Scaling Up Science-Based Care for Depression and Unhealthy Alcohol Use in Colombia: An Implementation Science Project

Mental disorders are increasingly recognized as a major cause of the global burden of disease, accounting for an estimated 7% of the disease burden worldwide and significantly contributing to disability and death (1, 2). The overall annual economic cost for mental disorders is estimated at more than US$2.5 trillion and is expected to exceed $16 trillion by 2030 (3, 4). Rapidly expanding access to mental health care on a large scale to effect a substantial population-wide impact is a significant global challenge (5, 6).

This challenge is particularly evident in low- and middle-income countries (LMICs) where >85% of the world’s population live (7–9). Between 76% and 85% of persons with severe mental disorders in LMICs receive no treatment for these problems. The health care workforce in LMICs is also grossly insufficient, with on average only one psychiatrist serving every 200,000 people and even fewer mental health providers having training in the delivery of psychosocial interventions. Only 36% of persons in LMICs are covered by mental health legislation, in stark contrast to 92% of persons in high-income countries.

Latin America is one region of the world comprising LMICs where the burden of mental health problems is high and services for mental health are low and where such services account for <2% of the health budget in the region (10, 11). In the Latin American country of Colombia, noncommunicable diseases account for 71% of the total disability-adjusted life years (DALYs) (12), with mental disorders accounting for almost 18% of those DALYs (13). The most recent National Mental Health Survey in Colombia (14, 15) showed that 10.1% of adult Colombians have a depressive or anxiety disorder, 4.3% have major depression, and approximately 12% have unhealthy alcohol use. Contributing to the mental health problems in Colombia is the long-standing violence in the country. Indeed, before the recent bilateral peace negotiations, Colombia had one of the longest internal armed conflicts in the world (16). Only 11% of persons with a mental disorder in Colombia receive mental health care (15). In cases where mental health care is provided, it can be highly effective (17–19).

In 2008, the World Health Organization (WHO) launched the Mental Health Gap Action Program, which includes strategies to scale up care for mental, neurological, and substance use disorders (20). And in 2018, the WHO launched the global SAFER alcohol initiative to reduce alcohol-related death and disability (21). The Pan American Health Organization within the WHO also adopted a Plan of Action on Mental Health (22). Colombia has embraced this plan and has developed a strategy to reinforce primary care–based care for common mental disorders (23).

A key element for achieving the stated goals of the plan is to leverage empirically supported mobile health technology to create a new model of mental health service delivery within primary care (22). Advances in digital technologies have created unprecedented opportunities to facilitate the rapid and widespread scaling-up of evidence-based mental health care. Most of the world’s population has a mobile phone (24), including >90% of persons in Colombia (25). Growing evidence suggests that increased access to these technologies is also evident in many traditionally underserved populations where disparities are prevalent (26–28).

Another key element of Colombia’s mental health care model is human resource development focused on the creation of mental health skills within primary care. Given the strikingly limited availability of the trained mental health workforce in Colombia, raising awareness of evidence-based screening and interventions for mental health within the primary care workforce is critical to integrated care. Mental health policy in Colombia has established mental health care as a fundamental right (16, 29), and similar patterns are evident in other Latin American countries, including Chile and Peru (30, 31).

In response to the mental health treatment service delivery needs in Latin America, an interdisciplinary team launched an implementation research project called “Scaling Up Science-Based Mental Health Interventions in Latin America” (also known as the DIADA [Detection and Integrated Care for Depression and Alcohol Use in Primary Care] project). This initiative, funded via a scale-up hubs cooperative agreement from the National Institute of Mental Health (NIMH) (32, 33), seeks to implement and assess the impact of a new model for promoting widespread access to mental health care (focused on care for depression and unhealthy alcohol use) within primary care settings and building an associated infrastructure to support research capacity and sustainability of the new service delivery model in Colombia. This initiative reflects a collaboration among partners at Pontificia Universidad Javeriana in Colombia, Dartmouth College in the United States, NIMH, as well as many governmental, nongovernmental, academic, industry, and multilateral organizations in Latin America.

