Treatment of Depression and PTSD in Primary Care Clinics Serving Uninsured Low-Income Mostly Latina/o Immigrants: A Naturalistic Prospective Evaluation

Abstract

Objectives:

Uninsured immigrants to the US are psychologically vulnerable due to limited access to mental health services. Latina/o immigrants from Central and South America are further at risk due to high levels of trauma exposure, both in their country of origin and adopted country. Effective behavioral interventions in primary care are needed to address this services gap for common trauma-related mental disorders, including depression and posttraumatic stress disorder (PTSD). A naturalistic prospective study compared depression and PTSD outcomes for uninsured primary care patients, mostly Latina/o immigrants, in clinics with distinct models for integrating behavioral health services. One clinic had a collaborative care program, i.e., a multi-component, system-level intervention with the goals of facilitating increased screening, improving accuracy, increasing uptake of evidence-based treatment, and utilizing measurement-based treatment to target. The other had co-located services, with an on-site therapist.

Method:

One hundred thirty-eight patients with presumptive depression were interviewed at baseline. Follow-up data were collected 8-months later. Care received between the two assessments was ascertained by chart review.

Results:

There was a significant decrease in depression symptoms regardless of clinic. Factors associated with improved depression outcomes included absence of comorbid PTSD and lack of additional trauma exposures. Although there was overall improvement in PTSD, there was significantly more improvement in the collaborative care clinic.

Conclusions:

Results suggest that both models were effective for depression and PTSD; however, the collaborative care model had added benefits for those patients with PTSD.

There is increasing awareness of the need for effective and efficient treatment strategies for common mental disorders (e.g., depression and anxiety) in primary care settings. This is particularly true for low-income, uninsured minority populations that experience significant mental health care disparities (Ault-Brutus & Alegria, 2016; Cook, McGuire, & Miranda, 2007; Cook, Trinh, Li, Hou, & Progovac, 2017; Safran et al., 2009). Census projections suggest that by 2060, the number of Latina/os in the United States will grow to approximately 119 million, representing 29% of the US population (Colby & Ortman, 2014). Central and South Americans are the third and fourth largest groups of Latina/os in the US (López & Patten, 2015).

Immigrants from Central and South America, while understudied, may warrant particular focus given their high level of exposure to traumatic experiences (Fortuna, Porche, & Alegria, 2008; Kaltman, Hurtado de Mendoza, Gonzales, Serrano, & Guarnaccia, 2011). Trauma exposure is a significant risk factor for common mental disorders, including depression and posttraumatic stress disorder (PTSD) (Hovens et al., 2009; Schell & Marshall, 2008). Further, trauma and PTSD predict medical morbidity (Burg et al., 2016; Schnurr, Wachen, Green, & Kaltman, 2014; Wolf & Schnurr, 2016), and the sequelae of trauma exposure are costly to health care systems (Eekhout, Geuze, & Vermetten, 2016; Walker et al., 2003). PTSD and depression often co-occur; when they do, treatment of depression is more complicated and may take longer to achieve remission (Campbell et al., 2007; Green et al., 2006; Hegel et al., 2005).

The research on the mental health of Latina/os in the US, specifically the prevalence of depression and posttraumatic stress disorder, has yielded inconsistent findings in comparisons with non-Latino Whites, with studies alternatively finding higher, lower, and equivalent prevalence of disorder (Alegria et al., 2013; González, Tarrag, Whitfield, & Vega, 2010; Jimenez, Alegria, Chen, Chan, & Laderman, 2010; Pole, Best, Metzler, & Marmar, 2005; Roberts, Gilman, Breslau, Breslau, & Koenen, 2011). This inconsistency may be in part derived from the fact that Latina/os are not a singular group, but rather are from diverse countries with different contextual factors and influences (Alegria, Mulvaney-Day, et al., 2007). Complexity is also derived from length of time spent in the US (Alegria, Mulvaney-Day, et al., 2007; Alegria, Sribney, Woo, Torres, & Guarnaccia, 2007). Other factors that have been found to influence the mental health of Latina/os in the US include: family support and conflict; exposure to discrimination, and perceived neighborhood safety (Alegria, Shrout, et al., 2007; Cook, Alegria, Lin, & Guo, 2009).

