To enhance the recognition of posttraumatic stress disorder (PTSD) among its patients, the Veterans Health Administration (VHA) mandates behavioral health screening for the disorder (
1,
2). Such screening generally occurs within the context of primary care via the Primary Care PTSD Screen (PC-PTSD) (
3). When screens are positive, timely follow-up diagnostic assessment and appropriate treatment for those diagnosed are critical elements of high-quality mental health services delivery. In particular, prompt diagnoses may foster recognition of service needs among those with the disorder and promote initiation and engagement in treatment (
4–
6). Nonetheless, few studies have examined factors associated with diagnosis and treatment among VHA patients who screen positive for PTSD.
Collaborative care services within the PC setting may be especially relevant for diagnosis and treatment initiation among patients who present with PTSD symptoms (
7); however, much of the prior research in this area has focused on depression and depression-related outcomes (
8,
9). Within VHA, the mandated, systemwide implementation of integrated care, commonly referred to as primary care–mental health integration (PC-MHI), began in fiscal year (FY) 2008. As previously described in greater detail, the PC-MHI program includes colocation of collaborative mental health providers in primary care and disorder-specific care management for individuals with psychiatric conditions (
10,
11). PC-MHI services may be delivered on the same day that initial need is detected and may include prescribing appropriate medications, delivering psychotherapy, and managing referrals to more specialized mental health clinics, including those treating PTSD.
The implementation of PC-MHI in VHA provides an opportunity to better understand the potential salutary impact of integrated care on mental health services delivery for patients with PTSD. For example, a prior facility-level evaluation found significantly greater increases in psychiatric diagnoses among facilities with documented PC-MHI activity compared with facilities without such activity (
12); however, a subsequent report found no evidence of differential rates of treatment initiation at specialty mental health clinics between facilities with and without PC-MHI (
13). Other single-site VHA analyses also lend support for the finding of a potentially positive role of PC-MHI services for patients with PTSD. For instance, Brawer and colleagues (
14) found increased completion of PTSD clinic consultations among patients referred by PC-MHI team members compared with those referred by primary care providers.
Despite these promising initial findings, to our knowledge no systemwide evaluation of the potential impact of PC-MHI on PTSD diagnosis and treatment initiation after screening has been conducted in VHA. Therefore, with a national sample of VHA patients who screened positive for PTSD this study compared rates of diagnosis and treatment initiation according to patients’ service setting on the screening day, namely, primary care only, PC-MHI, or specialty mental health clinics located outside of primary care. Using multivariable modeling, we estimated the association between setting of screening-day services and same-day PTSD diagnosis, as well as PTSD diagnosis that occurred within one year of screening. We also modeled the relationship between service setting on screening day and initiation of PTSD treatment, including psychotherapy, PTSD clinic visit, and antidepressant medication prescription receipt, within 12 weeks of screening. On the basis of prior studies, we hypothesized that receipt of same-day PC-MHI would be associated with increased odds of PTSD diagnosis and treatment initiation compared with primary care only.
Methods
Sample and Data
Study data were a 30% random sample of all VHA primary care patients in FY 2010. The sample included all patients who were in the 30% random sample, who had a positive PTSD screen recorded in FY 2010 (specifically, a score of ≥3 on the PC-PTSD), who received primary care services on the screening date, who did not have a recorded PTSD diagnosis or PC-MHI or specialty mental health services in the 12 months before the screening, and who were at least 18 years old, for a total of 21,427 veterans. PTSD screening assessments were from the VHA Corporate Data Warehouse (CDW) and were matched to patient administrative health records from VHA National Patient Care Database (NPCD) and pharmacy records from VHA Decision Support Services. The VHA Ann Arbor Healthcare System Institutional Review Board approved this study.
