Primary Care-Mental Health Integration in the Veterans Affairs Health System: Program Characteristics and Performance

Integrating mental health care services within primary care may facilitate recognition of mental health treatment needs (1), enhance access and treatment engagement (2,3), and improve outcomes while reducing costs (4,5). In 2007, the Veterans Health Administration (VHA) began national implementation of integrated mental health services within primary care settings, known as primary care–mental health integration (PC-MHI) (6). These services are understood to be part of a patient-centered medical home, referred to in VHA as patient aligned care teams (PACTs).

VHA policy requires implementation of PC-MHI services, which must include both care management (CM) and colocated collaborative care (CCC). The CM should include ongoing structured assessments and monitoring of patient mental health needs, with supervision of care managers by mental health providers (7). CCC should involve colocation of mental health and primary care practitioners in primary care settings as part of the medical home, with shared responsibility for mental health evaluation and treatment and availability for consultation when needed (7,8). CM and CCC are considered complementary and fundamental elements of integrated mental health services (7). CM involves longitudinal follow-up for mental health conditions, including medication management, guideline-based treatment support, and patient activation and education. CCC focuses on same-day mental health access, with brief problem-focused assessments and interventions and referral to specialty mental health clinics for individuals requiring complex mental health treatment (6,7,9). Site leadership determines whether these service components are delivered by the same staff, and patterns of staffing assignment vary across sites.

Implementation of PC-MHI has been substantial (2,10), with over 7.1 million encounters provided between October 1, 2007, and April 30, 2017 (unpublished data, VHA PC-MHI Evaluation, 2017). However, overall implementation has varied (2). In fall 2013, 12% of sites required to provide these services reported they were not available, and 47% of sites reported the absence of one or both required components, CM and CCC (unpublished data, VHA PC-MHI Evaluation, 2013).

The quality of integrated care services at the VHA has been examined by measuring the association between integrated care implementation with prevalence of diagnoses of common mental health conditions (1) as well as associations between receipt of integrated care services on the same day as a primary care encounter and engagement in subsequent mental health services (11) and receipt of timely treatment for depression (3). Both PC-MHI and other collaborative care models have been found to improve access to mental health services, continuity of care, and detection and treatment of mental health conditions (1,3–5,11,12). To date, little is known regarding the correlation between various characteristics of integrated care programs and whether key correlated characteristics are associated with performance. Understanding integrated care service structures, their concordance with the service model, and their associations with performance-related indicators could help to improve the organization and delivery of these services, both within VHA and in other health systems.

This study examined site reports from the 2013 National PC-MHI Evaluation Survey. The goals of the study were to identify the underlying structures of these PC-MHI service reports and to use administrative data to assess associations between the identified structures and performance indicators. This work also provides information for validating site reports and performance measures and presents new findings regarding implementation of integrated care services. These findings may focus attention on the association between key service elements and performance. On the basis of policy guidelines (7), we hypothesized that greater consistency between program structures and VHA’s model for integrated care services, such as care manager supervision, brief assessments, and adequate staffing for service demand, would be associated with better performance.

Methods

PC-MHI Survey

The VHA includes U.S. Department of Veterans Affairs (VA) medical centers (VAMCs), which provide inpatient and outpatient services, and community-based outpatient clinics (CBOCs), which generally are smaller and do not offer inpatient services. The 2013 National VHA PC-MHI Evaluation survey was completed by all VHA facilities required to provide PC-MHI services. These constituted 349 sites, including 166 VAMC divisions and 183 CBOCs serving at least 5,000 patients per year (8). The survey was fielded in November 2013. Each site provided a single self-report, often including multiple individuals. We conducted extensive validation, including follow-up to resolve reporting inconsistencies.

Survey Variables

The survey had 263 measures, excluding free-response items. It included questions regarding CM and CCC, integrated care implementation, activities, treatment availability, referrals, staffing, colocation in primary care, policies, issues affecting implementation, and participation of integrated care staff in specialty patient-centered medical home teams (PACTs). The survey used skip logic. For example, if sites reported that CM was not available, they were not asked follow-up questions regarding CM operations. In all such instances, missing responses were coded as a separate “not applicable” category.

