Treatment Moderators and Effectiveness of Engagement and Counseling for Latinos Intervention on Worry Reduction in a Low-Income Primary Care Sample

Abstract

Objective

We conducted a secondary analysis of randomized controlled trial data to determine if the Engagement and Counseling for Latinos intervention (ECLA), a brief, evidence-based, and culturally adapted cognitive behavioral intervention specifically designed for and effective at treating depression, also reduced co-occurring worry symptoms. We also explored whether delivery modality (telephone, face-to-face), and sociodemographic patient characteristics moderated treatment effectiveness.

Method

Between May 2011 and September 2012, low-income Latino primary care patients (N=257) with depression from Boston and San Juan were randomized to usual care (n=86), face-to-face ECLA (n=84), or telephone ECLA (n=87) and completed a psychosocial assessment at baseline and 4-months after randomization. We used intention-to-treat analyses with linear regression models with change in worry (four-months from randomization), as the primary outcome, and treatment condition as the primary predictor.

Results

Patients in ECLA experienced significant reductions in worry at 4-months from randomization than those in usual care (PSWQΔ=−3.28, p<.05). Among patients receiving ECLA, those in the telephone condition exhibited greater worry reductions than those in the face-to-face condition (telephone: M=−7.83, SD=11.45; face-to-face: M=−6.73, SD=12.23; p<.05). Employment status was the only significant treatment moderator. Unemployed patients did not exhibit any changes in worry irrespective of condition, whereas employed patients exhibited the greatest worry reductions across conditions.

Conclusions

Although worry was not a treatment target in ECLA, it also reduced worry among low-income Latinos, which suggests ECLA may have transdiagnostic clinical implications. Telephone-delivered ECLA might hold promise for increasing the uptake of mental health care among employed low-income Latinos.

While Latinos and non-Latino Whites have equal prevalence rates of psychiatric disorders, Latinos initiate mental health care less often, and are more likely to receive lower quality mental health care than non-Latino Whites (Cook et al., 2014; Miranda, McGuire, Williams, & Wang, 2008). Proposals to address mental health care disparities underscore the delivery of high quality and effective psychosocial treatments, such as cognitive behavioral therapy (CBT), through alternate means (e.g., telephone), and their implementation in non-specialty settings (e.g., primary care) (Miranda et al., 2008; Simon, Ludman, Tutty, Operskalski, & Von Korff, 2004). Yet, most of the implementation research has not included Latino communities, or other underserved groups (Institute of Medicine), 2015). When sizeable samples of Latinos were included, and a culturally adapted treatment was offered, randomized controlled trial (RCT) designs with a control group were not used (Castro, Barrera, & Holleran Steiker, 2010). Further, clinical practice norms continue to privilege the treatment of a single psychiatric disorder at a time, despite high rates of co-occurrence. Thus, it is unknown if interventions proven to be effective for one disorder have transdiagnostic clinical implications in so far as these same interventions can be used to address underlying maladaptive self-regulatory processes common across psychiatric disorder classes (Barlow, Allen, & Choate, 2004).

We conducted a secondary analysis of participant data from the Comparative Effectiveness Research to Eliminate Disparities study, a multi-site RCT designed to compare the effectiveness of a brief, evidence-based, and culturally adapted CBT intervention for depression, called “Engagement and Counseling for Latinos (ECLA)” (face-to-face vs. telephone) to usual care in Latino primary care patients (Alegria et al., 2014). ECLA compared to usual care resulted in improved depression and 30-day functioning, with moderate to large effect sizes (Alegria et al., 2014). Our objectives were to: (a) examine whether ECLA also reduced worry symptoms when compared to usual care; and (b) explore whether delivery modality (telephone vs. face-to-face) and patient sociodemographic characteristics, namely, socioeconomic status (SES; education, income, employment status), gender, and nativity status (immigrant vs. US-born) moderated the effectiveness of ECLA on worry reduction.

