Scale-Up Study Protocol of the Implementation of a Mobile Health SBIRT Approach for Alcohol Use Reduction in Mozambique

Year 1 will be devoted to SBIRT implementation (32), which will involve training and supervising staff, monitoring services and fidelity, providing implementation feedback to providers, and interviewing providers about implementation. In year 2, we will assess the sustainability of mSBIRT and SBIRT-CTS delivery (32). These implementation and sustainability phases will occur within the context of the cluster-randomized trial. In year 3—the scale-up phase—the more cost-effective strategy will be implemented in districts that received the less effective strategy.

Regardless of the trial arm, Ministry of Health procedures require that individuals who are reached, in need, and provided with care be seen weekly and screened with SBIRT. Those who are reached and engaged in follow-up constitute a retained population. Each weekly contact with a provider is documented by the first two items from the AUDIT, the 12-Item Short Form Health Survey (SF-12) (33), and the Mental Wellness Tool (mwTool, described below).

All patients and providers will provide written informed consent. Even though risks to provider participants are minimal, patient participants may experience increased distress and suicidal ideation or report physical or sexual abuse. In accordance with the Ministry of Health safety procedures, local coinvestigators (all clinicians) will be notified of an acute crisis within 24 hours. Additionally, we will convene a five-member data safety and monitoring board that will include an implementation scientist, statistician, mental health researcher with expertise in low-resource settings, Mozambique context expert, and ethicist. They will assist in monitoring adverse events and meet annually.

In the initial 6 months of the implementation phase, SBIRT and motivational interviewing experts will provide a 1-week training for Ministry Mental Health specialists. Trainees will present five cases in a group modality to become certified national trainers and supervisors. These newly minted trainers will then train and supervise the CHWs. We, the research team, will work with the trainers to adapt and pilot mSBIRT. We will also conduct a mixed-methods evaluation of the acceptability and feasibility of tablet use and capture CHWs’ intentions of using mSBIRT.

As part of the larger mental health study (31), all CHWs will be trained to use a tablet-based, comprehensive mental disorders assessment tool—the mwTool—that identifies presence of substance use disorders, common mental disorders, severe mental disorders, and acute risk for suicide. We will measure trainees’ knowledge (information), attitudes toward people with hazardous drinking (motivation), and self-efficacy (skills) in providing SBIRT before and after training (25, 29).

During the following 6 months of the implementation phase, the research team will provide active supervision through observation of role-play and case presentation sessions. CHWs will complete three cases and submit transcripts of the case sessions to their supervisor. Supervisors will evaluate the transcripts by using a treatment fidelity checklist comprising core intervention components and provide feedback to CHWs. If supervisors determine that a CHW conducts the SBIRT with insufficient fidelity or lacks competency, supervision and clinical support will continue until fidelity and competency are achieved.

The mSBIRT application software guides providers in following standardized treatment steps involving interactive methods, including knowledge tabs with tips and sample scripts, and automated processes to enable providers to recognize any clinical “red flags.” Metadata from the application will track individual provider implementation outcomes (adoption) and fidelity. Provider-level user data will indicate patients seen, key activity completion, and sessions per patient. If the user data of any provider appear to deviate from the standard sequence of activities, this will be flagged for supervisor intervention and corrective feedback to maximize fidelity. We will capture similar data in the SBIRT-CTS via provider and supervision checklists.

The 12-month sustainability phase will begin once all CHWs and supervisors have achieved competency in either mSBIRT or SBIRT-CTS. We will monitor the CHWs, supervisors, and any Ministry of Health personnel involved in the study, but we will not provide supervision unless requested or deemed necessary on the basis of performance. We will sample six clinics in each arm from rural (N=2), suburban (N=2), and urban (N=2) areas. We will conduct 12 audio-recorded focus groups (1–1.5 hours each) with CHWs (one group per catchment area) and 24 semistructured household interviews with patients or their family members (two interviews per catchment area).

In the scale-up phase, interviews will continue only in the six clinics of the superior arm: one focus group with CHWs at each clinic (N=6) and two household interviews per catchment area (N=12).

