Correlates & Predictors of Mental Healthcare Utilization for Veterans with PTSD: A Systematic Review

Abstract

Objective:

Treatment utilization among veterans with PTSD is low. Understanding correlates and predictors of mental healthcare for veterans with PTSD is critical to facilitating treatment utilization. However, given the size of the literature base and disparate findings, it is difficult to interpret available research. Correlates and predictors of mental healthcare for veterans with PTSD published between 2012–2016 were examined to (1) define the scope of recent literature and (2) summarize predictive evidence.

Method:

This two-phase systematic review conducted scientific database searches. Phase 1 defined the scope of recent literature (n=51) and tabulated types of correlates, outcomes, and sources of bias. Phase 2 summarized results from prospective studies (n=17) evaluating mental healthcare utilization. PROSPERO ID# CRD42017082686.

Results:

Demographics/Social Network Characteristics (61%) and Evaluated Need (61%) were the most common correlates of mental health utilization. Facilitators with the strongest evidence for predicting utilization (both initiation and retention) were Characteristics of the Episode of Care (e.g., mental health in primary care) and greater PTSD symptom severity. Study biases favoring VA enrollees (92%) and excluding outside VA care outcomes (86%) limit generalizability of results. Conclusions: Efforts focused on providing mental health in primary care, interventions to address mental health beliefs, and outreach to racial/ethnic minority veterans and those with mild to moderate PTSD could increase mental healthcare utilization.

Posttraumatic stress disorder (PTSD) is common among the approximately 20 million United States military veterans (Westat, 2010); prevalence estimates range from 8.1–15.2% depending on gender and service era, which is higher than the general population (Gradus, 2017). Veterans with PTSD experience impairment in mental health, physical, social, and occupational functioning (Erbes, Kaler, Schult, Polusny, & Arbisi, 2011; Magruder et al., 2004; Possemato, Wade, Andersen, & Ouimette, 2010). Fortunately, both psychotherapy and pharmacotherapy are effective (Haagen, Smid, Knipscheer, & Kleber, 2015; Puetz, Youngstedt, & Herring, 2015).

Large national studies show low utilization of mental healthcare among veterans with PTSD. For example, among veterans from Operation Enduring Freedom and Operation Iraqi Freedom (OEF/OIF) conflicts with new diagnoses of PTSD, only 9.5% received an adequate dose of mental healthcare and 20% attended no mental health appointments within the first year (Seal et al., 2010). Similarly, among veterans over age 49 with new diagnoses of PTSD, 26% did not attend a mental health appointment within a year (Smith, Cook, Pietrzak, Hoff, & Harpaz-Rotem, 2016). Neither of these estimates include veterans who have existing diagnoses or remain undiagnosed suggesting that estimates of untreated PTSD are likely higher. These numbers also suggest problems with both initiation (given a sizable proportion of veterans receiving no treatment) and retention (given the small proportion of veterans receiving an adequate does), two key elements of mental health treatment utilization.

While research about veterans is pertinent to the Veterans Administration (VA), it also has relevance elsewhere. Seventy three percent of veterans report never being enrolled in VA healthcare (Westat, 2010). Veterans often receive community healthcare services because many have alternate healthcare coverage through private insurance, Medicaid, or Medicare (Shen, Hendricks, Zhang, & Kazis, 2003), and others may not be eligible for VA healthcare. Even among VA enrollees, 19% reported only non-VA providers in the past year, and 36% reported both VA and non-VA care (Shen et al., 2003). Thus, it is important for both community and VA healthcare systems to engage veterans with PTSD in treatment.

The literature suggests unique barriers and facilitators affect former military service members, veterans, in the treatment seeking process. Two recent reviews suggest some military-related barriers and facilitators which likely persist following military separation (e.g., military culture, stigma, social support) (Coleman, Stevelink, Hatch, Denny, & Greenberg, 2017; Hom, Stanley, Schneider, & Joiner, 2017). However, they also identified some barriers and facilitators which are likely relevant only to actively serving military personnel (e.g., impact on military career, military leadership) suggesting the veteran population is likely distinct from actively serving service members (Coleman et al., 2017; Hom et al., 2017). Given these factors which make veterans different from civilian and active duty military populations, it is important to consider veterans treatment utilization independently.

Based on another recent review on mental healthcare utilization, individuals with PTSD also face unique treatment barriers. Kantor, Knefel, and Lueger-Schuster (2016) highlight trauma-specific factors as a unique barrier to care for trauma-survivors (individuals having experienced or witnessed a traumatic life event involving death, serious injury, or sexual violence). Examples cited within the review include concerns about re-experiencing traumatic event, avoidance of trauma reminders, and concerns about talking about past traumas in treatment (Kantor et al., 2016). Given that trauma-specific factors make PTSD different from other mental health concerns, we believe it is crucial to consider mental healthcare barriers and facilitators specifically for individuals with PTSD.

