Women's Health Issues 25-5 (2015) 535–541
www.whijournal.com
Original article
Posttraumatic Stress Disorder Symptom Severity and Socioeconomic Factors Associated with Veterans Health Administration Use among Women Veterans Keren Lehavot, PhD a,b,c,*, Ruth O’Hara, PhD d,e, Donna L. Washington, MD, MPH f,g, Elizabeth M. Yano, PhD, MSPH f,h, Tracy L. Simpson, PhD b,c,i a
Health Services Research & Development (HSR&D) Center of Innovation (COIN), VA Puget Sound Health Care System, Seattle, Washington Mental Illness Research, Education, and Clinical Center (MIRECC), VA Puget Sound Health Care System, Seattle, Washington Department of Psychiatry and Behavioral Sciences, University of Washington, Seattle, Washington d Sierra Pacific Mental Illness Research, Education, and Clinical Center (MIRECC), Palo Alto VA Health Care System, Palo Alto, California e Department of Psychiatry and Behavioral Sciences, Stanford University, Palo Alto, California f VA Health Services Research and Development (HSR&D) Center for the Study of Healthcare Innovation, Implementation & Policy, VA Greater Los Angeles Healthcare System, Sepulveda and Los Angeles, California g Department of Medicine, University of California Los Angeles (UCLA) David Geffen School of Medicine, Los Angeles, California h Department of Health Policy and Management, UCLA Fielding School of Public Health, Los Angeles, California i Center of Excellence in Substance Abuse Treatment and Education (CESATE), VA Puget Sound Health Care System, Seattle, Washington b c
Article history: Received 12 November 2014; Received in revised form 23 April 2015; Accepted 11 May 2015
a b s t r a c t Background: The Veterans Health Administration (VA) has historically focused on treating men. Although women veterans’ VA use is increasing, they remain more likely than male veterans to receive their care in non-VA settings. To date, there is limited research on factors associated with VA use among women. We examined the relationship between demographic, civilian, military, and health-related variables with past-year VA use among women veterans. Methods: Women veterans were recruited over the internet to participate in an anonymous national survey (n ¼ 617) in 2013. An empirically derived decision tree was computed using signal detection software for iterative receiver operator characteristics (ROC) to identify variables with the best sensitivity/specificity balance associated with past-year VA use. Results: ROC analysis indicated that 85% of participants with high posttraumatic stress disorder (PTSD) and depressive symptoms and who were younger than 54 years of age used VA in the past year. Of those who were 54 years of age or older and had very high PTSD symptoms, 94% used the VA in the last year. By contrast, only 40% of participants with relatively lower PTSD symptoms had VA past-year use, although among these individuals, VA past-year use increased to 65% for those with a relatively lower income. Conclusions: Findings suggest that greater PTSD symptoms, depressive symptoms, and low income correlate with VA use, with very high PTSD symptoms in older groups, high PTSD symptoms coupled with high depressive symptoms in younger groups, and low income in those with lower PTSD symptoms each associated with greater past-year VA use. Ensuring PTSD assessment and treatment, and addressing socioeconomic factors, may be key strategies for health care delivered directly or through contract with VA facilities. Published by Elsevier Inc.
The Veterans Health Administration (VA) is one of the largest integrated health care systems in the United States, with the mission of providing patient-centered, comprehensive services * Correspondence to: Keren Lehavot, PhD, VA Puget Sound Health Care System, 116-POC, 1660 S. Columbian Way, Seattle, WA 98108. Phone: 206-277-1511; fax: 206-764-2293. E-mail address:
[email protected] (K. Lehavot). 1049-3867/$ - see front matter Published by Elsevier Inc. http://dx.doi.org/10.1016/j.whi.2015.05.003
for veterans (Kizer, Demakis, & Feussner, 2000). Although the VA historically focused on treating men, women veterans’ VA use is increasing (Frayne et al., 2010). Nonetheless, they remain more likely than male veterans to receive care in non-VA settings (National Center for Veterans Analysis and Statistics, 2012). Women who obtain care outside the VA tend to have higher income, knowledge gaps about VA care, incorrect assumptions that VA services are unavailable to women, and
536
K. Lehavot et al. / Women's Health Issues 25-5 (2015) 535–541
perceived worse quality of VA care (Shen, Hendricks, Wang, Gardner, & Kazis, 2008; Vogt et al., 2006; Washington, Yano, Simon, & Sun, 2006). On the other hand, screening positive for posttraumatic stress disorder (PTSD) or reporting greater PTSD symptom severity correlate with increased rates of VA use among women veterans (Dobie et al., 2006; Lehavot et al., 2013; Suris, Lind, Kashner, Borman, & Petty, 2004; Washington, Davis, Der-Martirosian, &Yano, 2013a). Low income, poor health, racial/ethnic minority status, and military service-connected disability have been found to be associated with VA use among veterans in general (Mooney, Zwanziger, Phibbs, & Schmitt, 2000; Ouimette, Wolfe, Daley, & Gima, 2003; Skinner & Furey, 1998; Washington, Villa, Brown, Damron-Rodriguez, & Harada, 2005). With the recent passing of the Veterans Access, Choice and Accountability Act of 2014, legislation that allows veterans to seek VA-funded medical care outside the VA in certain circumstances (e.g., if they cannot get an appointment within a certain time period; if they live >40 miles from a VA facility), community providers may see an increase in the number of women veterans referred to them from the VA. Thus, it is increasingly important for VA and community providers to become familiar with the key characteristics that differentiate women veterans who use VA delivered care or VA-funded community care, from women veterans who receive all of their health care outside VA. Although information gleaned from multiple studies contribute insight about those who use VA care and those who do not, most recent studies have focused their examination of VA use around a single variable of interest. One exception is a study that examined the relationship of multiple patient and