Impact of Attitudes and Rurality on Veterans’ Use of Veterans Health Administration Mental Health Services

Abstract

Objective:

Veterans, especially those residing in rural areas, continue to underutilize mental health care. This longitudinal study assessed attitudes relevant to seeking mental health care services from the Veterans Health Administration (VHA) over 12 months, adjusting for residence.

Methods:

A questionnaire addressing attitudes, sociodemographic factors, residence, place identity, perceived health status and needs, and structural barriers was administered by telephone to 752 veterans with previous VHA service use. Service use data were obtained from a VHA database.

Results:

In adjusted models, four attitudes were significantly associated with underuse of VHA mental health care (no use vs. any use; no use vs. nonsustained use vs. sustained use). Higher levels of mistrust of others (adjusted odds ratio [AOR]=1.06, p=0.046), emotional stoicism (AOR=1.08, p=0.003), belief in the self-resolving nature of mental health problems (AOR=1.91, p=0.015), and belief in the efficacy of religious counseling for such problems (AOR=1.09, p=0.022) were associated with no subsequent service use versus any use. Place identity (suburban), older age, and greater need were associated with greater odds of VHA use. For the comparison of no use versus sustained use, women had lower odds of no use (AOR=0.49, p<0.001); similarly,women had lower odds of nonsustained use versus sustained use (AOR=0.45, p<0.001).

Conclusions:

The association of potentially modifiable attitudes with underuse of VHA mental health services suggests that attitudes offer useful targets for efforts to increase mental health care use. That these attitudes were influential regardless of residence suggests that programs addressing attitudinal barriers can be broadly targeted.

The Veterans Health Administration (VHA) of the U.S. Department of Veterans Affairs (VA) provides health care to >9 million veterans, nearly a third of whom (2.8 million) live in rural areas (1). Veterans, including rural-residing veterans, have underused VHA mental health care (2, 3).

In recent decades, the VHA has improved access to mental health care by creating a network of community-based outpatient clinics (CBOCs) (4), investing in telehealth (5–7), deploying mobile clinics (8), and paying for community-based services (9, 10). An emerging literature suggests that these efforts are also reducing the gap in mental health care utilization between rural and urban veterans (11, 12). However, findings have been inconsistent, especially when non–VHA-enrolled veterans are considered (13), and mental health care utilization remains suboptimal. The residual gap suggests that the VHA will also need to target influences other than service accessibility and availability. We undertook this research to explore the utility of VHA’s targeting individual attitudes and beliefs, overall and for veterans living in rural regions.

Both the theoretical (14) and the empirical literatures (15–18) highlight the influence of attitudes and beliefs on mental health care use among veterans, especially among veterans in rural areas. Several studies have suggested that attitudes, beliefs, and behavioral norms influence mental health care utilization more strongly than do structural barriers (19–22). Despite the high diversity within both rural and military populations (23, 24), several attitudes and beliefs relevant to mental health care utilization are frequently reported for each group, including an emphasis on independence, self-reliance, and stoicism and concerns about stigma (15, 25–35). Rural residence has also been associated with a distrust of “outsiders” (25, 26, 36) and, in some cases, trust in the efficacy of religious counseling for mental and emotional problems (36–38). Negative beliefs about mental health care or mental health care providers have also been found to be associated with reduced treatment seeking in both populations (15, 32, 33, 39–41).

The literature on associations between attitudes and beliefs and mental health care utilization among veterans is relatively limited (15), and the subset of studies exploring attitudes, mental health service use, and rurality is even smaller. In addition, much of this research is cross-sectional or retrospective and looks at the association of attitudes with past service use or hypothetical willingness to use services in the future, thus leaving the direction of association unclear (15). This article presents findings from a longitudinal study designed to assess the association of attitudes with subsequent use of VHA mental health care by veterans with a history of VHA use. Participants completed a telephone interview asking about their attitudes toward mental health care and mental health treatment seeking. Analyses combined interview data with administrative data on VHA mental health services use in the 12 months after interview completion.

