Linking Primary Care Patients to Mental Health Care via Behavioral Health Social Workers: A Stepped-Wedge Study

The U.S. Preventive Services Task Force recommends that health care systems screen patients for depression to improve identification of this underdiagnosed and undertreated condition (1, 2). However, identifying depression does not ensure treatment and subsequent improvement in health outcomes (3). Health systems need a systematic approach to connect patients with a new diagnosis of depression to therapy (4, 5). Furthermore, systems built to link patients with depression to treatment may also benefit patients with other behavioral health conditions (e.g., anxiety).

Implementing systematic treatment for depression can be challenging because of inadequate funding for and access to behavioral health resources (6). In research settings, the collaborative care model has been shown to improve care outcomes (7). The original model, published more than 20 years ago, colocated behavioral health providers within the primary care practice (8). However, health systems that want to colocate primary and behavioral health providers face financial, training, and process barriers (9, 10). New strategies to link patients to behavioral health services have emerged. The University of Washington pilot tested a centralized remote care management program that employed clinicians to monitor patients’ progress via chart review and facilitate engagement when appropriate. Patients in the program were more likely to have a follow-up Patient Health Questionnaire (PHQ-9) score recorded and to have improved PHQ-9 scores (11). Another study found that using a smartphone application that allowed the health care team to send appointment reminders, monitor symptoms, and share educational material reduced follow-up time, compared with usual care (12). Solutions do not always require advanced technology. One novel approach is to employ licensed independent social workers, previously used to facilitate in-person collaborative care (13), to instead facilitate care across sites.

With the widespread implementation of electronic health records (EHRs), it is easier to share information (14). EHRs enable clinicians to collaborate even when physically distant (15). Given the increased demand for behavioral health care necessitated by depression screening, the need for primary care physicians and behavioral health specialists to collaborate is pressing. In 2016, Cleveland Clinic implemented systematic depression screening, which increased identification of patients with depression (16). To link these patients to appropriate resources, the health system deployed behavioral health social workers (BHSWs)—licensed independent social workers with specialized behavioral health training—in primary care practices. Although the impetus to deploy BHSWs was depression screening, patients with anxiety often have similar needs (17), and physicians could also refer them to BHSWs. The objective of this study was to identify the impact of BHSWs on use of behavioral health services and pharmacotherapy in the 30 days after a patient’s initial diagnosis with depression or anxiety.

Methods

This quasi-experimental, stepped-wedge study (18, 19) included adults who received a new diagnosis of depression or anxiety between January 2016 and December 2019. We included patients who had at least one primary care visit at a location that was supported by BHSWs by February 2019. Because we were interested only in patients with a new diagnosis, we excluded anyone who had a diagnosis of depression or anxiety in 2015 or received a prescription for an antianxiety or antidepressant medication in 2015. EHR data were collected through February 2020, enabling a 35-day follow-up period after a visit. The Cleveland Clinic Institutional Review Board approved this study.

BHSW Intervention

Over 15 months (October 2017–January 2019), 10 BHSWs were hired to support 40 primary care practices (i.e., the staff-led practices within our large, integrated health system). Each BHSW supported internal and family medicine physicians in person at the BHSW’s primary site and via telephone at satellite sites. On average, each BHSW supported 45 physicians. The BHSWs worked full-time in this role and did not have additional responsibilities. Funding for BHSWs was through Cleveland Clinic’s Medicine Institute, but the BHSWs were supervised by a manager in the Behavioral Health Department of the Neurological Institute of Cleveland Clinic.

As BHSWs were hired, they were assigned to their preferred location. Often, the BHSWs chose their site according to where they lived. Hence, the order of BHSW deployment was not predetermined. Once a BHSW had been assigned, physicians within that group of sites (called “group”) could refer patients to the BHSW through a telephone encounter. In May 2018, an order set was implemented within the EHR to allow physicians to more easily refer patients to the BHSW. After receiving a referral through the EHR—or, less often, via warm handoff—the BHSW called the patient. The expectation was that BHSWs connected with a patient within 7 days. The BHSW administered a depression screener (i.e., the PHQ-9), asked about substance use and anxiety, and took a psychiatric history. The BHSW spoke with the patient for up to three times, offered community resources, or referred the patient to a behavioral health provider (i.e., a licensed independent social worker, psychiatrist, or psychologist). The resources available at each site (i.e., psychiatrist onsite) varied, which necessitated different approaches by each BHSW.

