Depression Treatment Uptake in Integrated Primary Care: How a “Warm Handoff” and Other Factors Affect Decision Making by Latinos

Latinos experience similar or higher rates of depression compared with non-Latino whites but are less likely to engage in treatment (1–5). Low utilization of mental health treatment among Latinos has been attributed to stigma related to mental illness and treatment, language barriers, time burden of treatment, low health literacy, and lack of insurance (1–5). Data show that Latinos who seek treatment do so in primary care settings (5–7), and two studies of older Latinos found that integration of depression care into primary care resulted in better engagement and treatment outcomes (8,9). Integrated behavioral health care is central to the Patient Protection and Affordable Care Act of 2010 (10), owing to the cost-efficiencies associated with its use and its clinical effectiveness for treatment of depression, anxiety, and a variety of comorbid chronic general medical conditions in the general population (11–19).

The ability to engage patients in treatment is central to the success of integrated care. For Latinos, establishing a connection with a primary care provider (PCP) and receiving an in-person referral, or “warm handoff,” to a behavioral health provider may be particularly relevant to whether patients seek behavioral treatment. A warm handoff builds on the cultural component of personalismo—the emphasis on warm personal relationships in traditional Latino culture. A warm handoff is believed to reduce mental health stigma and improve the chances of early engagement in treatment (20). The purpose of this study was to investigate factors presumed to enhance engagement in behavioral health treatment for depression in primary care. The study examined the role of a warm handoff and other factors influencing the decision to follow through with referrals.

Methods

A two-phase, retrospective cohort study design was used. Phase 1 consisted of review of quantitative data from patients’ medical charts. Phase 2 consisted of review of qualitative data derived from interviews with a subsample of patients referred for depression treatment (21,22) in a federally qualified health center in Napa, California. The clinic provides integrated mental health care to over 25,000 patients annually, of whom 60% are low-income Latinos and 90% are publicly insured or uninsured (23).

Chart Review Procedure

Medical charts of Latino patients who were between the ages of 18 and 65, had scores of 9 or higher on the Patient Health Questionnaire–9 (PHQ-9 [24]), and were referred by a general medical provider to a behavioral health provider for treatment of depression between 2009 and 2011 were selected from the clinic’s electronic medical records system for review. Data were abstracted electronically, deidentified, and recorded in a Microsoft Excel spreadsheet by a research assistant who was blinded to the study’s aims (25). Variables included service-related and demographic predictors of whether patients followed up on the referral within two months (follow-through). Charts with missing data were excluded from the analysis.

Referral type (warm handoff or prescribed referral), a service-related predictor, was identified via referral codes notated in medical charts. All charts that included a code for referral to a behavioral health provider, as well as same-day notes from behavioral health services, were categorized as warm handoffs. Charts that included only a referral code but no same-day meeting notes from behavioral health services were categorized as cold handoffs. Other predictor variables collected from the medical charts included: patient’s gender match and racial-ethnic match with the PCP and the behavioral health clinician and whether the referring PCP was a patient’s regular PCP or a different provider.

Demographic predictors were primary language, sex, age, insurance status (insured or uninsured), level of copayment (high, medium, or low), and eligibility for a federal sliding-scale fee.

The outcome variable, follow-through, was defined as attending an initial behavioral health visit for depression within two months of referral.

Clinical characteristics were depression severity on the PHQ-9, the presence of comorbid anxiety, and a prescription for antidepressants.

Qualitative Interviews

Following analysis of data collected during the chart review, semistructured qualitative interviews were used to explore factors influencing the decision whether to follow through with referrals (26,27). Purposive sampling was used to recruit a subsample of 16 patients who were referred for depression treatment on the basis of a PHQ-9 score of 9 or higher (28). The sample included patients who did and did not attend an initial visit after receiving a referral for depression treatment. Participants were selected on the basis of a range of characteristics that appeared to play a critical role in the decision to follow through with behavioral treatment, according to the quantitative findings. These characteristics included primary language (English or Spanish), referral type (warm handoff or prescribed), gender, and age. A total of 37 patients consented to be interviewed, but ten could not be reached or did not return messages and 11 decided not to participate. Participants were offered a $20 gift card to Target for their time. Interviews lasted 60–90 minutes, were conducted by the first author, and were recorded and transcribed verbatim. Spanish-language interviews were translated by a professional translator and back-translated to ensure language equivalency. The interview data were coded by using DEDOOSE and were analyzed by the first author and a bilingual-bicultural research assistant by using categorical, thematic summaries from individual interviews. [A copy of the interview guide is available as an online supplement to this article.]