Searches of electronic databases with the title of this review returns tens of thousands of results with various topic areas including stigma, mental health-related beliefs, logistic barriers, etc. This amount of literature and wide range of topics makes interpretation difficult for providers and administrators seeking to change practices, and researchers hoping to advance the field.

Objectives & Review Questions.

The ultimate aim of this comprehensive systematic review is to inform efforts to improve access to and utilization of mental healthcare for United States veterans with PTSD. For the purposes of this manuscript, both initiation and retention phases of mental health treatment are included in the conceptualization of utilization. We did not focus specifically on VA enrollees or VA users. There were two key research questions. First, what is the scope of the literature on correlates of mental healthcare utilization for veterans with PTSD? The objective of Phase 1 of this review was to identify gaps in the current knowledge base and recommend areas for ongoing research. Second, which mental healthcare predictors have the strongest evidence base and are the most promising targets for intervention to improve mental healthcare utilization in this population? The objective of Phase 2 of this review was to summarize results from prospective studies to inform policy, interventions, and implementation.

Methods

This review protocol is registered at the PROSPERO international prospective register of systematic reviews (CRD42017082686) and complies with PRISMA reporting guidelines (Liberati et al., 2009; Moher, Liberati, Tetzlaff, Altman, & The, 2009). The review questions, search strategy, and inclusion/exclusion criteria were informed by the PICOS framework (Moher et al., 2009). The following search terms were used: (“PTSD” OR “post-traumatic stress disorder” OR “posttraumatic stress disorder”) AND (“care” OR “treatment” OR “services”) AND (“utilization” OR “access” OR “barrier” OR “facilitator”) AND veteran. A comprehensive literature search was conducted using PubMed, PsycInfo, MEDLINE, CINAHL Plus, and Scopus databases; these databases were selected to cover a broad range of multidisciplinary literature. Primary literature searches were conducted January 10–13, 2017. An additional search was conducted in the PILOT database on October 6, 2017. These searches resulted in 1,238 search results; once duplicates were removed, this reflected 628 unique manuscripts. Authors also reviewed references of included articles and personal libraries resulting in 16 additional articles, for a total of 644 unique records (see Online Supplement).

Articles were included if they were English language, peer-reviewed original research articles about mental healthcare utilization in a United States veteran population with PTSD or probable PTSD published in the past five years (2012–2016). We limited the search to articles within the last five years to focus on findings that are most relevant to the current healthcare and sociocultural environment. Articles had to describe correlates of mental healthcare and present associated mental healthcare utilization outcome data. Mental healthcare utilization was operationalized to include both initiation and retention. Any mental healthcare (including general mental health, substance use, PTSD specific treatment, etc.) was the outcome of focus rather than PTSD services specifically because much of the literature does not distinguish PTSD treatment from general mental healthcare, practical difficulties distinguishing PTSD specific treatment, and because veterans with PTSD often have comorbid concerns which warrant non-PTSD focused treatment. We did not place inclusion/exclusion restrictions on study designs, comparison conditions, or treatment types; however, given that studies had to describe both correlates and associated outcomes, this limited the sample to quantitative designs. Articles about current military service members were excluded due to the likelihood that barriers to mental healthcare would differ between active duty military and veteran populations. Articles about health behaviors (e.g., weight management) were excluded because of the likelihood that barriers to treatment would be different for physical health and mental health concerns.

Initial eligibility was determined by screening titles and abstracts, followed by a full text review for potentially eligible articles. Data were extracted by trained research staff using a structured form with key data elements including study population characteristics, healthcare correlates, and utilization outcomes. Following data extraction, the first author reviewed all records and checked data accuracy.

Synthesis of results involved two phases corresponding to the two objectives. In phase 1, correlates of mental healthcare utilization among veterans with PTSD, utilization outcomes, and study quality and risk of bias were documented to evaluate the scope of the literature base. A general inductive approach (Thomas, 2006) was used to classify correlates and outcomes in the literature to identify gaps and strengths in the literature.

Given that the review spans both clinical trials and observational studies, classification of risk of bias as typically reported for either design was not applicable. Therefore, we classified the risk of bias based on commonly cited limitations and our own assessment of methodological limitations using observation and clinical trials tools as guidelines. Following an initial review of methods and limitations noted in articles in the sample, we generated a list of limitations and other potential risks that would significantly affect the quality of results. These quality indicators were summarized using descriptive statistics to estimate quality and risk of bias across studies in the review. The second author acted as auditor to this process to ensure that findings were replicable. Following review of the results, individual risk of bias summaries and overall literature search results were considered and summarized to identify risks of bias affecting the cumulative evidence.

