- BACKGROUND 1.1 Migration and Health Migration is a dynamic and complex global phenomenon that shapes the political, economic, demographic, and socio-cultural situations of nations, communities, and individuals [1]. According to the latest available evidence, there were 281 million international migrants worldwide in 2020 [2], accounting for 3.6% of the global population. Migration is triggered by several factors such as political instability, economic difficulties, and environmental disasters [3]. Modernisation, globalisation, stronger social networks, and the desire for a better life are also key facilitators of migration [4]. The effects of migration are multifaceted and can have significant consequences for both the sending and receiving countries [5]. One of the sectors shaped by migration is domestic work [6]. Reports showed that there are 11.5 million migrant domestic workers (MDWs) worldwide [7], representing 7.7% of the global estimate of 150.3 million migrant workers. The Arab States are among the major global destinations for MDWs, accounting for 27.4% of the 11.5 million MDWs worldwide [7]. Nearly three-quarters of the MDWs in the Arab States are women. Most of them are from Asian and African countries such as Sri Lanka, the Philippines, Bangladesh, Nepal, Indonesia, Ethiopia, and Kenya [8]. MDWs provide indispensable services in the destination countries [9]; nonetheless, they are confronted with vulnerabilities, leading to violations of their human and labor rights [10]. Evidence also shows that domestic workers often lack access to timely and appropriate health services.
Ethiopia is one of the major source countries for MDWs to the Middle East. Due to the economic crisis, high unemployment rate, internal conflict, and war, many are still forced to leave Ethiopia to seek employment abroad [11]. In the stakeholder engagement and involvement activity for Phase I grant of this project, the Ministry of Labour and Skills (MoLS) of Ethiopia reported that between July 2023 and May 2024, over 314,000 people, most of them were women, migrated to the Middle East using the legal channel and a lot more using illegal routes. MoLS also has a plan to send more than 700,000 domestic workers to different Middle East countries per year.
A recent phenomenon shaping the migration dynamics in Ethiopia is return migration [12]. According to official reports from the Ethiopian Ministry of Women and Social Affairs, over 70,000 MDWs are being deported from the different Gulf States every year. A significant number of other MDWs are also voluntarily returning. Several factors contributed to return migration [13], including adverse social and cultural environments, emotional maladjustment, serious health problems, and lack of skills to fulfil employers' needs. Most Ethiopian MDWs return before or without achieving their migration objectives.
Several studies documented the adverse experiences of Ethiopian MDWs in the Middle East. These include human rights violations [14], sexual abuse [15], denial of salary, lack of access to healthcare, overwork, and sleep deprivation [16], deportation [17], detention and confinement [18], physical and mental health problems, and stigma and discrimination [19,20]. Migration has also led to social, psychological, and economic crises, affecting not only the migrants themselves but also their families and communities [21]. A preliminary analysis of the scoping review was conducted in Phase I of this project and showed a higher prevalence of both mental (24%-51%) and physical (20%-40%) health issues. Higher prevalence of anxiety, depression, stress-related disorders and physical ailments were reported. During our stakeholder meeting, participants with lived experience of MDW shared their experience, stating, "The emotional, mental and physical challenges we experience, including depression, sleep deprivation, overwork, and inadequate meals, are overwhelming." Most of the studies conducted in MDWs are either qualitative or cross-sectional surveys, mainly among migrant returnees. To the best of our knowledge, there are no studies on Ethiopian MDWs in destination countries nor on interventions that mitigate the negative experiences of MDWs.
In the last few years, a new bilateral agreement was signed between the government of Ethiopia (GoE) and some of the Gulf States on recruiting MDWs. The agreement requires the GoE to offer qualified MDWs who meet certain specific criteria, including medical and physical fitness, to carry out specific tasks. Host countries then imposed comprehensive pre-migration medical screening, covering a wide range of conditions such as tuberculosis (TB), HIV, mental, sexual and reproductive health (MSRH). Twelve private health facilities in Addis Ababa were certified by the Gulf Medical Council to do the screening. Only those deemed medically fit are allowed to travel. Our preliminary analysis of the pre-screening electronic medical data in Phase I showed substantial variability of MDWs in their socio-demographic characteristics such as age, marital status, geographical origin, education, previous migration and domestic work experience, suggesting variable risk profile among MDWs. Regardless, limited awareness of their legal and medical entitlements, coupled with their socioeconomic standing and nature of their work renders MDWs a vulnerable, hard-to-reach population [22].