One finding that is clear is that Latina/os in the US experience significant mental health disparities related to access to care and mental health outcomes (Blanco et al., 2007; Cook, Manning, & Alegria, 2013; Cook et al., 2007; Jimenez, Cook, Bartels, & Alegria, 2013; McGuire & Miranda, 2008; USDHHS, 2001). Latina/os are more likely to be uninsured, compared to non-Latino whites as well as other minority groups (Alegria et al., 2006: DeNavas-Walt, Proctor, and Lee, 2005), and those with a mental disorder are less likely to utilize mental health services. When they do seek care, Latina/os often delay treatment and are more likely to receive inadequate and lower quality care, compared to non-Latino whites (Alegria et al., 2002; Young, Klap, Sherbourne, & Wells, 2001).

Research suggests that Latina/os in the U.S. often prefer to seek mental health care for common trauma-related mental disorders in primary care clinics (Eisenman et al., 2008; Kaltman, Hurtado de Mendoza, Gonzales, & Serrano, 2014). There are a number of existing models for integrating mental health care into the primary care setting (Collins, Hewson, Munger, & Wade, 2010; NCCBH, 2003). One specific model with a strong evidence-based is the collaborative care model (Katon et al., 1997; Unützer et al., 2002). Collaborative care is a multi-component, system-level intervention with the goals of facilitating increased screening and accurate diagnosis, and increased uptake of evidence-based treatment (medications, psychotherapy or both) (Bower, Gilbody, Richards, Fletcher, & Sutton, 2006; Katon et al., 1995). In this model, a patient-centered team includes a care manager, often a nurse or social worker, who facilitates communication between the patient, primary care provider, and a consulting psychiatrist. The primary care provider assumes responsibility for the patient’s mental health care, the care manager provides supportive services (e.g., assessment, care coordination, regular follow-up contact with the patient, psychoeducation), and a psychiatrist provides consultation to the team via regular meetings with the care manager, rarely providing direct service to patients.

Multiple reviews of collaborative care for depression have shown that this specific care delivery model is effective in improving depression symptoms, increasing adherence to care, and increasing patient satisfaction (Gerrity, 2016; Gilbody, Bower, Fletcher, Richards, & Sutton, 2006; Gilbody, Whitty, Grimshaw, & Thomas, 2003; Neumeyer-Gromen, Lampert, Stark, & Kallischnigg, 2004; Sighinolfi et al., 2014; Thota et al., 2012). This is likely due to a number of important characteristics of the collaborative care model, including a patient-centered care team, focus on population-based care, measurement-based treatment to target, evidence-based care, and accountable care (AIMS Center, 2013). Collaborative care is effective in reducing PTSD symptoms as well (Fortney et al., 2015; Meredith et al., 2016).

The collaborative care model holds particular promise for the treatment of uninsured or under-insured populations that do not have access to specialty mental health care. Toward this end, a network of community clinics in greater Washington, DC metropolitan area began an implementation of collaborative care in 2005. The network of clinics serves an uninsured population, many of whom are under-resourced immigrants from Central and South America. A number of adaptations to the collaborative care model were necessitated by characteristics of the target patient population (Kaltman, Pauk, & Alter, 2011). Because of the high level of trauma exposure in the clinic populations, the model included an assessment of trauma exposure as well as multiple disease targets, including depression, PTSD, and generalized anxiety disorder. In addition to the care manager who performs many essential functions in a traditional collaborative care program, a family support worker was added to the intervention team to help address patients’ basic needs related to housing, employment, and legal issues, and to increase patient engagement.

The current study was a naturalistic prospective study of this adapted collaborative care program in a primary care clinic that serves primarily poor and uninsured immigrants from Central and South America. The treatment received and mental health outcomes (depression, PTSD) of patients in the adapted collaborative care were compared to those of patients in another network clinic that served a similar patient population and had an on-site therapist (75% FTE) to provide mental health services. While several randomized controlled trials have demonstrated the effectiveness of psychotherapy for depression in primary care (Linde et al., 2015; Schulberg, Raue, & Rollman, 2002), we are unaware of any studies that have examined the real-world effectiveness of having a therapist as part of the clinic staff as a model of care for meeting the mental health needs of the clinic population. It was unclear whether employing a therapist in the absence of the added structure of a psychotherapy trial (e.g., diagnostic rigor, use of manualized treatments) would yield the same results.

The current study was originally designed to compare a clinic with the collaborative care model to a clinic without integrated behavioral health care. At the time of study implementation, however, each clinic in the network had some form of behavioral health treatment available. The clinic with an on-site therapist was chosen as the comparison clinic because of the sharp distinction between the two models of care integration and the similarity between the two patient populations. The clinic with an on-site therapist provided psychotherapy to as many patients as time allowed, using the therapist’s best judgment as to what psychotherapy strategies to employ. In contrast, the collaborative care clinic had a focus on serving the clinic’s population of patients with depression and/or PTSD. This was done through screening and diagnostic assessment, care management, use of evidence-based interventions, including pharmacotherapy algorithms and behavioral activation (e.g., Dimidjian et al., 2006; Jakupcak et al., 2010), and frequent symptom assessment to ensure treatment to target.