In phase 2, all prospective studies were used to summarize evidence for mental health utilization predictors in veterans with PTSD. We thus focused on studies that evaluated veterans prior to enrollment in specialty mental health. Because of the broad range of methodologies and outcome variables, we focused on summarizing outcomes of each study based on the results of the primary analyses and the objectives and hypotheses the study was designed to address. We did not include preliminary, post-hoc, or other exploratory analyses. We organized findings by the content domains identified in Phase 1. We did not attempt a quantitative synthesis (e.g., meta-analysis) because varying study designs made direct comparisons meaningless and at times obscured the studies’ main intentions.

Results

Fifty-one articles met inclusion criteria and were included in the current review. The Online Supplement lists studies included in the review and key characteristics.

Goal 1: Define literature scope.

To define the scope of recent literature, we identified themes in mental healthcare correlates and evaluated the methodological rigor of the studies.

Inductive analysis of correlates identified nine domains (Table 1). Upon review, we recognized that many of these domains were similar to those included in Andersen’s Behavioral Model, which is appropriate given that much of the literature on healthcare utilization uses this model as a guiding framework (Andersen, 1995). To be consistent with other literature in the field, where appropriate we used the same terms as the Behavioral Model for these domains. However, given that the themes were guided by an inductive approach, these labels did not always fit consistently and were modified as needed.

Table 1.

Phase 1 domains of mental healthcare correlates and utilization outcomes.

Domains	n	%	Examples
Mental healthcare correlatesa
Demographics & Social Network Characteristics	31	61%	Age, Gender, Race/Ethnicity, Marital status, Education, Occupation, Military Characteristics, Income, Legal problems, Social support
Evaluated Need	31	61%	PTSD severity, Overall health status, Types of PTSD symptoms, Comorbid conditions
Characteristics of the Episode of Care when Referred/Seeking Treatment	25	49%	Index clinic type of diagnosis/referral, Treatment type offered, Provider characteristics
Access & Enabling Resources	22	43%	Health insurance, Transportation, Travel time, Time to attend appointments, Service Connection status, Rural/Urban residence, Region
Previous Healthcare Utilization	18	35%	Previous mental healthcare, Medical healthcare utilization
Perceived Need	10	20%	Perceived need or desire for treatment
Mental Health Beliefs & Stigma	7	14%	Attitudes about mental healthcare, Beliefs about mental illness and treatment efficacy, Knowledge, Self-Stigma, Public Stigma
Personal Characteristics & Health Practices	2	4%	Coping skills, Coping styles, Resiliency, Personality
Satisfaction & Trust	2	4%	VA and provider trust, Privacy, Treatment preferences
Mental health utilization outcomes
Treatment initiation	31	61%	Attended at least one session, Received prescription of psychotropic medication
Amount of utilization	26	51%	Number of sessions attended, Number of prescription refills, Dosage achieved
Treatment completion	18	35%	Completed a specified course of treatment, Attended a minimum “dose” of sessions (e.g., 9 or more sessions), Achieved adequate length and dose of prescription medication
Type of treatment	13	25%	Psychotherapy/Medication management/Combined, Type of psychotherapy, Type of medication

^aMental healthcare correlate domain labels informed by the Behavioral Model (Andersen, 1995)

The nine domains reflect a broad scope including person factors (e.g., demographics, beliefs), environmental factors (e.g., social support, access), and healthcare system factors (e.g., clinic type where diagnosed). Mental health utilization outcomes were narrower and focused primarily on amount of mental healthcare utilization, status as a treatment initiator or completer, and types of treatment offered. Some outcomes were highly specific (e.g., completion of a specified psychotherapy protocol) and reflect specific constructs within utilization, while others were very broad and combine multiple utilization constructs (e.g., count of any mental health appointments reflecting both initiation and retention). Even within similar outcomes, utilization constructs (i.e., initiation, retention, completion, treatment selection) were operationalized differently across studies.

Prevalent correlates of mental health care utilization included Demographics & Social Network Characteristics (n=31, 61%), Evaluated Need (n=31, 61%), Characteristics of the Episode of Care when Referred/Seeking Treatment (n=25, 49%), Access & Enabling Resources (n=22, 43%), and Previous Healthcare Utilization (n=18, 35%). Prevalent outcomes included Treatment Initiation (n=31, 61%) and Amount of Utilization (n=26, 51%). Table 2 shows the prevalence of articles reporting on each correlate domain and provides examples of specific outcomes within each domain.

Table 2.

Number and percent of articles reporting on specific utilization outcomes within each correlate domain for the total sample and final sample.