METHODS

Eligibility and Sampling

Potential participants were identified from the VHA national administrative database, the Corporate Data Warehouse (CDW). To be eligible, veterans had to be 18–70 years old; to be living in one of four VHA administrative regions (Veterans Integrated Service Networks [VISNs])—VISN 1 (New England), VISN 16 (south central states), VISN 19 (Rocky Mountain states), and VISN 23 (north central states); and to have had a positive screen for depression or posttraumatic stress disorder recorded in their VHA medical records between October 1, 2009, and September 30, 2012. We excluded veterans with a diagnosis of dementia, schizophrenia, schizoaffective disorder, or psychosis not otherwise specified, because these conditions may affect memory and insight. Sampling was stratified by VISN, distance between a veteran’s residence and the nearest VA medical center (VAMC) (≤15 miles, 16–49 miles, and ≥50 miles), and gender. We oversampled female veterans to ensure that 25% of participants would be women.

Recruitment and Data Collection

Using CDW contact data, we mailed opt-out recruitment packets to potential participants (42). Two weeks later, project personnel called those who had not opted out to discuss the study in detail, assess interest, and confirm eligibility. We obtained verbal informed consent and HIPAA authorization from interested, eligible veterans. After participants consented, trained interviewers administered a structured questionnaire, using computer-assisted interviewing technology. Questionnaires were administered between July 2014 and April 2016. The Central Arkansas Veterans Healthcare System institutional review board approved the study procedures.

Data Sources and Measures

Data on the use of VHA outpatient mental health care in the 12 months after the survey interview were obtained from the CDW. Our primary outcome measure dichotomized mental health care use as any or none. We also divided use into sustained and nonsustained, where sustained use was defined as at least four outpatient mental health visits on separate days to the same VAMC (including its affiliated CBOCs) during the follow-up year. The four-visit criterion was based on the frequency with which appointments were commonly scheduled for stable patients. The questionnaire addressed sociodemographic data, perceived health status and need for mental health care, attitudes relevant to mental health care use, structural barriers, residence, and perceived rurality. We used the Veterans RAND 12-item Health Survey Mental Component Summary score (VR-12 MCS) as an indicator of need for mental health care (43).

The choice of attitudes to assess was based on the literature and findings from qualitative interviews with rural veterans and providers conducted earlier in this study (37). Whenever possible, attitudes were assessed with existing scales or subscales, specifically the Endorsed and Anticipated Stigma Inventory (beliefs about mental health treatment subscale) (44), Network Orientation Scale (independence and mistrust subscales) (45), Pain Attitudes Questionnaire (46), and Perceived Stigma and Barriers to Care for Psychological Problems (PSBCPP) (stigma [self-stigma of mental health treatment seeking] subscale) (47) (Table 1). We based our four-category classification of relative trust in the VHA for mental health care on dichotomized responses to two questions developed for the study: “If you had an emotional or mental health problem that disrupted your daily life, how likely is it that the people who are important May you would encourage you to seek mental health care at the VA” and, separately, “from a non-VHA provider?” To assess perceived structural barriers to mental health care use (e.g., cost and transportation), we used the PSBCPP barriers subscale (47), with the addition of one item developed for the survey (“mental health services are too far away”). The 13 attitude indicators (Table 1) and the barriers index were scored so that higher scores indicated more of the attitude or more barriers.

TABLE 1.

