more disabling course.

We do not know whether these findings pertain to other disability systems, such as Social Security, Worker’s Compensation, or tort claims. Also, we do not know which veterans in the present cohort appealed a denied claim or submitted new claims after their initial denial or whether they intended to do so. For those whose service connection for PTSD status changed, we do not have administrative or clinical information to better elucidate the reasons for change. Over the past few years the VBA has engaged in extensive training to improve processing of PTSD claims related to in-service sexual trauma. Unfortunately, we cannot directly link our findings to these efforts.