Measures
The primary outcomes of interest were PTSD diagnosis and PTSD treatment initiation. PTSD diagnosis was identified via an ICD-9-CM diagnosis code of 309.81 in NPCD records. PTSD treatment initiation outcomes included the following: receipt of psychotherapy, PTSD clinic visit, prescription for an antidepressant medication, and any treatment initiation (any combination of psychotherapy, PTSD clinic visit, or antidepressant prescription). Receipt of psychotherapy was identified by Current Procedural Terminology codes 90804–90815, 90845, 90847, 90853, and 90857 in any field of the outpatient or inpatient NPCD. PTSD clinic visit was assessed via clinic stop code in the NPCD. Antidepressant medications included citalopram, fluoxetine, paroxetine, sertraline, venlafaxine, escitalopram, fluvoxamine, and desvenlafaxine. Patients were assessed for PTSD diagnosis and treatment initiation outcomes on the screening day and ≤7 days, ≤12 weeks, ≤6 months, and ≤1 year after positive screen.
The focal predictor was setting of services on the date of PTSD screening, which was assessed with established identifiers in the outpatient encounters data set of the NPCD. Service setting on the screening date was categorized as follows: primary care only, same-day PC-MHI, and same-day specialty mental health care at a specialty mental health clinic.
The following covariates were also included in multivariable analyses: demographic characteristics (age group, sex, race, Hispanic ethnicity, and marital status), service-connected disability status of ≥50%, any prior psychiatric diagnosis other than PTSD, any prior VHA outpatient use, and positive PC-PTSD score (a score of 3 or 4) (
3). Covariates were assessed from the NPCD, with the exception of the PC-PTSD screen scores, which were assessed from CDW data.
Statistical Analysis
First, basic frequencies and percentages of demographic and clinical variables were calculated for the sample overall and by category of services received on the date of PTSD screening. Second, for each category of services, cumulative percentages of PTSD diagnoses and treatment initiation outcomes were computed according to the specified time intervals (≤1 day, ≤7 days, ≤12 weeks, ≤6 months, and ≤1 year of screening positive). Third, multivariable generalized estimating equations (GEE) logistic regression models were fit to estimate associations between category of screening day services and the odds of PTSD diagnosis on the screening day and ≤1 year of screening positive among the entire sample, and the odds of PTSD treatment initiation ≤12 weeks of screening positive among patients who also received a diagnosis on the date of their positive screen (
15). GEE modeling was used to take into account the clustered nature of the data, with patients nested within VHA facilities. We specified exchangeable working correlation structures and used robust variance estimators that produce valid estimates of standard errors even when the correlation structures are incorrectly specified. All GEE models also adjusted for patient demographic characteristics, service-connected disability status, prior psychiatric diagnosis, prior VHA outpatient use, and PC-PTSD score. Analyses were performed with SAS 9.2.
Results
Table 1 provides demographic and clinical information for the 21,427 patients in the sample overall with positive PTSD screens and by setting of services received on the PTSD screening date. As shown in the table, the three categories of services had significantly different distributions for all of the characteristics except race (all p<.05).
In
Table 2, the cumulative incidence of PTSD-related diagnoses and treatment initiation, which included psychotherapy, PTSD clinic visit, antidepressant prescription, and any of the three categories of treatment, are shown at specified time intervals over one year for each of the three categories of screening-day services. Overall, of the 21,427 patients, 10,809 (50.4%) received a diagnosis within one year of screening positive. Compared with patients who received only primary care services, a greater percentage of patients who received PC-MHI services on the date of their positive PTSD screen were also diagnosed as having PTSD on that same day (41.7% versus 24.8%). An estimated 47.5% of those who also received specialty mental health services on their PTSD screening date received a same-day PTSD diagnosis. Moreover, lower rates of PTSD diagnosis were observed at every specified time interval for the primary care–only group. At the end of one year, the cumulative incidence of PTSD diagnosis had increased in all groups; however, the lowest rate of PTSD diagnosis was observed for patients seen in standard primary care (47.9%). Similar patterns were also noted with respect to PTSD treatment initiation outcomes (
Table 2). For example, .5%, 40.0%, and 33.9% initiated psychotherapy on the same day as their PTSD screening among those who received primary care, PC-MHI care, and specialty mental health care, respectively.
