The LIFE-DM intervention was developed through a process of adaptation, stakeholder engagement, and pilot testing. The psychotherapy component was adapted from the Community Partners in Care project [31], which includes modules on cognitions, behaviors, and social skills. Based on our formative work through the multicomponent collaborative care model for depression and stakeholder input regarding feasible approaches for task-sharing in Vietnam [32], the team selected only BA and social skills or support modules, prioritizing pragmatic, concrete skills suitable for delivery by nonspecialist providers.

A review of microfinance interventions identified five components for integration with depression care: (1) livelihood assessment, (2) livelihood planning, (3) livelihood goal setting, (4) incorporation of financial stressors into psychoeducation, and (5) integration of livelihood activities into BA exercises. The adapted curriculum was reviewed with Da Nang Psychiatric Hospital staff and CHS workers for content relevance, cultural appropriateness, and delivery feasibility.

Two pilot groups were conducted over 4 months at 2 CHSs under psychiatrist supervision to refine content and delivery. Each group completed 12 sessions, with group sizes ranging from 6 to 10 and 4 to 9 participants, respectively (21 participants total across both groups). The microfinance loan component was not included during piloting. Attendance and 9-item Patient Health Questionnaire (PHQ-9) scores were collected to assess symptom change. The pilot groups also served as the primary training vehicle for local clinical supervisors.

Participant and supervisor ratings were collected throughout the pilot phase to evaluate intervention quality. At every session, participants rated the intervention on a 3-point Likert scale across domains of understanding (eg, “the session was easy to understand”), relevance (eg, “the lessons in this session are relevant to me”), helpfulness (eg, “I found this session useful”), feeling heard and engaged by facilitators (eg, “the facilitator understood my problems”), and skill development (eg, “this session taught me how to take small steps toward my livelihood goals”). Supervisors rated sessions on adherence across multiple domains, including group facilitation (eg, set the agenda), tracking (eg, checked homework), and session-specific content (eg, reviewed lessons from previous sessions), as well as overall quality (eg, the session was structured).

Participant ratings indicated high perceived relevance and understanding, and PHQ-9 scores showed preliminary symptom improvement. Based on these pilot findings, the intervention protocol was finalized for the main trial without major modifications. Pilot participants were not included in the main trial sample.

The study was conducted at CHSs, which are the most accessible and basic health care setting in Vietnam. They are the lowest administrative level in the health care system, serving a catchment area of approximately 10,000 people each. CHSs provide nonspecialized health care and refer to provincial hospitals for more complex medical care. They are generally managed by general practitioners and staffed by nurses, physician assistants, and midwives. CHS personnel typically have no training in mental health services. For the LIFE-DM program, the CHS collaborated with the Vietnamese WU, which is a well-established quasi-governmental organization focused on promoting the well-being and livelihood of Vietnamese women throughout the country. Four CHSs previously involved in the national mental health program, which offered free antidepressant medications in line with guidelines, were recruited for the study. These 4 CHSs were also similar in a number of key characteristics, including commune-level socioeconomic status (serving a high percentage of individuals considered poor to near poor according to government assignment), number of providers (4‐6 staff), composition of staff (usually run by a physician or physician assistant and supported by 3‐5 nurses), and clinic size (ie, approximately 500‐1000 patient load). Two were allocated to LIFE-DM intervention, and 2 were allocated to E-TAU.

Women residing in Da Nang, Vietnam, were recruited using multiple methods. Women approaching the 4 participating CHSs for health concerns were assessed for program eligibility during enrollment periods of 1 week at the start of each group. In addition, women seeking support for life stressors or exhibiting emotional distress were also referred by CHS or WU staff to be screened for the study during enrollment periods. Study inclusion criteria included (1) female gender, (2) aged 18 to 55 years, (3) monthly family income of less than 2,000,000 VNĐ (roughly US $91) per family member, (4) positive screen for depression using the PHQ-9 [38] (score of 10 or higher), and (5) no significant physical health problems or disabilities (eg, heart disease, cancer, chronic pain, blindness). Exclusion criteria include (1) positive screen on alcohol use (measured by the Alcohol Use Disorders Identification Test) [39], (2) elevated suicide risk, and (3) active psychotic and manic symptoms. Such patients were referred to the Da Nang Psychiatric Hospital for services.

