Primary Care–Mental Health Integration in the VA Health System: Associations Between Provider Staffing and Quality of Depression Care

The integration of mental health services in primary care settings is associated with improved treatment processes and outcomes for depression. These include greater treatment initiation and coordination of care with primary care physicians (PCPs) (1,2), increased likelihood that patients will engage in specialty mental health treatment following referral (3,4), and greater improvement in depression symptoms (5,6).

Given the benefits in the processes and outcomes of depression treatment associated with integration, the Department of Veterans Affairs (VA) began national implementation of Primary Care–Mental Health Integration (PCMHI) services in 2007, requiring all VA medical centers (VAMCs) and large community-based outpatient clinics (CBOCs) to have PCMHI services. Evaluation of this VA initiative indicated that implementation of PCMHI was associated with many of the benefits described above, as well as with increased capacity to diagnose depression among primary care patients (4,7,8). In addition to considering processes and outcomes of PCMHI services, it is also important to consider the structural elements that contribute to depression treatment quality. To date, little is known regarding whether and how PCMHI staffing and the discipline-specific mix of PCMHI staff are related to indicators of depression treatment initiation and continuation for individuals with a new episode of depression. This study investigated how these staffing care structures are associated with mental health treatment receipt in VA.

VA policies require PCMHI services to include care management and colocated collaborative care components (9). Care management includes routine monitoring of medication effectiveness, adherence, and treatment needs, provided by a care manager in coordination with PCPs. Colocated collaborative care involves mental health practitioners’ working in the primary care clinic setting, with shared responsibility for evaluation and treatment of mental health conditions (10). VA health system implementation of PCMHI has been substantial, with 88% of sites in 2013 indicating that they have any PCMHI service, 83% reporting implementation of collocated collaborative care, 58% reporting implementation of care management, and 54% reporting implementation of both collocated collaborative care and care management (11). The size and composition of PCMHI provider staffing are largely determined by the leadership at individual VAMCs (including any associated CBOCs), with input from regional and national leadership, particularly when a facility’s PCMHI activity is low. Thus there is the potential for large variation in the amount and type of mental health professionals employed (typically psychologists, psychiatrists, nurses, and social workers) by PCMHI services across VA facilities (11), yet little is known about the associations between PCMHI staffing and processes of care.

This study assessed whether facility PCMHI provider staffing was associated with greater performance on indicators of depression care quality at the facility level (including care provided in primary care and specialty care). We hypothesized that higher total staffing ratios would be associated with greater performance on facility-level indicators related to receipt of antidepressant medication and psychotherapy. We also hypothesized that when the analysis controlled for overall PCMHI staffing, the mix of discipline-specific staffing would differentially affect the likelihood of receipt of antidepressant medication and psychotherapy treatment. Psychologists and social workers typically provide psychotherapy treatments; therefore, we hypothesized that having more of these providers in PCMHI services would be associated with increased receipt of psychotherapy treatment and greater likelihood of engagement in psychotherapy. Social workers and nurses also often fill medication care manager roles, and so we also hypothesized that PCMHI staffing ratios for these professions would predict greater likelihood of any or adequate antidepressant medication treatment. Finally, because psychiatrists and nurse practitioners typically provide medication management, we hypothesized that having more of these prescribers would also be associated with the amount and adequacy of antidepressant medication treatment. Understanding how facility PCMHI clinical staffing ratios and staffing mix may affect treatment receipt may inform quality improvement and service enhancements.