However, MDWs maintain their pre-migration social connections with their recruitment agencies, peers from the same origin, and families through mobile calls or social media [23], also highlighted during our visit to local employment agencies. These social networks, in combination with the big health data generated during the migration process and technology [24], could be used to provide targeted health and psychosocial interventions tailored to the needs of MDWs. MDWs depart from their residence, interact with employment agencies, complete the medical check-up in Addis Ababa, and then move to their employer in the Middle East.
1.2 The INNOVETH study Fostering community identity and empowering MDWs can be difficult, particularly where MDWs are marginalised, hidden and have low social cohesion. MDWs are heterogeneous in age, education, marital status, domestic work experience and work transitions; all these factors are likely to influence their levels of peer support, behavioural risk, and health service use.
The INNOVETH intervention is specifically designed to reach women while abroad for domestic work and isolated from their families and community. The study is proposed to evaluate the INNOVETH intervention in a pragmatic cluster randomised controlled trial (RCT) nested within the labour migration programme in Ethiopia. The study aimed to determine the population impact and cost-effectiveness of the intervention among MDWs, in terms of the proportion of women who reported psychological distress using a validated psychological distress scale.
During Phase I of this project, investigators developed an intervention which aims to provide risk-differentiated support to engage with prevention and care services as well as to strengthen social cohesion and community empowerment among MDWs. It aims to empower MDWs with:
- Regular, risk-differentiated and structured one-to-one contact with peers (through mobile Apps) that nudges individuals and their networks to stay healthy.
- Strengthened solidarity and space to discuss problems, issues, and norms (through strengthening their social networks and participation in self-help groups)
- Conduit to facilitate transfer of information through the system and supporting programmatic response.
- RESEARCH OBJECTIVES AND QUESTIONS 2.1 Research Objectives The overarching aim of this pragmatic implementation-oriented project is to develop and evaluate a complex mental, sexual and reproductive health (MSRH) intervention package including peer-support, digital skills training, self-help groups, counselling, and mobile health application for Ethiopian MDWs in the Middle East.
The specific objectives of the project include:
- Develop, refine, and validate a risk-stratification tool for MDWs on MSRH to tailor the intervention.
- Refine, pilot test and implement the intervention, and evaluate its feasibility and acceptability.
- Evaluate the impact of the intervention in improving MSRH of MDWs and strengthening their empowerment and social networks.
- Evaluate the cost and cost-effectiveness of the INNOVETH intervention.
2.2 Research Questions The project has the following four research questions.
- What individual and work-related characteristics predict the risk of MSRH in MDWs post-migration?
- How effective is the INNOVETH intervention compared to the usual care (a 21-day pre-departure training (certificate of competence) with respect to MSRH post-migration?
- What is needed, in terms of support among MDWs to achieve optimal MSRH post-migration? To what extent does our intervention achieve this?
- What is the cost and cost-effectiveness of the interventions? 2.3 Overall Hypothesis The study hypothesise that MDWs' MSRH challenges vary based on sociodemographic characteristics; use of MSRH services; and host family characteristics. For example, younger, less educated, first-time migrants in larger or abusive host families may face higher mental health risks. Clusters of MDWs with specific health challenges could be identified and risk-stratified based on their socio-demographic characteristics utilising data science methods. Based on previous experience and evidence, the study also hypothesise that a theory-based complex intervention, adapted to the Ethiopian migrant context, combining skills training, peer support, counselling, MDW community empowerment, and innovative technology, can enhance adherence to MSRH services that will lead to improved health and well-being cost-effectively.
<!-- -->
- Primary endpoint defined as % of MDWs who have psychological distress after 24 months measured using the 21-item Depression, Anxiety and Stress Scale.
- Secondary endpoints defined as 1) % MDWs who have psychological distress after 12 months; 2) % MDWs who have a) depression, b) anxiety and c) stress at 24 months; 3) %MDWs who have self-reported sexual and/or reproductive health problems and sought healthcare services; 4) %MDWs who have self-reported sexual, emotional or physical violence; 5) %MDWs who have self-reported physical injury 6) % MDWs who self- reported pregnancy; 7) knowledge and attitude of MDWs on MSRH and services 8) % MDWs who changed their employer during the two years due to conflict, and 9) the overall well-being of MDWs, measured in terms of quality of life (QoL) at 24 months.