Composition and scoring of attitudinal measures used in the study^a

Attitude	Measure	No. of items	Item scoring	Score range
Included in multivariable models
Efficacy of mental health care	EASI subscale	8	1–5, strongly disagree to strongly agree	8–40
Efficacy of religious counseling	Study specific	3	1–5, strongly disagree to strongly agree	3–15
Mistrust of others	NOS subscale, modified	7	1–5, strongly disagree to strongly agree	7–35
Self-resolving nature of emotional problems	Study specific	1	0–4, no one gets better without professional help to everyone does	0–4
Self-stigma about seeking mental health care	PSBCPP subscale, modified	6	1–5, strongly disagree to strongly agree	6–30
Stoicism, emotional	Study specific	6	1–5, strongly disagree to strongly agree	6–30
Stoicism, physical	Pain Attitude Questionnaire	6	1–5, strongly disagree to strongly agree	6–30
Trust in the VA for mental health care	Study specific	2	4 categories (trust or encourage use of both VA and non-VA providers, trust VA only, trust non-VA only, and trust or encourage use of neither)
Not included in multivariable models because of lack of bivariate significance
Independence	NOS subscale, modified	10	1–5, strongly disagree to strongly agree	10–50
Mistrust of providers	Study specific	2	1–5, strongly disagree to strongly agree	2–10
Mistrust of VA system	Study specific	2	3 categories
Public stigma about seeking health care	Study specific	3	1–4, not at all to very embarrassed	3–12
Willingness to rely on others	Study specific	3	Two categories

^aEASI, Endorsed and Anticipated Stigma Inventory; NOS, Network Orientation Scale; PSBCPP, Perceived Stigma and Barriers to Care for Psychological Problems; VA, U.S. Department of Veterans Affairs.

We constructed two indicators of rurality that tapped distinct constructs. The first, geographic distance from the participant’s residence to the nearest VAMC, was calculated as the distance between the centroids of the zip codes in which each was located. The second, place identity, was self-identification as a rural, suburban, or urban person, assessed through the question, “If you were talking about yourself to others, how would you describe yourself?” Data on place identity were collected as part of the telephone survey. Distance from the participant’s current residence to the nearest VAMC was determined with the VA Planning Systems Support Group zip code database.

Statistical Analysis

Logistic regression was used to test the hypothesis that attitude scores would be associated with VHA mental health care use after adjustment for sociodemographic factors, residence, place identity, need, and structural barriers. Binomial logistic regression was used when the outcome was no or any mental health service use; multinomial logistic regression was used when it was no use, nonsustained use, and sustained use. Because we used stratified probability sampling in recruiting survey participants, sampling weights were used in all analyses. We used multiple imputation (SAS PROC MIANALYZE) with 12 rounds of imputation to address missing survey data. Using a model-building structure informed by Hosmer and Lemeshow’s purposeful selection algorithm (48), we entered variables into regression models in three sequential blocks. We first entered predictors of primary interest (attitudes), followed by confounders of primary interest (rurality indicators), followed by other potential confounders (demographic factors, need, and structural barriers). Because distance and place identity tap distinct aspects of rurality (geography vs. culture) and did not categorize participants identically (data available from the first author), we included both indicators in the final model. To avoid problems of multicollinearity due to the recursive relationship between attitudes and service utilization (14), we did not include previous service use in the regression models. All analyses were performed with SAS, version 9.3.

RESULTS

Participants

We mailed opt-out invitation letters to 3,490 veterans, of whom 1,513 (43%) declined participation, 173 (5%) were found ineligible, and 1,043 (30%) could not be contacted. We interviewed 761 (22%) veterans who were eligible and gave verbal informed consent. Data from nine veterans who did not complete the entire interview were excluded from the analyses.

Table 2 presents the unweighted characteristics of the 752 veterans who completed interviews. Most interview participants were men and currently married, self-identified as non-Hispanic White, and had an education beyond high school. The mean age was 52.4 (range 24–70) years, and the mean distance to the nearest VAMC was 43.6 miles (range <1–226 miles). (A table comparing characteristics of veterans who participated, who refused, and who could not be contacted is included in an online supplement to this article.)

TABLE 2.