In this new mental health service delivery model, we are harnessing mobile behavioral health technology for mental health (with a focus on depression and unhealthy alcohol use) and launching new mental health workforce training and service delivery models (including the integration of technology into service delivery). We are launching this project at multiple primary care sites in various parts of Colombia, with a plan to inform subsequent adoption in several other Latin American countries, including Chile and Peru. This article describes the design and methodological considerations of this multisite implementation science project.

Methods

Overview

At the launch of this project (in year 1), we conducted a pilot test of this new mental health care model for patients with depression and unhealthy alcohol use at a single urban primary care site in Bogotá. This allowed us to refine implementation procedures and measurement before the launch of the main implementation science study. Specifically, we reduced the scope of research assessments to reduce participant burden and increased the frequency with which we provided remote expert consultations with mental health experts on our team to complement our structured onsite trainings in mental health care. We are now expanding implementation across six primary care health care systems in urban and rural communities across Colombia on a staggered basis (in years 2–5). This study does not use a true stepped wedge design (34), but we will implement across primary care systems on a staggered basis (launching in a new primary care system about every 6 months) and expand the number of sites in which we implement over time. In this design, the order of launch of the sites was randomized after starting with a single site we identified in advance (in a nonrandomized way) because it was a site close to the location of our research team in Colombia and also functioned as our pilot site. By conducting this multisite implementation research project, we can assess the extent to which the implementation model and associated outcomes are replicable across sites and how the model needs to be modified for differing contexts.

As detailed in the Measures section, our primary evaluation of implementation outcomes was conceptually informed by the outcomes for implementation research taxonomy developed by Proctor and colleagues (35) (e.g., acceptability, adoption, appropriateness, feasibility, and penetration), and the consolidated framework for implementation research was used to guide the evaluation of the determinants of implementation success (including barriers and facilitators to implementation such as organizational climate and organizational leadership) (36). We will also assess patient outcomes, including depression, quality of life, and alcohol use (the outcomes data are securely stored in REDCap) (37). Most data will be collected via quantitative measures, but we will also conduct qualitative interviews with providers, administrative staff, and patients preimplementation and approximately every 6 months thereafter. Qualitative data will be analyzed via deductive analysis with a matrix based on the relevant scientific literature and goals of this project, followed by thematic data analysis that seeks to understand stakeholder perspectives within their sociocultural contexts. All study procedures were approved by the institutional review boards at Dartmouth College in the United States and Pontificia Universidad Javeriana in Colombia and the NIMH Data and Safety Monitoring Board. See Box 1 for an overview of the project challenges and key advantages and study design solutions.

Care Model

Patients will be offered a multicomponent model of science-based care. Providers at each site will undergo initial in-person training and periodic (in-person, virtual, and online) refresher training and consultation sessions in mental health care conducted by psychiatrists from Javeriana University, Bogotá, Colombia. All patients who consent to participate will be given access to the mobile therapeutic tool. A trained staff member at the primary care site will show them how to use the tool. As discussed above—given the ubiquity of access to technology worldwide—digital therapeutic tools delivered on mobile platforms may enable widespread reach and scalability of evidence-based mental health interventions (28, 38–40).

The mobile therapeutic tool provided to patient participants (Laddr, Square2 Systems) offers science-based self-regulation monitoring and health behavior change tools. Laddr includes tools for activating behavior change, solving problems, overcoming obstacles to effective behavior change, and developing skills for maintenance of health behavior change. These tools provide guidance on the execution of behavior change and on maintaining the end user’s motivation to change. The uniqueness of Laddr lies in the fact that it integrates tools that have been developed via an iterative patient-centered approach and shown (in more than a dozen National Institutes of Health–supported studies) to be highly effective for a wide array of clinically relevant phenomena, including substance use, alcohol use, mental health, risk-taking, chronic pain management, medication adherence, diet, exercise, diabetes management, and smoking (41–47). Laddr is available on multiple platforms (including desktop, Android, iPad, and tablets). To our knowledge, Laddr is the only mobile ecosystem that employs the core science-based therapeutic processes that promote behavior change for a broad array of disorders to flexibly apply to a broad array of populations according to their goals and needs.