The study was intended to address the following research questions: (1) were patients in the adapted collaborative care program more likely to achieve improved depression and PTSD outcomes than patients in the clinic with an on-site therapist?; (2) were patients in the adapted collaborative care program more likely to receive satisfactory care than patients in the clinic with an on-site therapist?; and (3) what additional factors were associated with improved depression and PTSD at both clinic sites? We hypothesized that patients in the collaborative care program would be both more likely to experience improved depression and PTSD outcomes and to receive satisfactory care as compared to patients in the clinic with an on-site therapist.

Method

Study Design

This study identified patients with presumptive depression in two primary care clinics that serve uninsured patients with the two models of service delivery described above. Study participants were assessed at baseline and then again at approximately eight months. A comprehensive chart review identified treatment that occurred in the interval between the two assessments.

Setting

The study was conducted in partnership with two primary care clinics in the same clinic system that serves low-income, uninsured communities in the greater Washington, DC metropolitan area. The majority of patients served by the two clinics were monolingual Spanish speaking immigrants. The collaborative care clinic served approximately 2700 unduplicated patients per year and the on-site therapist clinic served approximately1200 patients.

Participants

Participants were a convenience sample of patients receiving primary care, between the ages of 18-70, who met presumptive criteria for depression based on having a Patient Health Questionnaire-9 (PHQ-9) (Kroenke, Spitzer, & Williams, 2001) score ≥10, the cut-off for a moderate level of symptoms. Exclusion criteria included: 1) apparent incoherence, disorientation, or other impaired mental status at the time of recruitment; 2) documented substance abuse or dependence within the last 6 months; 3) bipolar disorder or schizophrenia; and 4) suicidal intent at the time of recruitment.

A total of 681 patients were screened over a period of seven months. Of these, 278 (41%) screened positive on the Patient Health Questionnaire-2 (PHQ-2). Of those who screened positive with the PHQ-2, 260 consented to participate in the study and be screened further with the PHQ-9. One hundred and ninety eight met depression inclusion criteria based on baseline PHQ-9 assessment, of which 5 patients were deemed ineligible due to cognitive impairment, other serious mental illness (e.g., schizophrenia, bipolar disorder), or having previously participated in the study.

There was a significant difference in the baseline depression screen-in rate by clinic, with 48.5% screening positive at the on-site therapist clinic and 35.8% at the collaborative care clinic (χ²(1) =10.95, p < .0001). With 55 participants lost to follow-up, the final sample, with baseline and follow-up data, comprised 138 participants (72% follow up rate). Ten individuals had missing data for the follow-up assessment of PTSD.

Individuals lost to follow-up were more likely to be male (χ²(1) =6.31, p < .05), never married (χ²(3) =17.07, p < .01), and employed (χ²(2) =25.76, p < .0001). Those lost to follow-up also had higher PHQ-9 scores, (t(191)= 2.03, p < .05). Although statistically significant, the difference represented only a 1.2-point difference on the PHQ-9. Those lost to follow-up did not differ by clinic on demographics or baseline values.

The average age of the sample of 138 participants was 47.6 (SD = 12.2) years. More than 80% (84.8) were women, and 51.4% were married/living with a partner. The majority of the sample was Latina/o (94.2%) and 89.9% were monolingual Spanish speakers. Of those who were foreign-born (94.2%), 66.2% had been in the US 10 years or more. The most frequently represented countries of origin were: El Salvador (40.6%), Guatemala (18.1%), Honduras (8.0%) and Mexico (8.0%). Sixty percent had completed less than a high school education and 58.0% were working part- or full-time. Table 1 presents participant demographics by clinic. Participants in the on-site therapist clinic were significantly more likely to be women, married/living together, foreign-born and Latina/o than those in the collaborative care clinic, which served a slightly more diverse immigrant population.

Table 1.

