The Sub-Saharan Africa Regional Partnership (SHARP) for Mental Health Capacity-Building Scale-Up Trial: Study Design and Protocol

Depression is a leading cause of death and disability worldwide, including in low- and middle-income countries (LMICs) (1–3), with mortality rates of those affected, and not limited to patients with severe cases or who commit suicide, twice the rates in the general population (4). Despite these high societal, economic, and quality-of-life costs, those with depressive illness do not receive the same quality of general medical health care as people in the general population (4). Depression often coexists with increasingly prevalent chronic medical conditions in LMICs, such as hypertension and diabetes, and is associated with worse clinical outcomes of these comorbid conditions (5). This confluence has led to calls for integrating mental health treatment with expanding chronic disease care systems in LMICs (6, 7).

Despite the high prevalence and negative consequences of depression, most sub-Saharan African health systems have severely limited capacity to treat patients with depression. The median number of mental health professionals per 100,000 population in sub-Saharan African nations is less than one-fiftieth of that in the United States (8, 9), and more than three-quarters of those needing mental health treatment in these countries have no access to such care (10). This enormous mental health treatment gap demands a focus on task-sharing models that equip nonspecialists, such as primary care providers or community workers, with the means to address common mental health problems (10).

Strong evidence exists that both psychosocial counseling and antidepressant treatment are effective for managing depression and that they can be effectively delivered by general practitioners and community workers in research trials conducted in LMICs. For example, with psychosocial counseling, the “Friendship Bench” intervention—a culturally adapted, manualized, problem-solving, therapy-focused counseling intervention developed by Chibanda and colleagues (11) in Zimbabwe—utilizes lay health care workers or community members to deliver structured problem-solving therapy and effectively manage common mental disorders in primary health care settings (11, 12).

Similarly, algorithm-based care for depression (ABCD), a task-sharing depression treatment model that uses antidepressants, has a strong evidence base and high potential for scale-up and impact in low-resource settings. ABCD enables nonpsychiatric staff to manage depression effectively in clinical settings through antidepressant prescription using algorithms that guide dose titration depending on systematically assessed depression symptom severity and adverse effects of antidepressants. The availability and demonstrated efficacy of low-cost antidepressants in low-resource settings make approaches such as ABCD part of a highly promising strategy that efficiently integrates depression treatment into existing clinical care systems. We have conducted research for >10 years on the effectiveness of ABCD in both primary (13, 14) and HIV (15) care and more recently have demonstrated the safety, feasibility, and acceptability of ABCD in Africa (16). Other research has demonstrated the effectiveness of ABCD integration into HIV care in Uganda (17) and its integration into chronic disease care in high-income countries (18, 19).

Despite this strong evidence that ABCD can be effectively integrated into primary clinical care, scale-up of such services in low-resource settings is impeded by a lack of knowledge of the level of resource investment required to achieve acceptable fidelity to the intervention and effectiveness in real-world practice. Implementation science research to identify optimal implementation strategies is therefore critical to advance scale-up efforts. The Sub-Saharan Africa Regional Partnership (SHARP) for mental health capacity building aims to address this knowledge gap. Our objective is to expand mental health treatment in the region by implementing high-quality research; developing research and implementation capacity among governmental, academic, and nongovernmental partners; rand enhancing dialogue between those partners.

One part of SHARP is a scale-up study whose goal is to compare the implementation and effectiveness outcomes of two implementation packages for the integration of depression screening and treatment into standard care in noncommunicable disease (NCD) clinics in Malawi. This clinical environment is ripe for scaling up. The Malawi Ministry of Health (MOH) has strongly prioritized an expansion of depression treatment in its chronic disease clinics (20, 21). However, in a setting with limited resources and few mental health professionals, and where the country’s first psychiatrist in a generation completed medical training only 2 years ago, the most efficient way to implement such a treatment model is unclear. A key knowledge gap impeding the scale-up of Friendship Bench and the ABCD is the level of resource investment required to achieve acceptable fidelity and effectiveness in various clinics.