Data Analysis

Chart review.

Chi square and t tests assessed bivariate relationships between the key predictor (type of referral), other predictors, and the outcome variable. Only variables that were statistically significant (p<.05) in initial univariate analyses, or that significantly improved overall model fit, were included in the final multivariate logistic regression model, although key predictors of interest (referral type and primary language) were retained, given their theoretical importance in the literature. A stepwise procedure was used for model building. Model fit was assessed with goodness-of-fit tests, including Akaike information criterion and –2 log likelihood. Statistical interactions between four key predictors (referral type, primary language, gender or racial-ethnic match between patient and provider, and depression severity) were assessed in multivariate analyses that included main effects and the cross-product term considered significant (p<.05). Only significant interactions were included in the final model. All tests were two sided and were based on a p value of <.05. Analyses were conducted by using SPSS, version 18.

Qualitative data.

Inductive and deductive methods were used to analyze the qualitative data (29). Themes were identified by using line-by-line textual analysis and social science queries or by searching for textual data related to the research question (28). We created summary overviews for each participant so that we could view the narrative in a larger social context (30). We examined the relationship between several codes (evaluation of the referral experience, primary language, physician-patient relationship, and confidence in the clinic) among those who attended or did not attend an initial behavioral health visit. We examined how a positive, negative, or neutral experience during the referral may have differed between primarily Spanish and English speakers.

Results

Chart Review

Sample.

A total of 1,537 Latino patients with qualifying PHQ-9 scores visited the clinic during the study period. Of these, 470 (31%) were referred to behavioral health services for depression treatment; the charts of 39 (8%) patients who were referred for treatment were missing data for key predictors and were excluded. Given that the proportion of charts with missing data was less than 10% of the referred sample, their exclusion from the final sample (N=431) most likely introduced minimal bias (1). Mean age was 43.5 years, and females constituted nearly three-quarters of the sample. Spanish was the primary language of a majority of patients, and over half were uninsured. The mean PHQ-9 score was 15.64, indicating moderately severe depression. At the time of the general medical visit, 16% of the sample had a concurrent diagnosis of anxiety or panic disorder and over half had been prescribed antidepressants (Table 1).

Predictors of follow-through.

Approximately one-third of referrals were warm handoffs, and 56% of all referrals were made by the patients’ regular PCP. Data analysis revealed no main effect for hypothesized predictors of follow-through. Neither referral type, gender or racial-ethnic match between patients and PCPs or behavioral health providers, nor primary language was associated with treatment follow-through (Table 2). However, an interaction effect was found between referral type and primary language (95% confidence interval=1.29–11.98) (Table 3). Compared with Spanish-speaking Latinos, the odds of follow-through with behavioral health treatment were approximately 75% lower among English-speaking Latinos who received a warm handoff versus a prescribed referral. In other words, for every four English speakers with a cold handoff who attended a follow-up, only one English speaker with a warm handoff attended a follow-up. In contrast, for Spanish speakers, the odds ratio for a warm handoff was .27 × 3.93, or 1.06, so their odds of treatment follow-through were about even. That is, for every one Spanish speaker who got a cold handoff and attended follow-up, one Spanish speaker with a warm handoff attended a follow-up. The final model also included income levels (sliding-fee scale) because the inclusion of this variable greatly improved model fit (Table 3).

Qualitative Interviews

Table 4 describes the demographic characteristics of the 16 patients who participated in qualitative interviews. Most were female, and half spoke Spanish as their primary language. All participants were referred for depression treatment, but only half attended an initial behavioral health visit. Purposive sampling provided the opportunity to explore decisions about whether to follow through with treatment and to elucidate the results of the quantitative phase of the study.

Inductive and deductive qualitative analyses were used to expand upon quantitative results by searching for themes related to how the experience of referral and related factors may have influenced the decision to follow through with behavioral health treatment (2,3). Four primary themes were identified: illness narrative, which refers to participants’ understanding of the causes and treatment preferences for depression; sense of connection to the clinic, PCP, and referral experience itself; readiness to engage in treatment for depression; and everyday barriers, including poverty, scheduling issues, and adequate understanding of the services offered. [A table of participants’ comments pertaining to each theme is available in an online supplement to this article.]