Domains	Outcomes																Total Across All Outcomes
	Amount of utilization				Treatment initiation				Treatment completion				Type of treatment
	Total		Final		Total		Final		Total		Final		Total		Final		Total		Final
	n	%	n	%	n	%	n	%	n	%	n	%	n	%	n	%	n	%	n	%
Demographics & Social Network Characteristics	13	42%	5	45%	19	61%	6	55%	11	35%	3	27%	8	26%	2	18%	31	100%	11	100%
Mental Health Beliefs & Stigma	1	14%	1	25%	5	71%	2	50%	1	14%	1	25%	2	29%	2	50%	7	100%	4	100%
Access & Enabling Resources	11	50%	2	29%	13	59%	5	71%	8	36%	2	29%	8	36%	2	29%	22	100%	7	100%
Perceived Need	2	20%	2	29%	8	80%	7	100%	3	30%	3	43%	1	10%	1	14%	10	100%	7	100%
Evaluated Need	13	42%	4	36%	20	65%	7	64%	11	35%	4	36%	8	26%	3	27%	31	100%	11	100%
Personal Characteristics & Health Practices	1	50%	1	50%	1	50%	1	50%	0	0%	0	0%	2	100%	2	100%	2	100%	2	100%
Previous Healthcare Utilization	8	44%	2	33%	9	50%	4	67%	9	50%	3	50%	3	17%	0	0%	18	100%	6	100%
Satisfaction & Trust	0	0%	0	0%	2	100%	1	100%	0	0%	0	0%	0	0%	0	0%	2	100%	1	100%
Characteristics of the Episode of Care when Referred/Seeking Treatment	10	40%	5	45%	16	64%	8	73%	11	44%	4	36%	6	24%	4	36%	25	100%	11	100%
Total Across All Domains	26	51%	8	47%	31	61%	11	64%	18	35%	5	29%	13	25%	5	29%	51	100%	17	100%

Note. Total reflects all studies included in Phase 1 of this review and Final reflects all studies included in Phase 2 of this review.

To evaluate the methodological rigor and risk of bias, we assessed each study for limitations which could contribute to bias (Table 3). Due to the nature of the topic, observational designs are often appropriate so this was not considered an indicator of bias for this review. The literature overall demonstrates a bias toward evaluating utilization among veterans already enrolled in VA healthcare (n=47, 92%) and VA healthcare outcomes exclusively (n=44, 86%), which limits generalizability to veterans seeking care outside of VA. Many studies also focused on specific subgroups (n=35, 69%). For example, 37% (n=19) of articles focused on veterans who had already sought mental healthcare and enrolled for treatment. Another common sample was OEF/OIF and Operation New Dawn (OND) era veterans (n=17, 33%). Across studies, there was also a high reliance on retrospective (n=25, 49%) and administrative (n=30, 59%) data. A relative strength is that very few studies report on small samples (defined as less than 125 participants, see Table 3 for justification) (n=8, 16%).

Table 3.

Number of studies demonstrating common risk factors for bias

	Whole review sample (N=51)		Final review sample (n=17)
	n	%	n	%
Only included veterans enrolled in VA	47	92%	16	94%
Did not include non-VA care in outcomes	44	86%	13	76%
Selective Sample	35	69%	9	53%
-Treatment seekingb	19	37%	N/Aa	N/Aa
-OEF/OIF/ONDc only	17	33%	7	41%
Heavily reliant on administrative data	30	59%	3	18%
Retrospective datad	25	49%	N/Aa	N/Aa
Focused only on a single predictor	20	39%	5	29%
Very small sample (≤125)e	8	16%	5	29%

^aThese are not applicable as the final sample was selected based on these quality indicators.

^bTreatment-seeking is defined as veterans who had already sought mental healthcare and enrolled for treatment in a specialty mental health clinic or research study. We did not count veterans engaged in interventions or services (e.g., care management or primary care mental health integration) delivered in medical settings designed to expand access to specialty mental health as treatment-seeking.

^cOperation Enduring Freedom (OEF), Operation Iraqi Freedom (OIF), and/or Operation New Dawn (OND) era Veterans.

^dProspective was defined as studies either collecting data prospectively as part of a study monitoring veterans over time or administrative data which was sampled in a way which allowed prospective analysis (e.g., data referenced reflected only characteristics of the veteran’s status prior to the mental healthcare utilization outcomes of interest)

^eA sample size of 125 or less was defined as a small sample because general power calculations show that these studies have less than 80% power to find a medium or larger effect size between two samples. Although specific studies have different power based on different analytic plans, sampling strategies, and proposed effects, this was selected as a general benchmark.