Methods

Data were obtained from the VA Corporate Data Warehouse (CDW) for the 349 VAMCs (N=166) or large CBOCs (N=183) falling under the policy requirement to implement PCMHI. The CDW contains patient demographic, diagnosis, procedure, clinical encounter, clinic type, and pharmacy data for all VA patients. Depression treatment measures were calculated for each facility from a cohort of 279,199 patients across these facilities who had a new episode of depression, defined by an index diagnosis of depression (ICD-9-CM codes 293.83, 296.2, 296.3, 296.9, 298.0, 300.4, 301.12, 309, and 311) (12) at a primary care or PCMHI clinic encounter in fiscal year 2013 and who did not have a depression diagnosis or antidepressant treatment in the 12 months prior to the index diagnosis date. Patients with a bipolar disorder diagnosis (296.0, 296.1, 296.4, 296.5, 296.6, 296.7, and 296.8) were excluded because antidepressant medications are not recommended as first-line treatments for bipolar depression, and thus indicators of antidepressant treatment do not apply to this population (13). Average patient characteristics per facility were used in analyses. These included average patient age, percentage of patients at the facility with an urban address, percentage of male patients, percentage of non-Hispanic white patients, percentage with a ≥70% military service–connected disability status, and percentage with a psychiatric comorbidity of a substance use disorder, posttraumatic stress disorder (PTSD), or other anxiety disorder. Facility characteristics used in analyses included the number of unique patients in primary care and facility type (VAMC or CBOC).

Measures

Receipt of any antidepressant medication was defined as having at least one fill of any antidepressant medication in the three months following the index diagnosis date for the new episode of depression. Adequate medication receipt was defined by using the nationally reported Healthcare Effectiveness Data and Information Set (HEDIS) antidepressant medication measure, which assesses whether patients with new episodes of depression receive antidepressant coverage for 84 of the first 114 days of treatment (14). However, because initial prescriptions for a 90-day supply satisfy the HEDIS requirements, sensitivity analyses in which at least one refill was also required were conducted (15).

Any psychotherapy was defined as receipt of at least one psychotherapy encounter (CPT codes 90804–90819, 90821–90824, 90826–90829, 90832–90834, 90836–90840, 90845–90847, 90849, and 90853) in the three months following the index depression diagnosis. We excluded from the main analyses procedure codes for brief medication visits (20–30 minutes) that included psychotherapy, because these may consist primarily of medication management with relatively little emphasis on psychotherapy. However, these were included in sensitivity analyses. Psychotherapy engagement was defined as receipt of three or more psychotherapy visits in the three months following the index depression diagnosis. Three visits was selected as the metric for psychotherapy engagement for several reasons. First, the number of sessions for psychotherapy treatments may differ between PCMHI and mental health specialty clinics in the VA; PCMHI models typically include brief psychotherapy treatments, whereas specialty clinics often offer more traditional psychotherapy protocols (eight to 12 sessions). Currently, the VA does not specify a minimum number of sessions for adequate receipt of psychotherapy in PCMHI. However, although there is no recommended minimum number of sessions for brief psychotherapy in primary care, a meta-analysis has suggested that brief psychotherapy in primary care can be effective when delivered in as few as three sessions (16).

Quality-of-care measures were not limited to treatment provided in primary care because one of the primary aims of the study was to determine the extent to which PCMHI staffing is associated with receipt of care beyond what might otherwise be delivered in the absence of PCMHI. If the measures were limited to only treatment provided in primary care, it would be difficult to determine whether the treatment provided in primary care more than offset potential reductions in care in a facility’s specialty care clinics.

Staffing ratios were calculated by using data from site self-reports in the 2013 VA PACT PCMHI Survey, an annual survey conducted by the VA National PCMHI Evaluation. The survey included VAMC divisions and CBOCs that serve 5,000 or more patients annually, and these sites are mandated to provide integrated mental health services in primary care clinics. All 349 sites completed the survey, and 308 reported providing integrated mental health services (11). Facilities with PCMHI services were asked in the survey to list all PCMHI staff located at the facility currently providing PCMHI services, including name, discipline, and the average hours per week the staff member works in PCMHI. Total clinical PCMHI staffing ratios were calculated by summing the reported PCMHI hours and computing the fraction of a full-time equivalent (FTE) for each PCMHI clinical provider at a facility and dividing by the size of the facility’s primary care patient population. Sites’ clinical PCMHI provider mix was assessed by using discipline-specific proportions of total PCMHI staffing, categorized as psychologists, social workers, psychiatric prescribers (psychiatrists and midlevel providers, such as nurse practitioners and physician assistants with prescriptive authority), and registered nurses. Sites that did not report PCMHI staff were excluded from the analyses that included provider mix.