Characteristics of 752 veterans who participated in the telephone survey

Characteristic	N	%
Demographic
Female	188	25
Race-ethnicity
Hispanic	46	6
Non-Hispanic Black	86	11
Non-Hispanic White	521	69
Other	99	13
Marital status
Currently married	386	51
Divorced	220	29
Never married	85	11
Other	61	8
Education
High school or less	127	17
More than high school	446	59
Bachelor’s degree or more	179	24
Age (M±SD years)	52.4±12.5
Residence
Regiona
VISN 1	182	24
VISN 16	194	26
VISN 19	192	26
VISN 23	184	24
Distance to VAMC (miles)b
M±SD	43.6±43.6
Medianc	29.9

^aVISN, Veterans Integrated Service Network. VISN 1, New England; VISN 16, south central states; VISN 19, Rocky Mountain states; VISN 23, north central states.

^bVAMC, Department of Veterans Affairs medical center.

^cInterquartile range=10.0–63.5.

Attitude Variables Included in Logistic Regression Models

We regressed each of the service use variables (no use or any use, and no use or nonsustained use or sustained use) separately on each of the attitude variables. Eight attitude variables that were statistically significant at p<0.10 in at least one of the bivariate analyses were included in all multivariable logistic models.

No Use Versus Any Use

When dichotomous use (no use vs. any use [the reference category]) was regressed on the set of the eight attitude variables, four were statistically significant (Table 3). Higher scores on mistrust of others, emotional stoicism, belief in the self-resolving nature of mental health problems, and belief in the efficacy of religious counseling for mental health problems were associated with increased odds of no use of VHA mental health care. These significant associations remained after the models were adjusted for each of the rurality indicators separately (Table 4). In these models, increasing distance was independently associated with increased odds of no service use, whereas suburban place identity was associated with decreased odds of no service use.

TABLE 3.

Associations of predictor variables with no postsurvey use (versus any use) of Veterans Health Administration (VHA) mental health care among 752 veterans

	Attitudes only				Fully adjusted modela
Variable	OR	95% CI	Z	p	OR	95% CI	Z	p
Attitude indicator
Mistrust of othersb	1.07	1.02–1.12	2.61	.009	1.06	1.00–1.11	2.00	.046
Self-stigmab	.99	.95–1.03	−.57	.566	.99	.95–1.04	−.29	.774
Stoicism, physicalb	1.00	.96–1.04	−.23	.815	.99	.95–1.04	−.21	.835
Stoicism, emotionalb	1.05	1.00–1.10	2.11	.035	1.08	1.03–1.14	2.93	.003
Trust in the VHA for mental health care (reference: trust in neither)
VHA only	.71	.32–1.60	−.83	.407	.62	.27–1.42	−1.13	.258
Both	1.10	.51–2.36	.24	.814	1.11	.51–2.45	.27	.789
Non-VHA only	.97	.41–2.32	−.07	.944	1.10	.45–2.72	.22	.829
Belief in self-resolving nature of emotional problemsb	1.97	1.21–3.20	2.72	.007	1.91	1.13–3.21	2.43	.015
Belief in efficacy of mental health careb	.97	.94–1.01	−1.47	.141	.98	.95–1.02	−.88	.380
Belief in efficacy of religious counselingb	1.07	1.01–1.14	2.23	.026	1.09	1.01–1.16	2.31	.022
Rurality indicator
Distance to VAMC (miles)c					1.003	1.00–1.01	1.70	.089
Place identity (reference: urban)
Rural					.92	.61–1.38	−.42	.677
Suburban					.61	.39–.96	−2.15	.032
Control variable
Age in years					1.03	1.01–1.04	3.80	<.001
Female (reference: male)					.74	.53–1.04	−1.75	.079
Race-ethnicity (reference: non-Hispanic White)
Hispanic					1.18	.59–2.35	.47	.641
Non-Hispanic Black					.79	.42–1.48	−.74	.462
Other					1.01	.62–1.62	.03	.978
VISN (reference: VISN 16)d
1					.75	.43–1.30	−1.03	.304
19					.64	.37–1.11	−1.58	.114
23					.97	.55–1.70	−.12	.908
Need (VR-12 MCS)b,e					1.05	1.02–1.07	3.68	<.001
Barriers index (structural and logistic)b					1.04	.81–1.34	.31	.760

^aAdjusted for rurality (distance and place identity), personal characteristics, need, and structural barriers.