A matched-pair cluster nonrandomized controlled trial was conducted, comparing the LIFE-DM intervention with E-TAU. Two CHSs were allocated to the LIFE-DM group intervention and 2 to the E-TAU arm, which offered guideline-based antidepressant medications. Cluster-level randomization was not feasible for two reasons: (1) the small number of available sites limited the use of randomization for achieving balance and (2) the WU, whose infrastructure was essential for delivering the microfinance component, imposed operational constraints on random assignment of microfinance services across sites. To strengthen comparability in the absence of randomization, CHSs were matched on key characteristics prior to allocation (see the Study Site Characteristics and Recruitment section), and within each matched pair, 1 site was assigned to LIFE-DM and the other to E-TAU.

The trial was designed to approximate real-world implementation conditions under which the program would be disseminated, in order to assess the feasibility and effectiveness in routine care settings. Although propensity score matching will be used to reduce baseline imbalances, we acknowledge that selection bias may still be present due to the nonrandomized cluster allocation and potential unmeasured confounding at both individual and cluster levels.

The primary outcomes are treatment engagement, depression symptom change in PHQ-9 score, and depression remission on the Mini International Neuropsychiatric Interview (MINI). We compare the means in a regression framework, after taking into account the design effect induced by clustering, to determine the power when comparing the intervention group to the control group. We note that the end-status analysis assumed for the power calculations can be conservative because it does not use all 3 waves of data and uses only the projected retained sample at month 12. For the power analyses, we used the conservative estimate of the projected retention rate of 79% at the 12-month follow-up. This translates to approximately 64 patients per condition at the 12-month follow-up. A recent meta-analysis found an effect size of 0.78 favoring therapies using BA over control conditions using professionals [40]. However, given the use of paraprofessionals with limited mental health training, we expect smaller effect sizes. For these power analyses, we used conservative effect sizes of 0.3 and 0.4. Since patients are clustered within intervention groups, patients’ outcomes may be correlated, even though other randomized mental health studies have found intraclass correlation coefficients that are essentially zero [41]. Using this effect size, a power of 0.85, a conservative intracluster correlation of 0.1, a cluster size of 10, and a design effect of at least 2, we determined that 142 participants were needed for the study. Calculations were performed using sampsi and sampclus commands on STATA [42]. To account for loss to follow-up, we planned to enroll 160 and ended up with a final sample size of 166 participants.

All study measures (Table 1) were adapted using a consensus approach to ensure linguistic and cultural appropriateness for use in Vietnam. Primary outcomes focused on depression severity and diagnosis, measured by the PHQ-9 [38], which assesses depressive symptom severity, and severe mental illness screener, which was based on items from the MINI [43]. Secondary outcomes included program participation indicators, such as group attendance, engagement in livelihood activities, and acceptance of depression care. The data were gathered through face-to-face surveys, clinic logs, and program records.

Anxiety symptoms were assessed using the 7-item Generalized Anxiety Disorder 7 scale (GAD-7) [44]. Health functioning was measured by the Short Form-12 Health Survey [45], capturing both physical and mental health status. Quality of life was assessed with the Quality of Life Enjoyment and Satisfaction Questionnaire [46], while BA was evaluated using the Brief Activation for Depression Scale–Short Form [47], which captures levels of engagement in goal-directed activities. Social capital was measured using an adapted version of the Short Social Capital Assessment Tool [48], which includes indicators of community participation and support. Economic and employment outcomes included measures of work status, income, household assets, loan and savings activities, and perceived performance of small businesses. Additional sociodemographic covariates such as age, marital status, education, mental health literacy, household size, and number of dependents were also collected to adjust for potential confounders.

The data will be analyzed with an intention-to-treat approach for all outcomes. We will conduct a longitudinal regression analysis [49] to compare the two interventions on (1) primary outcomes, including depression scores based on the PHQ-9 and remission from depression (based on MINI diagnoses), and (2) secondary outcomes, including BA, functioning, quality of life, treatment acceptance, and economic outcomes. To address internal validity concerns of a nonrandomized design, we will use propensity score-based weighting, which is a recommended statistical technique to equate groups on a large number of baseline covariates to adjust for potential baseline differences and strengthen the inferential pathway toward causation in nonrandomized design [50-52]. Propensity score weights will be calculated using baseline demographic and clinical characteristics such as the PHQ-9, the GAD-7, physical health functioning, mental health functioning, BA, social support, self-efficacy, mental health literacy, age, education, economic status, work status, marital status, and loan exposure.