Goal 2: Summarize evidence.

To better understand mental healthcare utilization predictors, we more closely reviewed prospective studies (Table 4). Seventeen articles were eligible. Two articles reported on the same sample from the same study using many of the same predictor variables and similar models with different outcomes so they were considered as one study (Spoont et al., 2014; Spoont et al., 2015).

Table 4.

Summary of the evidence for predictor constructs.

Domain	Summary
Demographics & Social Network Characteristics 9 of 10 studies have significant findings but low consistency	Gender 3 of 6 significant/consistent: Studies suggested higher utilization (i.e., number of visits, initiation, psychotherapy retention) among women veterans (Blais et al, 2014; Gallegos et al, 2015; Spoont et al., 2015). One study found different predictors for male and female Veterans (Breland et al., 2015). null findings (Grubbs et al., 2015; Jain, Greenbaum, & Rosen, 2012; Spoont et al., 2014) Age 2 of 4 significant: One showed some younger veterans are more likely to initiate/remain engaged in mental healthcare than veterans age 65+ (Spoont et al., 2014; Spoont et al., 2015); another associated older age with increased number of psychotherapy sessions but not initiation (total sample and women) (Breland et al., 2015). null findings (Harpaz-Rotem et al., 2016; Jain et al., 2012) Race/Ethnicity 4 of 6 significant/consistent: Significant studies showed White veterans were more likely to attend mental healthcare and be prescribed psychotropic medications, half were specific to certain racial/ethnic groups (Blais et al., 2014; Breland et al., 2015; Harpaz-Rotem et al., 2016; Spoont et al., 2014; Spoont et al., 2015) null findings (Grubbs et al, 2015; Jain et al, 2012) Employment 2 of 3 significant/inconsistent: One study showed employed veterans were less likely to initiate mental healthcare (Johnson et al., 2016), another showed occupational impairment predicted mental health medications (Kaier, Possemato, Lantinga, Maisto, & Ouimette, 2014), null findings (Grubbs et al, 2015) Education 1 of 1 significant: higher education predicted higher CPT engagement (Grubbs et al., 2015) Income 0 of 2 significant: Generally null (Grubbs et al., 2015; Spoont et al., 2015) Military history 2 of 3 significant/inconsistent: Three studies report on military history characteristics: one found more mental health visits with combat exposure (military component was null) (Blais et al., 2014), and two evaluated era with one finding OEF/OIF veterans were less likely to receive an SSRI/SNRI therapeutic dose (Jain et al., 2012) the other showed null results (Spoont et al., 2015) Social network: 5 of 7 significant/inconsistent: Two studies evaluated social support: one showed encouragement to seek PTSD treatment predicted increased initiation (Spoont et al., 2014); the other had null findings (DeViva, Sheerin, et al., 2016). Two studies evaluated social functioning: one showed impairment was associated with increased initiation (Johnson et al., 2016); the other showed null results (Kaier et al., 2014). Three studies evaluated marital status: one demonstrated increased psychotropic prescriptions among married veterans, (Harpaz-Rotem et al., 2016); one demonstrated decreased psychotherapy visits among married female Veterans but not male Veterans (Breland et al., 2015); one null results (Grubbs et al., 2015). One study demonstrating that being a parent was associated with more mental health visits and medication days (Kaier et al., 2014)
Mental Health Beliefs & Stigma 3 of 3 studies have significant findings and high consistency	All studies showed positive beliefs about mental healthcare predicts higher mental healthcare utilization (including initiation and retention) (DeViva, Sheerin, et al., 2016; Johnson et al., 2016; Spoont et al., 2014; Spoont et al., 2015). One study suggested beliefs about types of treatment (e.g., psychotherapy/medication) predicts utilization of that type of treatment (Spoont et al., 2014; Spoont et al., 2015). Two studies showed higher stigma predicts higher mental health initiation (DeViva, Sheerin, et al., 2016; Spoont et al, 2014)
Access & Enabling Resources 3 of 5 studies have significant findings and there is low consistency	Two studies showed anticipated access barriers and drive time associated with lower psychotherapy and overall mental health utilization, but not pharmacotherapy with the exception of predicting lower initiation in pharmacotherapy (DeViva, Sheerin, et al., 2016; Spoont et al., 2014; Spoont et al., 2015) One study showed urban residents were less likely to be prescribed psychotropic medications (Harpaz-Rotem et al., 2016) Two studies associated service connected status with higher mental healthcare utilization (veterans with a service connection status of 50% or higher were more likely to be prescribed psychotropic medications and veterans pending a service connection decision were more likely to engage in CPT) (Grubbs et al., 2015; Harpaz-Rotem et al., 2016), one showed the opposite effect in initiation and retention (Spoont et al., 2014; Spoont et al., 2015), null findings (Breland et al, 2015)
Perceived Need 3 of 6 studies have significant findings and there is moderate consistency	One study showed perceived treatment need predicted psychotherapy and pharmacotherapy initiation and retention (Spoont et al., 2014; Spoont et al., 2015); another showed interest in mental healthcare was associated with psychotherapy initiation but not number of sessions (Breland et al., 2015) One study showed passive study enrollment correlated with lower CPT initiation (Grubbs et al., 2015) One study showed that interest in physical healthcare was associated with decreased psychotherapy initiation (Breland et al., 2015) null findings (Jain et al, 2012; Johnson et al, 2016; Kaier et al, 2014)
Evaluated Need 9 of 10 studies have significant results but there is low consistency	PTSD severity was associated with increased mental healthcare utilization (i.e., initiation, retention, & adequate dose across psychotherapy and pharmacotherapy) (DeViva, Sheerin, et al., 2016; Harpaz-Rotem et al., 2016; Shiner et al., 2014; Spoont et al., 2014; Spoont et al., 2015); one study showed the opposite (Grubbs et al., 2015) null findings (Breland et al, 2015; Jain et al, 2012; Johnson et al, 2016) One study showed worse general mental health symptoms were associated with higher initiation and retention. (Spoont et al., 2014; Spoont et al., 2015) null findings (Johnson et al, 2016) Five studies evaluated comorbid conditions: two found that comorbid conditions increase psychotropic medication initiation (Harpaz-Rotem et al., 2016; Jain et al., 2012); one found medical but not mental health comorbidities decreased psychotherapy visits among the total sample and women veterans (not men) (Breland et al., 2015); one found comorbid depression decreases engagement in CPT (Grubbs et al., 2015), but another was null for depression (Breland et al., 2015); null findings (Kaier et al., 2014) Two studies evaluated the impact of specific symptom clusters and both showed null results for hyperarousal, one showed a lower number of visits associated with avoidance symptoms, and they had opposite findings regarding re-experiencing symptoms (one predicted more visits the other predicted less visits) (Blais et al., 2014; Johnson et al., 2016; Kaier et al., 2014)
Personal Characteristics & Health Practices 1 of 2 studies has significant results	One study demonstrated that self-efficacy to follow treatment recommendations predicted initiation of psychotherapy or pharmacotherapy (Spoont et al., 2014) null findings (DeViva, Sheerin, et al., 2016)
Previous Healthcare Utilization 4 of 6 studies has significant results with moderate consistency	One study showed a correlation between specialty medical and mental health visits (Kaier et al., 2014) Two studies show veterans who have received psychotherapy are more likely to be prescribed psychotropic medications (Harpaz-Rotem et al., 2016; Jain et al., 2012); another demonstrated that the total sample was more likely to initiate psychotherapy and male Veterans (not female Veterans) attended more psychotherapy sessions if they had an antidepressant prescription (Breland et al., 2015) null findings (Grubbs et al., 2015; Spoont et al., 2015)
Satisfaction & Trust 0 of 1 studies has significant results	null findings (Johnson et al, 2016)
Characteristics of the Episode of Care when Referred/Seeking Treatment 8 of 10 studies have significant results with moderate consistency	Generally primary care collaborative care management interventions (TOPS, TIDES/PTSD, RESPECT-PTSD) increased mental healthcare utilization (including psychotropic prescriptions; initiation, number of visits, and completion of psychotherapy; initiation, number of visits, and completion of CPT) (Fortney et al., 2015; Hoerster et al., 2015; Schnurr et al., 2013) Pilot support for two interventions to increase treatment utilization. (Possemato et al., 2016; Watts et al., 2015). A third supports an intervention for women but not men veterans (Gallegos et al., 2015) One of three studies evaluating impact of setting demonstrate that diagnosis in a medical setting compared with a mental health setting is associated with lower initiation and retention (Spoont et al., 2014; Spoont et al., 2015); one demonstrates that an adequate number of sessions is most likely if care is in both PC-MHI and specialty mental health settings, followed by specialty mental health settings (Shiner et al., 2014); null findings (Breland et al., 2015; Jain et al., 2012)