^bScale or subscale score. Odds ratios indicate the change in the odds of the outcome associated with a one-unit increase in the corresponding scale or subscale score.

^cVAMC, Department of Veterans Affairs medical center.

^dVISN, Veterans Integrated Service Network. VISN 1, New England; VISN 16, south central states; VISN 19, Rocky Mountain states; and VISN 23, north central states.

^eVeterans RAND 12-item Health Survey Mental Component Summary score.

TABLE 4.

Associations of predictor variables with no postsurvey use (versus any use) of Veterans Health Administration (VHA) mental health care among 752 veterans, adjusted for rurality indicator^a

	Rurality indicator
	Distance to VAMCb				Place identity
Variable	OR	95% CI	Z	p	OR	95% CI	Z	p
Attitude indicator
Mistrust of othersc	1.07	1.02–1.12	2.58	.006	1.07	1.02–1.23	2.66	.008
Self-stigmac	.99	.95–1.03	−.65	.517	.99	.96–1.03	−.37	.715
Stoicism, physicalc	1.00	.96–1.04	−.20	.839	.99	.95–1.03	−.59	.553
Stoicism, emotionalc	1.05	1.00–1.10	2.15	.031	1.05	1.00–1.10	2.05	.040
Trust in the VHA for mental health care (reference: trust in neither)
VHA only	.66	.29–1.50	−1.00	.319	.81	.32–1.59	−.84	.402
Both	1.03	.48–2.25	.09	.927	1.16	.54–2.48	.38	.704
Non-VHA only	.91	.38–2.18	−.22	.826	.99	.41–2.36	−.03	.979
Belief in self-resolving nature of emotional problemsc	1.99	1.22–3.25	2.76	.006	2.01	1.23–3.28	2.78	.005
Belief in efficacy of mental health carec	.98	.94–1.01	−1.37	.171	.97	.94–1.01	−1.50	.134
Belief in efficacy of religious counselingc	1.08	1.02–1.15	2.45	.014	1.08	1.02–1.15	2.43	.015
Rurality indicator
Distance to VAMC (miles)b	1.004	1.00–1.01	2.09	.037
Place identity (reference: urban)
Rural					1.10	.76–1.60	.51	.613
Suburban					.57	.37-.87	−2.60	.009

^aTwo separate models; one adjusting for distance and one for place identity (reference: urban). Outcome reference group: any use (0);outcome comparison group, no use (1).

^bVAMC, Department of Veterans Affairs medical center.

^cScale or subscale score (see Table 1). Odds ratios indicate the change in the odds of the outcome associated with a one-unit increase in the corresponding scale or subscale score.

In the fully adjusted model (Table 3), which included the attitude variables, distance, place identity, age, gender, race-ethnicity, VISN, need, and structural barriers, the four attitude variables remained statistically significant. Not only did adjustment not change which attitudes were significant, it had little impact on the magnitude of effects. Suburban place identity also remained significant; veterans who self-identified as suburban had lower odds of no service use than veterans who self-identified as rural or urban. Among control variables, only age and need were significantly associated with service utilization; the odds of no service use increased as age increased and as VR-12 MCS scores increased (higher MCS scores indicate lower need).

Although gender did not reach statistical significance in the fully adjusted model (p=0.079), because of the priority given to women’s health by VHA, we stratified findings by gender to explore potential differences. The magnitude of the effect of mistrust of others was similar for men and women. However, the magnitude of the effect of emotional stoicism was greater for male veterans, and the magnitude of the effects for belief in the self-resolving nature of emotional problems and in the efficacy of religious counseling was greater among female veterans. Suburban place identity was also more strongly associated with service use among female veterans than among male veterans (see online supplement).