Accordingly, we will conduct a randomized controlled trial in several clinics to determine the best way to integrate evidence-based depression treatment into NCD clinics in Malawi. We will compare depression treatment involving a basic implementation package with an enhanced implementation package. These two implementation packages were informed by previous implementation science strategies involving external and internal facilitation to implement evidence-based mental health treatment programs in clinical care (22–24). They were adapted through consultation with the Malawi MOH to be most relevant to the clinical context in Malawi. The basic implementation package will involve identifying an internal coordinator (IC) who is one of the full-time onsite providers at the clinic, reflecting an approach often used by the Malawian public health system. The IC provides mentoring to peers and support to leadership in implementing the treatment program and aligning it with clinical priorities. The enhanced implementation package will consist of the IC together with an external quality assurance committee (EQAC), a scalable and feasible package based on previous work in Malawi with HIV patients. This committee will complete quarterly audits at the facility to evaluate compliance with the depression treatment protocol and provide high-level support in implementing the treatment program through clinical expertise and limited onsite presence. Key challenges informed our study design and are listed in Table 1.

Objectives and Hypotheses

This article details a protocol for a scale-up study that compares the clinical effectiveness and cost-effectiveness of two different implementation packages in achieving integration of depression management into NCD clinics in Malawi: a basic implementation package (IC only) versus IC and EQAC (IC+EQAC).

Our hypotheses are as follows. Compared with providers at clinics receiving the basic implementation package (IC only), providers at clinics receiving the enhanced implementation package (IC+EQAC) will deliver depression treatment with greater fidelity to the intervention. Compared with patients receiving care at clinics with the basic IC-only package, patients receiving care at clinics with the IC+EQAC package will be more likely to achieve depression remission at 6 and 12 months. Compared with patients receiving care at clinics with the IC-only package, patients receiving care at clinics with the IC+EQAC package will be more likely to have well-controlled hypertension and diabetes at 6 and 12 months.

Methods

Overall Study Design

Our goal is to compare the clinical effectiveness and cost-effectiveness of two approaches for integrating depression screening and treatment into 10 NCD clinics in Malawi by implementing a clinical trial with 1:1 constrained randomization by clinic, such that five clinics receive the IC-only package and five receive the IC+EQAC package. This constrained design will ensure balance of group-level prognostic factors across the two compared arms, allowing us to avoid imbalance due to geographical region, number of providers, number of patients per month, quality of engagement in NCD service delivery, quality of engagement in initial study site visit, and quality of record-keeping. We will assess outcomes for those receiving care at clinics through either the IC-only or the IC+EQAC package, as indicated by providers’ fidelity to depression treatment; patients’ depression remission at 3, 6, and 12 months; and patients’ well-controlled hypertension and diabetes at 3, 6, and 12 months. We will also compare the cost-effectiveness of the two packages in achieving depression remission and NCD control.

Settings and Site Selection

The trial focuses on secondary-level district hospitals, the setting for the recent establishment of NCD clinics nationwide. After an assessment of key contextual and structural characteristics of each clinic, including organizational readiness for change, management support, patient panel size, patient demographic characteristics, and storage facilities for medications, we selected 10 hospital-based NCD clinics representing the three main regions of Malawi: northern (two clinics), central (four clinics), and southern (four clinics). We will enroll 116 patients from each of the 10 clinics, yielding a total sample of 1,160 patients.

Target Population and Selection Criteria

We will enroll patients meeting the following criteria in the research outcome interview component of the trial: eligible participants will be ages 18–65 years; be current or new patients receiving care for hypertension, diabetes, or both from a participating NCD clinic; and have elevated depressive symptoms (Patient Health Questionnaire–9 [PHQ-9] score ≥5). The PHQ-9 screens for and monitors depressive symptom severity (25) and has been validated in Malawi (26). Patients will be excluded if they have a history of bipolar or psychotic disorder or exhibit emergent threat of self-harm.

A research assistant will evaluate all NCD patients for potential eligibility. Data from all patients, regardless of whether they are enrolled in the research outcome interview component, will be included in the primary fidelity outcomes because the data come from aggregation of deidentified clinical records of all patients.

Study Procedures and Implementation Strategy Design

Depression screening and treatment program provided at both sites.

In both study arms, a depression screening and treatment program is being integrated into routine NCD care. This integration is being supported by the IC-only model in half of the clinics and by the IC+EQAC model in the other half. The depression screening and treatment program at all sites consists of the components described in the following.

Training.

Given empirical support for a “train the trainers” model (27), we trained three clinicians from each site as ICs, a role that includes training their colleagues. These trainers then trained all NCD care providers (clinical officers and nurses) at their facilities in administrating and scoring the PHQ-9 and using the ABCD to screen for suicidal thoughts, antidepressant management, and follow-up treatment decisions. NCD care providers at participating sites include nurses who handle vital signs monitoring and weight assessment and conduct health talks, and one or more clinical officers handle patient consultations and have prescribing authority. Both groups have received a brief orientation to mental health issues, primarily serious mental illness, as part of their nursing or medical training but have not received specialized mental health training.