Demographics & Social Network Characteristics, Evaluated Need, and Characteristics of the Episode of Care when Referred/Seeking Treatment were most frequently evaluated. Demographics & Social Network Characteristics research was mixed with discrepant findings between studies overall. However, the literature does suggest that male veterans and veterans who identify as racial and/or ethnic minorities are potentially at risk for failure to utilize (both initiate and continue) mental healthcare (Blais, Hoerster, Malte, Hunt, & Jakupcak, 2014; Gallegos et al., 2015; Harpaz-Rotem, Rosenheck, Mohamed, Pietrzak, & Hoff, 2016; Spoont et al., 2014; Spoont et al., 2015). PTSD symptom severity (Evaluated Need) was generally correlated with mental health service utilization across a broad range of outcomes including initiation, retention, and adequate dose across psychotherapy and pharmacotherapy suggesting that the veterans with the highest needs were successfully engaging in treatment (DeViva, Sheerin, et al., 2016; Harpaz-Rotem et al., 2016; Spoont et al., 2014; Spoont et al., 2015). Within the Characteristics of the Episode of Care when Referred/Seeking Treatment category, there was consistent evidence for services which can improve treatment utilization (both initiation and retention) among this population including primary care mental health integration services (such as care management) and decision aids (Fortney, Pyne, Kimbrell, & et al., 2015; Hoerster et al., 2015; Possemato et al., 2016; Schnurr et al., 2013; Watts et al., 2015).