No Use, Nonsustained Use, and Sustained Use

The 752 participants were fairly evenly distributed among service use categories: sustained use (N=293, 39%), nonsustained use (N=232, 31%), and no use (N=227, 30%). Table 5 summarizes the results of regressing the trichotomous service use variable on the full model. Not surprisingly, the pattern for no use versus sustained use was nearly identical, in terms of the magnitude of odds ratios (but not in terms of statistical significance), to the pattern for no use versus any use. Two attitudes were significantly associated with the odds of nonsustained versus sustained use: the odds of nonsustained use decreased as self-stigma scores increased but increased with increasing belief in the self-resolving nature of emotional problems. Women had significantly lower odds of being nonusers and nonsustained users versus sustained users.

TABLE 5.

Associations of predictor variables with postsurvey use (no use, nonsustained use, or sustained use) of Veterans Health Administration (VHA) mental health care among 752 veterans

	No use vs. nonsustained use				Nonsustained use vs. sustained use				No use vs. sustained use
Variable	OR	95% CI	Z	p	OR	95% CI	Z	p	OR	95% CI	Z	p
Attitude indicator
Mistrust of othersa	1.05	.98–1.11	1.41	.159	1.02	.95–1.09	.54	.588	1.07	1.00–1.14	1.94	.052
Self-stigmaa	1.03	.98–1.08	1.10	.273	.94	.89–.99	−2.35	.019	.97	.92–1.01	−1.41	.158
Stoicism, physicala	1.02	.97–1.07	.75	.454	.95	.91–1.00	−1.78	.074	.97	.92–1.02	−1.08	.281
Stoicism, emotionala	1.08	1.01–1.14	2.39	.017	1.08	1.02–1.15	.14	.892	1.08	1.02–1.15	2.49	.013
Trust in the VHA for mental health care (reference: trust in neither)
VHA only	.64	.25–1.66	−.91	.362	.59	.22–1.62	−1.02	.308	.92	.33–2.55	−.16	.876
Both	1.38	.56–3.40	.70	.485	.65	.24–1.73	−.87	.384	.89	.34–2.32	−.24	.812
Non-VHA only	.97	.35–2.67	−.07	.948	1.39	.45–4.32	.57	.572	1.44	.45–4.55	.61	.540
Belief in self-resolving nature of emotional problemsa	1.71	.92–3.18	1.68	.092	2.16	1.11–4.24	2.25	.024	1.27	.59–2.71	.61	.542
Belief in efficacy of mental health carea	.99	.95–1.04	−.41	.679	.99	.94–1.04	−.55	.579	.98	.93–1.02	−.98	.326
Belief in efficacy of religious counselinga	1.12	1.03–1.21	2.88	.004	.93	.85–1.01	−1.68	.094	1.04	.96–1.14	.96	.336
Rurality indicator
Distance to VAMC (miles)b	1.00	1.00–1.01	1.35	.177	1.00	.99–1.01	.19	.852	1.00	.99–1.01	1.46	.143
Place identity (reference: urban)
Rural	.87	.54–1.41	−.55	.582	1.10	.64–1.87	.33	.738	.96	.58–1.58	−.17	.864
Suburban	.66	.39–1.10	−1.59	.111	.86	.50–1.49	−.53	.596	.57	.33–.96	−2.11	.035
Control variable
Age in years	1.04	1.02–1.05	4.23	<.001	.98	.97–1.00	−1.88	.060	1.02	1.00–1.04	2.03	.043
Female (reference: male)	1.08	.73–1.60	.39	.700	.45	.30–.69	−3.64	<.001	.49	.33–.74	−3.43	<.001
Race-ethnicity (reference: non-Hispanic White)
Hispanic	1.04	.48–2.28	.10	.919	1.32	.53–3.30	.60	.548	1.38	.57–3.32	.72	.474
Non-Hispanic Black	.86	.41–1.79	−.40	.686	.83	.38–1.81	−.47	.636	.71	.34–1.51	−.89	.375
Other	1.12	.63–1.98	.39	.697	.81	.44–1.49	−.67	.505	.91	.52–1.60	−.33	.745
VISN (reference: VISN 16)c
1	.95	.49–1.81	−.17	.868	.63	.31–1.31	−1.23	.219	.60	.31–1.18	−1.47	.141
19	.64	.34–1.21	−1.37	.169	1.01	.49–2.07	.03	.977	.65	.33–1.29	−1.24	.216
23	1.08	.56–2.07	.24	.813	.78	.37–1.63	−.66	.506	.84	.42–1.69	−.48	.629
Need (VR-12 MCS)a,d	1.04	1.01–1.07	2.78	.006	1.01	.98–1.04	.90	.367	1.05	1.02–1.08	3.62	<.001
Barriers index (structural and logistic)a	.91	.68–1.22	−.62	.534	1.28	.93–1.75	1.53	.126	1.16	.86–1.58	.99	.322

