Impact of Treatment Beliefs and Social Network Encouragement on Initiation of Care by VA Service Users With PTSD

Posttraumatic stress disorder (PTSD) is one of the most common mental health problems among veterans, and approximately 8% of the five million veterans who use U.S. Department of Veterans Affairs (VA) health care have a diagnosis of PTSD (1). The VA spends substantial resources providing mental health care to veterans with PTSD and has greatly expanded its mental health staffing to meet increasing demands for services (1,2). Despite these efforts, VA health care users with PTSD often fail to receive timely treatment (3,4). Untreated PTSD can become chronic, with lifelong deleterious consequences, including hospitalization, unemployment, poverty, suicide, and poor health (5–8). Given the tremendous economic and human costs of chronic PTSD, facilitating mental health treatment engagement is a VA priority.

Numerous factors have been found to be associated with mental health treatment initiation, including age, illness severity, travel distance, and whether the diagnosis was made by a mental health specialist (3,4,9,10). As care becomes more patient centered, there has been greater interest in the role of treatment-related beliefs and attitudes (10–17). In cross-sectional and retrospective studies, negative beliefs about treatment are prevalent among individuals with mental illnesses who do not get mental health treatment (13,14,18–21). The relevance of this is unclear, however, because even though negative treatment-related beliefs may deter treatment seeking, many who initiate mental health treatment often hold comparably negative beliefs (16,21,22). Moreover, beliefs can change as a result of treatment, so that a greater prevalence of negative attitudes among those who never initiate treatment may be a consequence rather than a cause of noninitiation (21,23). If treatment-related beliefs were found to prospectively affect the odds of treatment initiation, however, these beliefs may be useful targets for interventions aimed at engaging veterans in mental health care.

Beliefs about mental illness and mental health care develop within social and cultural contexts (24,25). In a qualitative study of treatment seeking for PTSD (22), veterans reported that people in their social networks often facilitated or deterred treatment seeking by validating or invalidating their beliefs about their symptoms and potential treatments and by providing instrumental support and additional motivations to seek help (22,26). Although the influences of a person’s social network on participation in mental health care are complex (27–29), we wondered whether encouragement to seek mental health treatment by significant people in veterans' social networks would affect the odds of treatment initiation. For those who believe that they have a mental health problem, encouragement from significant people to seek treatment would validate their beliefs that the symptoms they experience reflect a problem requiring treatment—and for those who are uncertain, such encouragement would provide evidence that the sensations they are experiencing are in fact symptoms of an illness (26,30,31). Although social network influences can facilitate mental health care for those with other mental health conditions (31), it is not known whether people in veterans' social networks can facilitate treatment initiation for PTSD. Because VA service users with PTSD often have more limited social networks than those without PTSD (32) and may perceive lower levels of support from them (32,33), facilitating treatment engagement may be the most effective help family members and friends can provide.

To identify targets for future treatment engagement initiatives, we investigated whether treatment-related beliefs predict future receipt of mental health care. We hypothesized that VA service users who believe that they needed help, who have positive beliefs about mental health treatment, and who believe that they can follow providers’ recommendations would be more likely to initiate treatment, whereas those whose mental illness negatively affects their self-concept would be less likely (13,18,20). Our second question was whether encouragement by people in VA service users’ social networks facilitates initiation of mental health care. Although our study could not address the complex interplay of treatment-related beliefs and social network influences, we examined the contribution of social encouragement after controlling for treatment-related beliefs. We examined these research questions as part of a large prospective study of a national sample of VA service users recently diagnosed as having PTSD.

Methods

Overview

A national sample of VA service users recently diagnosed as having PTSD (defined below) were surveyed immediately after they had received a PTSD diagnosis from a VA clinician. Data from surveys and VA administrative databases were used to identify individual- and facility-level determinants of VA mental health treatment initiation during a six month follow-up period. The study was approved by the Minneapolis VA Institutional Review Board.