Mental Health Beliefs & Stigma, Access & Enabling Resources, Perceived Need, Personal Characteristics & Health Practices, and Satisfaction & Trust were less commonly evaluated. The research on Mental Health Beliefs & Stigma, suggests that positive beliefs about mental healthcare facilitate treatment initiation and retention, but also shows that veterans with higher levels of stigma were more likely to initiate treatment (DeViva, Sheerin, et al., 2016; Johnson et al., 2016; Spoont et al., 2014; Spoont et al., 2015). Consistent differences in directionality of findings between beliefs and stigma suggests that these are separate constructs with different implications for treatment utilization. More studies evaluated Perceived Need but the studies with positive findings are diverse which makes interpretation challenging; however, positive findings were generally in the same direction suggesting that a self-identified or self-perceived need might increase utilization (Breland, Greenbaum, Zulman, & Rosen, 2015; Grubbs et al., 2015; Spoont et al., 2014; Spoont et al., 2015).

Discussion

This systematic review covers correlates and predictors of mental healthcare utilization in veterans with PTSD. Many of the domains identified in the literature (Goal 1) demonstrated significant overlap with the Behavioral Model highlighting this model’s influence, utility, and broad scope (Andersen, 1995). Characteristics of the Episode of Care when Referred/Seeking Treatment, Evaluated Need, and Demographics & Social Network Characteristics were the most studied categories. Findings indicate that veterans with the most severe symptoms both initiate and continue mental healthcare, and that specific mental health programming can improve initiation and retention in mental health treatment. Generally, results from studies in this review didn’t differentiate predictors for the treatment initiation and retention phases of utilization.

Characteristics of the Episode of Care when Referred/Seeking Treatment has a relatively large evidence base and consistent evidence across a broad spectrum of outcomes; outcomes reported in the studies included in this review include initiation, number of visits, and completion across treatment types including psychotherapy, medications, and specific psychotherapy protocols (i.e., Cognitive Processing Therapy [CPT]). Many of these studies reflect efforts to improve healthcare systems and increase mental healthcare utilization. Several interventions build on implementation of mental health services into primary care. Primary care behavioral health and collaborative care models increase access through same day access, and providing behavioral healthcare in a convenient setting where patients area already receiving care (Dollar, Kearney, Pomerantz, & Wray, 2018; Post, Metzger, Dumas, & Lehmann, 2010).

There was a wealth of literature regarding demographic predictors of mental healthcare among veterans with PTSD but findings were discrepant. Although some studies suggest male veterans and veterans who identify as a racial/ethnic minority with PTSD are potentially at risk for under-utilization of mental healthcare, these findings were not consistent. The lack of consensus could suggest inequity is not always the case, especially in light of another finding from this review, that veterans with more severe PTSD (those most in need) were most likely to engage in mental healthcare. However, veterans with mild to moderate PTSD symptoms still experience considerable functional impairment (Brancu et al., 2016); therefore, outreach efforts directed at veterans with lower PTSD severity are indicated. Additional research is needed to understand whether there are healthcare disparities between groups of veterans, to help target outreach efforts to groups of underserved veterans, and tailor interventions to those most likely to benefit.

Some categories with a smaller literature show promise for further development. One is the Mental Health Beliefs & Stigma category, which showed consistent research support as a predictor of mental health utilization in the few studies that evaluated this domain. The valence of beliefs about mental healthcare consistently predicted healthcare utilization in the expected direction such that more positive beliefs about mental healthcare predicted more utilization and negative beliefs about mental healthcare predicted lower utilization. In fact, the study evaluating both initiation and retention in mental health among veterans with PTSD showed some evidence that the nature of beliefs (e.g., positive beliefs about psychotherapy vs. medication) might predict the type of treatment a veteran subsequently engages in (Spoont et al., 2014; Spoont et al., 2015). This warrants replication and extension, but provides evidence that beliefs about mental healthcare are related to utilization and suggests further research on interventions designed to address negative beliefs about mental healthcare (e.g., cognitive behavioral therapy for treatment seeking (Stecker, Fortney, & Sherbourne, 2011)) is warranted.