In addition, five or six peer NCD patients at each site have been trained in the Friendship Bench intervention, which is designed for low-resource settings (11) and which we have previously adapted for use in Malawi (28). The team of Dr. Chibanda, who helped develop Friendship Bench, trained a group of Malawian trainers, who have now trained these peer patients as counselors. Two counselors who were the top performers at each site as determined by their Malawian trainers were identified as local peer supervisors.

Clinical care.

All patients presenting for care at participating NCD clinics will be given a PHQ-9 sheet at registration along with their NCD medical chart (called mastercard). Patients will take the PHQ-9 into the consultation room with a health care worker (either a clinician or nurse) who will complete and score the PHQ-9. Patients with a PHQ-9 score of 0–4 are recommended for no mental health treatment. Patients with a positive screen for the core depression symptoms of low mood or anhedonia (i.e., a score of >0 on the first two items of the PHQ-9, also known as a PHQ-2 score) and with a score of 5–9 are recommended for the Friendship Bench intervention, and patients with a score of ≥10 are recommended for antidepressant initiation. The actual treatment plan is at the clinician’s discretion (patients can both be initiated on antidepressants and referred for counseling). The treatment plan is recorded on a “mental health mastercard” medical chart, which is filed with the PHQ-9 and the NCD mastercard for retrieval at the next visit when the treatment plan is reassessed and adjusted if needed.

Basic implementation package.

The IC-only package consists of identifying a clinical coordinator (internal champion) at each site. This coordinator is an existing onsite NCD clinician and facility NCD coordinator who is responsible for ensuring the smooth integration of the treatment program. One coordinator and two alternates from each site were identified and trained as clinician trainers as described above. These coordinators have the following responsibilities: train all health care workers at the health facility in the depression treatment program (both initial training and as-needed additional training of new arrivals); maintain a log of current health care workers and training status; ensure accurate record-keeping, including numbers of patients screened, of patients with depression, and of those receiving treatment, and conduct follow-up of patients previously started on treatment; lead depression care supervision monthly; supervise Friendship Bench supervisor monthly; ensure NCD clinic coverage; ensure stable antidepressant medication supply; raise implementation challenges with facility leadership; and compile monthly reports to submit to the MOH. The Malawian public health system often uses this basic implementation strategy, including identification of a local coordinator with the aforementioned responsibilities, in the roll out of new health care programs.

Enhanced implementation package.

The IC+EQAC package includes all the elements of the IC package, with the addition of an EQAC that visits each facility on a quarterly basis each year to audit paperwork, review service delivery, and provide feedback and recommendations to the coordinator, the facility leadership, and the MOH. Each EQAC will consist of two to three MOH officials, one Friendship Bench master trainer, and one to two additional staff members for administrative and logistical support. The site visit will last 2 days and will include an initial meeting with the coordinator and facility leadership to discuss successes and challenges, review of record-keeping, and training, supervision, and review of duty logs of clinicians and counselors. It will also include a review of service delivery indicators, completeness of medical charts, and the appropriateness of treatment decisions as recorded on the medical charts. During the visit, clinical flow and delivery of care, including depression screening and treatment plan discussion, will be directly observed. A meeting will be held with all Friendship Bench counselors that will include listening to at least one recorded counseling session and a discussion of successes and challenges. A debrief meeting will be conducted with the coordinator and facility leadership to discuss findings and recommendations, and a written report will be submitted to the facility’s director for health and social welfare, the NCD clinics, and the mental health unit of the MOH.

Measures

Our key measures (Table 2) will allow us to address our hypotheses regarding fidelity, depression remission, and control of hypertension, diabetes, or both. We will monitor these parameters via periodic chart review fidelity measures for ABCD, including the number of completed supervision sessions and the percentages of new patients completing the PHQ-2 (target ≥80%), of PHQ-2–positive screens with the PHQ-9 (target ≥90%), of PHQ-9–positive screens with assessment for depression (target ≥90%), of patients with confirmed depression enrolled in the study who start antidepressant treatment or Friendship Bench counseling (target ≥80%), of follow-up encounters at which the clinical decision follows the algorithm (target ≥80%), and of algorithm divergences for which the reason for divergence is documented (target ≥90%).

