INTRODUCTION

About 7% of adults in the USA experience major depression each year and lifetime risk is nearly 30%.1,2 Depression is associated with lower quality of life, worse health outcomes, and higher mortality,3,4,5,6,7,8,9,10,11 yet more than a third of people with depression do not receive treatment.2 Since depression is most often treated in primary care, it is important to have primary care providers (PCPs) who are prepared to diagnose and manage depression.12,13,14 Systematic screening can improve detection,15,16 but screening alone does not improve care or outcomes.17 Many PCPs lack confidence in their abilities to assess depression severity, offer counseling or behavioral interventions, and provide ongoing management.18,19,20,21,22

Integration of behavioral health (BH) and primary care is one key to improving depression care. The goal of integration is not only to facilitate access to behavioral health providers (BHPs) such as therapists, social workers, and psychiatrists, but also to increase the primary care team’s capacity to identify, assess, and treat BH conditions.23 Several studies have found that collaborating with BHPs increases PCPs’ own knowledge and confidence in managing depression and other BH conditions.24,25,26,27,28,29,30 However, one study found that PCPs’ perceived competence did not differ at integrated and non-integrated clinics, with integration defined as having BHPs onsite at least once per week.31

One major challenge to understanding the impact of BH integration is that it is a broad term encompassing multiple approaches to care and varying degrees of interaction between PCPs and BHPs.32,33 Prior studies have not explored how different components of integration impact the care PCPs provide for patients with depression. Community health centers (CHCs) are an important setting to study this question because they serve a large number of patients with depression—about one in 11 people in the USA receive care at CHCs—and the majority of CHCs have implemented some level of integrated BH.34,35,36 The aim of this study was to examine provider- and organization-level factors, including specific elements of integrated BH (e.g., screening, treatment protocols, access to BHPs, collaboration), associated with higher PCP confidence in identifying and managing depression at CHCs.

METHODS

Design

We partnered with Midwest Clinicians’ Network (MWCN), a nonprofit corporation that provides education, research, and networking opportunities, because of its wide reach (>120 CHCs in 10 states) and our history of successful collaborative research.37,38,39,40 In summer 2016, we sent a survey to BH directors at MWCN-affiliated CHCs. If a CHC did not have a BH director, the survey was sent to the medical director or another leader knowledgeable about the CHC’s BH services. In fall 2016, we contacted the CHC leaders who had completed the survey (“CHC survey”) and requested names and mailing addresses of PCPs who cared for adult patients to send a second survey (“PCP survey”). The study was determined to be exempt by the University of Chicago Institutional Review Board.

Measures

BH leaders and PCPs answered questions about the CHC environment, including depression screening practices, systems and support for depression care, availability of BH services, and integration between primary care and BH. PCPs also answered questions about their attitudes and experiences related to depression and BH. The conceptual model for the study was informed by social cognitive theory, which posits a reciprocal relationship between person, behavior, and environment and emphasizes the role of self-efficacy or confidence in determining performance (see Fig. 1).41,42

Figure 1
figure 1

Conceptual model

Full size image

To measure PCP confidence in diagnosing, assessing severity, providing counseling, and prescribing medication for depression, PCPs rated their agreement with statements (e.g., I am confident in my ability to assess depression severity; 1 = strongly disagree to 5 = strongly agree), adapted from the Perceived Self-Efficacy in Diagnosing and Treating Depression Scale.20,43 An overall confidence score was calculated as the sum of PCPs’ responses to the four items (range 4–20; Cronbach’s alpha=0.79).

Publicly available data on CHCs’ patient populations were downloaded from the Bureau of Primary Health Care Uniform Data System.44

Analysis

This cross-sectional study had a hierarchical multilevel structure: PCP (level 1), CHC (level 2), and state (level 3). We used basic descriptive statistics to summarize the CHC- and PCP-level data. We used a multilevel model via a linear mixed model (LMM) to model PCP confidence and test associations with predictors. State and CHC nested within state were considered random intercepts. We first conducted unadjusted univariate analyses by adding one predictor at a time into an LMM. We then conducted adjusted multivariate analyses of those covariates with p-value <0.10 in unadjusted analyses. We applied the backward model selection procedure to eliminate non-significant predictors in the multivariate LMM until all p-values were <0.10.