Results from the multivariable GEE logistic regression models predicting same-day and one-year PTSD diagnoses are presented in
Table 3. As shown in the table, patients who received same-day PC-MHI services had significantly greater odds of receiving a PTSD diagnosis on the same day as their positive PTSD screen compared with patients who received solely primary care (adjusted odds ratio [AOR]=2.23). The AOR associated with same-day specialty mental health services was similar (AOR=2.56) and was not significantly different from the PC-MHI estimate (p>.05). Somewhat lower estimates were obtained from the model estimating one-year PTSD diagnosis (for same-day PC-MHI, AOR=1.67; for specialty mental health, AOR=2.03; the two estimates differed significantly, p<.05). Although not the primary interest, other findings with the models are noteworthy. With respect to the same-day diagnosis model, patients in the two older age groups, black patients, Hispanic patients, individuals who were not married, patients with a prior psychiatric diagnosis other than PTSD, and individuals who had used VHA outpatient services had lower odds of diagnosis than their respective comparison groups. Males, patients with service-connected disability of 50% or greater, and patients with a PC-PTSD score of 4 had greater odds of diagnosis compared with their respective comparison groups. The one-year model had similar findings, with the following exceptions: Hispanic ethnicity was not significantly associated with diagnosis; patients whose race was unknown and individuals who were never married had lower odds of PTSD diagnosis within one year of screening.
Table 4 shows the results from the GEE models estimating the association between location of services and treatment initiation among the group of patients who received a PTSD diagnosis on the same day as their positive screen (N=5,966). For each model, patients who received same-day PC-MHI had significantly greater odds of initiating PTSD treatment within 12 weeks of their diagnosis compared with those in primary care (AORs ranged from 1.64 to 4.12). Receipt of same-day specialty mental health care was also associated with greater odds of PTSD treatment initiation (AORs ranging from 2.05 to 4.45). No differences were detected between the AORs for same-day PC-MHI and specialty care for the treatment models, with the exception of the antidepressant model (p<.05). In addition, patients ages 45–64 had lower odds of receiving an antidepressant and any treatment, and patients ≥65 had lower odds of receiving an antidepressant, Hispanic patients had greater odds of receiving psychotherapy and any treatment. Patients who were never married had lower odds of receiving an antidepressant and any treatment. Patients with a service connection of 50% or greater had lower odds of psychotherapy, a PTSD clinic visit, and any treatment. Individuals with a prior psychiatric diagnosis other than PTSD had lower odds of a PTSD clinic visit. Also, any prior outpatient use was associated with lower odds of receiving psychotherapy and any treatment, and individuals who had a PC-PTSD score of 4 versus 3 had greater odds of receiving psychotherapy, having a PTSD clinic visit, and receiving an antidepressant prescription or any treatment.
Discussion
To our knowledge, this was the first systemwide evaluation of the relationship between setting of services on the day of positive PTSD screening and subsequent PTSD diagnosis and treatment initiation in VHA. We found that same-day PC-MHI services compared with primary care–only services were associated with greater odds of receiving a PTSD diagnosis, both on the same day and up to one year after screening positive. Among patients diagnosed as having PTSD on the same day as their positive screen, same-day PC-MHI services were also associated with increased odds of initiating PTSD treatment within 12 weeks of diagnosis, compared with those receiving only primary care at screening. Such treatment initiation included the outcomes of PTSD clinic encounter as well as receipt of psychotherapy and antidepressant medications, which are in line with VHA PTSD treatment guidelines of referral to specialized care and initial treatment options for PTSD, respectively (
1). Furthermore, same-day PC-MHI services had PTSD diagnosis and treatment initiation outcomes similar to those with same-day specialty mental health services.