- RQ1 - RESEARCH DESIGN AND METHODS What individual, sociodemographic and migration - related characteristics predict psychological distress among MDWs in the Middle East?
3.1 Risk stratification study The study team will develop a risk prediction model (risk stratification) that uses pre-departure socio demographic and migration-related characteristics to predict psychological distress during the first 2 months of post-emigration among Ethiopian MDWs in Middle East countries.
3.2 Methods 3.2.1 Design: Cross sectional survey The study aimed for at least 10-20 outcome events per predictor parameter (EPV). A minimum of 400 MDWs are required to run the predictive model. The potential participants will be recruited from licensed agencies in Addis Ababa with support from the MOLS, NGOs, and training centers. A 25-30-minute structured questionnaire will be used to collect socio-demographic data, migration experience, and the 21 DASS questions. A structured assessment will be conducted 2-3 months after arrival to collect predictors and outcome variables.
3.2.2 Inclusion criteria
- Age 18 and/or older
- Living or working on the study site (for at least 2 months)
3.2.3 Exclusion criteria
- Newly arriving MDWs who has not lived in the study site for 2 months
- Non-domestic migrant workers
3.2.4 Analysis: logistic regression model
Regression coefficients will be converted into a points-based system, where each predictor is assigned points and summed to generate a total individual risk score. Based on the total score, individuals will be classified into three risk categories:
- Moderate risk (15%-30%)
- High risk (>30%) 3.3 Ethical considerations in relation to RQ1 Informed consent will be obtained to participate in the study. Strict confidentiality measures will be implemented, with no data shared with recruitment agencies or employers. Safety protocols will be established to respond appropriately to cases of distress, suicidal ideation, or violence. To minimize harm, sensitive questions will be placed later in the data collection process, and privacy will be ensured at all stages.
- RQ2, RQ3 and RQ4- RESEARCH DESIGN AND METHODS RQ2: How effective is an added, adapted peer-led support and self-help groups intervention combined with mobile application (INNOVETH) compared to the usual 21-day COC (certificate of competence) training alone? 4.1 INNOVETH trial overview The trial will consist of two arms: an intervention arm and a control arm. Participants in the intervention arm will receive enhanced training packages on MSRH topics and digital literacy, enabling them to access resources and support through the INNOVETH mobile application while in destination countries. The INNOVETH App, which will contain information similar to that available on the existing "Mengede" app (developed by MoLS), will be installed for the intervention group before departure. This package is designed to be implemented at all stages of the migration journey: before, during, and after departure.
The control arm will receive the usual care provided by the government, along with basic training on mobile application use. This training will also facilitate post-migration data collection from the control group. All MDWs, regardless of trial arm, can access the "Mengede" app.
4.2 Specific trial objectives
- To determine the effectiveness of the INNOVETH intervention (peer support and self-help groups based on mobile application) to reduce the proportion of MDWs with MSRH problems post-migration.
- To identify the support needed for Ethiopian MDWs to achieve optimal MSRH post-migration and to find out the extent to which the intervention achieves this.
- To determine the cost and cost-effectiveness of the intervention. 4.3 Theory of Change This project draws inspiration from the peer-led microplanning model developed by the Avahan India AIDS Initiative, which optimised female sex workers (FSW) programme coverage through trained peer educators using a risk-based vulnerability algorithm [25,26]. Similarly, the AMETHIST intervention in Zimbabwe, a peer-led support initiative for FSWs through the structured risk algorithm, successfully reduced HIV transmission [24,27,28, 29, 30]. Building on this evidence and leveraging pre-migration medical screening and socio-demographic data among Ethiopian MDWs in the Middle East, the study proposes a theory-based, risk-differentiated intervention, tailored to Ethiopian MDWs. This approach aims to enhance social cohesion, empower communities, and improve MSRH outcomes among Ethiopian MDWs in the Middle East.
4.4 Trial Endpoints 4.4.1 PRIMARY ENDPOINT The primary outcome is a composite psychological distress index including depression, anxiety, and stress among MDWs. This is the proportion (%) of MDWs recruited to the survey who have psychological distress after 24 months, measured using the 21-item DASS [31]. This primary endpoint data will be collected through cross-sectional surveys conducted across all clusters of Ethiopian MDWs after 24 months of risk stratified intervention.
4.4.2 SECONDARY ENDPOINTS
- % MDWs who have psychological distress after 12 months.