The goal is for CHWs to use the mwTool to screen annually every household in their catchment area. For those who score positive for substance use disorders, CHWs will administer the AUDIT whose score will guide intervention. Patients with an AUDIT score of abstinence or low risk (<5) will receive feedback and follow-up in 4 weeks. Those with scores 5–15 will be considered to be at risk for or to have hazardous substance use and will receive a brief intervention weekly for 4 weeks. Patients who score in the harmful use range (AUDIT score ≥16) or in the dependent range (≥20) will also receive the four-session intervention plus medical monitoring by a primary care provider or psychiatric technician. Before the fourth session, CHWs will administer the AUDIT again. The fourth session reassessment will determine whether patients will be discharged (if they have an AUDIT score indicating abstinence or low risk), referred to an ongoing group intervention at the community level (AUDIT score is the same or improved, but not in low-risk range), or treated by a primary care provider or psychiatric technician (if they have worsening symptoms).

SBIRT sessions last 45–60 minutes and involve screening and a brief intervention consisting of providing feedback and health information, engaging patients in motivational conversations, and negotiating a plan for change (34). SBIRT incorporates principles of motivational interviewing—using an empathic style, supporting motivation and self-efficacy for change (35), building awareness about effects of alcohol use, and creating an action plan. We will establish a Mozambican SBIRT-mSBIRT workgroup with specialists who have expertise treating substance use disorders to adapt the treatment to the Mozambican context. Our adaptation workgroup will provide input into the mSBIRT application itself and suggestions for the general content and icon development.

To determine the effectiveness and implementation outcomes, we will use the AUDIT and the mwTool. In both arms, the CHWs will screen household members yearly for mental health and substance use problems with the mwTool, which incorporates some AUDIT items as well as screeners for other mental health problems. Each CHW has a catchment area of 250–400 households (36). To decrease CHW burden, the CHWs will administer the mwTool to only two household members (ages ≥16 years) per home but will ask about other household members. Local research has shown that surrogate respondents provide valid information on alcohol intake (8). CHWs will individually administer the full AUDIT (37) and SF-12 (33, 38–40) to anyone age ≥16 years screening positive for substance use disorders on the mwTool (Table 2). On the basis of AUDIT scores, CHWs will deliver either mSBIRT or SBIRT-CTS, according to the trial arm. The full AUDIT will be administered at sessions 1 and 4 to measure effectiveness. The study measures are organized by RE-AIM outcomes (Table 2) (41–55).

For aim 3, we will use a mixed-methods approach to identify organizational and clinician factors that affect SBIRT implementation. Interview guides and analyses will be guided by our conceptual model combining the theory of planned behavior with organizational constructs from the CFIR (28). We will triangulate quantitative and qualitative data to identify the factors influencing adoption and sustainability and facilitate translation of research findings into effective policy. In the “crossover” phase, adoption and causal model measures will be administered to all participants. In each focus group, we will present our quantitative results in easy-to-interpret graphics and descriptions. We will ask participants to consider the results, adoption rate, and any associations between clinician (intentions) and organizational (culture, climate) factors and adoption.

The cluster-randomized trial will address hypotheses 1 and 2 regarding the implementation and effectiveness of the interventions across the eight mSBIRT and four SBIRT-CTS districts. The hypotheses will be used to compare the interventions’ effect on outcomes measured for individuals and CHWs aggregated at the clinic level. We expect to include 62 clinics across the 12 districts, employing 417 CHWs who serve between 104,250 and 166,800 households. Clinic-level outcomes will be measured at baseline and then at 12, 18, and 24 months. Because clinics are nested within districts, we will use hierarchical linear models (HLMs), which will include random effects on each level and fixed effects for the intervention indicator and for the time variable.

Hypothesis 1 will test whether mSBIRT is superior to SBIRT-CTS on each of the RE-AIM implementation outcomes, by using HLM analysis, including an indicator for intervention, time (baseline, 12, 18, and 24 months), and intervention × time interaction. The magnitude and statistical significance of the interaction term will indicate whether a change in implementation outcomes over time significantly differs by intervention arm. If the arms differ on baseline measures, we will include those measures as control variables in the HLM.