^aScale or subscale score. Odds ratios indicate the change in the odds of the outcome associated with a 1-unit increase in the corresponding scale or subscale score.

^bVAMC, Department of Veterans Affairs medical center.

^cVISN, Veterans Integrated Service Network. VISN 1, New England; VISN 16, south central states; VISN 19, Rocky Mountain states; VISN 23, north central states.

^dVeterans RAND 12-item Health Survey Mental Component Summary score.

DISCUSSION

In this longitudinal study, we found that a subset of attitudes and beliefs was associated with no use of VHA mental health care over 12 months by veterans who had used VHA for general medical or mental health care in the past. This was the case both when utilization was dichotomized (none or any) and when it was trichotomized (no use, nonsustained use, or sustained use). Associations for those attitudes remained statistically significant after adjustments for rurality, demographic factors, need, and structural barriers. The set of attitudes significantly associated with lack of VHA mental health service use was not unexpected. It does not seem surprising that veterans with high scores on mistrust of others, emotional stoicism, belief in the self-resolving nature of mental problems, or belief in the efficacy of religious counseling in resolving such problems would be disinclined to seek mental health care.

One attitudinal barrier frequently endorsed by veterans and active military personnel (33, 35, 37, 49, 50), concern about public stigma and self-stigma, was not associated with reduced mental health service use in our study. Our single significant finding regarding stigma was that veterans with higher self-stigma scores had reduced odds of nonsustained versus sustained service use. Stigma is addressed in most of the published research on attitudes and mental health care utilization. Although the results of these studies are mixed (16, 51), our findings are consistent with an expanding body of research indicating that stigma either is not associated with actual service use or is associated with increased use (16, 32, 52–56). These counterintuitive findings may arise because the severity of the mental health problem outweighs concerns about stigma (51, 52), a need for services increases concerns about stigma (16), or receipt of services increases awareness of the social impact of help seeking (52, 55). They could also arise from inconsistencies and methodological challenges in the measurement of stigma (52).

This study contributes to the existing literature in several ways. Whereas much of the literature on mental health care utilization and attitudes is retrospective, assessing the relationship of current attitudes to previous service use (12, 16, 32, 33, 35, 40, 41, 51, 57), our study clarified the temporal sequence of attitudes and service use by looking at the influence of current attitudes on service use over the 12 months after an interview. To our knowledge, it is the first study that has looked at the influence of both attitudes and rural residence on future use of VHA mental health care. None of the six longitudinal studies we found that assessed the impact of attitudes and beliefs on mental health care use among U.S. military veterans also assessed the impact of rural residence (53, 56, 58–61).

Our study expanded the set of attitudinal barriers commonly assessed and, to our knowledge, it is the first that has studied attitudes and use of mental health care among veterans and includes a measure of beliefs regarding the efficacy of religious counseling. Inclusion of this variable arose from our initial qualitative work as well as from the literature (36–38, 62, 63). Its association with higher odds of no service use warrants additional investigation, especially for female veterans.