In this study, the Laddr components that have been shown to be effective in the management of depression and unhealthy alcohol use are offered to patients who meet diagnostic criteria for one or both of these disorders. This therapeutic content helps to activate and motivate behavior change according to an individual’s values (a process known as behavioral activation [BA]) (48, 49). In this process, individuals are provided with tools and strategies to help identify their values in areas such as health, parenting, family relations, social relations, work and career, leisure, and personal growth. BA helps individuals take steps to create an environment that supports healthy and goal-directed behavior consistent with their values. It also includes systematic tracking of behavior and consequences of behavior to help identify and disrupt self-defeating behavioral patterns.

It also provides problem-solving therapy (PST). PST is a practical and effective intervention for many disorders, including depression. PST’s goal is to teach individuals skills in solving problems to enable them to self-manage and control negative states and behaviors. Its treatment process focuses on participants’ appraisal of specific problems, their identification of the best possible solutions, and the practical implementation of these solutions, as well as increasing exposure to pleasant events (50). PST teaches skills and provides guidance in the execution of behavior change (through cognitive-behavioral therapy [CBT] [51, 52] and the community reinforcement approach [CRA] [53] to behavior change). CBT teaches a broad array of skills and behaviors to manage problematic emotions, behaviors, and cognitive processes and increase and maintain health behaviors. Examples include managing negative thinking, identifying and altering cognitive distortions, communication skills, decision-making skills, stress management, and time management. CRA is an extension of CBT, designed to help individuals establish new healthy patterns of behavior and maintain them and leverage social, recreational, family, and vocational reinforcers to support positive behavior change.

Note that the research team created a version of Laddr that was translated into Spanish and edited to be culturally appropriate to Colombia before its use in this project. During this process, the research team worked with translators and cultural experts in Colombia in forward and backward translation of app content as well as in the cultural adaptation of content (e.g., converting language regarding a standard drink of alcohol to include the local drink “chica”).

Participants diagnosed as having depression may also be offered access to antidepressant medications, as determined in consultation with the primary care physician (in accordance with the Colombian clinical guidelines). Medications for managing alcohol use disorders are not currently available in Colombia. The flow of participant activity is summarized in an online supplement to this article.

Study Sites

We are collaborating with six primary care networks spanning diverse rural, semirural and urban locations across Colombia. These sites also provide diversity in access to and usage of mobile technology. The sites serve between 14,000 and 200,000 patients. None of these primary care systems routinely screened for or treated depression or alcohol use disorders within their primary care programs before the launch of this project.

Study Procedures

Participant criteria.

Providers and administrative staff who provide implementation process and outcomes data in this study must be ≥18 years of age and have worked for the study site for at least 3 months. Patient participants must be ≥18 years old; in care at one of our collaborating primary care sites; screen positive for minor (score of 5–9), moderate (score of 10–14), moderately severe (score of 15–19), or severe (score of 20–27) depression on the Patient Health Questionnaire–9 (PHQ-9) (54) or screen positive for problematic alcohol use (score of ≥8) on the Alcohol Use Disorder Identification Test (AUDIT) developed by WHO (55); have a confirmed diagnosis of depression or alcohol use disorder after a clinical consultation at the primary care site; and be willing to provide informed consent to use the mobile intervention and complete study assessments. Patients are excluded if they have been diagnosed as having a co-occurring severe mental illness (e.g., schizophrenia, bipolar disorder, or depression with psychotic features), have alcohol withdrawal symptoms that require a higher level of care (e.g., emergency or inpatient treatment), or express suicidal intention or are intoxicated or otherwise incapable of informed consent.

Participant recruitment, screening, and informed consent.

Patients are asked to complete screeners (for depression and problematic alcohol use) at the time of check-in on arrival at the primary care site. Screeners are electronic and completed on a kiosk in the waiting rooms of the primary care sites. The depression screener starts with the two-item Whooley et al. assessment (56): “During the past 30 days, have you been bothered by feeling down, depressed or hopeless?” “During the past 30 days, have you been bothered by little interest or pleasure in doing things?” A positive response on either or both questions will then lead to a PHQ-9 (54) screener for depression.

