Using AI to reduce maternal and neonatal mortality in SSA.

Maternal mortality and morbidity are the most significant health challenges facing vulnerable women and adolescent girls in sub-Saharan Africa (SSA), exacerbated by the barriers they face in accessing Sexual Reproductive Health (SRH) care, including the impact of social norms on their personal agency, and the limited access they have to financial resources and health information. While the number of women and girls who died each year from complications of pregnancy and childbirth declined significantly between 2000 and 2017, from 451,000 to 295,000, over 800 women still die each day. However, the levels of maternal mortality remain unacceptably high in SSA – 533 maternal deaths per 100,000 live births, or 200,000 maternal deaths a year, constituting over two thirds (68%) of all maternal deaths per year worldwide. The tragedy is that almost all maternal deaths can be prevented, as is illustrated in the disparities across regions and between the richest and poorest countries. In high-income countries, the lifetime risk of maternal death is 1 in 5,400, compared to 1 in 45 in low-income countries. According to the World Health Organization, Sub-Saharan Africa also had the highest neonatal mortality rate in 2020 at 27 (25–­­32) deaths per 1000 live births, and a child born in this region is 10 times more likely to die in the first month than a child born in a high-income country. While country-level neonatal mortality rates in 2020 range from 1 death per 1,000 live births to 44, the risk of dying before the 28th day of life for a child born in the highest-mortality country was around 56 times higher than the lowest-mortality country.

These high rates of maternal and neonatal morbidity and mortality in SSA as a result of preventable causes (including infection, malnutrition, obstetric complications, inadequate antenatal and postnatal care, and a lack of financial resources) need to be addressed. The risk of adverse pregnancy outcomes is distributed highly unevenly across and within population groups, with women living in poverty, adolescent girls, women with advanced gestational age, rural women, and women living with certain health conditions (such as high blood pressure, hypertension, diabetes, or sexually transmitted infections, including HIV) disproportionately impacted. With 44% of its population below the age of 15 years, SSA is the youngest region in the world, with a large proportion of the population either having entered adolescence or due to enter adolescence within the next decade, SRH is one of the most significant health challenges facing adolescents in SSA. Pregnant adolescents do not benefit equally from routine antenatal care, so a more targeted approach is necessary to address their specific needs.

Goal:   To reduce maternal and neonatal morbidity and mortality, adverse events and their preventable causes including HIV, Syphilis, Hepatitis B and Cervical Cancer  

Objectives:  

1. To develop and implement an AI model to predict risk of adverse events in antenatal and postnatal women and their neonatal care 

2. To increase the knowledge of and access to antenatal, obstetric, postnatal and neonatal care 

3. Deploy scalable solutions that enable women to reach care during the pregnancy, childbirth and postpartum continuum, regardless of where they seek care 

4. To improve health system capacity to deliver optimal maternal health and appropriate obstetric emergencies and neonatal care 

Our solution will utilize advanced data science and AI modelling to develop an algorithm that predicts a pregnant woman’s risk of experiencing adverse outcomes in the antenatal and postnatal period including their new-borns.

m2m will integrate the risk stratification/assessment as part of essential mother/child services during pregnancy and the post-natal period until the child is 2 years old. We will deploy this risk assessment to support tailored services to increase awareness and ensure that all “high risk” pregnant and breastfeeding women (PBFW) are closely monitored by Peer Mentors, and when necessary, refer emergencies to health facilities. The solution will strengthen m2m’s use of digital health tools in educating and creating community networks to plan, lead and deliver lifesaving emergency obstetric and neonatal care (EmONC) interventions, through (virtual) training and mentoring of health care workers and community health workers on EmONC. m2m will use its current Peer via Phone (PvP) eservice delivery platform to provide comprehensive client-centred and needs-driven counselling, risk management and support to all antenatal and postnatal women and their families, and promote the use of the Virtual Mentor Mother Platform (VMMP), an interactive Whatsapp based platform that delivers on demand health messaging to all clients, including content on antenatal and postnatal risks and warning signs.