- % MDWs who have a) depression, b) anxiety and c) stress at 24 months.
- % MDWs who have self-reported sexual and/or reproductive health problems and sought healthcare services.
- % MDWs who have self-reported sexual, emotional, or physical violence.
- % MDWs who have self-reported physical injury.
- % MDWs who have self-reported pregnancy.
- Knowledge and attitude of MDWs on MSRH and services.
- % MDWs who changed their employer during the two years (contract period) due to conflict.
- The overall well-being of MDWs, measured in terms of quality of life (QoL) at 12 months and 24 months.
- Perceived levels of peer support among MDWs.
- Acceptability and perceived quality of health services.
4.4.3 PROCESS EVALUATION ENDPOINTS A comprehensive process evaluation will be conducted throughout the trial period.
Indicators of implementation Intervention 1) Number of empowerment workers hired and trained 2) Weekly individual supervision meeting and monthly collective support meeting with all empowerment workers.
- Number of modules, number of mobile applications, number of training sessions prepared, number of peer support provided Self-Help Groups (SHGs)
- Number of EWs trained to initiate and support 2x/month Self Help Group (SHG) in 6-month cycles.
- Number of EWs retained in post over 2 years.
- Number of Outreach Workers (Ows) employed to oversee and troubleshoot SHG programme.
- Biweekly supervision SHG meetings by outreach workers for the first three months with monthly supervision from 3 months to 12 months.
4.4.4 COST ANALYSIS ENDPOINTS
- Total and incremental economic cost
i. Estimate total and incremental costs for routine care alone vs. routine plus INNOVETH intervention.
ii. Costing will follow a health system perspective (or societal, if applicable).
b. Unit cost measures i. Average costs per person reached/ contact ii. Average costs per person per service delivered. c. Variation by service delivery sites and over time (program maturity) i. Assess whether total costs, cost profiles and unit costs evolve over time due to programme learning effects and as staff get more efficient at their roles.
ii. Longitudinal nature of INNOVETH provides an opportunity for econometric analysis of cost determinants.
d. Key cost drivers i. Identify major cost contributors including program management, service delivery inputs, and location.
e. Discounting i. Costs will be discounted at 3% annually, consistent with the cost-utility analysis 4.4.5 COST EFFECTIVENESS AND COST-UTILITY ENDPOINTS Cost-effectiveness Costs associated with developing, adapting, delivering, and maintaining the intervention will be included, as well as staff costs for training, moderating the online forum (application), and providing telephone counselling. To compare costs and outcomes between the intervention and control groups, an incremental cost-effectiveness ratio (ICER) will be calculated.
Cost-utility The incremental cost-effectiveness ratio (ICER) will be compared to relevant WTP thresholds to determine whether the intervention provides good value for money within the Ethiopian health system context. Further, the overall benefits of the intervention will be assessed in terms of quality-adjusted life years (QALYs), which will serve as the primary outcome for the cost-utility analysis. QALYs will be estimated using the EQ-5D-5L instrument, a generic utility-based health status measure developed by the EuroQol Group, applying the Ethiopian value set (32]. This approach captures the overall improvement in well-being resulting from reductions in the frequency and severity of psychological distress attributable to the intervention.
4.4.6 DETERMINANTS OF TARGET BEHAVIOURS This protocol examines the key behavioural determinants influencing the target outcome behaviors essential to the intervention's success: engagement with support services, trust in digital application, psychological resilience, and reduced vulnerability to health risks. Individual, social, and structural factors-such as self-efficacy, social support, digital literacy, and perceived institutional trust-that may act as barriers or facilitators to these outcomes will be assessed.