Because we examined individual attitudes rather than a composite of negative beliefs and attitudes, this study also allowed for a more nuanced understanding of the influence of individual attitudes. The association we observed between increased service utilization and self-identification as a suburban person (i.e., place identity) is intriguing. Its significance in models that included distance as well as measures of attitudes, need, and structural barriers and its greater salience for women suggest a need for additional qualitative research to better understand other aspects of rural, urban, and suburban cultures that influence mental health service use (64–66).

Our findings should be considered in the context of study limitations. First, all participants had used VHA general medical or mental health services in the past, and most used at least some VHA mental health care in the 12 months after their telephone interview. Findings may not generalize to veterans using services outside the VHA system. Second, we had no data on participants’ use of mental health services outside the VHA during the follow-up year. Although we expect that the significant attitudes would influence VHA and non-VHA mental health care utilization similarly, additional longitudinal research incorporating both VHA and non-VHA data is warranted to test that hypothesis. Third, although the response rate is respectable for a telephone survey (67, 68), two-thirds of the eligible participants whom we could contact (1,513 of 2,274) declined to take part. We used statistical weighting to ensure that our sample conformed to the study population with respect to VISN, distance to the nearest VAMC, and gender, and we found no significant differences in race-ethnicity between participants and nonparticipants. Although we cannot rule out nonresponse bias arising from differences in other characteristics, including age, we are encouraged by the Pew Research Center’s findings of little nonresponse bias for the types of questions included in our survey (68). Fourth, several of the attitude indicators used have not been validated, although similar items have been used in comparable surveys.

Fifth, we recognize that attitudes are dynamic and that some attitudes about mental health care may both reflect previous treatment experiences and influence future service utilization. Because our analyses focused on the impact of attitudes on subsequent service use rather than on attitude development, to avoid overcontrolling, we did not include presurvey treatment experiences in our analytical models. Sixth, because the study was not adequately powered for the trichotomous outcome analyses or for the gender-stratified dichotomous outcome analyses, nonsignificant findings in these analyses may not be stable. These limitations are offset in part by the study’s strengths, including its longitudinal nature, use of probability sampling, use of objective administrative data for VHA mental health care utilization, and separate consideration of attitudinal and structural barriers.

CONCLUSIONS

Our finding that participants’ attitudes and beliefs were independently associated with VHA service use after analyses that controlled for service need and distance indicates that addressing attitudes is another potentially useful target for the VHA in its efforts to increase veterans’ utilization of needed mental health care. Although several attitudes frequently ascribed to many rural cultures were associated with patterns of VHA mental health care utilization, taking attitudes into account did not eliminate the association of utilization with distance and with place identity, a cultural marker of rurality. The finding that attitudes are influential regardless of residence suggests that VHA programs that address attitudinal barriers can be broadly targeted.

Not all attitudes are amenable to intervention by the VHA—e.g., general mistrust of others. However, our findings suggest additional avenues by which the VHA might work to reduce the rural-urban gap in mental health care use. For example, this study’s findings regarding the impact of belief in the efficacy of religious counseling support expansion of ongoing VA efforts to increase veterans’ engagement in mental health care through collaborative relationships between VHA mental health care providers and community-based religious providers (62, 69). Similarly, the role of emotional stoicism and the influence of belief in the self-resolving nature of mental problems as barriers to VHA mental health care utilization suggest that expansion of the suicide prevention educational and behavioral campaign, epitomized by the slogan “It takes the strength and courage of a warrior to ask for help,” could contribute to reductions in the rural-urban utilization gap (70).

HIGHLIGHTS.

• The Veterans Health Administration (VHA) wants to better understand the relationship of attitudinal barriers to mental health service utilization among its users, especially among VHA users residing in rural regions.

• Individual beliefs and attitudes predicted 12-month use of VHA mental health care by veterans with a history of VHA service use, even after adjustment for demographic factors, rurality, need, and structural barriers.

• Veterans who endorsed a belief in the efficacy of religious counseling for mental health problems were more likely than other veterans to use no VHA services—a finding that supports VHA’s efforts to increase collaboration with community-based religious providers.