Facility Characteristics

Facilities were classified as VHA medical centers or outpatient clinics on the basis of administrative records. Sites were also classified by the number of primary care service recipients at the site from October 1, 2013, to December 31, 2013. Primary care population size was categorized as 6,000 or fewer; 6,001–9,000; 9,001–14,000; or greater than 14,000 (range 2,083–49,489).

Performance Indicators

We assessed three integrated care performance indicators created by the VHA. These indicators were generated by using VHA’s Corporate Data Warehouse administrative data, and their correlation with patient outcomes has yet to be validated. First, we assessed prevalence of receipt of integrated care services among primary care patients. This was calculated as 100% multiplied by the number of patients who received integrated care encounters in 2013 divided by the number of patients who received nonintegrated care, primary care, or integrated care in 2013.

Second, we assessed same-day access to integrated care and primary care. A primary objective of integrated care is to provide immediate access to assessment and collaborative treatment for patients identified as having mental health symptoms (7). This measure was calculated as 100% multiplied by the number of patients with an initial in-person encounter for integrated care and a same-day primary care encounter in 2013 divided by the number of patients with an initial in-person encounter for integrated care in 2013. For patients with more than one encounter for integrated care, the most recent encounter was considered an initial encounter if it occurred at least two years after the most recent previous encounter (3). VHA program guidance recommends that most patients be connected to initial integrated care services through a warm handoff from a primary care provider to an integrated care provider (7). However, the measurement of same-day access is not limited to warm handoffs; it uses VHA administrative data to capture receipt of same-day encounters regardless of the mechanism for coordination.

Third, we assessed extended treatment duration, operationalized as receipt of integrated care services in four consecutive two-month periods following an initial integrated care encounter in 2013.

Receipt of integrated care services and same-day access are established VHA performance indicators. Extended duration of integrated care is used to assess the prevalence of long-term integrated care treatment as a practice pattern, and a higher score indicates lower adherence to brief treatment practices. Although a longer course of integrated care services can be helpful for some patients, the PC-MHI model advocates briefer appointment durations (30 minutes or less), with limits on the duration of PC-MHI services (number of sessions and weeks of services) (7) to ensure access to a broader reach of primary care patient population. Prior work suggests that easily accessible, rapid access to services and short-term treatment are the features of the colocated care model that best predict improvement in depression outcomes (13). This work was conducted as part of ongoing VHA operations with approval from the VHA Ann Arbor Institutional Review Board.

Principal-Components Analysis

We used principal-components analysis (PCA), a variance-focused data reduction approach, to identify integrated care service structures from the survey responses. PCA generates summary variables accounting for the maximal amount of response information (14). PCA was conducted on a correlation matrix to account for different scales used in the survey measures.

The initial PCA identified six summary variables, or factors, on the basis of the scree plot, eigenvalues greater than 1, and considerations regarding interpretability. The individual survey variables were regarded as loading significantly on a factor if their correlation with the factor had an absolute value greater than .50 (15). We calculated Cronbach’s alphas to assess internal consistency among variables correlating with the same factor. All factors except factor 2 had alphas of at least .70. Consequently, we conducted a second PCA for variables that correlated initially with factor 2. This identified two factors, and the final PCA solution included seven factors. To determine optimal regression weights, factor loadings were multiplied by the standardized survey variable responses. These products were summed to yield factor scores.

Analyses With Factor Scores

We evaluated Pearson correlations among the factors and between each factor and the three performance measures. We generated hypotheses regarding associations between factors and performance based on our review of the PCA-generated factor structures and recommendations related to integrated care services (7). We also calculated mean factor scores by category of facility characteristic. Associations between mean factor scores and facility characteristics were evaluated by using Wilcoxon rank sum tests and analyses of variance. All analyses were conducted with SAS software, version 9.3. This work was conducted as part of ongoing VHA operations with approval from the VHA Ann Arbor Institutional Review Board.