A regression analysis [49] with propensity weights will be conducted to estimate treatment effect with repeated measures of our outcomes at 3 time points accounting for demographic covariates, such as age, marital status, baseline income variables, work status, baseline mental health, functioning, and social predictors. Effects may vary over 6 and 12 months; therefore, we will also test for linear time trends in the regression models. Outcomes for patients enrolled from the same intervention group, cohort, and site may be correlated. Ignoring this clustering effect would lead to the underestimation of the variance of treatment effects. Thus, we will adjust for group clustering to account for the correlation among the measures of patients enrolled in the same intervention group, cohort, and site [53].

We will examine variable distributions from all time points to assess missing data patterns within each study wave and replace them with information gathered from clinic logs and program records. Based on previous similar studies in Vietnam in general and Da Nang in particular, we expect response rates of approximately 80% at 6-month and 12-month follow-ups, resulting in about 128 patients at the 2 follow-ups. All planned analyses will use likelihood methods, which are effective under missing at random assumptions [54]. This approach helps account for participants dropping out of the study while making the most efficient use of all the data collected. Because time will be a random effect within the mixed model analyses, participants without complete data will be included fully in the analyses. For all outcomes, Cohen d will be calculated to estimate the effect size using pooled SD of baseline scores. Two-sided P<.05 will be considered statistically significant.

This study was approved by the Da Nang Psychiatry Hospital Institutional Review Board (01/17/HDDD) and the RAND Corporation (2012‐0031 RAND IRB). The study was conducted in accordance with the Declaration of Helsinki. All participants provided written informed consent before participating in the study. Participants received 50,000 VNĐ (approximately US $2.30) for completing each survey as compensation for their time. All participant data will be deidentified and assigned unique study identification numbers prior to analysis.

This study was funded prior to implementation, and the data were collected from February 2014 to September 2015. A total of 198 women were screened for eligibility, of whom 166 met the inclusion criteria and were enrolled. Thirty-two women were excluded for the following reasons: outside the eligible age range (n=3), income above the threshold (n=4), PHQ-9 score below 10 (n=8), elevated suicide risk (n=1), declined participation prior to consent (n=14), or withdrew after consenting (n=2; Figure 1). Baseline characteristics of participants are presented in Table 2. Data analysis is in progress, and the findings are expected to be disseminated in 2027.

Among the 166 participants enrolled in the study (Table 2), the average age was 42 (SD 8) years. Most women were married (n=119, 71.69%). The majority of participants were able to easily read (n=135, 81.33%) and write (n=131, 78.92%), and 73.49% (n=122) completed primary school. Most women were employed (n=120, 81.63%). On average, the per capita family income was US $40.52 (SD 18.93) per month, and individual income was US $82.35 (SD 53.83) per month. In addition, 29.52% (n=49) lived below the poverty level according to World Bank 2010 criteria for Vietnam (653,000 VNĐ per person per month, or US $1 per person per day). Approximately 47.59% (n=79) reported having existing loans.

The mean score on the PHQ-9 scale for depression was 14.43 (SD 3.46), and 9.58 (SD 3.52) on the GAD-7 scale for anxiety. Most participants had a current diagnosis of depression (n=141, 85.45%). In terms of social capital support, 55.42% (n=92) of the participants had no active memberships in any social groups in the community, and 33.96% (n=54) of the participants reported having no source of social support (including both organizations and individuals). Participants reported an average of 4.27 (SD 2.19) structural barriers to care, such as transportation, money, and child-care responsibilities. In addition, participants reported a mean depression stigma score of 26.17 (SD 3; possible range is 0-40), where higher scores indicate higher perceptions of stigma.

In terms of economic characteristics, the groups differed with regard to business ownership and employment status, with women in the intervention group more likely to have prior experience owning a business (n=50, 54.35% vs n=18, 24.32%; P<.001), currently owning a business (n=43, 46.74% vs n=16, 21.62%; P<.001), and currently working (n=81, 90% vs n=39, 68.42%; P<.001). Those in the intervention group had higher individual income (US $91 per month vs US $71 in the control group; P=.02), but there were no significant differences between the groups by family income.