health characteristics with any use of VA care among women veterans in the past 5 years (Mengeling, Sadler, Torner, & Booth, 2011); results demonstrated that VA users were more likely to have served in a combat area, have a current PTSD diagnosis, and report poorer physical health than non-VA users. Nonetheless, the study population was composed of women who were enrolled in VA care, 52 years old or younger, and living in the Midwest, limiting generalizability of the findings and the ability to differentiate VA users from nonenrollee VA non-users. Moreover, neither this study nor others we are aware of that are recent enough to include veterans of the wars in Iraq and Afghanistan have examined interactions among variables associated with correlates of VA use, which would provide a more fine-grained depiction of those using VA care. Most available studies on women veterans’ VA utilization use survey data that are five years old or older. Given the rapid increase and quickly changing demographics of women veterans (Frayne et al., 2014), an examination of VA use with more recent data is called for. We therefore conducted an exploratory study in a recently collected national sample of women veterans to examine a variety of factors that may be associated with women veterans’ past-year VA use. We assessed a variety of patient characteristics, including demographic, civilian, military, and health-related variables using a data-driven method for modeling that utilizes an iterative receiver operating characteristic curve (ROC) approach. This approach extends prior work by providing information about interactions among variables as well as empirical cut-scores for individual variables, allowing us to better characterize the characteristics or combinations of characteristics of women veterans most strongly associated with VA use in the past year.
Methods Data Source and Study Sample The current study describes a secondary data analysis from a web-based, national survey of women veterans that was conducted from February to May 2013. These women participated in a parent study that aimed to oversample lesbian and bisexual female veterans, and thus advertisements targeted women veterans more broadly as well as lesbian and bisexual veterans (Lehavot & Simpson, 2014). Advertisements were disseminated over the Internet using Facebook advertising and through listservs and online groups focused on veterans, women veterans, and sexual minorities. The study was advertised as an anonymous survey focused on the unique life experiences of women who had served in the military, and survey items assessed a variety of life experiences, stressors, and health symptoms. The primary aim of the parent study was to compare stressors and health symptoms between heterosexual and sexual minority women veterans (Lehavot & Simpson, 2014). Those who followed the survey link were taken to a web-based information statement that explained the anonymous nature of the survey, purpose of the survey, risks and benefits, and eligibility criteria (age 18 years, woman who previously served in U.S. armed forces, living in the United States). This research was approved by the Institutional Review Board at the VA Puget Sound Health Care System. A total of 918 women participated in the study. Seventeen individuals were ineligible because they were transgender (n ¼ 10), underage (n ¼ 1), or did not identify their birth sex, current gender, or both (n ¼ 6). Of the remaining 901 eligible participants, 11 did not respond to the main outcome measure pertaining to VA use and 273 had incomplete data on the main study variables, leaving a final analytic sample of 617 women veterans. Among the 273 individuals who indicated whether they had used VA in the past year but had incomplete data on the main study variables, missing values ranged from less than 1% (n ¼ 2) on age and education to 22% (n ¼ 194) on PTSD symptoms. Before conducting primary analyses, for each of the main study variables, we compared the 617 completers with the individuals with incomplete data on some of the main study variables who had nonmissing data for that variable. There were no differences on rate of past-year VA use or any study variable for the two groups, with the exception that those with missing data were more likely to be a member of a racial/ethnic minority group. Because race/ethnicity was not associated with VA use (p > .05), those with missing data were otherwise similar to completers, and ROC analysis is based solely on complete cases; all subsequent analyses were conducted with the 617 participants who had complete data. Measures VA use The main outcome was whether participants had used VA health care within the last year (yes/no). Patient characteristics Demographic measures included age, racial/ethnic minority status (White or racial/ethnic minority), sexual minority status (heterosexual or sexual minority), marital status (married/in a domestic partnership or not), presence of children in the home
K. Lehavot et al. / Women's Health Issues 25-5 (2015) 535–541
(yes/no), employment (part- or full-time employed versus not employed), household income (categorical scale ranging from under $10,000 to $71,000), and education (categorical scale ranging from less than high school graduate to postgraduate studies). Civilian trauma measures included history of moderate to extreme child abuse or neglect, measured by the 25-item Childhood Trauma Questionnaire and indicated by its scoring guidelines (Bernstein & Fink, 1998; Bernstein et al., 1994); history of adult sexual assault before and/or after military service, indicated if the participant reported experiencing oral, vaginal, or anal sex without consent during this time period (Koss et al., 2007); and history of adult physical victimization before and/or after military service, indicated if the participant reported that she was 1) threatened with physical violence, 2) had objects thrown at her, 3) was chased, followed, or stalked, 4) was punched, hit, kicked, or beaten, or 5) was threatened or assaulted with a weapon during this time period (D’Augelli, 2005). Military-related measures included military service branch, presence of VA service connection pension, and service in Operation Enduring Freedom/Operation Iraqi Freedom/Operation New Dawn (OEF/OIF/OND). Unit