Independent Variable

In the stepped-wedge design, each site’s group served as its own internal control (pre- vs. postdeployment). We assigned the 40 primary practices to seven groups on the basis of when a BHSW started supporting the group of sites. (A figure in an online supplement to this article shows the time line.) The preintervention period lasted between 21 and 36 months.

Patients were assigned to the control or intervention group on the basis of the date of their initial diagnosis (the index date) of depression or anxiety during the study period. Patients were assigned to the intervention group if their condition was diagnosed after the BHSW started at the site, and patients whose index date was before the BHSW started were assigned to the control group. We excluded patients whose first diagnosis occurred during the month that the BHSW began working.

Diagnoses were based on ICD-10 codes assigned on the day of the visit (see table in online supplement). Patients who received a diagnosis of anxiety and depression on the same day were classified as having both.

Dependent Variables

Behavioral health services.

We collected dates of in-person and telephone visits with therapists (i.e., licensed independent social workers or psychologists) and in-person visits with psychiatrists from the EHR. We collected information on both telephone and in-person visits because cognitive-behavioral therapy is similarly effective when delivered by either modality (20). We determined whether a visit occurred within 30 days of the index visit, as a measure of the timeliness of care. We also assessed whether a patient had at least three visits with a therapist or psychiatrist within 90 days of his or her index date. We did not include visits to the BHSW as an outcome.

Composite treatment outcome.

We categorized patients’ treatment within 30 days into three groups: no treatment, medication only, and behavioral health visit. Patients included in the behavioral health visit group had a visit with a therapist or psychiatrist and may or may not have been prescribed a medication.

Medications.

Although the BHSWs were tasked with linking patients to therapy, we examined their impact on pharmacotherapy because many patients with newly diagnosed depression or anxiety initiate treatment with a medication, especially if behavioral therapy is not readily available (2, 21). From the EHR, we identified all psychotropic medications prescribed within 30 days of the index visit.

Confounders.

We collected information from the EHR on age (in years), sex (male or female), race (Black, White, or other), marital status (married or domestic partner vs. other), and health insurance (private, Medicare, Medicaid, or other).

Analysis

We described the change in the percentage of patients who received behavioral health treatment before or after BHSW deployment, stratified by diagnosis group (depression, anxiety, or both). We used multilevel logistic regression models to identify the association between BHSW deployment and receipt of treatment. Separately, we modeled the adjusted probability of having a visit with a therapist, having a visit with a psychiatrist, having a behavioral health visit, and receiving a prescription for pharmacotherapy. The regression models were adjusted for age, sex, race, insurance, and marital status as fixed effects. Intervention status (pre- versus postdeployment) was also included as a fixed effect. The site’s group and time were both included as random effects in the model (22). To understand whether a change in visits was due to deployment of the BHSWs or some other secular change in the health system, we plotted the adjusted probability of each event across time for each group. All analyses were conducted with Stata 14.0.

Secondary analysis.

We used the same analytic approach as in the main analysis to determine whether patients had higher odds of having at least three visits within 90 days of a diagnosis after BHSW deployment.

Sensitivity analysis.

We were concerned that we might miss behavioral health visits to therapists outside our health system. Hence, in a sensitivity analysis, we restricted the population to individuals insured by our employee health plan (EHP), because such individuals have strong financial incentives to stay within Cleveland Clinic’s health system.

Results

The study population included 68,659 persons who had at least one primary care visit between 2016 and 2019 and a diagnosis of depression (N=14,237, 21%), anxiety (N=33,377, 49%), or both (N=21,045, 31%). Patients seen before BHSW deployments were on average slightly older (49 vs. 48 years, p<0.01) and more likely to be female (66% vs. 64%, p<0.01) and Black (12% vs. 10%, p<0.01), compared with patients seen after BHSW deployment (Table 1). On average, 4,179 patients were diagnosed as having anxiety or depression each quarter during the predeployment period versus 4,755 during the postdeployment period (p=0.08).