Fidelity to Friendship Bench will be assessed through the domains of number of sessions delivered and received and adherence to counseling components. We will measure number of sessions delivered with process data indicating the number of sessions made available to participants by their counselors. Similarly, we will measure sessions received as the number of sessions that participants actually attend. Adherence to the intervention protocol will be measured with the Friendship Bench fidelity checklist, which rates a counselor’s adherence to core Friendship Bench counseling components by assessors’ listening to an audio recording of a counselor’s session. All sessions for which the client consents to recording will be recorded, and 10% of recordings per counselor, selected at random, will be reviewed and rated for fidelity during the course of the trial.

Depression remission will be defined as a PHQ-9 score <5, a score widely used and validated (29). Depression response, a common secondary indicator of treatment effectiveness, will be defined as a ≥50% decrease in the PHQ-9 score from that recorded at enrollment (29). Depression-free days is a cumulative measure of burden of depression over time calculated, according to a previously published algorithm (30–32), by translating depressive severity measures at the beginning and end of each time interval into the proportion of days spent depressed. Hypertension and diabetes outcomes will be measured at 0, 6, and 12 months by trained study data collectors recording the data on a tablet and submitting it electronically for storage and later analysis.

Analysis Strategy

Analysis of primary and secondary outcomes will be intent-to-treat comparisons between the two study arms (Table 2). These comparisons will be made with generalized linear regression models with a log link and binomial error distribution to compare individual-level or appointment-level probabilities. Outcomes (see below) will be measured at baseline and at 3, 6, and 12 months.

Primary outcomes: implementation outcomes (fidelity to intervention).

The study will evaluate three primary outcomes (Table 3). The first is the probability that each patient eligible for depression screening completes the screening. We define this outcome as completing the PHQ-2, and, if the PHQ-2 score is >0, also completing the PHQ-9. The second outcome is the probability that each patient eligible for depression treatment starts the treatment within 30 days of diagnosis. We define eligibility for depression treatment as a PHQ-9 total score of ≥5 and a final determination by the treating clinician that the patient is appropriate for depression treatment. The third primary outcome is the probability at each follow-up appointment in the first 3 months of depression treatment that the clinical treatment decision follows the depression treatment guidelines.

For these three primary fidelity outcomes, where depression screening and treatment actions are clustered within provider and clinic, the analysis will consist of a robust variance estimate clustered by provider and clinic.

Secondary outcomes: effectiveness outcomes (patient health).

Two secondary outcomes will be considered. The first is the probability that each patient achieves depression remission at 3 months. The second is the probability that each patient’s NCD is well controlled at 3 months. For the secondary patient-level effectiveness outcomes, the design effect introduced by clinic-level randomization will be addressed with a robust variance estimate clustered by clinic.

Supplemental analyses.

Planned supplemental analyses will consider patients with hypertension and diabetes separately and consider patients with uncontrolled versus well-controlled baseline NCD separately. Subgroup analyses will be conducted with regression models fit with and without an interaction term(s) between intervention arm and the stratifying variable.

Sample size calculations will be based on two-tailed statistical significance tests, a type I error probability (α) of 0.05, and desired power (1–β) of 0.80. The fidelity outcomes are considered event-specific probabilities. Events (depression screening, treatment initiation, and follow-up action) are nested within providers, who are nested within clinics, the unit of randomization. Assuming a moderately high within-provider correlation (intracluster correlation coefficient [ICC]=0.5) and a lower across-provider-within-clinic correlation (ICC=0.1), and given the large number of providers per clinic, the combination of provider- and clinic-level design effects will yield an effective sample size of N=92 for the primary outcomes. This effective sample size will confer 80% power to detect absolute improvements (risk differences) of 18–27 percentage points in the fidelity outcomes, depending on the outcome frequency (risk) in the comparison arm.

Patient-level outcomes (e.g., depression remission) are patient-specific probabilities that are nested within clinics but not within providers (patients frequently do not see the same provider across visits). Assuming a low intracluster correlation in this patient outcome (ICC=0.02), as has been observed in similar studies (33–36), our sample of 1,160 patients across 10 clinics is equivalent, after accounting for the clinic-level design effect, to an effective sample size of N=352. Assuming 90% retention at 3 months (in line with very high retention rates achieved in many previous large multisite trials by University of North Carolina Project–Malawi) (37, 38), this sample confers 80% power to detect an absolute improvement (risk difference) of 11–15 percentage points in the patient-level outcomes, depending on the outcome frequency (risk) in the comparison arm.