Demographic characteristics by clinic

Variable	Collaborative Care Clinic (n = 65)	On-Site Therapist Clinic (n = 73)
Age in years - M (SD)	47.7 (12.8)	47.6 (11.8)
Gender - n (%) Female	50 (76.9)	67 (91.8)*
Marital Status -n (%) Married/Living Together	23 (35.4)	48 (65.8)*
Ethnicity - n (%) Latina/o	58 (89.2)	72 (98.6)**
Nativity - n (%) Foreign Born	58 (89.2)	72 (98.6)*
Time in US - n (%) in US less than 10 years	23 (35.4)	21 (28.8)
Education - n (%) Less than high school education	36 (55.4)	47 (64.4)
Employment Status -n (%) Employed Full or Part-Time	37 (56.9)	43(58.9)

^*p < .01,

^**p < .001

Procedures

Patients were approached in clinic waiting rooms by bilingual/bicultural research assistants and asked for verbal assent to be screened for a study; once in a more private section of the reception area, the study goals were described. Patients who agreed were screened in Spanish or English, as preferred, using the PHQ-2 (Kroenke, Spitzer, & Williams, 2003) to determine preliminary eligibility. Patients who scored ≥ 4 were invited to participate further. A formal informed consent process included permission for both baseline and follow-up interviews and a medical record review. Patients were administered a baseline interview that confirmed eligibility using the PHQ-9 and included assessments of demographic variables, trauma exposure, and PTSD symptoms. Most participants chose the Spanish version of the assessments and were interviewed by bilingual/bicultural research assistants. If patients continued to meet study criteria for presumptive depression at the baseline assessment, defined as a PHQ-9 score ≥ 10, their primary care providers were informed of the assessment results related to depression and PTSD symptoms, for appropriate follow-up and provision of services.

For the follow-up assessment, participants were contacted by phone to schedule a telephone assessment an average of 8.2 (SD = 1.6) months following their baseline interviews, and follow-up timeframe did not differ significantly by clinic. The target time frame for the follow-up assessment was 6-9 months. This window was chosen to allow sufficient time for the interventions to have a sustained impact, and to reach as many participants as possible, given the mobility of the patient population. Participants were compensated with a $20 gift card to a local grocery store for each assessment completed. The Georgetown University Institutional Review Board and the primary care clinics’ review committees approved all study procedures.

At the time of the study, both clinics were in a transition phase from a paper record to an electronic health record. A clinic administrator from each site extracted data from the clinic’s electronic health record based on a list of variables related to mental health treatment during the study period. Data were then linked to participants’ study IDs and given to the research team with personal identifiers removed. For the collaborative care clinic, a study team member was also able to review that program’s additional records for collaborative care specific activities. This component of the data collection occurred at the time of the follow-up interview.

Because the treatment models at the two clinics were distinct in scope and in services offered, a strategy was needed to be able to compare the adequacy of care regardless of the treatment model. Thus, we defined satisfactory care separately for each clinic, in line with its treatment model. For the collaborative care clinic, satisfactory care was defined as participation in the collaborative care program - evidenced by an intake evaluation, presentation in the team meeting, and one or more services, including medication management, therapy, and/or follow-up assessment. Individuals treated with antidepressant medication for at least two months outside of the collaborative care program, a practice that some of the providers were comfortable with given their ongoing participation with the collaborative care program, were also deemed to meet the threshold of satisfactory care. For the on-site therapist clinic, satisfactory care for therapy was defined as having 3 or more sessions - based on a meta-analysis that suggested that brief psychotherapy for depression and anxiety in primary care could be effective when delivered in as few as three sessions (Cape, Whittington, Buszewicz, Wallace, & Underwood, 2010). For both sites, a course of antidepressant medication of at least two months was selected based on patient distribution, with at least 75% of those who took medications having taken it for at least 2 months, a reasonable amount of time for the medication to have an effect.

Measures

All assessments were administered in an interview format. This was done to maximize understanding by participants regardless of literacy level. Demographic variables included: age, gender, education level, marital status, employment status, self-reported income, country of origin, and years living in the US.

Current depression symptoms were assessed using the PHQ-9, a brief assessment of depression typically used in primary care settings (Kroenke et al., 2001). The PHQ-9 has adequate reliability, convergent validity, discriminant validity, and sensitivity to change (Cameron, Crawford, Lawton, & Reid, 2008; Kroenke et al., 2001; Löwe, Kroenke, Herzog, & Grafe, 2004; Martin, Rief, Klaiberg, & Braehler, 2006). The reliability and validity of the Spanish version of the PHQ-9 has also been previously established (Diez-Quevedo, Trangil, Sanchez-Planell, Kroenke, & Spitzer, 2001; Wulsin, Somoza, & Heck, 2002). Presumptive depression was defined by a score ≥10. Cronbach’s alpha for this sample was 0.45.