Results

PCA Factor Extraction

The seven-factor solution was composed of 129 variables, which represented 49% of the 263 survey variables included in the PCA. We interpreted the factors as core implementation, CM assessments and supervision, CM supervision receipt, colocated collaborative care (CCC) prescribing providers, CCC behavioral health providers, integrated care staff participation in special population primary care teams (PACTs), and complex mental health conditions treatment. Table 1 lists survey elements related to each factor and reports internal consistency among all variables correlating with the same factor (Cronbach’s alpha scores). There was some conceptual overlap between the variables that loaded significantly on a factor and those that did not, yet the nonloading variables did not contribute substantially to overall variance.

As stated previously, we expected to observe better performance at sites reporting service structures consistent with VHA’s model for integrated care. Following identification of the seven-factor solution, we hypothesized that program performance would be better at sites with higher scores for factors related to the principal elements of integrated care (core implementation, CM assessments and supervision, CM supervision receipt, CCC by prescribing providers, and CCC by behavioral health providers). We also anticipated that scores for these factors would be higher at VAMCs and larger facilities, which often have greater mental health staffing and resources (16). These expectations were supported. For example, greater scores indicated sites provided focused assessments and brief therapy for common mental health conditions, measurement-based services using structured assessment tools, and briefer wait times and encounters (7).

We did not have specific expectations regarding associations for factor 6 (engagement of integrated care staff in PACTs). In VHA, these teams are components of the overall PACT program with a clinical or patient subpopulation focus, including geriatrics, women’s health, postdeployment clinics, homeless programs, and home-based primary care. Participation in these teams may be indicative of a fully operational and functional program or, perhaps, an overextension of integrated care staff that could diminish the program’s focus. Finally, we anticipated that performance would be lower at sites with greater scores for factor 7 (treatment of complex mental health conditions). Provision of integrated care services for individuals with complex mental health needs may be inconsistent with the integrated care model of using brief services (encounters of 30 minutes or less) to address population health needs while referring individuals with complex needs to specialty mental health clinics (6,7).

Analyses With Factors From PCA

Table 2 presents results from Pearson correlations estimating associations among factors. CM assessments and supervision was positively associated with core implementation (r=.13) and negatively associated with treatment of complex mental health conditions (r=–.11). CM supervision receipt was negatively associated with CCC by prescribing providers (r=–.11).

Table 3 presents mean factor scores by facility characteristic (facility type and number of primary care patients over three months). Compared with CBOCs, the VAMCs had higher scores for CM assessments and supervision, CCC by prescribing providers, and special population primary care teams (PACTs). They had lower scores for CM supervision receipt. Those serving more patients had greater scores for core implementation, CM assessments and supervision, CCC by prescribing providers, and special population primary care teams (PACTs) and had lower scores for CM supervision receipt.

The mean±SD percentage of recipients for each performance indicator was 6.6%±4.8% (range .0%−27.7%) for integrated care, 33.4%±21.1% (range .0%−100.0%) for same-day integrated care access, and 4.6%±5.8% (range .0%−46.4%) for extended integrated care. Table 4 presents Pearson correlations evaluating associations between the factors and the performance indicators. Core implementation was positively associated with same-day access (r=.28). CM assessments and supervision was positively associated with receipt of PC-MHI services (r=.13) and same-day access (r=.18). Having CCC prescribing providers was positively associated with receipt of integrated care services (r=.23). Having CCC behavioral health providers was positively associated with same-day access (r=.12). PACTs were positively associated with same-day access (r=.17).

Discussion

This study used PCA to describe structures of integrated mental health services within VHA primary care settings. Findings document the importance of integrated care implementation and fidelity, and they offer new information regarding associations with facility characteristics and performance.

Factors and Facility Characteristics

Compared with outpatient clinics, VA medical centers had greater scores for CM assessments and supervision and CCC by prescribing providers. Larger primary care population size was positively associated with core implementation, CM assessments and supervision, and CCC by prescribing providers. These findings are consistent with previous observations that medical centers and larger VHA facilities were quicker than to implement integrated care, including the required components. This may reflect staffing capacity at these facilities (16). Our findings reinforce the importance of ensuring access to integrated care in smaller facilities.