We hypothesized that ECLA (telephone and face-to-face) would reduce worry symptoms compared to usual care. Our exploratory conceptual model was that access to telephone ECLA would: (1) remove logistical barriers to care such as lack of transportation, (2) accommodate variable hour work schedules, and (3) lessen the mental health stigma associated with receipt of psychotherapy. We hypothesized that the reduced barriers to care would result in greater participation in telephone ECLA, which in turn, would translate into greater worry reductions for those in telephone ECLA versus face-to-face ECLA or usual care. Because variable work schedules are common in the low income labor market, and mental health stigma is higher in men compared to women (Hahm, Cook, Ault-Brutus, & Alegria, 2015), we expected that telephone ECLA would result in higher doses of treatment and subsequently yield greater worry reductions for men and low SES patients. We also expected that immigrants would prefer face-to-face delivery because of cultural values such as personalismo (preference for face-to-face interaction) (Laria & Lewis-Fernández, 2006). In the context of receipt of a patient-preferred delivery format, we hypothesized that immigrants in the face-to-face condition would exhibit greater reductions in worry than their US-born counterparts in either condition.

Methods

Participants

Between May 2011-September 2012, 257 English or Spanish speaking Latinos (≥18 years) with moderate/severe depressive symptoms on the Patient Health Questionnaire-9 (PHQ-9 ≥ 10;Kroenke & Spitzer, 2002) were enrolled from 5 community clinics in Boston, Massachusetts, and 3 health clinics in San Juan, Puerto Rico. Those with a history of psychosis, < 3 months of specialty mental health treatment, and current suicidality were excluded.

Procedures

Participants completed a baseline, two-month, and four-month psychosocial interview and bi-weekly PHQ-9 assessments. ECLA is an effective telephone plus care management CBT intervention for depression (Dwight-Johnson et al., 2011). The ECLA used in this trial was adapted to be shorter (6–8 vs. 8–10 weekly/bi-weekly sessions), for use in low-literacy and limited English proficiency participants (Ramos & Alegria, 2014), and for face-to-face administration. ECLA consisted of psychoeducation, behavioral activation, cognitive reframing, and motivational interviewing delivered by a mental health professional. Sessions targeted avoidance behaviors and maladaptive cognitive schemas common across depressive and anxiety disorders (e.g., behavioral activation, emotion regulation, cognitive reframing). The Institutional Review Boards of the Cambridge Health Alliance in Boston, Massachusetts and the University of Puerto Rico, in San Juan, Puerto Rico approved the study.

Design

Eligible participants were randomized after the baseline into: (1) telephone ECLA, (2) face-to-face ECLA, or (3) usual care. Participants in usual care received the standard of care that is typical of their primary care clinic (e.g., prescription of antidepressants and/or possible referral to specialty care). Clinicians exhibited high fidelity to ECLA (Alegria et al., 2014).

Measures

Predictor

The primary predictor was treatment condition: Telephone ECLA, face-to-face ECLA, or usual care.

Outcome

The primary outcome was change in worry symptoms from baseline to four-month. Worry was measured with the Penn State Worry Questionnaire (PSWQ), a 16-item assessment with strong psychometric properties (Meyer, Miller, Metzger, & Borkovec, 1990), and proven validity in Spanish speakers (Novy, Stanley, Averill, & Daza, 2001). Each question is scored on a 5-point scale (1=not typical of me to 5=very typical of me). Responses to raw values were reverse coded (in some cases) and summed; higher values reflect greater worry (range: 16–80). Scores ≥40 reflect moderate to severe worry. The internal consistency reliability of the scale was excellent; overall (α=0.89), Puerto Rico (α=0.90) and Boston (α=0.87).

Engagement factors

Number of ECLA intervention sessions (total, missed, additional), and satisfaction with treatment were collected. Receipt of a prescription/medicine for emotions, nerves, or mental health from any type of professional, and scheduled appointments with a psychologist, counselor, or therapist (not part of ECLA) were also collected.

Sociodemographic factors

Participants’ self-reported gender, race, age, nativity status, educational attainment, employment status, and income were collected at baseline.

Covariates

Past month functioning was assessed at baseline with the World Health Organization Disability Assessment Schedule (WHODAS) (Ustun et al., 2010).