Trauma exposure was assessed with a set of five questions that addressed whether the patient had been exposed to physical violence, sexual violence, emotional abuse, had witnessed violence, or experienced any other extreme situation not covered by the previous questions. An example of a question is “Have you ever been physically attacked, like being hit, kicked, or beaten up?” The questions were already being utilized by the collaborative care clinic as part of the ongoing program’s intake evaluation (Kaltman, Pauk, et al., 2011). The emotional abuse question was added based on prior research with the same immigrant community, which highlighted the frequency of this type of exposure (Kaltman, Hurtado de Mendoza, et al., 2011). This measure was pilot tested with patients in the same study clinics using standard cognitive testing techniques to ensure comprehension by the target patient population. If the participants answered one or more of the trauma exposure questions in the affirmative, they were considered trauma exposed and were assessed for PTSD symptoms.

Current PTSD symptomatology was evaluated with the PTSD Checklist for DSM-IV (PCL, Weathers, Litz, Herman, Huska, & Keane, 1993) among only those participants who endorsed at least one trauma exposure. Respondents rated how much they were bothered in the past month by each of the PTSD symptoms included in the DSM-IV diagnosis (APA, 1994) on a 5-point Likert-type scale ranging from “Not at all” (1) to “Extremely” (5). The PCL has good reliability with structured interviews for PTSD (Blanchard, Jones-Alexander, Buckley, & Forneris, 1996). It has been used in studies of Latina/o immigrants (Eisenman, Gelberg, Liu, & Shapiro, 2003) and a psychometric study demonstrated general measurement equivalence across English- and Spanish-language versions (Grant, 2004). Presumptive PTSD was defined by a score ≥ 30 (VA National Center for PTSD, 2014). Change in PTSD symptoms from baseline to follow-up assessment was measured using a subset of six PCL items with good sensitivity and specificity with the longer version, and high correlations with the longer total score (Lang & Stein, 2005). Cronbach’s alpha for this sample was 0.87 for the full PCL and 0.67 for the PCL-6.

A comprehensive chart review was completed for all patients included in the follow-up data set, employing the electronic medical record (EMR) of each clinic. The chart data abstraction was implemented by an EMR officer at each clinic organization at the time of the follow-up telephone interview and spanned a one-year prior to that time, to include the baseline phase of the study. The variables collected included: medical visits, behavioral health visits, psychotropic medications utilization, and medical conditions. Data specific to the collaborative care program were also collected, including intake evaluation, presentation at the team meeting with the consulting psychiatrist, medication management, reassessment of depression and/or PTSD symptoms, and visits with the family support worker. EMR data were available for 98% of the patients in the final sample, with three patients not having available chart data for review.

Data Analysis

Data collected in person at the onset of the study, via a computer-assisted phone interview at 8-months follow-up, and via medical record abstraction were joined together into a spreadsheet where data were de-identified and password protected. Personal identifiers with the corresponding study identification number were stored separately in a locked cabinet. The de-identified patient data were imported into Stata 14™ (StataCorp, 2015) for statistical analysis.

T-test comparisons of means (for continuous variables) for two independent groups and Frequency Distribution Pearson chi-square (for categorical variables) were conducted to compare sociodemographic and mental health variables at baseline. Analyses also compared (1) the final sample that completed follow-up to the sample of participants lost to follow-up; and (2) the patients in the two clinic sites.

For each of the outcome measures studied (i.e., PHQ-9 and PCL-6), independent-samples t-tests were used to compare the clinics at baseline. The differences between the clinics over time were estimated for each outcome measure using linear mixed models, which included: a dummy variable for time, a dummy variable for clinic, and their interaction. These models were also adjusted by age, trauma experienced between baseline and follow up, and satisfactory care. Predicted margins were computed for each model to illustrate the differences over time and between clinics. The PTSD model was conducted on the 91 patients that met criteria for PTSD at baseline. Ten of these participants had missing data for PTSD at follow-up. This was due to an error in procedures that happened periodically through the study. We were unable to find anything systematic to explain why this happened (i.e., it was unrelated to time within the study, interviewer, etc.).

Results

Table 2 presents baseline and follow-up levels of depression and PTSD. No statistically significant differences between the clinics were observed on these variables at baseline. At baseline, the mean depression score for participants in both clinics fell within the upper limits of the moderate depression range on the PHQ-9 (moderate depression is defined as a score of 10-14). At follow-up (Table 2), 73.9% of participants in the collaborative care clinic and 68.5% in the on-site therapist clinic had reached the target of a PHQ-9 score ≤ 9 (OR (collaborative care vs. on-site therapist) = 1.30, CI = 0.62 – 2.73).