Participants

Included were VA service users who received a diagnosis of PTSD (ICD-9 code 309.81) during an outpatient visit at any VA facility from June 2008 to July 2009. Those who in the prior year had received antidepressants or antipsychotics or who had any mental health appointments except for appointments related to chemical dependency (34) were excluded. Also excluded were those with a moderate to severe cognitive or schizophrenia spectrum disorder or with no available mailing address. Women and persons from racial-ethnic minority groups were oversampled. We sampled approximately 30% of eligible women, 37% of Hispanic men, 40% of men of non–African-American nonwhite race, 9% of African-American men, 4% of white men, and 18% of men of unknown race.

Of the 13,974 VA service users with PTSD who were selected by our sampling frame and met inclusion criteria, 979 were eliminated because they were no longer eligible (N=16) or deceased (N=11) or had no valid address (N=952). Baseline surveys were sent to the remaining 12,952. Of these, 8,492 surveys were returned (response rate of 66%). Because appointment and pharmacy information lag in administrative databases, some who were surveyed were no longer eligible on the basis of the updated data, leaving 7,645 eligible service users who completed baseline surveys.

Data sources and procedures

Administrative databases.

The Veterans Integrated Service Network (VISN) and facility in which the diagnostic appointment occurred were abstracted from the National Patient Care Database (NPCD), as were demographic and diagnostic data and information about VA disability status (whether the disability was service connected) and VA health care utilization. Data on outpatient prescriptions were abstracted from the Decision Support System National Pharmacy Extract database. Drive time to the nearest VA facility was determined by using the Planning Systems Support Group zip code database. Data on outpatient encounters are uploaded daily to the NPCD from each VA facility. The NPCD creates SAS files of the encounter data, which are updated every two weeks. We identified new cases of PTSD every two weeks in order to survey individuals immediately after this defining encounter. Those whose administrative records were not complete within two weeks of the encounter were included in the target population but not surveyed.

Surveys.

Surveys were used to assess PTSD severity, quality of life related to mental health, anticipated access barriers, treatment-related beliefs, and whether VA service users received encouragement to seek mental health treatment (hereafter called social encouragement). We surveyed using a modified Dillman approach (35). First, an explanatory introductory letter was mailed to eligible VA service users within two weeks of a PTSD diagnosis. This was followed two days later by a packet containing a cover letter with informed consent information, the survey, and a $10 cash payment. Ten days later, nonrespondents were sent a second cover letter and survey. Nonrespondents to the second packet were sent a third packet via Federal Express. Surveys were sent between June 2008 and August 2009.

Study outcomes

We used three dichotomous outcome measures of treatment initiation in the six months after receipt of the PTSD diagnosis: receipt of any guideline-recommended antidepressant, receipt of any psychotherapy, and receipt of either antidepressants or psychotherapy.

Pharmacologic treatment initiation.

Records of prescriptions, including drug class and release date (fill date), were collected. Pharmacotherapy initiation was defined as the release of a new prescription for an antidepressant consistent with clinical practice guidelines to treat PTSD—that is, selective serotonin reuptake inhibitors or serotonin-norepinephrine reuptake inhibitors (36) in the six-month period after the index diagnostic appointment.

Psychotherapy.

Psychotherapy sessions were identified in administrative data by therapy-related procedure codes (CPT) for which the provider was a mental health specialist. Initiation of psychotherapy was defined as at least one psychotherapy visit in the six-month follow-up period, excluding the initial diagnostic appointment.

Explanatory variables

Variables assessed as predictors included demographic characteristics, measures of treatment need, access factors, treatment-related beliefs, and social encouragement. Demographic variables included age, gender, and race and ethnicity as defined by the NPCD.

Treatment need.

PTSD symptom severity was assessed by the PTSD Checklist–Military version (PCL-M) (37,38) (Cronbach’s α=.94). Mental health quality of life, an indicator of mental health functioning, was assessed by the mental health component score (MCS) of the Veterans 12-Item Short Form Health Survey (39).

Access.

Access factors included VA disability status, drive time from the veteran’s residence to the nearest VA facility, clinic in which PTSD was diagnosed, and what (if any) access barriers were anticipated by VA service users should they seek mental health treatment. VA disability status, drive time, and clinic type were abstracted from VA databases. Clinics where the recent PTSD diagnosis was made were classified as follows: PTSD specialty clinic, general mental health clinic, primary care clinic, specialty medicine clinic, urgent care setting, or ancillary clinic (for example, occupational therapy). Anticipated access barriers were assessed by a survey checklist: clinic appointment times, cost, distance from the clinic, reliability of transportation, need to take care of dependents, knowledge about how to get treatment, or no access problems or barriers. VA disability status was considered to be an access factor because it increases eligibility for VA services.