The brief screener for alcohol, the AUDIT–Consumption Questions (AUDIT-C) (57), consists of three questions: “How often do you have a drink containing alcohol?” “How many standard drinks containing alcohol do you have on a typical day?” and “How often do you have six or more drinks on one occasion?” The AUDIT-C is scored on a scale of 0–12. Scores ≥4 for men and ≥3 for women are considered positive for unhealthy alcohol use. Persons who screen positive on the AUDIT-C are asked to complete the full AUDIT (55).

Screening results will be provided to the primary care clinician (electronically or via a printout delivered by the patient or a clinical staff member) when the patient enters an examination room to see the primary care clinician. If a patient has a positive screen for depression or unhealthy alcohol use, the clinician will complete a more in-depth diagnostic interview. Each site has an internal protocol for evaluating and managing suicide risk. Patients with a positive screen for depression (PHQ-9 score of ≥5) or alcohol use problems (AUDIT-C score of ≥8) and meeting all other inclusion criteria will be informed about available treatment options. Patients who provide informed consent with research staff will then be asked to complete baseline assessments.

Clinicians are provided with an electronic decision aid tool (on a tablet) to guide them in offering evidence-based treatments to patients. This is designed as a matrix to help patients understand the options available to them as part of their care (including the advantages and disadvantages of the various options). We also offer videos and other educational resources (e.g., pamphlets) to help patients understand how the care options available to them work to support a process of shared decision making. All patients who join the study will be assigned a unique study identification number (which will be linked to all patient data collected from a given participant).

Measures.

A summary of the measures employed in this study is provided in Table 1, and the flow and timing of data collection for all measures is presented in the online supplement. The primary assessment tool is the Integrated Measure of Implementation Context and Outcomes in LMICs (58). This tool has instruments for the consumer (patient), provider, and organizational staff as well as an instrument that assesses sustainability. The consumer instrument measures an intervention’s acceptability (17 items), adoption (12 items), appropriateness (13 items), feasibility (14 items), and penetration (eight items). The provider instrument measures acceptability (16 items), adoption (nine items), appropriateness (16 items), feasibility (20 items), penetration (eight items), organizational climate (13 items), and organizational leadership (10 items). The organizational measure includes 10 acceptability items, 13 adoption items, 12 appropriateness items, 14 feasibility items, eight penetration items, 15 organizational climate items, and 10 organizational leadership items. Sample items from this assessment tool are provided in Box 2.

In addition, the sustainability instrument was adapted from the Program Sustainability Assessment Tool (PSAT) (59, 60). It assesses eight core domains that affect a program’s capacity for sustainability, including environmental support, funding stability, partnerships, organizational capacity, program evaluation, program adaptation, communications, and strategic planning (61). It has demonstrated internal consistency reliability, structural validity, and usability (59, 62). The Behavioral Health Integration in Medical Care (BHIMC) index is a quantitative organizational measure of the level of behavioral health integration in medical practice settings. It evaluates policy, clinical practice, and workforce dimensions of integration using mixed methods—a combination of document review and observation. The BHIMC instrument was translated into Spanish and adapted for use in Colombia (63).

The costs of implementing the care model will be measured with the time-driven activity-based costing (TDABC) approach (64). This method involves creating a detailed process map to illustrate every administrative and clinical process activated during the treatment for depression and alcohol use disorders over a complete care cycle.

Patient Outcomes

We will use the standardized Patient Health Questionnaire–8 (PHQ-8) as the primary patient outcome measure to assess our hypothesized reduction in depression (65). We will assess problematic alcohol use via the Quick Drinking Screen (66), assessing past–30-day alcohol use. We will use the 12-item WHO Disability Assessment Schedule 2.0 (67) to measure functional status and health-related quality of life and the General Anxiety Disorder (GAD-7) screener, a seven-item screening questionnaire that has been validated in outpatient care settings (68). The Non-Study Medical and Other Services instrument assesses patients’ medical resource use that is not part of the intervention (e.g., nontreatment therapy visits, physician visits, residential or hospital detoxification, and hospital and emergency department visits). The Non-Medical Expenses for Depression instrument assesses the nonmedical costs of depression (69). This measure is in Spanish and has been used in Latin America and by our team at Javeriana University. The Health and Work Performance Questionnaire, developed by WHO, assesses the impact of depression on work performance (including sickness absence, presenteeism, and critical incidents) (70).