support, or the quality of relationships with other military personnel during service, was measured by a 12-item scale from the Deployment Risk and Resilience Inventory (DRRI) with items ranging from 1 (strongly disagree) to 5 (strongly agree; King, King, & Vogt, 2003). History of combat exposure was measured using a 15-item scale from the DRRI. History of sexual harassment during military service was assessed with an item asking whether the participant was ever subjected to uninvited or unwanted sexual attention, such as touching, cornering, pressure or sexual favors, or verbal remarks. Sexual assault and physical victimization during military service were each assessed by the same items used for civilian sexual assault and physical victimization. Health factors included PTSD symptom severity in the past four weeks as indicated by the 17-item PTSD Checklist-Civilian Version (PCL-C; range, 17–85; Weathers & Ford, 1996) that is associated with the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV) PTSD diagnostic criteria (4th ed.; American Psychiatric Association, 1994), depressive symptoms in the past 2 weeks as indicated by the 8-item Patient Health Questionnaire (PHQ; range, 0–24; Kroenke et al., 2009), and alcohol misuse in the past year as indicated by the 10-item Alcohol Use Disorders Identification Test (range 0–40; Saunders, Aasland, Babor, de la Fuente, & Grant, 1993). For each of these scales, higher scores indicated greater symptoms. Physical health was measured using the General Health Rating item from the SF-12: “In general, you would say your health is.” Responses ranged from 1 (excellent) to 6 (very poor), with greater numbers indicating worse selfrated health (Ware, Kosinski, & Keller, 1996). Statistical Analysis We first compared demographic, civilian, military, and health characteristics between participants who had used VA in the past year versus not, using independent t tests for continuous variables and c2 tests for categorical variables. To determine the set of variables most strongly associated with past-year VA use, we computed an empirically derived decision tree with signal detection software for iterative ROC (Signal Detection Software ROC Software, 2007). Kraemer expanded ROC analysis to specifically identify the characteristics of individuals most likely to have a specified outcome (Kraemer,
537
2002). Unlike a traditional linear or logistic regression analysis, ROC is a data-driven, exploratory approach and is nonparametric in that it does not make restrictive linearity or additive assumptions. When the linear model assumptions of logistic analyses are satisfied and all potential interactions are included in the model, the results of ROC and logistic regression analysis are likely very similar. However, logistic regression models can assign the same risk score to participants who may have very different combinations of associated variables. Further, the iterative ROC analysis is entirely agnostic in that the researcher is only required to specify the independent variables and outcome of interest. There is no requirement to prespecify potential interactions, because this information is yielded by the ROC analytic process itself. As such, this approach is especially useful for hypothesis-generating (as opposed to hypothesis-testing) research and exploratory analyses. The ROC examines the independent impact of each variable and all possible interactions among variables, with their associated cut-points, on the outcome, identifying those with optimally balanced sensitivity and specificity. Specifically, this iterative ROC analytic approach first calculates the sensitivity and specificity for all of the independent variables and their associated cut-points against the outcome of interest. The ROC then identifies the variable and its associated cut-point with the optimal sensitivity and specificity for identifying the outcome of interest, as indicated by a kappa statistic. When the optimal variable and associated cut-point are identified, their association with the outcome of interest is tested with a 2 2 c2, with significance set at the p < .01 level (O’Hara et al., 2002). The ROC then splits the sample into two groups based on the cut-point for the variable that yielded the optimal sensitivity and specificity. The software reiterates the process in each of these two groups to further identify variables and the cut-points most strongly associated with past-year VA use, resulting in additional subgroups. The analysis continues until the remaining variables are not significant or there are no longer enough participants for further analysis. The ROC analysis thus identifies interactions among variables and results in a decision tree that divides the sample into subgroups with the best sensitivity and specificity for the outcome of interest (O’Hara et al., 2002). Using the variables and interactions identified by the ROC approach, we conducted a logistic regression as a secondary analysis to examine whether findings from the ROC could be replicated using this more traditional approach. Results Participant Characteristics Of the 617 women veterans in the analytic sample, 85% of participants were non-Hispanic White, 5% were Hispanic, 4% were African American, 4% were mixed race, and 3% were other. In addition, 38% identified as lesbian or bisexual and 26% served in OEF/OIF/OND. The mean age was 49.41 (SD ¼ 13.81) and most participants had at least a college degree (53%) and a household income more than $36,000 (65%). Fifty-five percent indicated past-year VA use. Table 1 shows the bivariate relationship between the independent variables and past-year VA use. Women veterans who used the VA in the past year (vs. those who did not) were younger and were significantly more likely to not be married or in a domestic partnership, not have full- or part-time employment, and to report lower income. They were also more likely to
538
K. Lehavot et al. / Women's Health Issues 25-5 (2015) 535–541
Table 1 Demographic, Civilian, Military, and Health-Related Variables by Past-Year VA Use among Women Veterans (N ¼ 617) Variable