Beliefs.

Treatment-related beliefs were those identified as likely contributors to treatment initiation. Perceived need for mental health care, defined according to previous studies (40,41), was assessed by patients’ agreement or disagreement with the statement, “At this time I feel I need help to deal with emotional problems, PTSD and/or stress in my life,” which was scored dichotomously. Self-efficacy to engage in mental health treatment (42) was evaluated by a 4-point Likert agreement-disagreement scale for the statement, “It is (or would be) easy for me to follow doctors' recommendations to treat PTSD symptoms or similar emotional problems.” Negative illness identity (13) was assessed by items from the Internalized Stigma of Mental Illness (43) (Cronbach’s α=.87), a scale developed with a VA population, with reasonable retest reliability (r=.68) and construct validity (for example, “I am disappointed in myself because I have PTSD”). Beliefs about the use of psychotherapy to address emotional problems were assessed with items from the Beliefs About Psychotherapy Scale (44) (Cronbach’s α=.72), which has good concurrent validity and sensitivity to cultural differences. Because avoidance is a core symptom of PTSD, we added the item, “I would be able to talk about what bothers me in therapy.” Beliefs about antidepressants were assessed by five items related to antidepressants from the Patient Attitudes Toward and Ratings of Care for Depression Scale adapted to PTSD (45) (Cronbach’s α=.77)—for example, “Antidepressants are effective in treating emotional problems or PTSD.”

Social encouragement.

Social encouragement to seek treatment was assessed by the survey item, “In the past year, have people in your life encouraged you to get treatment for PTSD or other emotional problems?” Responses were made by checklist: no one, spouse or significant other, other family members, other veterans, friends, medical providers, and employers. For this analysis, we used the following four groupings: family members (spouses or other family), friends or other veterans, family members and friends or other veterans, and no family members and no friends or other veterans.

Data analyses

We posited a superpopulation model in which individual characteristics, perceptions, beliefs, and social factors, as well as characteristics of the health care organization, play causal roles in the receipt of treatment. A hierarchical modeling approach to the analysis of complex, clustered data fits well in model-based approaches (46). In contrast to design-based approaches to finite population sampling, this approach does not view treatment outcomes as fixed. We based our inference about parameters in the superpopulation model by using the conditional distribution of the outcomes for the observed sample given the predictive covariates.

Among respondents, we used Markov chain Monte Carlo multiple imputation to impute values of missing dichotomous, ordinal, and interval survey items. For the remaining categorical variables, we constructed multinomial logistic regression models, using all other survey items as predictors, for imputing values of any missing survey items. Ten different sets of imputed values were constructed for each missing survey item.

Three hierarchical regression models were constructed for each of the treatment outcomes by using VISN and facility within VISN as random effects. Additional variables were entered as fixed effects into the models in blocks. The first block of variables included the demographic (age, gender, race, and ethnicity), need (PCL-M score and MCS), and access (drive time, service-connected disability status, and anticipated barriers) factors. To this base model we added a second block of belief measures. We then added the social encouragement variable as a third block to the beliefs model. These models were fit to the imputed data sets, and results were aggregated by using standard methods for multiple imputation.

To determine whether nonresponse bias may have affected our models, we examined whether the associations between the outcomes and the set of variables in the model from administrative databases varied among survey responders and nonresponders. We used models of the outcome measures similar to those described above, adding interactions between response status and each of the administrative variables. With only one exception, we did not see significant differences between responders and nonresponders. The one exception was a difference in associations among those whose initial encounter was in an emergency department or ancillary clinic. Because such encounters represent a small proportion of the population, we assume that the association between outcomes and survey-based predictors was the same among responders and nonresponders as it was for the administrative measures.

Results

Sample characteristics are presented in Table 1. Tables 2, 3, and 4 present multivariate models of receipt of any treatment (either pharmacotherapy or psychotherapy), receipt of psychotherapy, and receipt of pharmacotherapy, respectively. [A table presenting mean scores on the belief scale items is included in an online data supplement to this article.]