Illness narrative.

Participants identified psychosocial stressors, such as socioeconomic problems (loss of a job or injury on the job [N=8]) and poverty (N=14) as key causes of depression. Poverty was a particularly prominent theme for men (N=3), whereas for women, stressors related to gender roles, particularly motherhood and child rearing and marital strife, were the most commonly mentioned causes of depression (N=13).

Preference for depression treatment was related to perceived cause of depression. Treatment that focused on finding a new job was preferred by two of three men, and treatment that focused on talking about problems was preferred by 11 of 13 women. Use of medication for depression was the preferred focus of treatment for seven participants. Participants (N=9) indicated that they would be more likely to engage in treatment if there was a match between their preference for treatment and the services offered.

Sense of connection.

Most participants reported positive feelings about the clinic (N=12), quality of their health care (N=11), and aspects of the clinic that they considered Latino friendly (N=7). English speakers expressed more frustration with the clinic and greater distrust, especially regarding confidentiality (N=3). Overall, the English speakers’ narratives reflected greater apprehension about the quality of care.

Several patients were unaware that they had a regular PCP (N=4), having seen a different medical provider at each visit (N=3). Others (N=7) described a very close, trusting relationship with a PCP, and these participants were more likely to attend their behavioral health visit (N=6). Spanish speakers were more likely than English speakers to report a close relationship with a PCP.

Participants’ experience of referral to behavioral health services varied greatly. Of the nine participants who received a warm handoff, five described experiences that fell far short of a discussion of depression treatment options followed by a personal introduction to a behavioral health provider prepared to offer a brief intervention and reassurance. Instead, they described the referral as a neutral or negative experience. Three participants who received a warm handoff from a medical assistant tended to express frustration or confusion about the process or felt disconnected from the process and unsure why their PCP didn’t address their concerns about depression.

Others (N=4) experienced the referral as reassuring—the PCP addressed their depression and clearly explained the behavioral health program and how it could help. For these participants, a warm handoff added a sense of comfort and enhanced their readiness to engage in treatment. Participants who described their referral experience as positive felt comfortable meeting the behavioral health provider, better understood depression, and—typically—received a same-day intervention leading to symptom relief or increased hopefulness about symptom remission. Regardless of referral type or primary language, a careful description of the behavioral health program by the PCP, well matched with the patients' perceived cause of depression, affected participants’ attitude toward and understanding of behavioral health treatment. Participants (N=7) who described feeling rushed during visits or felt that the recommendations did not address their complaints were less likely to follow through on treatment recommendations, regardless of referral type. English speakers were more likely to question the treatment model and express concern over the brief nature of behavioral health services.

Readiness.

Readiness was directly affected by the perceived severity or acuity of depressive symptoms. Several participants (N=3) reported having previously received referrals to behavioral health treatment, but they waited to follow through until treatment was absolutely necessary.

Participants (N=4) who reported having talked with supportive family or friends about their depression were more likely to attend a visit. Conversely, participants (N=3) who kept their depression private from family or friends were less likely to have attended their behavioral health visit.

Everyday barriers.

Poverty and low health literacy were barriers to attending a behavioral health appointment, particularly for Spanish speakers. For many participants (N=8), paying to attend a behavioral health visit posed a barrier. Notably, these participants did not endorse cost as a reason for ever missing general medical visits, which are more expensive than behavioral health visits. For other participants (N=7), forgetting the appointment, a misunderstanding about the appointment time, or a request by the clinic to reschedule the appointment resulted in lack of treatment follow-through.

Discussion

The integration of behavioral health services in primary care is a promising model for improving access to depression treatment among Latinos who experience access barriers and low treatment uptake. A warm handoff is often described as important for maintaining treatment engagement among minority populations (20,31). In this study, only half of Latinos referred to behavioral health services for depression attended an initial visit, and practices believed to enhance engagement did not appear to improve uptake. In fact, English-speaking Latinos were significantly less likely to engage in treatment if they received a warm handoff rather than a prescribed referral. In addition, other factors felt to be important for engagement—in particular, racial-ethnic or gender match with a PCP or a behavioral health provider, were unrelated to follow-through.