Demographics Age, M (SD)*** Racial/ethnic minority, n (%) Sexual minority, n (%) Married/domestic partnership, n (%)* Children, n (%) Employed, n (%)* College graduate, n (%) Income, n (%)*** <10,000 10,000–15,000 16,000–20,000 21,000–25,000 26,000–35,000 36,000–50,000 51,000–70,000 71,000 Civilian Moderate to extreme child abuse, n (%) Adult sexual assault, n (%)*** Adult physical victimization, n (%)*** Military Branch, n (%)** Air Force Army Coast Guard Marines Navy Reserves VA SC pension, n (%)*** OEF/OIF/OND, n (%)*** Unit support, M (SD)y,*** Combat, n (%)*** Sexual harassment, n (%)*** Sexual assault, n (%)*** Physical victimization, n (%)*** Health PTSD symptoms, M (SD)z,*** Depressive symptoms, M (SD)z,*** Alcohol misuse, M (SD)z General health, n (%)*** Excellent Very good Good Fair Poor Very poor
Used VA in Last Year Yes (n ¼ 339)
No (n ¼ 278)
47.39 56 139 137 100 142 182
51.87 37 98 137 75 140 144
25 22 26 27 46 63 48 82
(13.29) (17) (41) (40) (30) (42) (54) (7) (7) (8) (8) (14) (19) (14) (24)
230 (68) 139 (41) 246 (73)
56 169 2 22 73 16 109 105 35.91 154 275 163 223
(17) (50) (1) (7) (22) (5) (32) (31) (12.45) (45) (81) (48) (66)
47.96 (20.35) 10.31 (7.33) 4.46 (5.53) 21 59 92 115 41 11
(6) (17) (27) (34) (12) (3)
9 11 8 13 30 49 44 114
(14.04) (13) (35) (49) (27) (50) (52) (3) (4) (3) (5) (11) (18) (16) (41)
185 (67) 79 (28) 152 (55)
58 105 5 18 64 28 27 54 40.22 78 188 78 119
(21) (38) (2) (7) (23) (10) (10) (19) (12.11) (28) (68) (28) (43)
32.56 (16.74) 5.82 (5.90) 4.19 (4.93) 32 74 98 57 15 2
(12) (27) (35) (21) (5) (1)
Abbreviations: OEF/OIF/OND, Operation Enduring Freedom/Operation Iraqi Freedom/Operation New Dawn; SC, service connection. * p < .05. **p < .01. ***p < .001. y Higher numbers indicate greater level of unit support during military service. z PTSD symptoms measured with PTSD Checklist (PCL; range, 17–85); depressive symptoms measured with Patient Health Questionnaire (PHQ; range, 0–24); alcohol misuse measured with Alcohol Use Disorders Identification Test (AUDIT; range, 0–40). Higher numbers indicate worse symptoms or greater alcohol misuse.
report both civilian and military sexual assault and physical victimization, military sexual harassment, combat exposure, VA service connection pension, less unit social support during military service, greater PTSD and depressive symptoms, and poorer general health. Empirically Derived ROC Decision Tree As shown in Figure 1, PTSD symptoms provided the best sensitivity and specificity for detecting past-year VA use. If a
woman veteran scored 42 or greater on the PCL, her likelihood of past-year VA use was 75%, whereas if she scored less than 42 her likelihood of past-year VA use was 40%. Among the subgroup of women with a PCL score of 42 or greater, being younger than 54 years of age was associated with a 80% likelihood of VA use, versus 64% for those who were 54 years or older. Among those who were younger than 54 years of age, a score of 9 or greater on the PHQ (assessing depressive symptoms) was associated with an 85% likelihood of VA use compared with 65% for those with a score of less than 9 on the PHQ. Among those who were 54 years or older, a score of 73 or greater on the PCL was associated with a 94% likelihood of VA use compared with 55% for those with a score of less than 73 (but 42 or greater) on the PCL. For women veterans who scored lower than 42 on the PCL, the next variable that maximized association with past-year VA use was income. Income less than $26,000 was associated with a 65% likelihood of past-year VA use, whereas those with an income of $26,000 or greater had a 34% likelihood. Among those with income $26,000 or greater, a PCL score of 24 or greater (but less than 42) was associated with a 45% likelihood of VA use, compared with a 24% likelihood if the PCL was less than 24. Table 2 provides the sensitivity and specificity values computed by the ROC analysis; these are computed in an iterative manner associated with the decision tree, such that each sensitivity and specificity value applies only to the subset of participants identified by the previous variable in the decision tree and its cutpoint. Secondary Analysis The initial variable most optimally associated with VA pastyear use identified by the ROC was PTSD symptoms. We created a categorical variable for PTSD symptoms based on the various cut-points identified in the ROC model (PCL <24, 24 to <42, 42 to <73, and 73). Using a logistic regression, PTSD symptoms were significantly associated with past-year VA use at the bivariate level (odds ratio [OR], 2.18 [95% CI, 1.41–3.37] for PCL 24 to <42; OR, 5.94 [95% CI, 3.78–9.33] for PCL 42 to <73; OR, 15.08 [95% CI, 6.68–34.06] for PCL 73; reference group, PCL <24). When all other main effects and interactions indicated by the ROC analysis were entered into the logistic regression (i.e., PTSD, age, income, depression, PTSD income, PTSD age, PTSD age depression, using cut-points identified by the ROC), the only significant interaction was between PTSD symptoms and income (Wald ¼ 12.91; p ¼ .005). The three-way interactions indicated by the ROC analysis could not be tested with the logistic regression analysis, given that the ROC used the PTSD symptom variable multiple times with different cut-off scores. Discussion The purpose of the current study was to identify a set of characteristics associated with past-year VA use among women, one of the most rapidly growing groups of veterans generally and among VA users specifically (Frayne et al., 2010; Yano et al., 2010). Results of the ROC analysis indicated that 85% of participants with a PCL of 42 who were younger than 54 years of age and had a of PHQ 9 used the VA in the last year. Of those who were 54 years of age or older and had very high scores on the PCL (73), 94% used the VA in the last year. On the other hand, only 40% of participants with a PCL of <42 used VA in the last year. Nonetheless, if those individuals reported an income of less than
K. Lehavot et al. / Women's Health Issues 25-5 (2015) 535–541
539
Figure 1. Decision tree for past-year Veteran’s Health Administration (VA) use based on empirically derived cut-scores. Red indicates greater likelihood and blue indicates lesser likelihood of past-year VA use within each subgroup. Posttraumatic stress disorder (PTSD) symptoms measured with PTSD Checklist (PCL, range 17–85); depressive symptoms measured with Patient Health Questionnaire (PHQ, range 0–24). Higher numbers indicate worse symptoms on each scale.