Table 2.

Baseline and follow-up levels of depression and PTSD by clinic

Variables	Collaborative Care Clinic (n = 65)	On-Site Therapist Clinic (n=73)	t-test (rank sum) p-value
PHQ-9 score
Baseline	14.8 (3.6)	14.8 (4.1)	0.91
Follow-up	5.9 (6.1)	7.2 (5.3)	0.20*
Full PCL score
Baseline1	46.4 (14.2)	44.4 (15.1)	0.47
PCL-62
Baseline	19.1 (4.7) n=44	19.1 (4.4) n=47	0.98
Follow-up	12.4 (5.9) n=38	14.7 (5.5) n=43	0.07*

¹Full PCL scores are reported for participants reporting trauma exposure, n =115

²PCL-6 scores are reported for participants meeting criteria for PTSD at baseline, n = 91

^*p-value of p less than or equal to .05 based on nonparametric test

Levels of trauma exposure were high, with 83.3% of participants (n = 115) reporting one or more exposures. The most frequently reported trauma exposure was physical violence (51.5%), followed by emotional abuse (42.8%), sexual violence (34.8%), and witnessed violence (33.3%). It is noteworthy that 80 participants (58.0%) reported experiencing an extreme situation not covered by the specific exposures surveyed. Fifty-three participants (81.5%) at the collaborative care clinic and 62 participants (84.9%) at the on-site therapist clinic reported exposure to trauma. Forty-four participants (67.7%) at the collaborative care clinic and 47 participants (64.4%) at the on-site therapist clinic met the cut-off for presumptive PTSD in civilian primary care clinics (score of 30 or higher) (VA National Center for PTSD, 2014). For later/multivariate analyses, those without trauma exposure were excluded from PTSD analyses.

Table 3 presents treatment received by participants at each clinic, including medication only (2 months), three or more therapy visits, collaborative care services, and combinations of care. Significantly more participants in the collaborative care clinic received satisfactory care than at the on-site therapist clinic. Forty-eight percent of participants in the collaborative care clinic and 25.4% of the on-site therapist clinic received satisfactory care (χ²(1) =7.76, p < .01) as defined for their respective clinics.

Table 3.

Treatment received by participants with chart review data by clinic

Variable	Collaborative Care Clinic (n = 64)	On-Site Therapist Clinic (n = 71)
	n (%)	n (%)
Less than satisfactory care	33 (51.6%)	53 (74.6%)
Satisfactory care	31 (48.4)	18 (25.4%)
Types of Treatment received:	n (proportion of clinic total)	n (proportion of clinic total)
Medications only	3 (4.7%)	6 (8.5%)
Collaborative care	17 (26.6%)	N/A
Therapy only	N/A	11 (15.5%)
Combined treatment	11 (17.2%)	1 (1.4%)

In the multivariate model of PHQ-9 scores over time, statistically significant effects were observed for time (an average decrease of 7.5, 95% Cl = −8.8 to −6.2); new trauma exposure during the study period (higher average PHQ-9 by 2.2 for those with a new trauma experience, 95% CI = 0.55 to 3.86); and having PTSD at the onset of the study (higher average PHQ-9 by 2.2 for those with PTSD at baseline, 95% CI = 0.85 to 5.55). No effects were observed for clinic or having received satisfactory care. No other interactions, including the time by clinic interaction were statistically significant.

In the multivariate model of PCL-6 score over time, statistically significant effects were observed for age (average decrease in PCL-6 score for each year by 0.07, 95% CI = −0.14 to −0.00) and having moderate/severe depression (higher PCL-6 score by 4.5, 95% CI = 3.0 to 6.1). The time by clinic interaction was significant (p = 0.006), with predictive margins indicating that patients in the collaborative care clinic experienced greater symptom reductions over time; that is a decrease of 2.0 in the PCL-6 score in the on-site therapist clinic (95% CI = −3.8 to −0.2) compared to a decrease of 5.2 in the collaborative care clinic (95% CI = −7.0 to −3.4), which resulted in a statistically lower PCL-6 mean score for the collaborative care clinic by 2.6 units (95% CI: −4.71 to −0.53. No other interactions tested were statistically significant.

