Mental Health Service Use in a Representative Sample of National Guard Soldiers

Data were drawn from the Ohio Army NG Mental Health Initiative (OHARNG MHI). The OHARNG MHI is a dynamic longitudinal cohort of Ohio Army NG soldiers who are interviewed annually to assess mental health, substance use, and life experiences. All soldiers were asked to participate in the study with the option to opt out. Between June 2008 and February 2009, a total of 11,212 soldiers did not opt out of the study, and accurate contact information was available for 6,514 (58%) participants. Of these 6,514 soldiers, 4,198 were contacted before we reached our target sample of eligible participants (N=2,616); therefore, 2,316 (36%) participants had not been contacted when the baseline cohort closed. Among the 6,514 for whom contact information was received, 187 (3%) were ineligible (for example, because they were too young or were retired), 31 (1%) were disqualified (did not speak English or had problems hearing), and 1,364 (21%) did not wish to participate. The overall response rate was 43%, calculated as the number of participants who completed a survey or who consented but were ineligible divided by the number of working telephone numbers minus those that were disqualified (N=2,616+187/6,514–31).

Baseline interviews were conducted between November 2008 and November 2009 (wave 1). Participants were contacted for follow-up interviews one year later during wave 2 (November 2009–November 2010), and again during wave 3 (November 2010–November 2011). A total of 1,770 (68%) soldiers from the original group of 2,616 responded to the wave 2 follow-up survey; we restricted the final sample to soldiers who participated in both waves 2 and 3 (N=1,189). As reported previously (9), our sample was demographically comparable to the entire OHARNG at time of sampling, although the sample participants were slightly older and less likely to be married. Factors associated with loss to follow-up between waves 1 and 2 included age (p<.001), with the youngest (18–25 years old) least likely to return; race (p<.05), with blacks least likely to return; marital status (p<.001), with those who were never married least likely to return; education (p<.001), with those having the lowest level of education least likely to return; and rank (p<.01), with enlisted individuals least likely to return. Factors associated with loss to follow-up between waves 2 and 3 included age (p<.01), with the youngest group least likely to return; marital status (p<.01), with those who were divorced, separated, or widowed least likely to return; education, (p<.001), with those having the lowest education (high school graduate, GED, or less) least likely to return; and psychopathology (p<.01), with those who had psychopathology less likely to return.

After giving informed consent, soldiers participated in telephone interviews that collected information on mental health, substance use, military experiences, and life events history. The study design was reviewed by the University of Michigan and the Case Western Institutional Review Board, and the informed consent of the participants was obtained after the nature of the procedures had been fully explained.

Sociodemographic factors included gender, age, race-ethnicity, income, health insurance status, education level, and marital status. Age was divided into categories of 18–24, 25–34, 35–44, and ≥45. Race categories included white, black, and other race, and ethnicity included Hispanic and non-Hispanic. Income was measured as ≤$60,000 or >$60,000. Insurance status was measured as a dichotomous variable indicating presence or absence of health insurance. Military factors included rank (officer or enlisted, cadet, or civilian employee status) and deployment experience (none, deployment to Iraq or Afghanistan, or other).

We recruited a random subsample of 500 participants who were compensated for participating in an in-person clinical validation interview in addition to their telephone interviews (10). In-depth, in-person interviews were conducted by clinicians at a location or setting familiar to the participant and lasted an average of two hours. Diagnoses of PTSD, depression, generalized anxiety disorder, and alcohol abuse and dependence as well as suicidal ideation were validated by comparing telephone interview diagnostic assessments with in-person diagnostic assessments.

We assessed PTSD at wave 2 consistent with DSM-IV criteria. To screen positive for PTSD, participants had to meet all DSM-IV criteria. Criterion A1 (directly experiencing the traumatic event) was assessed with a list of 25 potentially traumatic non–deployment-related experiences and a list of 45 potentially traumatic deployment-related experiences. The list of potentially traumatic non–deployment-related experiences included items from the Life Events Checklist in the Clinician-Administered PTSD Scale (CAPS) (11,12) and items commonly used in population-based studies (13). The list of potentially traumatic deployment-related experiences included the 25 nondeployment experiences asked in the context of deployment and an additional 20 experiences from the Deployment Risk and Resilience Inventory (14). Criterion A2 (witnessing the event as it occurred to others) was assessed by asking participants whether they felt terrified or helpless while watching the occurrence of an event that they self-identified as the “worst.” Criterion B (re-experiencing the event), criterion C (avoidance and numbing), and criterion D (hyperarousal) were assessed by using questions from the PTSD Checklist–Civilian Version (PCL-C) (15,16). Finally, we assessed criterion E (greater than one month’s duration of symptoms) and criterion F (clinically significant impairment) by asking participants about the duration of their illness and degree of impairment. PTSD was validated in this sample with CAPS results from the clinical in-person validation study; the sensitivity, specificity, and Cronbach’s alpha, respectively, were comparable across deployment-related (.50, .93, and .95) and non–deployment-related (.47, .94, and .93) assessments (10).