4.5 Study locations The study will be conducted both in Ethiopia and in destination countries including the Kingdom of Saudi Arabia, United Arab Emirates and Qatar. Potential migrants come to Addis Ababa, the capital city of Ethiopia, for pre-migration comprehensive medical screening, to meet their local recruitment agencies, and to travel to their respective destination country. Participants will be recruited in Addis Ababa, in collaboration with recruitment agencies, when they come for these services. The study will run through SPHMMC, Ethiopia which provides tertiary care in the country including MSRH services. Digital literacy training will be provided and the study will collect pre-migration data while MDWs are in Addis Ababa, before their travel. In the destination countries, baseline data will be collected, implement the intervention package (peer support, self-help group, mobile application, and remote specialist counselling services) and collect midline and endline data. Twenty-four districts designated as migration hotspots will be randomised to the INNOVETH intervention or usual care. Randomisation will be restricted on a range of key factors such as district population size, the proportion of MDWs migrating from the district each year, language and culture. The study sites will be drawn from purposively selected migration hotspot districts and zones in Ethiopia. Randomisation will be done among the following twenty-four migration hotspots: Assela, Shashemene, Jimma city, Limmu, Asendabo, Agaro,Kombolcha, Bati, Kemisse, Dessie, Hayke, Were ilu, Debre Birhan, Ataye, Worabe, Dalocha, Welkite, Butajira, Hossana, Wolaita-Sodo, Bodity, Assosa, Addis Ababa and Shagar City.
4.6 Standard of care and the INNOVETH intervention The implementation and impact evaluation will be based on a pragmatic, two-arm cluster RCT with a process evaluation. The standard of care (usual care) refers to the national migration program. MDWs in the same district speak the same language and have similar culture. This will enable the use of existing social connections, ease building new networks among MDWs, minimize culture and communication barriers, develop efficient peer-led SHGs, and facilitate intervention implementation. The recruitment strategy will ensure a representative and scalable approach, while addressing logistical challenges associated with a mobile population. Selected 24 districts will be randomly assigned to the intervention and usual care, with 12 clusters in each group.
From each cluster, MDWs (n=200 per cluster) willing to e-consent (electronic consent) will be recruited to participate in the study. MDWs in the intervention clusters will receive: i) a pre-departure skills training, ii) peer-led support, iii) SHG, and iv) a Mobile application with MSRH content, in addition to the basic information provided in the 'Mengede' app. MDWs in the control clusters receive pre-departure training on use of the mobile application and information included in the 'Mengede' app but without any MSRH information. Outcomes will be measured at pre-departure, baseline (2 months), midline (12 months), and endline (24 months) of recruitment. The mobile application's analytics functionality will track user engagement and collect data for impact evaluation.
4.6.1. USUAL CARE In the labour and skills office in the nearby district office, potential MDWs will provide their fingerprints, and a CV is created for them on the labour market information system (LMIS). Authorised institutions provide MDWs COC after completing a standard 21-day training consisting of how to navigate destination countries and a review of existing rules and procedures (the usual care). This mandatory 21-day COC skills training is provided by the technical and vocational Education and Training (TVET) colleges and certified private training institutions. Unlike first-time migrants, the duration of the COC training is 6 days for re-migrants. Upon completion of the training and passing the examination, they will be issued with a COC from the MoLS. With this certificate, they travel to Addis Ababa and connect with one of the foreign employment agencies, who had partner employment agency in the destination countries. When a host family chose the MDWs for employment, they will complete pre-departure medical screening in one of the 12 certified private hospitals or diagnostic centers in Addis Ababa. This certification is done by the Gulf Medical Council in the Middle East.
All MDWs undergo routine medical screening to ensure 'fitness' for travel. Those who are 'unfit' to travel, will receive tailored basic health consultation linked to nearby health institutions for follow-up treatment. The result of the pre-departure medical screening is electronically recorded and transmitted to the destination countries as well as the MoLS system. Those MDWs who are deemed "fit" to travel will complete the visa application through their employment agency and will travel to their destination. Upon arrival or within a few days, they will be subject to the same medical checkup, and this information is retained in the destination countries. In case of pregnancy or positive diagnosis for HIV or TB, they will be returned to Ethiopia. Currently, all MDWs can access the "Mengede" app, an app developed by the MOLS, post-migration. This app is integrated into the 'Telebirr' app, a mobile payment service app developed by Ethio Telecom, the state-owned telecommunication and internet service provider in Ethiopia.
As part of the INNOVETH project, INNOVETH digital skills training will be developed supported with mobile application that integrates MSRH information in text and audio. However, the app for the control cluster is similar to the intervention app but only providing information available on the "Mengede" app.
4.6.2. INNOVETH INTERVENTION The intervention starts with an enhanced training package of psychosocial education on resilience and mental health, comprehensive sexual and reproductive health (SRH) education, and digital literacy training to equip MDWs with access to resources and support through a mobile application in their destination.