Analytic Plan

Chi-squared tests or independent samples t-tests were used to evaluate differences in the distribution of the baseline demographic, engagement, and outcome variables across condition and site. We used intention-to-treat analysis with linear regression models with change in worry as the primary outcome, and study condition as the primary predictor. The usual care group served as the reference category. Model 1 adjusted for treatment condition and site. Model 2 further adjusted for gender, race, age, education, employment status, nativity status, and income. Participants who were born in Puerto Rico were classified as US-born. Model 3 additionally adjusted for health and disability status. To test for moderation by sociodemographic characteristics, and by delivery modality we entered a series of cross products (2-way or 3-way interactions) into separate models adjusted for covariates. Effect sizes for each delivery modality were calculated using Cohen’s d (Cohen, 1992). Two sensitivity analyses were conducted: (1) linear models with an additional adjustment for baseline PHQ-9 score, and (2) site by condition cross products were entered into site-adjusted models. To account for missing data, multiple imputation was conducted with the Imputation by Chained Equations (ICE) package in STATA (Stata Corporation, 2011). We imputed the dataset 10 times, and combined the analytic results using the Multiple Imputation of Missing values; results using imputed data are reported.

Results

There were 257 participants (Boston [n = 127], San Juan [n = 130]), 86 in usual care, 84 in face-to-face ECLA, and 87 in telephone ECLA. There were no significant differences in baseline demographic characteristics, engagement factors, psychosocial measures, or number of ECLA sessions by condition (Table 1). Level of satisfaction did not differ between the ECLA conditions. Receipt of a prescription/medicine for mental health concerns, and number of mental health appointments (non ECLA) were low, and did not differ by condition. Participants in San Juan were more likely to be 50 years or older, unemployed, to report annual income of <$15,000, and to have higher functional impairment than those in Boston. No site differences were found in the distribution of worry or depressive symptoms at baseline (Supplemental Table 1).

Table 1.

Distribution of Baseline Demographic Characteristics, Engagement Factors, and Psychosocial Measures by ECLA Treatment Condition among Low-income Latino Primary Care Patients (N = 257)

	Usual Care		Face-to-Face ECLA		Telephone ECLA		p-value
	N = 86	%	N =84	%	N = 87	%
Gender
Male	15	17.4%	19	22.6%	13	14.9%	0.417
Female	71	82.6%	65	77.4%	74	85.1%
Race
White	27	31.4%	22	26.2%	24	27.6%	0.697
Black/Trigueña/Morena	29	33.7%	27	32.1%	30	34.5%
Unreported	28	32.6%	28	33.3%	26	29.9%
Mixed/Mezcla/Indio	2	2.3%	7	8.3%	7	8.0%
Age
18–34	19	22.1%	20	23.8%	25	28.7%	0.842
35–49	32	37.2%	30	35.7%	32	36.8%
50+	35	40.7%	34	40.5%	30	34.5%
Education
6th grade or less	18	20.9%	20	23.8%	20	23.0%	0.240
7th – 11th grade	23	26.7%	31	36.9%	20	23.0%
12th grade or more	45	52.3%	33	39.3%	47	54.0%
Employment
Employed	36	41.9%	35	41.7%	31	35.6%	0.677
Unemployed/out of the labor force	50	58.1%	49	58.3%	55	63.2%
Nativity*
Foreign-born/immigrant	43	55.1%	41	51.9%	42	55.3%	0.957
US-born	3	3.8%	2	2.5%	3	3.9%
PR-born	32	41.0%	36	45.6%	31	40.8%
Income
$0 – $14,999	67	77.9%	61	72.6%	68	78.2%	0.632
$15,000 or more	19	22.1%	23	27.4%	19	21.8%
Engagement Factors
Number of total ECLA sessions	0	0	4.58	3.17	4.90	2.80	0.494
Missed ECLA sessions	N/A	N/A	2.01	2.60	1.66	2.30	0.343
Additional ECLA sessions	N/A	N/A	0.60	0.93	0.55	0.83	0.748
“Very satisfied’ with ECLA intervention	N/A	N/A	58	87.88%	55	82.09%	0.102
Received prescription for mental health condition**	12	16.22%	14	18.92%	7	9.72%	0.279
Mental health appointment (not ECLA)***	14	17.95%	7	8.86%	8	10.53%	0.187
Psychosocial Measures	M	SD	M	SD	M	SD	p-value
PHQ-9 at baseline	15.96	3.97	17.42	4.45	16.77	4.26	0.084
WHODAS at baseline	27.26	11.07	27.11	11.57	28.94	11.36	0.568
PSWQ at baseline	61.07	10.99	61.00	11.83	63.44	9.23	0.238
PSWQ at four-month	56.49	11.88	54.52	11.52	55.54	12.22	0.584

Note. N = 256 for employment variable; ECLA= Engagement and Counseling for Latinos Intervention (culturally-adapted cognitive behavioral therapy [CBT] intervention for depression). WHODAS= World Health Organization Disability Assessment Scale; PHQ-9= Patient Health Questionnaire; PSWQ= Penn State Worry Questionnaire.