Hypothesis 2 investigates the interventions’ effect on clinical outcomes (first and second subhypotheses) and on costs (third subhypothesis). For the first subhypothesis, we will use an HLM (controlling for intervention and time) to estimate changes in hazardous drinking and health-related quality of life at months 12, 18, and 24 after SBIRT exposure. Overall intervention effects will be tested over time, and interactions of the intervention by gender, age, and urbanicity of the patient will test for effect modification. For the second subhypothesis, we will again use an HLM (controlling for intervention and time) to predict changes in scores from baseline in health-related outcomes (viral load and HIV or TB care adherence) at months 12, 18, and 24 by using changes in hazardous drinking from baseline. We will assess effect modification by gender, age, and urbanicity by including them as interaction terms. Finally, the third subhypothesis will test whether mSBIRT is more cost-effective than SBIRT-CTS in improving both implementation (hypothesis 1) and clinical outcomes (first and second subhypotheses of hypothesis 2). Comparison of costs will be aggregated across the sustainability phase in year 2. We will assess health-related quality of life with the SF-12, which can be converted to construct quality-adjusted life years (QALYs) (56). QALY is a measure of health-related quality of life that incorporates estimates of mortality and morbidity rates into a single index for ease of comparison across conditions and interventions. Incremental cost-effectiveness ratios will identify the superior arm for the scale-up phase (numerator=difference in mean costs between arms, denominator=difference in reach and reduction in hazardous drinking). Uncertainty in costs and outcomes will be assessed with cost-effectiveness acceptability curves (57–59).

Power was calculated on the basis of participation of approximately 480 CHWs located across clinics in 12 districts and an anticipated dropout rate of about 20%. Each CHW visits between 250 and 400 families in the community at least once a year, resulting in a coverage of approximately 150,000 families (about 75,000 families within each randomized condition). For hypothesis 1, to test for differences in implementation outcomes between randomized study condition at the family level (e.g., percentage households screened), we will have 80% power to detect small to moderate effect sizes between the two conditions, with Cohen’s d=0.23–0.36, following standard power calculations for cluster-randomized trials (60). This power calculation was derived by assuming an intraclass correlation (ICC) of families within the 12 randomized districts in the typical range for health care outcomes across randomized regions, that is, an ICC of 0.02–0.05 (61, 62). For hypothesis 2, we conservatively assumed that 3%−5% of the population will be identified as having engaged in hazardous drinking and will be enrolled in either intervention. Thus, we expect that 4,500–7,500 clients will enroll across the two randomized study arms nested within 480 CHWs who will provide treatment. Given this sample of treated individuals, we will have 80% power to detect intervention differences of Cohen’s d ≥0.16. This power was calculated by assuming an ICC of 0.4 for client outcomes within CHWs, which is conservatively at the high end of ICCs found within health service providers in low-income settings (63). All power calculations assume an alpha of 0.05 and that all tests are two-tailed.

For hypothesis 3, we will examine organizational and clinician-level factors affecting implementation of the intervention with data from the implementation, sustainability, and scale-up phases. The first subhypothesis of hypothesis 3 is that organizational factors such as urbanicity, culture, and climate will affect clinicians’ attitudes, perceived norms, and self-efficacy regarding SBIRT use, which will in turn affect their intentions to use SBIRT. The second hypothesis is that clinicians’ demographic characteristics and intentions to use SBIRT will affect SBIRT adoption and use. To test both subhypotheses, we will use structural equation modeling (64, 65) to estimate the effect of organizational factors and clinicians’ demographic characteristics on determinants of intentions (attitudes, norms, and self-efficacy) and the effect on intentions (Figure 1). Organizational factors will include urbanicity, number of staff, patients’ characteristics and the ready-set-change measure (66). Consistent with best practice, we will aggregate these organizational measures at the clinic level. Intentions and determinants of intentions to adopt SBIRT will be measured with validated self-report measures (Table 2). The use of structural equation modeling allows direct incorporation of clinic-level random effects and provides straightforward estimation of indirect effects. We will also use structural equation modeling to estimate the effect of intentions on adoption, while controlling for the clinic and clinician characteristics described above. We will use interaction terms to further test for effect modification of the organizational factors on the association between intentions and adoption.

For qualitative analysis, interviews and focus groups will be digitally recorded, transcribed, and loaded into Nvivo software for analysis. To avoid bias, analysts will be blinded to implementation outcomes. Initial coding will draw on the CFIR, and an inductive process of iterative coding will identify themes and categories. Analysts will then be unblinded to implementation outcomes to determine distinguishing constructs. Finally, for mixed-methods data analysis, quantitative data are gathered before qualitative data and weighted equally: QUAN→QUAL (67, 68); the function is complementary, and the process is connecting (having the qualitative data set build on the quantitative data set) (67).