Discussion

This study was a naturalistic prospective evaluation of a collaborative care program treating depression and PTSD in a clinic that serves uninsured patients, primarily immigrants from Central and South America. Receipt of treatment and mental health outcomes (depression and PTSD symptoms) among patients with depression in a clinic with a collaborative care program were compared to those of patients in a clinic with an on-site 75% FTE therapist available. Patients at both study clinics, on average, experienced a significant decrease in depression and PTSD symptoms between the baseline and follow-up assessments. This finding is similar to a recent randomized controlled trial of collaborative care for PTSD in an underserved patient population (Meredith et al., 2016), in which patients in both the collaborative care arm and the “enhanced care as usual” arm improved.

At the follow-up assessment, more than 70% of participants at both clinics had a PHQ-9 score of less than 10 (73.9% of participants in the collaborative care clinic, 71.01% in the on-site therapist clinic), the treatment target typically used by collaborative care programs, a very positive outcome. This result is noteworthy given the setting of under-resourced clinics and the mostly uninsured and impoverished patient population. Improvement in depression was not associated with clinic site/model, suggesting that both treatment models were effective.

The receipt of satisfactory care was not associated with improvement in depression over time. However, significantly more participants in the collaborative care clinic received satisfactory care. This is similar to a recent study of collaborative care for depression for Latina/os in public-sector primary care clinics, which found that patients in the collaborative care arm were more likely to receive psychotherapy or antidepressant medication treatment that those in the enhanced usual care arm (Lagomasino et al., 2016). Services in the collaborative care clinic also tended to be more comprehensive, consistent with the treatment model there, as compared to the on-site therapist clinic, where patients tended to receive either therapy or medications, but rarely both. There are several possibilities as to why the receipt of satisfactory care did not confer better depression outcomes. First, the definition of satisfactory care, based on the relatively low levels of care rendered to this uninsured population, may not have been the right standard. However, other studies that have found sustained effects for brief behavioral interventions in primary care clinics (Kaner et al., 2007; Ray-Sannerud et al., 2012), as well as research suggesting that the greatest improvements in mental health interventions occur early in treatment (Haas, Hill, Lambert, & Morrell, 2002; Stiles et al., 2003). Second, it is possible that the added personal attention provided by a compassionate research staff member during the relatively intensive baseline assessment may have contributed to patient healing, again minimizing group differences. Third, a number of cases of depression may have resolved on their own, consistent with cyclic nature of the disorder (Burcusa & Iacono, 2007).

Of note, more than half of the patients in both clinics received either no care or a level of care below the defined threshold of satisfactory care (52.3% in the collaborative care clinic and 72.6% in the on-site therapist clinic). This is consistent with a recent study of the reach of interventions that integrate behavioral health into primary care. In that study of 11 practices with integrated care models with a total of 24,906 patients, 41% of eligible patients received integrated services (Balasubramanian et al., 2015). The low level of care received in our clinics may be due to patient factors such as refusal or patients not being able to work within the clinic’s schedule for treatment, a realistic concern for the working poor. In addition, providers may have recommended a watch and wait approach prior to initiating care, or they may not have been able to address the identified mental health concern among the host of other health concerns that patients frequently present. Finally, especially within the on-site therapist clinic, there were capacity issues that may have prevented patients from having a thorough mental health evaluation or accessing mental health care.

A high level of exposure to potentially traumatic events was observed in this study. More than 80% of participants in both clinics reported exposure to one or more of the potentially traumatic events assessed. Among general primary care populations, studies have found a prevalence of trauma exposure between 57-88% (Alim et al., 2006; Gillespie et al., 2009; Holman, Silver, & Waitzkin, 2000; Liebschutz et al., 2007; McQuaid, Pedrelli, McCahill, & Stein, 2001; Norman et al., 2006), with studies of urban poor samples yielding the higher estimates. For the depression outcome, experiencing additional trauma during the study period was associated with worse outcomes, which is not surprising given the strong relationship between trauma exposure and a range of negative mental health outcomes (Creamer, Burgess, & McFarlane, 2001; Hovens et al., 2009; Kessler et al., 1999; Schell & Marshall, 2008). This finding highlights the importance of screening for trauma exposure in primary care.

Comorbid PTSD was observed in 68% of patients in the collaborative care clinic and 64% in the on-site therapist clinic. Other studies have found the prevalence of PTSD among patients with depression to be closer to one-third (Campbell et al., 2007; Gerrity, Corson, & Dobscha, 2007; Green et al., 2006), much lower than what was observed in this sample. The higher prevalence of PTSD observed here is not surprising given the high level of trauma exposure experienced by impoverished immigrants from Central and South America. Identifying and treating PTSD alongside depression is important because PTSD has been shown to complicate and lengthen the treatment of depression (Green et al., 2006; Hegel et al., 2005).