^*N = 233 for nativity status variable in raw dataset.

^**Respondents indicated whether they received a prescription or medicine for their emotions, nerves, or mental health from any type of professional.

^***Respondents indicated whether they had an appointment with a psychologist, counselor, or therapist (not part of ECLA intervention).

Was ECLA for depression effective at reducing worry and for whom?

Participants who received either telephone ECLA or face-to-face ECLA demonstrated significant reductions in worry at four-months relative to their usual care counterparts in site-adjusted models (PSWQ Δ=−3.08, p <.05) and models adjusted for baseline demographics and site (PSWQ Δ=−3.38, p <.05) (Table 2). The reduction in worry attributable to ECLA (PSWQ Δ=−3.28, p < .05) remained significant even after adjustment for functional impairment. We found a significant interaction for employment status and treatment condition (p for interaction = .03; Table 2, Model 4). Employed participants in ECLA exhibited significant decreases in worry (PSWQ Δ=−9.83, p <.05) relative to their employed counterparts in usual care (PSWQ Δ=−2.38, p <.05), and their unemployed counterparts in ECLA (PSWQ Δ=−5.65, p <.05) or usual care (PSWQ Δ=−5.52, p <.05). Worry symptom scores were comparable for unemployed participants in either condition (Figure 1). None of the other interactions were statistically significant (p for interaction >.05).

Table 2.

Effect of ECLA on Change in Worry 4-Months from Randomization among Low-income Latino Primary Care Patients (N = 257)

	Model 1		Model 2		Model 3		Model 4
	β	SE	β	SE	β	SE	β	SE
Treatment condition
Usual Care	ref		ref		ref		ref
ECLA (telephone and face-to-face)	−3.08*	1.55	−3.38*	1.57	−3.28*	1.57	−7.84**	2.43
Gender
Male			ref		ref		ref
Female			1.81	2.07	1.75	2.07	1.71	2.05
Race
White			ref		ref		ref
Black/Trigueña/Morena			−1.66	1.95	−1.76	1.96	−2.10	1.95
Unreported			−2.69	2.05	−2.59	2.05	−2.78	2.03
Mixed/Mezcla/Indio			3.01	3.32	3.22	3.33	2.78	3.29
Age
18–34			ref		ref		ref
35–49			−2.22	2.01	−1.93	2.01	−1.77	2.00
50+			−2.49	2.07	−2.11	2.08	−2.10	2.05
Education
6th grade or less			ref		ref		ref
7th – 11th grade			−0.05	2.15	0.14	2.14	−0.29	2.13
12th grade or higher			0.18	2.09	0.46	2.11	−0.10	2.07
Employment
Employed			ref		ref		ref
Unemployed/out of labor force			1.67	1.85	1.92	1.88	−3.35	3.07
Nativity
Immigrant			ref		ref		ref
US-born			2.13	2.08	2.35	2.08	2.69	2.09
Income
$0 – $14,999			ref		ref		ref
$15,000 or more			−0.88	2.24	−0.95	2.23	−0.65	2.21
Functional Status
WHODAS score at baseline					−0.07	0.07
Interactions
ECLA treatment Condition*unemployed							7.56*	3.37
Constant	−4.22	2.61	−2.00	7.07	−2.35	7.35	0.49	7.17

Note

^***p<0.001,

^**p<0.01,

^*p<0.05;

ECLA= Engagement and Counseling for Latinos Intervention (culturally-adapted cognitive behavioral therapy intervention for depression); WHODAS= World Health Organization Disability Assessment Scale. All models adjust for site.

Ref= reference category. US-born includes those born on the US-mainland and in Puerto Rico.