The intervention package is designed to be implemented at all stages of the migration journey: before, during and after departure. The intervention will also leverage the mobile application to deliver educational materials in text and audio formats, which will only be available for the app in the intervention group. When MDWs visit their employment agency for routine pregnancy check in 48 hours before departure, the INNOVETH mobile application will be installed, provide MDWs with training on digital literacy and using the mobile application to access health education, and link them with EWs for peer support. EWs will receive an intensive one-week training to enable them to support MDWs based on their specific needs and risk profiles and take a diary to record relevant data. This ensures personalised and scalable support tailored to each MDW's risk level. A group of 40-50 MDWs in the intervention arm will be linked to a trained EW for individualised peer support.
EW will also invite some of the MDWs to join an SHG, which includes 5-10 MDWs who will meet every two weeks. Each SHG will identify their own priorities, discuss topics and activities such as saving money, and receive support towards these. The SHGs will strengthen social networks and MDWS' resilience, develop ownership of the intervention and community empowerment, and uptake of MSRH services. The risk stratification tool and the intervention package will be refined iteratively and validated through expert consensus meetings and MDWs community and stakeholder engagement and involvement.
Trained EWs will contact fellow MDWs regularly through virtual or in-person meetings, assess their vulnerability on a quarterly basis, record data on each contact, and adjust the frequency of contact with MDWs guided by data collected at their previous contact and MDW's level of risk: weekly for those at the highest risk and monthly for those at the lowest risk. EWs will also support MDWs on mobile applications and respond to crises such as workplace conflict, abuse, or health emergencies. EWs will work closely with their supervisors based in Ethiopia and discuss the data, risk stratification, and topics for discussion at the subsequent contact. They will also refer MDWs experiencing moderate to severe mental health issues, to specialised MSRH services at St. Paul's hospital Millennium Medical College in Ethiopia via telephone. Figure 5 summarises the intervention package.
4.7 Trial Design The trial is a cluster RCT. A two-arm, cluster RCT design will be used for this study to prevent information contamination (spillover effects).
4.7.1. CLUSTER DEFINITION Cluster refers to a district, administrative and geographical migration hotspot in Ethiopia where MDWs originate.
4.7.2. RANDOMISATION Twenty-four districts will be randomised (1:1) to receive the INNOVETH intervention or standard care. To maximise transparency and buy-in from key stakeholders, the research team will invite MoLS, district representatives and representatives of MDWs and employment agencies. Imbalance between arms will be minimised by using restricted randomisation and matching of clusters using 2025 district level migration data.
4.7.3. MEASUREMENT OF PRIMARY AND SECONDARY ENDPOINTS Endpoint data will be collected from the sources described in Table 5 below Primary Endpoint Data Source % of all MDWs recruited to the survey who have psychological distress after 24 months measured using the 21-item Depression, Anxiety and Stress Scale. Survey Secondary Endpoints Data Source(s)
- % MDWs who have psychological distress after 12 months. Survey
- % MDWs who have a) depression, b) anxiety and c) stress at 24 months. Survey
- % MDWs who have self-reported sexual and/or reproductive health problems and sought healthcare services. Survey
- % MDWs who have self-reported sexual, emotional or physical violence. Survey
- % MDWs who have self-reported physical injury. Survey
- % MDWs who self-reported pregnancy. Survey
- knowledge and attitude of MDWs on MSRH and services. Survey, Process data
- % MDWs who changed their employer during the two years due to conflict, and 9) the overall well-being of MDWs, measured in terms of quality of life (QoL) at 24 months.
4.8. STATISTICAL ANALYSIS Primary analysis will be based on intention-to-treat principle, using cluster summaries to account for clustering. Cluster summaries of individual MDWs data will be calculated, then the mean of the cluster summaries will be calculated for each group. For unadjusted analysis a linear regression model will be fitted on the cluster summaries at the endline, with treatment dummy and cluster summaries at baseline as regressors. The p-value and confidence intervals will be used for the coefficient of the treatment dummy. To address the possibility of confounding, adjusted analysis will be conducted using the 'two-step' method of Hayes and Moulton [36,37]: i) individual logistic regression model will be fitted with the primary outcome as the dependent variable and age (specified a priori) and other variables, that appeared imbalanced post-randomisation, as independent variables to generate predicted probabilities for each MDWs. The mean of the predicted probabilities will be interpreted as predicted prevalence for each cluster. ii) the 'residual' risk difference (RD) will be calculated by subtracting the predicted prevalence from the observed prevalence in each cluster. Then, a linear regression model will be fitted.