Qualitative findings illuminated contextual factors in the referral experience that influence treatment uptake. Patients were more likely to attend an initial behavioral health visit for depression if the prescribed treatment was in concert with their beliefs about the causes of depression. Addressing instrumental barriers to care, illness acuity and subsequent readiness for treatment, and family engagement may also enhance the referral process for Latinos. However, English-speaking Latinos seemed more skeptical of brief treatment models and expressed more distrust in the clinic. Overall, participants indicated that they would be more likely to attend an initial visit if their PCP was involved in the warm handoff. However, most participants experienced the warm handoff as rushed and confusing, particularly if the referral was made by a medical assistant.

The ideal of having a PCP explain behavioral services and make a referral is recommended by integrated care programs (31–35); however, pressure for productivity leads to time restrictions for providers. A lack of reimbursement for same-day visits (36) (warm handoffs) may lead to “task shifting” of referrals from the PCP to unlicensed, lower-cost staff. The qualitative findings from this study suggest that patients may find task shifting confusing. Recent studies have found that developing the patient-provider relationship and matching treatment with patient preference improve uptake and adherence among Latinos and other depressed populations (37,38). One potential way to enact these recommendations and reduce the likelihood for confusion would be to simply ask patients whether they are interested in counseling and whether they would like to be introduced to the behavioral health provider. This study suggests that English speakers and Spanish speakers may have different needs requiring tailored approaches to referral.

Addressing instrumental barriers to care through case management has also been shown to improve depression treatment uptake (16,39). In addition, treatment recommendations should be in line with patients’ beliefs about the causes for depression. For example, in this study, men were more likely to identify lack of employment or financial concerns as the cause of their depression. It is likely that a referral to behavioral health treatment that incorporates elements of this explanatory model—for example, meeting with someone who can address and engage in problem solving about employment needs rather than simply provide talk therapy—would be more effective in engaging such patients in care. Finally, follow-through was mitigated by low health literacy and by forgetting appointments, especially if they had been rescheduled. Recent research on the use of mobile reminder systems suggests that simple text messaging or voice mail reminders can overcome these barriers (40,41).

Findings should be considered with caution, given the reliance on medical records from one location and the small qualitative sample. In addition, of the 1,537 patients who screened positive for depression, less than a third were referred for treatment. Future studies should examine factors that affect who gets referred to behavioral health treatment and how providers decide whether to use a warm handoff or a prescribed referral when referring patients, although these topics were beyond the scope of this study. As highlighted by the qualitative findings, the construct validity of certain variables, such as the warm handoff, is limited, given that chart review could not account for the more nuanced characteristics related to the quality of the referral or whether a warm handoff was conducted by a general medical provider or a medical assistant. However, the qualitative follow-up allowed exploration beyond the limitations inherent in using cross-sectional medical records data and provided a foundation upon which future studies can build.

Conclusions

Our findings suggest that disparities in utilization of treatment among Latinos remain a problem, even when colocation of behavioral services removes structural barriers to follow-through on referrals for depression treatment. Qualitative findings from this study suggest that English and Spanish speakers may have different expectations of treatment; thus tailoring a warm handoff to an individual patient is critical. Improved referrals will involve building on the patient-provider relationship and eliciting the patient’s perceived causes of depression, carefully matching treatment recommendations with treatment preferences and addressing everyday barriers to care.

A lack of reimbursement for warm handoffs and the fast pace of primary care may lead providers to shift these critical tasks to unlicensed staff. Although the warm handoff has been touted as a best practice of integrated care, this study highlights a gap between ideal and real-world implementation of the practice. Given that a majority of interview participants did not receive a warm handoff and that many of those who received a warm handoff found it confusing, more research is needed to identify specific elements of the warm handoff that help engage patients in care, particularly among patients with varying levels of acculturation (English speakers versus Spanish speakers, for example).

Research on best practices for eliciting patient preferences and encouraging patient empowerment during the medical visit is gaining attention (42), and future research about how to incorporate these elements in integrated behavioral health settings in a time-efficient manner should be included in the health services research agenda. A randomized study incorporating these components in the referral process to determine their effect on treatment uptake is a necessary next step to better understand the effectiveness of the warm handoff and would help address the limitation of selection bias in this study.