We also conducted missing data analyses. The primary analysis was based on the complete dataset, due to only 1.9% missing in PCP confidence and 0–11.9% missing in one of the covariates. There were 30.9% missing across PCP confidence and covariates. We imputed all missing data, using the multiple imputation method, via the Markov chain Monte Carlo algorithm. We used all covariates for the multiple imputations and generated 10 imputed datasets, which were used to calculate a mean dataset, where all missing values were filled with mean values. SAS version 9.4 software was used for analyses.

RESULTS

Four of the 132 MWCN-affiliated CHCs were deemed ineligible: three offered limited primary care services and mail for one was undeliverable. Of the remaining CHCs (N=128), 77 returned completed surveys (60% response rate, see Supplemental Figure 1) and 73 provided contact information for PCPs. Of 1039 eligible PCPs, 538 from 71 CHCs returned surveys (55% adjusted response rate, see Supplemental Figure 2). States with the most responding CHCs were Illinois (N=15), Michigan (N=14), and Ohio (N=13). On average, 26% of patients at CHCs were Black/African American, 16% were Hispanic/Latino, and 68% had an income below the federal poverty level. Most PCP respondents were female (71%), were White (73%), and had been working at their CHC an average of 5.8 years (Table 1).

Table 1 Characteristics of Community Health Center (N=71) and Primary Care Provider (N=538) Survey Respondents
Full size table

The overall mean confidence score was 15.25±2.36. Most PCPs reported they were confident in diagnosing depression (93% agree or strongly agree (N=497), mean 4.17±0.59), assessing severity (74% (N=394), 3.90±0.78), and prescribing medication (87% (N=463), 4.03±0.68). Less than half were confident in their ability to provide counseling for depression (37% (N=199), 3.15±0.98). Among all PCPs, 39% (N=200) were interested in more training on medication management of depression and 29% (N=151) were interested in more training on counseling; among those who did not express confidence in these areas, 64% (N=43) desired training in medication management but only 33% (N=107) desired training in counseling. Notably, unlike depression counseling, the majority of PCPs were confident providing health behavior counseling (68% (N=362)).

Descriptive statistics for CHC- and PCP-level factors are summarized in Table 2. Nearly all PCPs (98%, N=526) reported that their CHCs had a depression screening protocol, with all CHCs using the PHQ-2 and/or PHQ-9.36,45,46 Other systems, tools, and policies to support depression care were less common. While more than half of PCPs (58%, N=306) indicated their CHCs had a sufficient number of non-psychiatrist BHPs, few (13%, N=70) thought they had a sufficient number of psychiatrists. The majority of PCPs always had access to BH treatment plans (71%, N=367), sometimes made joint decisions with BHPs (68%, N=350), but never co-signed treatment plans (65%, N=335). PCPs agreed that BH care was important but only half (53%, N=281) thought their patients would be willing to see BHPs.

Table 2 Community Health Center Environment, Primary Care Provider Experiences, and Primary Care Provider Attitudes Related to Depression and Behavioral Health Care
Full size table

Modeling results for overall confidence are shown in Table 3. Results for confidence in diagnosing, assessing, prescribing, and counseling are shown in Supplemental Table 1.

Table 3 Community Health Center and Primary Care Provider Factors Associated with Provider’s Self-Rated Confidence in Managing Depression
Full size table

PCP Characteristics

Compared to physicians, physician assistants and advanced practice nurses had lower confidence overall (−0.72, p=0.047 and −0.69, p=0.01, respectively). Overall confidence decreased with age (−0.02 per year, p=0.03), as did confidence providing counseling (−0.01, p=0.03). PCPs who cared for more patients with depression were more confident overall (0.29 per 1-level category increase, p=0.003) and in diagnosing (0.05, p=0.02), assessing severity (0.10, p=0.001), and prescribing medication (0.06, p=0.01).

Depression Screening

PCPs’ satisfaction with the accuracy of depression screening was associated with greater confidence overall (0.38 per 1-point higher rating of accuracy, p=0.01) and in diagnosing (0.12, p=0.002), assessing severity (0.13, p=0.01), and prescribing medication (0.08, p=0.04). Satisfaction with the efficiency of screening was associated with confidence in providing counseling (0.12 per 1-point higher rating of efficiency, p=0.046). At CHCs where PCPs performed depression screening, PCPs were more confident in their ability to provide counseling (0.25, p=0.03), compared to CHCs where PCPs were not responsible for screening.