The findings from this study are generally consistent with results from prior research demonstrating the utility of integrated mental health care services within the primary care setting. Such related prior work includes numerous randomized trials as well as several meta-analyses that support the efficacy of collaborative care for improving depression and depression-related outcomes (
8,
9,
16,
17). More similar to the specific outcomes studied in this investigation, results from another trial have shown that integrated care may improve mental health treatment access and engagement (
18). Although fewer studies have focused specifically on individuals with PTSD, trials in this area have indicated positive initial results (
7,
19). For example, Schnurr and colleagues (
7) found that patients with PTSD who were randomly assigned to a collaborative care intervention were more likely than those in a control condition to have a mental health visit, fill an antidepressant prescription, and have an antidepressant refill; however, no PTSD symptom differences were observed between groups at follow-up. With administrative data from systemwide PC-MHI implementation in VHA, results from our evaluation complement this growing PTSD-specific evidence base by providing empirical support for the potential benefit of integrated care on PTSD diagnosis and treatment initiation. Further research, however, is needed to examine whether such increases in diagnosis and treatment initiation associated with PC-MHI services are also associated with improvements in treatment outcomes among patients with the disorder.
This study also provided evidence for the utility of PC-MHI in particular as an integrated care model that can facilitate psychiatric diagnosis and treatment. Recently, Pomerantz and colleagues (
10) summarized key findings from published studies evaluating the VHA PC-MHI program. Results from our investigation parallel previous findings from such national and local evaluation efforts. Of note, the findings from this study are consistent with those from prior national evaluation studies, including a facility-level analysis that found increases in psychiatric diagnoses after PC-MHI implementation (
12), a study with primary care patients that demonstrated that use of same-day PC-MHI services was associated with an increased likelihood of a subsequent mental health–related encounter (
20), and a depression screening study that found that use of same-day PC-MHI services was associated with increased odds of initiating depression treatment within 12 weeks of screening (
21). Our findings also correspond to those from prior regional and single-site analyses that suggest that PC-MHI programs may facilitate more timely treatment for those with positive mental health screens (
22), increase access to mental health care (
23–
25), improve the likelihood of completion of PTSD specialty consultations (
14), and promote shorter- and longer-term retention in mental health treatment (
26). Collectively, these findings support the role of PC-MHI in enhancing access to and engagement with mental health care in VHA.
This study enables better understanding of PTSD screening within VHA, specifically regarding mental health care subsequent to screening positive, by using data that span the entire health system. In a prior regional study of PTSD screening, older veterans compared with younger veterans had lower odds of a specialty mental health visit in the year after screening positive for PTSD (
27). The oldest age group in that study was also less likely to receive a prescription for an antidepressant medication (
27). On the basis of results from our investigation, such age-related treatment differences may be explained by our finding that older age groups were less likely to receive a PTSD diagnosis after screening positive. Clearly, further investigation is needed to better understand potential age-related PTSD treatment disparities. In another regional study that examined behavioral health screening, including PTSD, Shiner and colleagues (
28) found that among patients who screened positive for PTSD, those treated in the PC-MHI or specialty setting were more likely than those in primary care to receive adequate treatment. Although we examined different, yet related, outcomes, findings from this study are broadly similar to those in the Shiner and colleagues study.
There were several potential limitations of our study. First, we conducted analyses of administrative data. Although models adjusted for several potential confounding variables, we cannot rule out the possibility that another unmeasured factor or incomplete adjustment might have accounted for the observed association linking same-day PC-MHI services (and same-day specialty mental health care) with increased odds of PTSD diagnosis and treatment initiation. For example, patients with more severe PTSD may have been more likely than others with PTSD to receive same-day PC-MHI; thus, analyses adjusted for PC-PTSD score. Second, some patients who had an initial positive screen in primary care may have left VHA before receiving a diagnosis. Third, it is expected that a percentage of patients in the sample who screened positive would not be diagnosed as having PTSD after a complete diagnostic assessment and would therefore either not require treatment or be diagnosed as having another psychiatric condition. Consequently, for a more conservative approach, we conducted analyses of treatment initiation among only patients who received a PTSD diagnosis on their screening date. Fourth, administrative codes for receipt of psychotherapy and for receipt of medication are general and not specific to PTSD, which may partially explain the higher cumulative rates of treatment initiation than diagnoses. Fifth, the study was conducted in VHA; therefore, findings may not be generalizable to other settings.