$26,000, VA use in the past year increased to 65%. In sum, the combination of PTSD symptoms, age, and depressive symptoms, or PTSD symptoms and income were the most optimal variables in our model associated with past-year VA use. PTSD symptom severity emerged as the initial variable with the best balance of sensitivity and specificity associated with VA use over and above other characteristics. This finding parallels those of several other studies that indicate significant associations between PTSD and VA utilization (Dobie et al., 2006; Lehavot et al., 2013; Suris et al., 2004; Washington et al., 2013a). It also strongly supports the routine PTSD screening of women veterans who come into VA care, which has been universally implemented in VA (The Management of Post-Traumatic Stress Working Group, 2004), as well as advocates for ensuring that VA’s community-based clinics and medical centers are robustly equipped to provide women veterans with specialty mental health care if needed. The optimal cut-point detected on the PCL for past-year use was 42, a score that is suggestive of possible PTSD diagnosis. Different cut-points on the PCL have been identified in the literature as indicating possible PTSD diagnosis, with some using a cut-point of 38 for women veterans and others using 50 (Dobie et al., 2002, 2004; Vogt, Pless, King, & King, 2005). The PCL score identified in this study as most strongly associated with past-year VA use falls within this range. Indeed, VA administrative data from fiscal 2012 showed that mental health conditions were in the top five domains of most frequent medical conditions in women VA patients, and of these mental health conditions PTSD ranked as second most common
Table 2 Sensitivity, Specificity, and c2 Values by Order of Computation in ROC Analysis for Past-Year VA Use Variable and Cut-point 1. 2. 3. 4. 5. 6.
PCL 42 Age < 54 Income < $26,000 PHQ 9 PCL 73 PCL 24
Sensitivity
Specificity
c2
0.58 0.75 0.30 0.81 0.33 0.66
0.76 0.43 0.89 0.41 0.96 0.58
76.62*** 7.94** 20.41*** 7.89** 8.76** 13.80***
Abbreviations: PCL, PTSD Checklist; PHQ, Patient Health Questionnaire; ROC, receiver operator characteristics; VA, Veteran’s Health Administration. The sensitivity and specificity values are computed in an iterative manner associated with the decision tree, such that each sensitivity and specificity value applies only to the subset of participants identified by the previous variable in the decision tree and its cut-point. * p < .05, **p < .01, ***p < .001.
among women aged 18 to 64 (Frayne et al., 2014). The fact that PTSD symptom severity emerged as the most optimal initial characteristic associated with past-year VA use may indicate that women veterans specifically seek VA for its specialty mental health services, because the VA is a recognized leader in treating PTSD. Nonetheless, our data cannot directly test this hypothesis. Among women with higher PTSD symptoms (i.e., PCL 42), age further refined prediction of past-year VA use. Women younger than 54 years of age with higher PTSD symptoms were more likely to have used VA (80%) than women who were 54 and older with higher PTSD symptoms (64%). In general, the age distribution of women veterans using VA has shifted in the last decade, with greater numbers of younger women now using VA; additionally, women VA patients are younger than their male counterparts and more likely to have served in OEF/OIF/OND (Frayne et al., 2014). Our data further suggest that there is an interaction between PTSD symptoms and age in characterizing VA users, such that those who are younger are more likely than older veterans to use VA services if their PTSD symptoms are moderately high (PCL 42). However, among the older women veterans, even more severe PTSD symptoms (PCL 73) were associated with a greater likelihood of VA use. Among women with relatively lower PTSD symptoms (i.e., PCL < 42), income further refined prediction of past-year VA use. Specifically, lower income was associated with greater likelihood of past-year VA use among these women. Like PTSD, income has also been identified by prior research as being associated with VA use, indicating that the VA is more likely to be used by those who cannot afford other care or have a financial disincentive to do so (Mooney et al., 2000; Washington et al., 2013a). This reflects on the relatively more disadvantaged economic situation of those patients coming to the VA, and that VA may serve as a health care “safety net.” Limitations The current study had several limitations. Although we included several variables potentially associated with VA use, measures of women’s knowledge regarding VA eligibility, perceptions of VA quality, and other reported barriers to care that may impact VA use were unavailable (Washington et al., 2006). Furthermore, eligibility for VA care was not measured directly. Eligibility for VA health benefits is affected by a number of factors; most veterans who separated under any condition other than dishonorable are eligible, and different priority groups are
540
K. Lehavot et al. / Women's Health Issues 25-5 (2015) 535–541