It is surprising that research also consistently showed that veterans reporting higher perceived stigma were more likely to engage in mental healthcare. Potential reasons could be that stigma is more salient for veterans who are considering mental healthcare, or veterans experiencing more severe symptoms are more likely to experience stigma. An in depth quantitative review specifically focused on stigma and help-seeking among service members demonstrated that stigma is frequently reported but not associated with quantitative utilization patterns as would be expected (Sharp et al., 2015). A qualitative review describes how stigma can affect the timing and subjective experience of mental healthcare utilization in military populations (Coleman et al., 2017). These findings help us better contextualize the findings from our own study and suggest that stigma is present and affects the experience of care but not quantitative mental healthcare utilization outcomes.

Several methodological characteristics stood out as risk factors for bias in the field. Particularly notable is the emphasis on research only involving veterans already enrolled in VA and limited research on non-VA mental healthcare. Given that 73% of veterans report no VA healthcare use at all (Westat, 2010), and among VA enrollees 73% of veterans have alternative healthcare coverage, and 55% use at least some healthcare outside of the VA (Shen et al., 2003), there is a large amount of mental healthcare utilization not adequately captured. Another characteristic which affects bias and overall generalizability of the research is the heavy emphasis on OEF/OIF/OND veterans. Although this is a growing population, the large majority of veterans are from other war eras and peacetime. Post 9/11 Gulf War era veterans account for only approximately 23% of the veteran population (National Center for Veterans Analysis and Statistics, 2015). Veterans from other eras might face different mental healthcare barriers.

One of the strengths of this review is the breadth. This breadth allowed us to consider research findings in context with studies of related but different constructs. This approach allows analysis of studies with widely different predictive models and research at both the system level (e.g., population-based outreach to at risk groups, primary care behavioral health programming) and individual level (e.g., targeted interventions for individual barriers). The breadth also allowed for inclusion of a variety of methodologies (e.g., randomized controlled trials, administrative chart reviews, and prospective surveys) which strengthens implications by being inclusive of articles with different strengths and weaknesses.

Despite its strengths, some limitations of this review warrant discussion. First, since this review focused on quantitative studies, all barriers examined were defined by researchers a priori which might have limited the range of identified barriers. Second, this review did not attempt a quantitative synthesis, which limits the ability to draw strong conclusions about the sizes of effects. The review provides a structured qualitative review of the literature due to large differences in predictors and methodological designs evaluated. These differences made direct comparisons problematic because statistical analyses (including appropriate effect size metrics) are vastly different for randomized controlled trials and observational designs as well as dichotomous and continuous outcomes. Because the different effect size metrics used were not directly comparable, we focused on significance tests since this was directly comparable between studies but this does limit our findings since significance does not equate to relevance. Further, implications of findings from these designs is entirely different causing direct comparisons of findings to be misleading. Also, our design mixed findings across studies with bi-variate and multivariate models which may have different implications and patterns. Further research should compare studies with these different models more closely to better understand how these constructs together and individually affect treatment utilization patterns. Another limitation is that we were not always able to obtain a detailed understanding of the impact of predictors and correlates because some studies were highly specific (e.g., focused specifically on initiation or retention, or specifically on psychotherapy or medication management) and others combined constructs within outcome measures.

Conclusions

The finding that veterans with the most severe symptoms are utilizing mental healthcare is encouraging, but ongoing efforts to increase mental healthcare initiation and retention are warranted. Research suggests that veterans identifying as racial and ethnic minorities and veterans with mild to moderate symptoms of PTSD might be at risk of low utilization, thus efforts to increase utilization targeting these groups could be warranted. Also, it is encouraging that incorporation of mental health services into primary care settings is one of the predictors with the most consistent research support for increasing both mental healthcare initiation and retention; this predictor is addressable through health services programming. Additional research should focus on addressing potential sources of bias including the heavy emphasis on VA healthcare enrollees and VA healthcare outcomes and pursue domains which have been understudied but show promise including the impact of mental health beliefs on utilization.

Clinical Impact Statement:

This review suggests efforts to increase mental healthcare in Veterans with PTSD should focus on Veterans identifying as racial/ethnic minorities and Veterans with mild/moderate symptoms. Additionally, research suggests mental health programming in primary care settings increases mental healthcare utilization for Veterans with PTSD.