Figure 1.

Change in worry as a function of treatment condition and employment status among low-income Latino primary care patients (N=257).

Does delivery modality matter?

Participants in telephone ECLA experienced a statistically significant reduction in worry when compared to those in usual care, in site-adjusted models (PSWQ Δ= −3.62, p <.05), models that further adjusted for demographic characteristics (PSWQ Δ= −4.13, p<.05), and models that further adjusted for functional status (PSWQ Δ= −3.98, p <.05). Participants in face-to-face ECLA were not significantly different from those in usual care in their change in worry symptom scores when the intervention groups were analyzed separately. Among participants receiving ECLA, those in telephone ECLA had greater reductions in worry than those in face-to-face ECLA (telephone: M = −7.83, SD = 11.45; face-to-face: M = −6.73, SD = 12.23; p = 0.046).

Are certain delivery modalities better at lowering worry for specific subgroups of people?

We observed a trend for moderation by employment and treatment condition (p for interaction =. 08). Employed participants in telephone ECLA appeared to exhibit the greatest decreases in worry when compared to their employed and unemployed counterparts (Table 3). None of the other interactions were statistically significant (p for interaction >.05).

Table 3.

Comparative Effectiveness of Face-to-Face ECLA, Telephone ECLA, vs. Usual Care on Change in Worry 4-Months from Randomization among Low-income Latino Primary Care Patients (N = 257)

	Model 1		Model 2		Model 3		Model 4
	β	SE	β	SE	β	SE	β	SE
Treatment condition
Usual Care	ref		ref		ref		ref
ECLA face-to-face	−2.53	1.78	−2.59	1.81	−2.54	1.81	−6.57*	2.80
ECLA telephone	−3.62*	1.81	−4.13*	1.83	−3.98*	1.83	−9.14**	2.89
Gender
Male			ref		ref		ref
Female			1.92	2.08	1.86	2.08	1.87	2.07
Race
White			ref		ref		ref
Black/Trigueña/Morena			−1.65	1.95	−1.75	1.96	−2.10	1.96
Unreported			−2.77	2.05	−2.66	2.05	−2.96	2.04
Mixed/Mezcla/Indio			3.06	3.32	3.26	3.33	2.71	3.31
Age
18–34			ref		ref		ref
35–49			−2.24	2.01	−1.96	2.01	−1.75	2.01
50+			−2.58	2.07	−2.21	2.08	−2.16	2.06
Education
6th grade or less			ref		ref		ref
7–11th grade			−0.15	2.15	0.03	2.15	−0.29	2.15
12th grade or higher			0.32	2.11	0.57	2.11	0.14	2.10
Employment
Employed			ref		ref		ref
Unemployed/out of labor force			1.74	1.85	1.97	1.88	−3.34	3.08
Nativity
Immigrant			ref		ref		ref
US-born			1.99	2.09	2.22	2.09	2.64	2.11
Income
$0 – $14,999			ref		ref		ref
$15,000 or more			−0.96	2.24	−1.02	2.23	−0.74	2.21
Functional Status
WHODAS score at baseline					−0.07	0.07
Treatment Condition*Employment							p = 0.08
telephone*unemployed							6.80 ŧ	3.78
face to face*unemployed							8.41*	3.95
Constant	−4.22	2.61	−2.05	7.07	−0.85	7.06	0.39	7.18

Note

^***p<0.001,

^**p<0.01,

^*p<0.05,

^ŧp < 0.10;

ECLA= Engagement and Counseling for Latinos Intervention (culturally-adapted cognitive behavioral therapy intervention for depression). All models adjust for site. Ref= reference category. US-born includes those born on the US-mainland and in Puerto Rico.

Effect size

The effect size of telephone ECLA on worry reduction relative to usual care was found to be small-to- medium, Cohen’s d = .41. In contrast, the effect size of face-to-face ECLA on worry reduction relative to usual care was found to be small, Cohen’s d =.25.