Sample Size and Analysis Strategy

We expect that approximately 10 providers and 10 staff at each of the six sites will complete the implementation context and outcomes measure for two to five assessment time points per site (depending on when they launched). We will also collect data on these implementation measures from approximately 25 patients at each of the six sites for two to five assessment time points per site. Our organizational implementation measure (our primary outcome) will yield 80% power at the two-sided 0.05 significance level to detect differences pre- versus postimplementation that are at least 0.55–0.73 times the size of a standard deviation. The BHIMC and the TDABC are collected at the site level, once per site at each of our six sites at each assessment time interval. Assessment of change across time will be primarily used to evaluate qualitative changes pre- and postimplementation, and large effect sizes (1.25 standardized effect or greater) will be observed with 80% power at the two-sided 0.05 significance level. This analysis is independent of the number of patients at each site who are exposed to this model of care.

At a sample size of 1,200 patients completing patient-level clinical outcomes assessments (with at least two measures per participant, a highly conservative estimate of data collection), we will have 80% power at the two-sided 0.05 significance level to detect small pre- versus postimplementation effect sizes of 0.13–0.17. Organizational implementation outcomes and patient-level implementation outcomes will be analyzed with linear mixed-effects models (71), with the primary comparison being mean outcome before and after implementation of the new care model. To account for potential correlations of observations within site, the model will include a random site effect; thus, all statistical tests comparing outcomes pre- and postimplementation will take this within-site nonindependence into account.

Patient-level clinical outcomes will be evaluated with linear mixed-effects models that include fixed effects for time from enrollment to evaluate whether patient-level outcomes improve over time at sites implementing the novel care model. The models will also include a random site effect to account for similarities of outcomes among individuals within the same site and random individual-level intercept and slope terms to account for nonindependence of repeated assessments of each individual. We will also conduct cross-site analyses to examine the extent to which patterns of results are similar or differ across populations and contexts. To integrate data across studies to explore patterns and generality of outcomes, we will use structural equation modeling–based meta-analysis and meta-analytic structural equation modeling (72, 73).

Results

Results to date from baseline data analyses have indicated that we can quantitatively detect the degree to which primary care systems offer integrated mental health care and identify areas for increasing capacity for care (63). Further, our early experience implementing technology-assisted screening and decision support for depression and unhealthy alcohol use into the workflow of Colombian primary care systems has underscored the feasibility and acceptability of this model of care. And we have seen that this system of care markedly increases rates of depression screening and diagnosis (74). Indeed, after having launched at only four of our partnering sites, our partnering primary care sites have gone from not routinely screening any patients for depression or alcohol use disorders to screening >13,000 individuals. And qualitative data from in-depth interviews and focus groups conducted with health professionals, administrative professionals, patients, and community organization representatives in our partnering primary care institutions in Colombia have indicated that digital technology is perceived as useful in evaluating, diagnosing, and treating patients with depression and unhealthy alcohol use in primary care. Perceived potential challenges include technology access limitations and literacy challenges in certain communities (75).

Next Steps

We now plan to complete all participant recruitment in this study in Colombia by the end of 2020 and to broadly disseminate results in scientific and nonscientific outlets (including with our diverse partners in Latin America). We also plan to conduct training sessions in this model of care in primary care systems in Chile and Peru (who have been participating in our project in Colombia from its inception). Overall, this project will create knowledge to inform new science-based approaches to scaling up mental health implementation research in Latin America. This project will also gather information on how to best implement and expand mental health care capacity with innovative digital technologies. The project will also build significant capacity in Latin America for delivering science-based mental health care to meet a large unmet need as well as for conducting systematic mental health research. If successful, this approach could be expanded over time to include other areas of mental health (e.g., severe mental illness), chronic disease management, and health-promoting interventions based on community needs and priorities in Latin America. This project may also serve as an important demonstration project to low-resource contexts globally as they tackle the significant burden of mental disorders and scale up access to evidence-based models of mental health service delivery.