assigned based on variables such as having a service-connected disability, being low income, recent military service, and military sexual trauma. Despite not measuring VA eligibility directly, VA non-users in our study included women veterans with each of these VA eligibility factors. Our study oversampled lesbian and bisexual women veterans and recruitment took place over the Internet. As a result, the sample may not be generalizable to all women veterans (although sexual orientation was not a significant corrolate in the model). For instance, the rate of VA use in our sample was greater than what has been indicated by prior research (Washington et al., 2006), although this may be explained by the relatively greater proportion of OEF/OIF/OND women veterans in our sample compared with some of this previous research, because they are the military service era cohort with the greatest use of VA health care (Washington, Bean-Mayberry, Hamilton, Cordasco, & Yano 2013b). Nonetheless, our sample demographics are similar in many ways to those reported for women veterans in the National Survey of Women veterans, a nationally representative sample of this population (Washington, Bean-Mayberry, Riopelle, & Yano, 2011), with approximately one-half of the women in each of these samples being college graduates and employed, and the majority being White. Nonetheless, the ROC statistical approach is inherently sample specific and exploratory, with findings meant to generate hypotheses that can be tested with future samples. Implications for Practice and/or Policy The data presented from this recent, national, online study with women veterans contribute to the literature on women veterans and VA health care use. We examined the potential impact of a variety of demographic, civilian, military, and healthrelated characteristics using an exploratory approach. Although previous studies on women veterans have highlighted the independent roles of PTSD and income on VA use, our findings illustrate the complex interactions among PTSD symptoms, depressive symptoms, and age, as well as between PTSD symptoms and income, in association with VA use. Results from our exploratory study suggest that women with high PTSD and depressive symptoms and who are relatively younger (or with very high PTSD symptoms and who are older) are most likely to use the VA. These interactions should be further evaluated a priori in large, representative surveys of women veterans. Also, it is unclear whether these women are currently being treated for PTSD. Follow-up studies could investigate whether those who report PTSD symptoms in a study are likely to be screened by their health care provider for these symptoms and referred to treatment, and whether this differs between VA and non-VA providers. Although the VA has instituted mandatory screenings for PTSD and depression among all veterans, community providers may now also serve women veterans who are referred from the VA under the Veterans Access, Choice and Accountability Act of 2014. Even before implementation of the Veterans Choice Act, VA has been using non-VA contracts and fee-basis care to provide access to health care services through community providers, for services that are not available on-site at VA facilities, with twice as many women veterans as male veterans utilizing this form of care (Frayne et al., 2012). Community providers serving VAreferred women veterans may consider implementing such routine mental health screenings, many of which are brief (e.g., the 4-item Primary Care-PTSD screener and 2-item PHQ screener for depression; Kroenke, Spitzer, & Williams, 2003; Prins et al.,
2003). Among women with lower PTSD symptoms, our data suggest that lower income is associated with VA use, and thus an assessment of patients’ needs for social services (e.g., housing, financial stressors or assistance) may be beneficial. Offering patients these types of services based on assessed need may improve therapeutic rapport and engagement with treatment. In conclusion, the decision tree model improves our ability to distinguish which women veterans use VA services and provides information to help prepare clinicians to recognize their unique characteristics and needs. Addressing mental health needs and socioeconomic factors may be key strategies for VA and for community providers serving women veterans who are referred from the VA, to effectively care for our nation’s women warriors. Acknowledgments This research was supported by a research grant to Drs. Lehavot and Simpson from the VISN-20 Mental Illness Research, Education, and Clinical Center (MIRECC). Dr. Lehavot’s effort was supported by a VA CSR&D Career Development Award (IK2 CX000867) and Dr. Yano’s effort was supported by a VA HSR&D Service Senior Research Career Scientist award (Project # RCS 05–195). The authors have no conflicts of interest. The material in this manuscript is the result of work supported by resources from the VA Puget Sound Health Care System (KL, TS), Palo Alto VA Health Care System (RO), and VA Greater Los Angeles Health Services Research and Development Center (DW, EY). The views expressed in this article are those of the authors and do not necessarily reflect the position or policy of the Department of Veterans Affairs or the United States government. References American Psychiatric Association (APA). (1994). Diagnostic and statistical manual of mental disorders (4th ed). Washington, DC: Author. Bernstein, D. P., & Fink, L. (1998). Childhood trauma questionnaire manual. San Antonia: The Psychological Corporation. Bernstein, D. P., Fink, L., Handelsman, L., Foote, L., Lovejoy, M., Wenzelm, K., . Ruggiero, J. (1994). Initial reliability and validity of a new retrospective measure of child abuse and neglect. American Journal of Psychiatry, 151, 1132– 1136. D’Augelli, A. R. (2005). Developmental and contextual factors and mental health among lesbian, gay, and bisexual youths. In A. M. Omoto, & H. S. Kutzman (Eds.), Sexual orientation and mental health: Examining identity and development in lesbian, gay, and bisexual people.. Washington, DC: American Psychological Association. Dobie, D. J., Kivlahan, D. R., Maynard, C., Bush, K. R., McFall, M., Epler, A. J., & Bradley, K. A. (2002). Screening for post-traumatic stress disorder in female Veteran’s Affairs patients: Validation of the PTSD checklist. General Hospital Psychiatry, 24, 367–374. Dobie, D. J., Kivlahan, D. R., Maynard, C., Bush, K. R., Davis, T. M., & Bradley, K. A. (2004). Posttraumatic stress disorder in female veterans: Association with self-reported health problems and functional impairment. Archives of Internal Medicine, 164, 394–400. Dobie, D. J., Maynard, C., Kivlahan, D. R., Johnson, K. M., Simpson, T., David, A. C., & Bradle, K. (2006). Posttraumatic stress disorder screening status is associated with increased VA medical and surgical utilization in women. Journal of General Internal Medicine, 21, S58–S64. Frayne, S. M., Phibbs, C. S., Friedman, S. A., Berg, E., Ananth, L., Iqbal, S., . Herrera, L. (2010). Sourcebook: Women veterans in the Veterans Health Administration. Volume 1. Sociodemographic characteristics and use of VHA care. Washington DC: Women’s Health Evaluation Initiative, Women Veterans Health Strategic Health Care Group, Veterans Health Administration, Department of Veterans Affairs. Frayne, S. M., Phibbs, C. S., Friedman, S. A., Saechao, F., Berg, E., Balasubramanian, V., . Hayes, P. M. (2012). Sourcebook: Women veterans in the Veterans Health Administration. Volume 2. Sociodemographics and Use of VHA and Non-VA Care (Fee). Washington DC: Women’s Health Evaluation Initiative, Womens Health Services, Veterans Health Administration, Department of Veterans Affairs.
K. Lehavot et al. / Women's Health Issues 25-5 (2015) 535–541 Frayne, S. M., Phibbs, C. S., Saechao, F., Maisel, N. C., Friedman, S. A., Finlay, A., . Haskell, S. (2014). Sourcebook: Women veterans in the Veterans Health Administration. Volume 3. Sociodemographics, utilization, costs of care, and health profile. Washington DC: Women’s Health Evaluation Initiative, Women’s Health Services, Veterans Health Administration, Department of Veterans Affairs. King, D. W., King, L. A., & Vogt, D. S. (2003). Manual for the Deployment Risk and Resilience Inventory (DRRI): A collection of measures for studying deploymentrelated experiences of military veterans. Boston: National Center for PTSD. Kizer, K. W., Demakis, J. G., & Feussner, J. R. (2000). Reinventing VA health care: Systematizing quality improvement and quality innovation. Medical Care, 38, 7–16. Koss, M. P., Abbey, A., Campbell, R., Cook, S., Norris, J., Testa, M., . White, J. (2007). Revising the SES: A collaborative process to improve assessment of sexual aggression and victimization. Psychology of Women Quarterly, 31, 357– 370. Kraemer, H. C. (2002). Evaluating medical tests: Objective and quantitative guidelines. Newbury Park, CA: Sage. Kroenke, K., Spitzer, R. L., & Williams, J. B. W. (2003). The Patient Health Questionnaire-2: Validity of a two-item depression screener. Medical Care, 41, 1284–1292. Kroenke, K., Strine, T. W., Spitzer, R. L., Williams, J. B., Berry, J. T., & Mokdad, A. H. (2009). The PHQ-8 as a measure of current depression in the general population. Journal of Affective Disorders, 114, 163–173. Lehavot, K., Der-Martirosian, C., Simpson, T. L., Briggs, E. C., Ippen, C. G., Ostrowski, S. A., . Pynoos, R. S. (2013). The role of military social support in understanding the relationship between PTSD, physical health, and healthcare utilization in women veterans. Journal of Traumatic Stress, 26, 1–4. Lehavot, K., & Simpson, T. L. (2014). Trauma, posttraumatic stress disorder, and depression among sexual minority and heterosexual women veterans. Journal of Counseling Psychology, 61, 392–403. Mengeling, M. A., Sadler, A. G., Torner, J., & Booth, B. M. (2011). Evolving comprehensive VA women’s health care: Patient characteristics, needs, and preferences. Women’s Health Issues, 21, S120–S129. Mooney, C., Zwanziger, J., Phibbs, C. S., & Schmitt, S. (2000). Is travel distance a barrier to veterans’ use of VA hospitals for medical surgical care? Social Science and Medicine, 50, 1743–1755. National Center for Veterans Analysis and Statistics (2012). FY 2012 profile of unique veteran users. Available: www.va.gov/vetdata/docs/SpecialReports/ Profile_of_Unique_Veteran_Users.pdf. Accessed October 6, 2014. O’Hara, R., Thompson, J. M., Kraemer, H. C., Fenn, C., Taylor, J. L., Ross, L., . Tinklenberg, J. R. (2002). Which Alzheimer patients are at risk for rapid cognitive decline? Journal of Geriatric Psychiatry and Neurology, 15, 233–238. Ouimette, P., Wolfe, J., Daley, J., & Gima, K. (2003). Use of VA health care services by women