Sensitivity analyses

Adjustment for baseline depressive symptoms did not change the observed estimates appreciably (Supplemental Table 2).¹ Site modified both the effectiveness of the ECLA intervention, and the effectiveness of the ECLA delivery modality (p for interaction <.05). In the aggregate, ECLA was more effective in Boston than in Puerto Rico (Supplemental Figure 1). However, respondents in face-to-face ECLA in Puerto Rico had significantly smaller reductions in worry than respondents in the face-to-face ECLA in Boston; worry reductions for those in the telephone condition were comparable across sites (Supplemental Figure 2).

Discussion

Our secondary analysis of RCT data of ECLA, a brief and effective CBT intervention for depression (Alegria et al., 2014), shows that ECLA also reduces worry symptoms in low-income Latino primary care patients with depression, and these reductions were independent of baseline depressive severity. These results suggest that this version of ECLA might have broad transdiagnostic implications, particularly because these gains were observed in the absence of any anxiety related modifications to the treatment. Because of the excess prognostic risks conferred by anxiety among those with depression (Fava et al., 2008), future research should examine whether treatment of worry results in improved medical prognosis among those with co-occurring depression and anxiety. Additional research that tests whether ECLA would be effective at treating patients with primary worry and no depression, or patients with primary depression and no worry, would help clarify its transdiagnostic uses.

Our findings also suggest that telephone ECLA may be an effective and acceptable alternative to its face-to-face variant for worry reduction, given the effect sizes associated with telephone ECLA (d = .41) versus face-to-face ECLA (d =.25), the overall equal levels of high satisfaction with the ECLA treatment, and the comparable effectiveness of telephone ECLA in Boston and Puerto Rico. Importantly, in contrast to our hypotheses, telephone ECLA did not reduce barriers to treatment participation. As such, greater doses of treatment (number of sessions) do not appear to explain the increased effectiveness of telephone ECLA. Future research that seeks to identify the specific mechanisms that contribute to the differential effectiveness by delivery modality would help identify the active ingredients.

We also found that employment status modified the effectiveness of ECLA. While employed participants in the ECLA condition experienced clinically significant reductions in worry, participants who were unemployed and in either condition did not exhibit any natural or treatment-related worry remission. Participants who were unemployed compared to those who were employed might have had greater exposures to chronic stressors and fewer social and environmental resources (Falconnier, 2009), which in turn may have prevented them from being in a position to learn the therapeutic skills in a short-time frame and in turn reap the benefits from short-term behavioral therapy. Structural interventions that focus on securing employment and financial stability might have the greatest impact on worry among the unemployed.

We did not find gender or nativity differences in treatment effectiveness. The small sample of men and the confounding of nativity status by site likely prevented the detection of statistical differences between subgroups. We were also unable to test for language moderation because of the confounding of language by site. Future research adequately powered to detect gender differences and explore gender variation in help seeking beliefs would prove beneficial.

Limitations

Our study is not without its limitations. First, this was a mostly low-income Latino sample, which limits the generalizability of these findings. The constricted range in SES might also have prevented our ability to detect significant moderation effects by education or income. Second, the numerical differences in worry reduction observed between telephone ECLA and face-to-face ECLA though statistically significant, were generally small and may not translate into clinically significant differences. Only 9.21% – 13.92% of respondents experienced remission of worry symptoms (PSWQ score <40). Future studies adequately powered to detect clinical remission of worry are needed to definitely test the causal relationship between ECLA participation and remission of worry to normative levels.

Conclusion

Our findings provide preliminary evidence that ECLA, a brief, evidence-based, and culturally adapted CBT intervention for depression, also reduces worry symptoms in Latino primary care patients with depression. As such, ECLA might have broad transdiagnostic implications for psychosocial treatments for Latinos. Our findings also indicate that the telephone might be a particularly promising modality to deliver effective mental health care to employed low-income Latinos considering the medium effect sizes observed and high levels of engagement and satisfaction with care. Future research should examine how to implement these telehealth interventions with potential transdiagnostic implications in the communities shown to exhibit the greatest public health need and the greatest potential for therapeutic gain.

Public Health Significance.

This study suggests that a brief, culturally adapted, cognitive behavioral intervention (ECLA) for depression also reduces worry, especially among employed and low-income Latino primary care patients. Our results also suggest that telephone-delivered ECLA, in comparison to face-to-face-delivered ECLA, may hold strong promise as a method to increase access to effective and high quality mental health treatments for Latino populations.