In terms of mental health characteristics, the intervention group was found to have a higher percentage who met the criteria for depression (n=84, 91.30% vs n=57, 78.08%; P=.02) according to the MINI Diagnostic Interview. However, severity scores on the PHQ-9 did not differ between the 2 groups. Participants in the intervention group also reported a lower mean number of structural barriers to care (3.87 vs 4.78; P=.008) and higher goal self-efficacy (10.15 vs 9.23; P=.01), as well as business self-efficacy (6.66 vs 5.74; P=.005).

The LIFE-DM trial aims to evaluate whether an integrated depression care and livelihood intervention can increase treatment engagement and is more effective than E-TAU in a collaborative care approach. The study addresses a critical gap in global mental health: how to effectively deliver evidence-based depression care to populations burdened by both mental illness and poverty. Despite the strong relationship between depression and poverty [3], this combined approach to care has not been previously evaluated in a controlled study. Many individuals in resource-poor settings face barriers to mental health care [15,16], including lack of awareness and knowledge about depression and depression care, stigma about mental health treatments, cost, economic stressors, and lack of value placed on mental health treatment. The LIFE-DM model is designed to integrate psychosocial support with financial empowerment, through microfinance activities, vocational training, and savings programs, to increase the acceptability and perceived value of depression treatment. This integrated approach is hypothesized to improve mental health, functional status, and economic well-being. If effective, this model may offer a scalable approach for similar low-resource settings.

This model is informed by principles of “stepped care” and “community-partnered process,” which emphasize accessible, acceptable, and contextually grounded interventions [55]. By incorporating income-generating activities, vocational support, and microfinance opportunities, the intervention may help reframe depression care as a means of improving broader life circumstances, thus reducing stigma and enhancing uptake [22]. It is anticipated that this integrated approach will increase the perceived relevance and acceptability of mental health interventions, especially among populations that view mental illness through social and economic lenses rather than clinical ones.

Furthermore, our findings are expected to contribute to the growing field of task-shifting in global mental health [17]. By engaging paraprofessional providers, WU, and CHS staff, the intervention explores the potential for extending the collaborative care models to settings with limited specialist resources. This approach aligns with broader efforts to strengthen health systems through decentralization, community-based services, and the use of nonspecialist providers. The findings from this study may help inform the development of scalable models for delivering evidence-based depression care in resource-limited contexts.

This study has several limitations. The lack of randomization was the primary limitation of the study due to the pragmatic design of the trial. As the baseline data show, there were a number of baseline mental health and economic differences between the LIFE-DM group and the E-TAU group. Although both groups were low-income and had elevated levels of depressive symptoms, the LIFE-DM intervention group reported a higher rate of business experience and ownership as well as higher individual income. They also tended to have higher levels of depressive symptoms but no differences in the rate of depression than the E-TAU group. This suggests that there may be some level of selection bias particularly as referrals for LIFE-DM appear to be motivated by interest in the microfinance program. As these are common sample bias issues in pragmatic trials, we will statistically equate groups using propensity weights, which is a recommended solution for pragmatic or real-world trials where randomization may not be possible. However, residual confounding may remain due to unmeasured differences between the groups. We therefore interpret the findings as estimates of effectiveness under real-world conditions rather than causal effects equivalent to those derived from randomized controlled trials. A further limitation is that the study was conducted in only 4 urban community health stations in Da Nang, which may limit generalizability to rural or other low-resource settings. Finally, reliance on self-reported measures may introduce bias, particularly for sensitive outcomes such as income and mental health symptoms.

The LIFE-DM trial is designed to examine an integrated approach to addressing the dual burden of depression and poverty in low-resource settings. By integrating mental health care with economic empowerment and leveraging community-based human resources, the intervention has the potential to offer a scalable, contextually relevant model that may be adaptable across diverse settings. As global health systems seek sustainable, multifaceted solutions to the dual burdens of mental illness and poverty, LIFE-DM may help inform future policy, programming, and research efforts, including potential integration into Vietnam’s National Health Insurance scheme through reimbursement mechanisms for community-based psychosocial care delivered in primary care settings. Ongoing analyses are expected to provide further insight into its effectiveness and feasibility for broader implementation.