The need for adjusted analyses will be evaluated by comparing the characteristics of participants enrolled in the two arms. Although constrained randomization is expected to yield clinics with approximately balanced characteristics between the arms, clinics in one arm may have more effective screening and therefore may identify different types of patients with depression; thus, patient characteristics might not be balanced between the arms. We will evaluate balance between arms in demographic (age, sex), NCD (hypertension vs. diabetes, controlled vs. uncontrolled) and mental health characteristics (depressive symptom severity, anxiety symptoms, posttraumatic stress disorder symptoms, and alcohol use). We will assess potential confounding by comparing the unadjusted primary analysis to an analysis controlling for potential confounders. Both unadjusted and adjusted results will be presented.

Cost-effectiveness analyses.

We will estimate and compare the average and total costs of the IC-only and IC+EQAC interventions, and the incremental cost per additional case of remitted depression at 3 months of IC+EQAC relative to IC only. These health and economic impact data will inform stakeholders about the sustainability, value for money, and opportunities for efficiency gains of the two interventions. We will measure nonresearch costs of the intervention by using globally accepted best practices. Costs will be captured throughout the trial, and we will measure total delivery costs of each approach deterministically (39). We will use probabilistic sensitivity analysis to account for stochastic uncertainty and use league tables to display results (40).

Trial Status and Timeline

We launched this trial in May 2019. We are planning for 24 months of recruitment (through spring 2021) and 12 months of follow-up (through spring 2022).

Results

The trial will generate critical knowledge to inform policy makers, administrators, and clinicians on which strategies may be more effective at implementing and sustaining standard evidence-based depression screening and management approaches in busy community NCD clinics with limited resources. With fidelity, depression remission, and control of hypertension or diabetes as the main outcomes, along with an assessment of the cost-effectiveness of the interventions, stakeholders will be able to use evidence to inform selection and implementation of depression management packages. Various scenarios can be envisioned. For example, should each package implement and sustain depression screening and management with fidelity and with few differences in clinical outcomes, stakeholders would likely opt for the basic IC-only package. Should both be implemented with fidelity, but the IC+EQAC version produces better clinical outcomes, stakeholders could discuss whether the increased benefit is worth the increased cost. In any case, meaningful information will be created to help stakeholders make a well-informed decision.

Other key results will evolve from the project’s capacity-building efforts. As part of the overall SHARP project, which involves trainees in both Malawi and Tanzania, we will train young researchers to design, implement, and analyze studies of how to implement evidence-based mental health interventions in low-resource settings. Small grant opportunities will allow them to propose their own pilot projects and to participate in the current study. Further, we are partnering with another global mental health hub, Scaling-Up Partnerships in Research to Implement and Disseminate Sustainable and Scalable Evidence-Based Practices in Mozambique (PRIDE) (41), to jointly mentor other young investigators from their hub for related projects in Malawi and Tanzania.

Next Steps

The Malawi and Tanzania MOHs both strongly support the design of the proposed scale-up study because of the perceived high value and relevance of the proposed depression treatment approach to the mental health agendas in both countries and the importance of identifying the most therapeutically effective and cost-effective strategies for integrating depression screening and management into NCD care. Integration of antidepressants and psychosocial counseling into existing NCD care is viewed as an important component of an overall mental health response because the high prevalence of depression and limited availability of clinical providers in Malawi necessitate an efficient, low-cost approach. The Malawi MOH supports our design and views it as scalable and sustainable within the Malawi health care system. The study is seen as a logical next step of ongoing MOH efforts to expand access to depression management and as an important complement to other efforts to expand counseling capacity. Should these findings reveal benefits in such resource-limited settings as in Malawi, we believe these findings will be generalizable to other countries in sub-Saharan Africa.

On its completion, this scale-up study will provide key information to guide the Malawi MOH in furthering its intention of scaling up ABCD in existing chronic disease care settings. In collaboration with the MOHs and other SHARP partners, we will use the study results as the basis for drafting a detailed revision to the national Mental Health Action Plans in Malawi and Tanzania. Further, the execution of this study will offer SHARP mentees the opportunity to gain invaluable experience in mental health–related implementation science research by participating in the design and conduct of specific aspects and procedures of the study under the mentorship of senior investigators. Overall, this component of the SHARP U19 application to the National Institute of Mental Health will substantially contribute to the SHARP goal of enhancing research capacity in both mental health treatment and implementation science in Malawi and Tanzania.