veterans: findings from a national sample. Women and Health, 38, 77–91. Prins, A., Ouimette, P., Kimerling, R., Cameron, R. P., Hugelshofer, D. S., Shaw-Hegwer, J., . Sheikh, J. I. (2003). The primary care PTSD screen (PC-PTSD): Development and operating characteristics. Primary Care Psychiatry, 9, 9–14. Saunders, J. B., Aasland, O. G., Babor, T. F., de la Fuente, J. R., & Grant, M. (1993). Development of the Alcohol Use Disorders Identification Test (AUDIT): WHO Collaborative Project on Early Detection of Persons with Harmful Alcohol ConsumptiondII. Addiction, 88, 791–804. Shen, Y., Hendricks, A., Wang, F., Gardner, J., & Kazis, L. E. (2008). The impact of private insurance coverage on veterans’ use of VA care: Insurance and selection effects. Health Services and Research, 43, 267–286. Signal Detection Software ROC Software. (2007). Version 4.22. Stanford and Palo Alto, CA: Stanford University School of Medicine and the Sierra Pacific Mental Illness Research Education and Clinical Centers (MIRECC). Skinner, K. M., & Furey, J. (1998). The focus on women veterans who use Veterans Administration health care: The VA women’s health project. Military Medicine, 163, 761–766. Suris, A., Lind, L., Kashner, T. M., Borman, P. D., & Petty, F. (2004). Sexual assault in women veterans: An examination of PTSD risk, health care utilization, and cost of care. Psychosomatic Medicine, 66, 749–756. The Management of Post-Traumatic Stress Working Group. (2004, January). VA/ DOD clinical practice guidelines for the management of post-traumatic
541
stress. Version 1. Available: www.healthquality.va.gov/ptsd/ptsd_full.pdf. Accessed October 6, 2014. Vogt, D. S., Bergeron, A., Salgado, D., Daley, J., Ouimette, P., & Wolfe, J. (2006). Barriers to Veterans Health Administration care in a nationally representative sample of women veterans. Journal of General Internal Medicine, 21, S19– S25. Vogt, D. S., Pless, A. P., King, L. A., & King, D. W. (2005). Deployment stressors, gender, and mental health outcomes among Gulf War I veterans. Journal of Traumatic Stress, 18, 115–127. Ware, J., Kosinski, M., & Keller, S. D. (1996). A 12-item short form health survey: construction of scales and preliminary tests of reliability and validity. Medical Care, 34, 220–233. Washington, D. L., Bean-Mayberry, B., Hamilton, A. B., Cordasco, K. M., & Yano, E. M. (2013b). Women veterans’ healthcare delivery preferences and use by military service era: Findings from the National Survey of Women Veterans. Journal of General Internal Medicine, 28, S571–576. Washington, D. L., Bean-Mayberry, B., Riopelle, D., & Yano, E. M. (2011). Access to care for women veterans: Delayed healthcare and unmet need. Journal of General Internal Medicine, 26(Suppl 2), 655–661. Washington, D. L., Davis, T. D., Der-Martirosian, C., & Yano, E. M. (2013a). PTSD risk and mental health care engagement in a multi-war era community sample of women veterans. Journal of General Internal Medicine, 28, 894– 900. Washington, D. L., Villa, V., Brown, A., Damron-Rodriguez, J., & Harada, N. (2005). Racial/ethnic variations in patterns of VA ambulatory care use. American Journal of Public Health, 95, 2231–2237. Washington, D. L., Yano, E. M., Simon, B., & Sun, S. (2006). To use or not to use: What influences why women veterans choose VA health care. Journal of General Internal Medicine, 21, S11–S18. Weathers, F., & Ford, J. (1996). Psychometric properties of the PTSD Checklist (PCL-C, PCL-S, PCL-M, PCL-PR). In B. H. Stamm (Ed.), Measurement of stress, trauma, and adaptation.. Lutherville, MD: Sidran Press. Yano, E. M., Hayes, P., Wright, S., Schnurr, P. P., Lipson, L., Bean-Mayberry, B., & Washington, D. L. (2010). Integration of women veterans into VA quality improvement research efforts: What researchers need to know. Journal of General Internal Medicine, 25, S56–S61.
Author Descriptions Keren Lehavot, PhD, is a Core Investigator at the Health Services Research & Development (HSR&D) Center of Innovation (COIN) at VA Puget Sound Health Care System and Assistant Professor at the University of Washington Department of Psychiatry and Behavioral Sciences. Her research focuses on women veterans’ mental health and health disparities.
Ruth O’Hara, PhD, is Associate Professor at the Stanford Department of Psychiatry and Behavioral Sciences. Her research focuses on the physiological markers of neurocognitive impairment in a range of late-life disorders. She has expertise in the use of receiver operator characteristics analysis.
Donna L. Washington, MD, MPH, is a Professor of Medicine at the VA Greater Los Angeles Healthcare System and UCLA David Geffen School of Medicine. Her research examines health care access and quality for women and racial/ethnic minorities, with a focus on veterans.
Elizabeth M. Yano, PhD, MSPH, is Co-Director and a Research Career Scientist at the VA Greater Los Angeles Healthcare System and Adjunct Professor of Health Services at the UCLA School of Public Health. Her work focuses organizational influences on quality.
Tracy L. Simpson, PhD, is a Clinician Investigator at the VA Puget Sound Health Care System and Associate Professor at the University of Washington Department of Psychiatry and Behavioral Sciences. Her work focuses on PTSD and substance use comorbidity and randomized clinical trials.