Systems, Tools, and Policies for Depression Care

At CHCs with depression tracking systems, PCPs were more confident overall (0.48, p=0.045) and prescribing medication (0.14, p=0.03), compared to CHCs without tracking systems. Having stepped care protocols was associated with greater confidence assessing severity (0.18, p=0.01) and providing counseling (0.27, p=0.01). In final models, confidence was not associated with automated reminders to screen or assess severity, algorithms for antidepressant initiation or titration, depression treatment order sets, or required BH training.

Availability of BHPs

PCPs who agreed that their CHC had a sufficient number of psychiatrists were more confident diagnosing depression (0.20, p=0.02) and assessing severity (0.24, p=0.03), compared to PCPs who did not agree. PCPs who had an easier time doing warm hand-offs were more confident diagnosing depression (0.08 per 1-level increase in rating of ease, p=0.01). Confidence was not significantly higher among PCPs who reported their CHC had a sufficient number of non-psychiatrist BHPs.

Regular Collaboration with BHPs

PCPs who always had access to patients’ BH treatment plans were more confident overall in managing depression (1.59, p=0.002), compared to those who did not have access. PCPs who sometimes or always made joint decisions on treatment plans with BHPs were more confident prescribing medication for depression (0.47, p<0.001 and 0.38, p=0.01, respectively), compared to those who did not make joint decisions. Sharing offices and co-signing treatment plans with BHPs were not significantly associated with confidence in the final model.

Attitudes and Expectations About BH Care

PCPs’ attitudes and expectations about BH care were not significantly associated with confidence in managing depression.

Patient Population

PCPs working at CHCs with a higher percentage of patients living below poverty were less confident overall, 1.53 points lower per 100% increase in patients below poverty (p=0.03). In particular, those working at CHCs with higher poverty were less confident prescribing medication (−0.66, p<0.001). PCPs working at CHCs with a higher percentage of Black/African American patients were less confident diagnosing depression, 0.20 points lower per 100% increase in Black/African American patients (p=0.03).

Missing Data Analysis

No missing patterns were found in the data. Modeling results for overall confidence using imputed data are shown in Supplemental Table 2. Results using complete and imputed data were consistent although models with imputed data had more significant covariates.

DISCUSSION

We surveyed >500 PCPs and found that most were confident in diagnosing, assessing, and prescribing medication for depression, but less than half were confident in providing depression counseling. PCPs’ confidence was positively associated with having access to BH treatment plans, a system for tracking patients with depression, an accurate and efficient screening protocol that involved the PCP, stepped care protocols, and a sufficient number of psychiatrists on staff. PCPs’ confidence was lower at CHCs with a greater proportion of patients living below poverty, and PCPs’ confidence diagnosing depression was lower at CHCs with a greater proportion of Black/African American patients.

These findings suggest that system-level efforts to integrate primary care and BH can bolster PCPs’ self-efficacy in caring for patients with depression. In integrated practices, BHPs often provide formal and informal training to PCPs and staff.23 PCPs gain knowledge and confidence from the experience of regularly collaborating with BHPs and seeing patients’ BH treatment plans. Conversely, we found that simply requiring PCPs to have training in BH was not associated with confidence. This is consistent with prior research showing that training alone does not improve depression detection, treatment, or outcomes; it must be combined with organizational changes.17,47,48,49

Our findings also align with previous research emphasizing the importance of a systematic approach to screening. Studies have shown that, although PCPs express confidence in their ability to diagnose depression and prefer to rely on clinical intuition over standardized tools, in real-world practice, they do not accurately identify cases.22,48,50,51,52,53 In our study, an accurate and efficient protocol was associated with greater PCP confidence in identifying and treating depression. In addition, at CHCs where PCPs were responsible for screening, PCPs were more confident providing counseling for depression. Given that studies have found higher screening rates when screening is conducted by staff rather than PCPs, a protocol that relies on other team members and involves the PCP may be most helpful.54,55,56

Interestingly, we found that confidence was higher when PCPs felt their CHC had a sufficient number of psychiatrists, but was not related to their perception of the number of other BHPs. This distinction might be due to psychiatrists’ status as physicians, their expertise in psychotropic medication, the relative rarity of psychiatrists, or some other factors. Difficulty recruiting and retaining psychiatrists has been one of the most commonly cited barriers to BH integration in CHCs.36,57 CHCs are often located in areas without psychiatrists, and resource-limited CHCs may favor hiring other BHPs with lower salaries.58,59 However, psychiatric consultation is a key component of the collaborative care model, which is the integrated BH model with the most robust research support.60