Behavioral Health Visits

Visits with a therapist increased from 2% in the predeployment period to 11% postdeployment for patients with depression (p<0.01) and from 2% to 10% (p<0.01) for patients with anxiety (Table 2). The adjusted odds of having a visit with a therapist increased in the postdeployment period for patients with depression (adjusted odds ratio [AOR]=4.97) and anxiety (AOR=5.00). The percentage of patients with a psychiatrist visit increased from 1% to 2% for patients with depression (p<0.01). In the regression model, the AOR for having a psychiatry visit was 1.82 for patients with depression and 1.58 for patients with anxiety (Table 2).

In the postdeployment period, the percentage of patients with at least one behavioral health visit increased by approximately 10 percentage points (p<0.01 for all comparisons) (Figure 1). For patients with depression only, the proportion of those with a visit increased from 4% predeployment to 14% postdeployment (p<0.01) and for patients with anxiety, from 3% to 12% (p<0.01). The likelihood of having any behavioral health visit (therapist or psychiatrist) was significantly higher in the postdeployment period for patients with depression (AOR=4.35) or anxiety (AOR=4.27) (Table 2).

Trends in Behavioral Health Care

Figure 2 shows the adjusted probability of having a visit with a therapist or a psychiatrist by group over time. In the BHSW postdeployment period, therapist visits clearly increased, and most groups’ adjusted probability of having a visit significantly increased after BHSW deployment (p<0.01). The exception was the group of sites that transitioned in the third quarter of 2018 (group 3), for which the likelihood of a therapist visit remained relatively flat for the entire study period. As Figure 2 also shows, the increase in the adjusted probability of having a visit with a psychiatrist was overall slight and substantially varied across groups.

Medications

The adjusted odds of receiving a medication among patients with depression indicated no changes after BHSW deployment (Table 2). For patients with anxiety, medication prescriptions overall decreased by 4 percentage points in the postdeployment period. This change was driven by a decrease in prescribing benzodiazepines (predeployment, 19%, N=3,207; postdeployment, 13%, N=2,089; p<0.01). Patients with anxiety had significantly reduced odds of receiving a prescription in the postdeployment period (AOR=0.80). Similarly, patients with both anxiety and depression had lower odds of receiving a prescription in the regression model (AOR=0.89).

Secondary Analysis

We also calculated the percentage of patients who received at least three behavioral health visits within 90 days. Before BHSW deployment, 1% (N=504 of 38,022) of patients had at least three visits within 90 days of the index visit. After deployment, 3% of patients (N=1,000 of 30,637) had at least three visits (p<0.01). The likelihood of having at least three behavioral health visits within 90 days was significantly higher in the postdeployment period (AOR=2.56, 95% confidence interval=2.13–3.07).

Sensitivity Analysis

When we restricted our analysis to individuals with EHP insurance (N=6,785), our results were similar to the results of the main analysis (Table 3). Among patients with depression in this group, 17% received a behavioral health visit in the 30 days after diagnosis, versus 6% before BHSW deployment. For patients with depression, the likelihood of receiving any behavioral health treatment was higher after the deployment (AOR=3.84).

Discussion

We found that connections to behavioral health care increased substantially after the deployment of BHSWs in primary care. The odds of having at least one visit with a behavioral health specialist in the 30 days after diagnosis increased more than threefold postdeployment for patients with depression, anxiety, or both. This increase was primarily due to visits with therapists. Of note, the absolute increase in behavioral health visits was approximately 10 percentage points, and the absolute increase in psychiatry visits was smaller (1 percentage point). Findings from our study support the notion that BHSWs can connect patients to therapists and psychiatrists even when they are based in separate locations.

Although patients with depression can be treated with pharmacotherapy, linking patients to behavioral therapy is important because patients often prefer it to pharmacotherapy (23). In our study, 13% of patients with a diagnosis of depression attended a behavioral health visit. In comparison, 4% of patients with depression attended a behavioral health visit after the implementation of a quality improvement project at 16 primary care practices (21). The project included training sessions and a Web-based application to screen, monitor, and guide treatment decisions (21). Patients with mild depression should be initially offered therapy plus symptom monitoring, and patients with moderate depression should be offered therapy, pharmacotherapy, or both (24). For patients who require both, there is no order effect on eventual outcomes (25), and initial treatment with behavioral therapy, which has no known adverse effects, may be therefore preferable.