The Patient Health Questionnaire–9 (PHQ-9) (17) was used to evaluate depression symptoms at wave 2 on the basis of DSM-IV criteria. Each of the nine questions is scored as 0 (not at all) to 3 (nearly every day), with total scores ranging from 0 to 27. Individuals who met criteria for either major depressive disorder or other depressive disorder and reported that his or her symptoms seemed to have occurred together were identified as having depression. Other depressive disorder was diagnosed if two to four depressive symptoms were present for at least “more than half the days” in the past two weeks, and one of the symptoms was depressed mood or anhedonia. Major depressive disorder was diagnosed if five or more of the nine depressive symptoms that constitute DSM-IV criteria were present for at least more than half the days in the past two weeks, and one of the symptoms was depressed mood or anhedonia. This algorithm was validated against the depression module in the Structured Clinical Interview for DSM-IV Axis I Disorders (SCID I) (18) and yielded moderate sensitivity (.51), high specificity (.83), and a Cronbach’s alpha of .66 (10).

To assess generalized anxiety disorder, participants were administered the Generalized Anxiety Disorder Assessment (GAD-7) (19) at wave 2. The GAD-7 evaluates co-occurring symptoms on a scale of 0, not at all, to 3, nearly every day, with possible scores ranging from 0 to 21. To screen positive for generalized anxiety disorder, a person must have a score of ≥10, have symptoms for at least six months, and report functional impairment. Validated against the generalized anxiety disorder module in the SCID I, generalized anxiety disorder had low sensitivity (.04), high specificity (.96), and a Cronbach’s alpha of .72 (10).

A single variable called alcohol use disorders included both alcohol abuse and alcohol dependence, both of which were assessed by using the Mini-International Neuropsychiatric Interview (20) at wave 2 to evaluate the presence of DSM-IV criteria. Participants who displayed one or more of the maladaptive patterns that constitute criterion A of substance abuse and whose symptoms led to impairment or distress screened positive for alcohol abuse. Alcohol dependence was classified as displaying three or more of the maladaptive patterns that constitute criterion A of substance dependence and having symptoms that have led to impairment or distress. Validated against the SCID I, alcohol abuse had moderate sensitivity (.40), high specificity (.80), and a Cronbach’s alpha of .57; alcohol dependence had higher sensitivity (.60), comparably high specificity (.81), and a Cronbach’s alpha of .76 (10).

To assess past-year use of mental health services at wave 3, participants were asked, “Since we last spoke, did you ever go to see any of the following professionals or self-help groups for problems with your emotions or nerves or for problems with your use of alcohol or drugs?” Participants were then asked whether they received such services through a physician; a psychiatrist; a psychologist; a minister, priest, rabbi, or other spiritual advisor; “another type of mental health professional”; or a self-help group, such as Alcoholics Anonymous and Narcotics Anonymous. Participants were allowed to endorse multiple provider categories. This set of questions was derived from the National Comorbidity Survey (21).

First, we computed chi square statistics examining the relationship between independent variables that were measured at wave 2 and mental health service utilization at wave 3. In addition, we calculated crude odds ratios (ORs) and confidence intervals for mental health service use at wave 3 after adjustment for each independent variable. The variable for “mental health need” was coded positive if any of the mental health conditions described above were present at wave 2.

Second, we performed a multivariable logistic regression analyzing predictors of mental health service use within the entire sample. We included as predictors the independent variables from wave 2 that were significantly related to use of mental health services at wave 3. Third, we performed multivariable logistic regression with the same model but only among participants who had a mental disorder. We performed sequential logistic regressions in the order indicated by Tables 2 and 3. All comparisons and statistical tests were performed in SAS 9.2.

Fewer than one in five soldiers (16%) used mental health services in the past 12 months. Several variables were associated with significantly increased odds of using mental health services within the past 12 months at wave 3: being female; insured; divorced, separated, or widowed; and non-Hispanic black or other non-Hispanic race-ethnicity; having poor general medical health; and having mental health need, including suicidal ideation, at wave 2 (Table 1). Age, income, rank, and deployment experience were not significantly associated with mental health service use.

After adjusting for all factors that were significantly associated with mental health service use, being female (OR=1.7), non-Hispanic black (OR=2.3), Hispanic (OR=2.6), and insured (OR=2.8) and having poor general medical health (OR=1.9) and mental health need (OR=4.8) were associated with increased service use (Table 2).

Among the 227 participants with mental health need at wave 2, 85 (37%) used mental health services in the 12 months prior to wave 3. Being non-Hispanic black (OR=4.1) was the sole significant predictor of mental health service use at wave 3 when all other variables were entered into the model (Table 3).

Of those with mental health problems who used any mental health service, over half saw a physician (59%) or a psychologist (51%), and nearly half (45%) saw a psychiatrist (Table 4). Over a quarter (28%) reported seeing clergy or a spiritual advisor, and a relatively small percentage (10%) reported use of self-help groups.

As shown in Figure 1, mental health need at wave 2 was associated with differential use of services during the following 12 months. Participants with depression or alcohol use disorders were less likely than participants with other mental health problems to access mental health services. There was also a dose-response relationship between number of disorders and prior-year use of mental health services. [A figure illustrating this relationship is available online as a data supplement to this article.]