On average, patients in both clinics saw significant improvements in their PTSD symptoms. In contrast to the findings for depression however, and consistent with the study hypothesis, patients in the collaborative care clinic evidenced more improved PTSD symptoms than those in the on-site therapist clinic. This finding suggests that for patients with comorbid depression and PTSD, the collaborative care model, especially one including a trauma focus, may result in improved outcomes.

A further finding of interest was the significantly differential screen-in rate observed in the two clinics: more participants in the on-site therapist clinic (48.5%) screened positive for depression than those in the collaborative care clinic (35.8%). One possible explanation for this finding given the roughly similar patient populations served by the two clinics is that more patients in the collaborative care clinic had already been identified and treated successfully for depression. Outreach screening is an important part of a collaborative care model and this result provides support for this systemic approach, which may identify more patients in need of care than other models of integrated care. However, other explanations (e.g., chance sampling differences) could account for the same finding, so further study is warranted.

As mentioned earlier, this sample also had a high proportion of patients with various chronic medical conditions (77%), which have the potential to impact mood and overall quality of life. Although this study did not focus on the relationship between mental health and co-existing chronic diseases, future studies should explore baseline depression levels and outcome improvement in relationship to conditions that impair general health.

Some study limitations warrant mention. First, only one clinic with each treatment model was evaluated. A study with multiple clinics randomized to each treatment model would allow for more rigorous analysis and generalizable findings. Second, this study was originally designed to compare a collaborative care program to a clinic without mental health services. However, this was not feasible by the time of study implementation. Fortunately for the patients, there were no longer any potential comparison clinics that did not have some component of mental health treatment. Third, both clinics were implementing electronic health records during the study period, increasing the likelihood that some treatment data may have been missing (missing data would most likely be random). The medical record review was also challenged by a lack of systematic and consistent documentation in the records. Fourth, there was a 28% loss to follow-up and those lost to follow-up differed in important ways from the original sample. However, given the multi-stage consenting process and multiple contacts required to participate in the study, we were encouraged that 72% of those recruited were able to fully participate. This rate of retention was similar to other studies of collaborative care programs (Meredith et al., 2016).

In conclusion, this prospective study adds to the literature on integrating behavioral health into primary care, and specifically the collaborative care model, by examining the impact of a community-implemented collaborative care program for a particularly vulnerable and understudied patient population that is uninsured and ineligible for Federally-funded health programs. Our results suggest that both treatment models were effective for both depression and PTSD within a context of screening for these disorders. The similarity in findings in this study across the two treatment models does not provide support for adoption of one treatment model over the other. Clearly further research is needed. An additional research question for future studies could examine the costs of the two models as well as their cost-effectiveness.

Our study did find added benefits of the collaborative care model for those patients with PTSD and appeared to be more effective in identifying patients in need of care. Key characteristics of the collaborative care model, including outreach screening, a focus on population management, and measurement-based treatment to target, as well as the adaptations made to the collaborative care program specific to this patient population, may all work together toward ameliorating the disparities that underserved groups such as Latina/o immigrants in the US experience with regard to access to and quality of mental health services.

Given the level of need for mental health care, particularly among uninsured and underserved patient populations, it is imperative to continue the discussion regarding the best models of integrated care and processes for identifying and treating common mental disorders in primary care, including trauma-related outcomes. Understanding the relationship between characteristics of the target patient population and the services offered, particularly with an understanding of the level of care needed to achieve an impact, will be important to support mental health care that is accessible, relevant and effective.

Table 4.

Linear mixed models changes in depression and PTSD

Variable	PHQ-9 Coefficient (95% CI)	PCL-6
Age	−0.02 (−0.07 to 0.03)	−0.07 (−0.14 to 0.00)
New Trauma	2.20 (0.55 to 3.55)**	−0.19 (−2.20 to 1.81)
PTSD (Baseline)	2.20 (0.85 to 3.55)***	N/A
Depression (Baseline)	N/A	4.52 (2.96 to 6.08)***
Received satisfactory care	0.83 (−0.51 to 2.17)	−0.17 (−1.84 to 1.50)
Time	−7.49 (−8.82 to 6.16)***	−1.98 (−3.77 to −0.19)*
Clinic	0.02 (−1.54 to 1.59)	0.58 (−1.41 to 2.58)
Time*Clinic	−1.54 (−3.47 to 0.39)	−3.21 (−5.51 to −0.90)**
N	270	166
Wald chi2 (df)	311.2 (7)	135.85 (7)
p-val	<0.001	<0.001

^*p < .05,

^**p < .01,

^***p < .001