Data Analysis

All analyses were conducted at the facility level with SAS version 9.3. Descriptive statistics were calculated and compared by using t tests and chi-square tests. Two types of multiple regression analyses were run for each of the four treatment outcomes. In the first set of analyses, total staffing FTE per 10,000 unique primary care patients was the main predictor variable of interest. In the second set of analyses, separate regression models were run for each provider discipline. Each of these models included as the main predictor variable of interest the proportion of total staffing accounted for by the provider discipline specified by the model; total staffing FTE per 10,000 unique PC patients was included in these models as a control variable. All analyses controlled for the average prevalence per facility of the following characteristics among primary care patients with depression: age, sex, race-ethnicity, urban residential status, and diagnosis of PTSD, anxiety disorder, and substance use disorder. Additional control variables included number of unique primary care patients per facility and facility type (VAMC or CBOC). All regression coefficients are unstandardized.

Results

Across all 349 sites, 65% of the primary care patients with a new episode of depression received any antidepressant medication and 45% received adequate antidepressant medication. Twenty-seven percent of cohort patients received any psychotherapy in the three months following their depression diagnosis, and 11% engaged in psychotherapy (defined as attending three or more psychotherapy sessions (Table 1). Table 1 also presents facility characteristics and average patient characteristics for primary care patients with a new episode of depression. On average, facilities reported 2.3 clinical PCMHI FTEs per 10,000 primary care patients.

Total Staffing

Table 2 presents results from linear regression models estimating the association between total staffing ratios (FTEs per 10,000 patients) and mental health treatment receipt. Total PCMHI staffing FTEs were not associated with any or adequate antidepressant medication receipt. Total PCMHI staffing was associated with higher levels of any psychotherapy (B=1.16, p<.01) and psychotherapy engagement (.39, p<.01).

Discipline-Specific Staffing Proportions

Table 3 presents models that included total PCMHI clinical staffing FTEs per 10,000 patients and proportions of site PCMHI teams by category of provider staffing. In these models, higher proportions of social workers were associated with a greater percentage of patients receiving adequate antidepressant medication (3.16, p=.03). Higher proportions of nurses were associated with lower levels of psychotherapy engagement (−2.83, p=.02). Proportions of psychologists and psychiatric prescribers were not associated with depression care process measures.

Sensitivity analyses using modified measures of antidepressant treatment and psychotherapy did not yield any differences in significance across findings.

Discussion

This study investigated the relationship between PCMHI staffing characteristics and the quality-of-care processes among VA patients given a diagnosis of depression by PCPs or PCMHI providers. Overall, higher levels of PCMHI clinical provider FTEs per 10,000 patients were associated with higher rates of psychotherapy receipt but were not associated with antidepressant medication receipt. The positive association between quality of psychotherapy for depression and total PCMHI provider staffing is particularly notable considering that PCMHI is provided in addition to the VA health system’s comprehensive specialty mental health services that are often located in the same facility as primary care. Higher PCMHI proportions of social workers were associated with higher receipt of adequate antidepressant medication, and higher proportions of nurses in PCMHI were associated with less psychotherapy engagement. These results suggest that PCMHI staffing may influence the extent to which improvements in the quality of care for depression are realized.

The finding that total PCMHI FTEs were not associated with higher rates of medication use may be related to the fact that the proportion of staff who were psychiatric prescribers (for example, psychiatrists and nurse practitioners) was not significantly related to receipt of or adherence to antidepressant medication. These findings were surprising and not consistent with initial hypotheses. One explanation may be that the function of psychiatric prescribers may overlap with functions of primary care providers, who often prescribe antidepressant medications without specialty prescriber input (17,18). It is also possible that the addition of a psychiatric prescriber to PCMHI does not provide a marginal benefit relative to the availability of psychiatric prescribers in VA specialty mental health settings. Psychiatric medication prescribers also often have a supervisory or consultative role in collaborative care models, which may limit their ability to directly affect patient adherence compared with the other disciplines. This finding is consistent with a prior VA study, which found that a psychiatric resident consultation service was not as effective as a care management model in improving depression care (6). This finding, however, is in contrast to a meta-regression analysis of collaborative care trials, which found that care management that included supervision by a psychiatrist was associated with better outcomes (5). Further investigation is needed to determine the role of psychiatric prescribers in VA PCMHI services and the extent to which they operate in a traditional referral-based consultant model or serve as supervisors of care managers, who care for a larger number of patients.