Incidentally, we found that prescriptions for benzodiazepines decreased after BHSW deployment. This decrease may be linked to benzodiazepines’ association with opioid overdose deaths (26) and physicians’ changing prescribing habits (27) in response to the public’s perception that these habits were partly responsible for the opioid epidemic (28). Benzodiazepines are safe when prescribed for short periods, but dependence can develop after 4 weeks for half of benzodiazepines users (29). For patients with a new diagnosis of anxiety, physicians may have been reluctant to withhold pharmacotherapy if they did not have an alternative. Given that behavioral health visits increased after BHSW deployment, their availability to link patients to treatment may have encouraged physicians not to prescribe medications.

One of our study’s strengths was the inclusion of both depression and anxiety. Both disorders are common concerns (30) and frequently co-occur (31). A study of primary care patients in Germany found that patients with both depression and anxiety were more likely to receive behavioral health treatment (OR=5.16), compared with patients with either disease (32). Similarly, we found that the proportion of patients who received behavioral health treatment was larger among those who had both diagnoses than among those with only one. We also found that a slightly higher percentage of patients with only depression were linked to treatment, compared with patients with only anxiety. Ideally, primary care patients with behavioral health needs, such as anxiety, will also benefit from systematic supports linking patients to treatment. Thus, quality of care can be improved for patients beyond those with a specific disease for which a quality measure has been incorporated into value-based payment programs.

To meet quality measures within pay-for-quality programs, practices are increasingly screening their patients for depression and are identifying new cases of this disorder (33). After diagnosis, patients with depression should be linked to care to prevent unnecessary emergency department and hospital visits (34), which are key drivers of cost for accountable care organizations (ACOs) (35). A study found that incentivizing behavioral health care through ACO payments did not meaningfully improve receipt of care, in part because of limited provider capacity (36). The findings of our study suggest that health systems and ACOs should consider employing BHSWs to increase short-term capacity and link patients to appropriate care. Virtual visits (including via telephone) can increase capacity, because a clinician is not restricted to a specific location or usual office hours. Instead, a behavioral health specialist can meet with patients anywhere and may choose to offer evening or weekend hours. Increased care access outside routine business hours may be important, given that patients with financial instability are more likely to miss behavioral health appointments (37). Further, a meta-analysis found that interventions led by a remote therapist are as effective as in-person ones (38), and clinicians and patients’ ability to use remote interventions has been expanded during the COVID-19 pandemic (39).

This observational study had several limitations. First, sites were not randomly assigned to receive the intervention. Unmeasured bias may have been present in the order and timing of the BHSW deployment. Importantly, the order in which each group transitioned was independent of the group’s eagerness to work with a BHSW. Second, depression and anxiety are relapsing and remitting disorders. Patients who received a diagnosis earlier than 2015 might have been included in the data set if they did not receive a diagnosis of, or medication for, depression or anxiety in 2015. However, our approach has been used previously (2). Further, patients who experience a relapse of depression that has been in remission for ≥1 year should be treated similarly to those who have received a new diagnosis. Third, it is possible that the introduction of the BHSWs led to a change in physicians’ pattern of identifying and diagnosing depression or anxiety. Fourth, patients might have sought mental health care outside our integrated health system. Our sensitivity analyses of patients likely to stay within our health system found a similar increase in the adjusted odds of treatment after the deployment of the BHSW, suggesting that our findings are likely reliable and that more patients received treatment in the first month after diagnosis. Finally, we know anecdotally that the BHSWs varied in their approach to linking patients to care on the basis of patients’ needs and location-specific resources. Although these differences in approach decreased the consistency of the intervention across practice sites, these varied approaches likely increased the external generalizability of our study.

Conclusions

The results of our large, retrospective, stepped-wedge study indicate that deploying BHSWs in primary care practices increased linkages to behavioral health care for patients with depression and anxiety. Other health systems that have advanced EHR systems should consider introducing social workers trained in behavioral health to triage patients receiving new diagnoses of depression, anxiety, or both and systematically link them to appropriate care.