Similar to earlier studies, we found that PCPs’ confidence in providing counseling for depression was lower than their confidence in diagnosing and prescribing medication.22,52,61 Brief evidence-based behavioral and psychosocial interventions for depression can be delivered by PCPs.62,63,64 The skills and strategies involved have considerable overlap with health behavior counseling, which PCPs expressed greater confidence doing. Younger PCPs were more confident in providing counseling for depression, which may reflect changes over time in health professionals’ attitudes toward mental illness, public acceptance of mental health treatment, and efforts to integrate BH into primary care training.65,66,67

We found that PCPs were less confident overall, and in particular less confident prescribing medication, at CHCs with more patients living below poverty. PCPs may feel that treatment will have limited effect on outcomes if patients have substantial unmet social needs, or that socioeconomic or insurance barriers will prevent patients from engaging in treatment. Additionally, we found that PCPs were less confident diagnosing depression at CHCs with more Black/African American patients. This aligns with prior research showing disproportionate under- and misdiagnosis of depression among Black/African American patients.68,69,70,71

An important caveat is that we measured PCP confidence, which may not reflect actual knowledge, skill, or behavior—though prior research has found that greater confidence is associated with more guideline-adherent care for depression.72 While we did not assess quality of care or patient outcomes, previous studies have found that screening, stepped care, and patient tracking improve depression outcomes,73,74,75,76,77,78,79 and the number of BHPs in a clinic is associated with degree of integration.35 Effects of access to BH treatment plans are under investigation.80 Factors that we did not find to be associated with PCP confidence could still affect quality of care and patient outcomes. For example, studies have shown that reminders increase screening and treatment rates81,82,83 and that use of treatment algorithms is associated with greater likelihood of receiving adequate care for depression.84,85 A combination of components and effective implementation are likely key.86

There are a few additional limitations to this study. Findings may not be generalizable to all primary care settings given our focus on midwestern CHCs and the predominantly White female respondents. It is possible that CHCs and PCPs with a greater interest in BH may have been more likely to respond to the survey. The cross-sectional design precludes causal inferences, and results reflect PCPs’ comfort managing depression when the survey was conducted in 2016. We likely did not consider all factors related to confidence. In particular, we did not assess PCPs’ beliefs in regard to what their role in depression care should be or mental health stigma.

The results of this study may be helpful for CHCs and other health care organizations interested in improving depression care. Having clinic-level systems in place for depression screening, treatment, and follow-up may be instrumental to increasing PCPs’ confidence in managing depression. Tools like registries and treatment protocols support the PCP in proactively managing patients and ensuring they receive adequate treatment.87,88 Shared access to BH treatment plans is also helpful for PCPs and could be made the default in a clinic’s EHR; only psychotherapy notes need to be kept private according to HIPAA rules.89 In regard to personnel, organizations may want to consider the value of access to psychiatrists for both patients and PCPs. Psychiatric nurse practitioners could also play a role in integrated BH.90 When on-site psychiatry is not feasible, telemedicine-based collaborative care for depression has been shown to be effective in the CHC setting.91 Moreover, collaborative care has been shown to improve treatment engagement among underserved racial/ethnic populations.92

Policy changes could also support depression care. Current depression quality measures focus on screening and remission 1 year after a positive screen.44 Evaluating intermediary processes and outcomes (e.g., follow-up assessments, dose and duration of medication and/or therapy, significant symptom reduction) could be useful. Increasing reimbursement rates, establishing CHCs as training sites, and offering financial incentives and mentorship for those employed in underserved areas could encourage more psychiatrists to work in CHCs.93,94,95 These changes are important for promoting health equity, given the shortage of BHPs is more severe in areas with higher concentrations of racial/ethnic minorities.96 Additional efforts are needed within and beyond the health care system to address economic and racial/ethnic disparities in depression prevalence, care, and outcomes.93,97,98 Future research should examine why confidence was lower among advanced practice nurses and physician assistants, and assess PCPs’ confidence in managing other BH conditions.

In conclusion, low PCP self-efficacy is a barrier to effective depression care. Primary care clinics should be encouraged and incentivized to implement systems and tools that support PCPs and facilitate depression care. Integrated BH may be an important strategy to increase PCPs’ confidence in caring for patients with depression. Additional efforts to address socioeconomic barriers and structural racism are needed to improve depression care and reduce disparities among vulnerable populations.