Medical centers and larger sites also had greater levels of participation by integrated care providers in PACTs. This may reflect greater presence of such teams at these sites and greater staff resources, which enable integrated care practitioners to contribute to these teams.

Factors and Integrated Care Performance Indicators

Ideally, when primary care providers identify a patient with a need for mental health services, the services are provided on the same day (7). Consistent with this expectation, we observed a positive association between core implementation and same-day access, and warm handoffs to integrated care providers loaded positively on the core implementation factor. Sites utilizing handoff methods may be better able to provide needed same-day care than sites using other referral methods, such as written referrals.

Sites with greater scores for CM assessments and supervision and CCC by prescribing providers had greater program reach, providing services to a larger portion of primary care patients. Greater scores for CM assessments and supervision included reviewing and supervising care manager patient panels and medication management, as well as using structured tools in initial and follow-up assessments. For CCC by prescribing providers, greater scores indicated having prescribing providers’ treatment offices interspersed in primary care clinics, appointments of 30 minutes or less, and briefer wait times for appointments.

Factors drawn from the site survey reports were not significantly associated with extended duration of integrated care services.

Factors and Depression Treatment

Prior studies indicate that regular supervision of care managers, particularly by a member of the mental health staff, is associated with improved depression outcomes (17,18). In this study, survey variables loading on the CM assessments and supervision factor included CM supervision (review of the patient panel, activities, and medication management) and greater staffing of care manager supervisors by mental health staff. These are important service components for sites to develop because they may enhance depression care quality within the integrated care setting.

Collaborative care has also been found to be associated with improved depression outcomes and greater use of antidepressants (17,18). Moreover, psychological interventions, with or without antidepressant treatment, have been found to be associated with greater depression symptom improvement, compared with antidepressant treatment alone (17). Survey variables loading significantly on factor 4 (CCC by prescribing providers) and factor 5 (CCC by behavioral health providers) included whether these providers prescribed psychotropic medications and indicators regarding whether these providers provide short- and long-term nonpharmacological treatment. Thus greater scores for these factors may indicate sites that were better able to accommodate patient treatment preferences (antidepressants or nonpharmacological treatment), which may result in greater depression symptom improvement.

Associations Among Site Factors

In exploratory analyses regarding associations among factors, the negative association observed between treatment of complex mental health conditions and CM assessments and supervision suggests that sites allocating integrated care resources to treatment of severe depression, bipolar disorder, schizophrenia, and other severe conditions were less likely to provide essential CM services. A policy expectation for integrated care is providing brief, problem-focused services for common mental health conditions. This analysis suggests that sites with less fidelity to this brief treatment model also have less effective CM.

Limitations

We note several limitations of the study. First, there is a possibility that errors occurred in survey self-reports. However, extensive validation was conducted to enhance accuracy. Further, site visits conducted as part of VHA’s national evaluation of integrated care have supported the validity of prior reporting. Second, we observed a high frequency of missing responses because of skip logic; however, the uniform way in which missing responses were coded may minimize bias. Third, subjectivity was involved in determining which factors to retain and the conceptual interpretation of these factors. In this analysis, these decisions were based on statistical tests and literature. Finally, this study examined associations between site reports and three performance indicators. These indicators were created internally and selected on the basis of their ongoing use in VHA system monitoring and the importance of brief treatment. Future analyses should validate these indicators through associations with clinical outcomes and consider additional dimensions of performance.

Conclusions

This study contributes important new information regarding the structures of the PC-MHI program, particularly their interrelation, their concordance with the integrated care model, and their associations with performance. Key findings underscore the importance of implementation of core components of integrated care for ensuring mental health care access and treatment engagement. Enhancing specific integrated care components, including CM case supervision, colocation of collaborative care generally, and the provision of both pharmacological and nonpharmacological interventions within the integrated care setting, may confer greater benefits for patients (17,18). Study findings also document lower levels of implementation of PC-MHI services at smaller facilities. Further study investigating implementation barriers at smaller facilities is warranted. Finally, these findings from the largest integrated health system in the United States offer an important example for other health systems regarding mental health integration in primary care settings.