Proportions of registered nurses in PCMHI services were also not associated with receipt of antidepressant medication. One explanation for this finding may be that in 2010 the VA began national implementation of patient-centered medical homes (19), of which nurse care management is a key component (19,20). Thus primary care nurses may provide care management functions for psychiatric medications that overlap with services provided by PCMHI nurses. However, the proportion of social workers, who also often serve as care managers, was associated with greater antidepressant treatment adequacy. Although we did not specifically test for differences between the two disciplines, the contrast in findings could indicate that nurses in PCMHI programs may need more specific training to engage patients in continued mental health treatment. Data from the Canadian Community Health Survey indicated that patient dropout (defined by leaving mental health treatment before the provider agreed to termination) over a 12-month period was most likely for those who had seen a nurse but was less likely for those who had seen a social worker or psychologist (21). In addition, in a meta-analysis of collaborative care interventions, care managers with a specific mental health background were more effective than those without (5). The degree of prior mental health experience, if different across disciplines of care managers, could influence the relationship between discipline-specific staffing and care quality.

The finding that PCMHI program staffing was associated with greater psychotherapy receipt but not with greater antidepressant treatment adequacy could be related to the fact that psychologists are the most common discipline within VA PCMHI programs. Social workers, the second most common discipline, also often provide psychotherapy. The proportion of psychologists or social workers was not associated with greater receipt of psychotherapy, suggesting that programs in which most staff were either psychologists or social workers performed similarly in terms of delivering psychotherapy. Staffing by nurses, who generally do not provide psychotherapy, was associated with less receipt of psychotherapy. Increased delivery of psychotherapy in primary care may represent a more patient-centered approach to depression treatment (22). However, although there is evidence to support the effectiveness of brief psychotherapy for depression outside a PCMHI model, meta-analysis of controlled trials has not shown psychotherapy to be an effective augmentation to antidepressant treatment provided as part of collaborative depression care (5). The incorporation of depression outcome data into quality measurement could help clarify the balance in staffing between psychotherapy providers versus medication care managers (and their supervisors) to optimally improve population-level depression outcomes.

This study had several limitations, and certain conclusions should be circumscribed. First, although we investigated psychotherapy receipt, we could not measure the type of psychotherapy received or fidelity to evidence-based protocols. We note, however, that psychotherapy in the VA is increasingly delivered according to evidence-based protocols (23,24). Second, the measure of antidepressant treatment quality assessed only the quantity of medication provided and did not assess other aspects of antidepressant medication management, such as adequate dosing, intensification, or management of side effects. Finally, the cross-sectional observational nature of the study limits the extent to which conclusions may be drawn about the causal role of PCMHI staffing in improving care. Although we adjusted for available patient and facility characteristics that may have confounded the relationship between staffing and care, it is possible that the analyses did not control for other important confounders. Future research should examine the implementation factors that account for differences in staffing and the causal effects of staffing interventions.

Conclusions

Integration of mental health services into primary care is an important way to improve mental health care access, engagement, and outcomes. This study found that total PCMHI staffing was associated with higher levels of psychotherapy receipt, and higher proportions of PCMHI social worker staffing were associated with better performance on measures of antidepressant treatment adequacy in VA integrated care programs. Future quality improvement efforts should examine fidelity to depression care management protocols given the lack of increase in antidepressant treatment associated with total staffing or staffing specifically by medication prescribers and nurse care managers. Findings from the VA’s extensive national implementation of PCMHI services may also inform work in other health systems.