Although the focus of this study was on social and belief factors, we also note that for all outcomes, older VA service users, those who had a service-connected disability, and those who received a diagnosis in a non–mental health clinic were less likely to receive treatment, and those who were seen in PTSD specialty clinics were preferentially treated with psychotherapy. Although longer drive times decreased the odds of receiving any treatment, anticipated access barriers decreased only the odds of initiating psychotherapy but not pharmacotherapy.

VA service users who perceived a need for treatment were more likely to initiate care (odds ratio [OR]=3.75). Positive beliefs about psychotherapy (OR=1.07) or about antidepressants (OR=1.18) were each associated with increased odds of receiving the respective treatment. VA service users who believed it would be easy for them to follow mental health providers' recommendations to treat PTSD were more likely to receive some treatment. Social encouragement to get mental health care increased the odds of treatment receipt, even after the analysis controlled for beliefs, particularly if encouragement was received both from family members and from friends or other veterans (OR=1.85). In contrast, negative illness identity had little effect on treatment receipt, and to the extent that it did, it increased the odds of receipt.

Discussion

In a prospective study, we found that having treatment-related beliefs that were positive increased the odds of treatment initiation in the six-month period after receipt of a new diagnosis of PTSD, even after we controlled for variation in treatment need, access, and demographic characteristics. Our specific hypotheses regarding treatment-related beliefs were largely substantiated, and our results were consistent with those of others (14,40).

As in previous studies (40), VA service users who perceived a need for treatment were much more likely than those who did not to get mental health care. Many who suffer from mental health problems do not perceive a need for help (18,22), and others grapple with uncertainty about whether their distress is severe enough to warrant it (26). Therefore, helping VA service users identify mental health problems when they arise may be a key factor in successfully engaging them in treatment before they develop any of the deleterious consequences associated with untreated mental illness.

Our hypothesis about whether social encouragement to get mental health care would facilitate treatment initiation was supported: those who received encouragement to seek treatment were more likely to receive mental health care than those who were not encouraged. Despite having more limited social networks (32), VA service users with PTSD are significantly influenced by the social networks that they do have. Because the influence of treatment-related beliefs on treatment initiation was diminished only slightly when social encouragement was included in the models, both may be useful targets for interventions designed to engage VA service users in PTSD treatment.

In contrast to expectations, negative illness identity functioned as a facilitator of treatment initiation. Perhaps in this context, negative illness identity reflected the negative self-appraisals associated with PTSD, now included in DSM-5 under criterion D (47,48) and not captured by the PCL-M.

The study had several limitations. First, we do not know for certain that veterans given a diagnosis of PTSD by VA clinicians actually had the disorder. The extent to which misdiagnoses occur or how they may affect mental health treatment initiation is unknown; however, given the high PCL-M mean score for the sample and the fact that 95% of the sample had PCL-M scores above cutoffs associated with high specificity (49,50), it is likely that most of the veterans in the sample truly had PTSD.

Another limitation is that we had no information about treatment veterans may have received at Vet Centers or from non-VA sources. Evidence suggests that veterans with PTSD view the VA as having expertise in the treatment of PTSD (22,51), but results may have differed if outside treatment sources were known. Although we attempted to adjust for item nonresponse bias by using multiple imputation and to assess, via modeling, the possibility of survey nonresponse bias, we cannot be sure that these adjustments eliminated bias. Further, self-report measures may not have fully captured the study domains of interest. Finally, these findings may not apply to veterans who do not use VA services.

Conclusions

Whether VA service users initiate mental health care after being diagnosed as having PTSD depends not only on symptom severity and treatment accessibility but also on facilitation by those in their social networks, whether they perceive a need for treatment, how they view mental health treatments for PTSD, and how they view their ability to follow treatment recommendations. Involving family members and others in veterans' social networks in encouraging them to initiate mental health care may be an effective treatment marketing strategy. Helping veterans identify when their symptoms of distress reflect a problem that needs treatment and providing them with accurate information about the effectiveness of the newer PTSD treatments may further improve treatment engagement efforts.