We will develop an AI and machine learning model trained on data to predict which women are most likely to experience adverse birth outcomes and postnatal complications through inputs that identify the highest risks on: 

• Likelihood to experience adverse events, for example using m2m risk stratification criteria on nutritional status, age, time between pregnancies and lifestyle factors (Gestational Diabetes, Hypertension etc.) 
• When and how to seek antenatal and postnatal services, for example, women’s or community health workers’ ability to recognize illness, knowledge and education on risks and social, legal or empowerment issues 
• When a women can reach a health facility in time to receive lifesaving EmONC services, for example distance, availability of mother shelters or transport, preparedness and awareness of existing services and community support 
• Likelihood of receiving adequate and appropriate care at community and facility level, for example, factors around Family Centered Care, Competence of health workers an effective referral system and availability of essential supplies. 
• Develop a stratification model for safe motherhood and obstetric and new-born care  
• Integrate draft AI into existing services (integrated into a client centric app to flag clients for referral) to support identifying risk factors, suggesting prevention strategies and treatment options that benefits early detection and linkage to care  
• Utilization of the developed AI model to incorporate the expressed needs of women and girls who are most at risk of maternal mortality by making recommendations based on risk factors and emergencies, prevention strategies and treatment options.  

Through this risk assessment, support tailored services and increase awareness: 

• At health facility level, Mentor Mothers will scale up current activities around conducting group health education talks on safe motherhood, signs of obstetric emergencies, newborn care integrated into RMNCH topics 
• m2m Mentor Mothers will conduct outreach services to conduct screening, diagnostic testing, referrals, counselling and support using digital health platforms/trackers as an integral part of safe motherhood, obstetric and newborn care.  
• At community level, integrating safe motherhood/ emergency obstetric and newborn care topics into household visits and interactive community meetings. 

AI risk stratification will support targeted community awareness programs such as: 

• Partnership with Facility Safe Mother Child Health Teams to lead and deliver Combination Prevention Campaigns, for example through face-to-face outreach OR through SMS messaging, social media and radio 

Through a peer-based approach using virtual learning (such as VMMP), m2m Mentor Mothers will support women and girls in the communities, working through mobilization, sensitization, group health education, support groups and awareness raising activities at facility and community levels.  

• Supporting routine strategic data collection, reporting, and monitoring, assure data-driven strategic changes to the health systems, and support facilities to integrate safe motherhood /obstetric emergencies/new-born care services provision within existing routinely collected RMNCH data. 
• Integrating routine weekly facility data audits on all maternal and neonatal deaths reported and actions to be taken 
• Including all maternal and neonatal mortality indicators into country HMIS reports and performance monitoring in line with national targets 

Reduction in maternal and neonatal morbidity and mortality can be achieved most efficiently through the identification of those individuals most at risk, pro-active follow-up, and their linkage to care, including antenatal services. m2m’s digital health platforms have successfully used technology to maintain strong human relationships and drive behavior change for disease prevention and the promotion of health seeking practices. We understand that digital solutions are complementary to face-to-face service delivery. m2m has developed and utilizes mobile health (mHealth) apps for comprehensive client management:

Commcare App 1: A custom-developed electronic client appointment diary used at facility level for comprehensive client management and follow-up for improved retention in care. Commcare App 2: A custom-developed electronic family folder used for case management at community level. The application facilitates referrals for key services to health facilities by linking directly with App1. Mentor Mothers can capture data on their devices while offline and without internet connectivity, and later sync data to the server once an internet connection becomes available, uploading to DHIS2, a centralized database that is used as a standard data repository for the organization’s programmatic data. Routine data are collated from the paper-based client management tools through the use of tally sheets and summary sheets. DHIS2 also uses trackers that are used to collect data on group based program activities.

In response to the COVID-19 pandemic, we rapidly adapted to continue delivering services to clients using digital tools. Our hybrid (in-person and virtual) delivery model offers an innovative suite of eServices including: peer-via-Phone services (a combination of bespoke and scripted scheduled phone calls) and m2m’s Virtual Mentor Mother Platform (VMMP)—an interactive WhatsApp-based chatbot providing clients with critical health information, including on healthy pregnancy and motherhood, 24 hours a day, in 28 languages.

Building on our existing digital tools and systems to include an AI component will enable client data to be analyzed and used to guide recommendations towards health services and reduce or eliminate risk of adverse maternal, neonatal and child outcomes. We will equip community health workers, who collect the data for client management, with valuable transferable skills—digital literacy, online case management, client support, data use, and analysis for client care. The proposed activities build on our existing mHealth platforms, local strengths, and leverage pre-existing programs, they respond to community needs and advance more equitable maternal health outcomes. Advancing data science and AI modelling to develop an algorithm that predicts the risk of experiencing adverse outcomes in antenatal and postnatal women and their newborns will enable risk assessment to be carried out early during the antenatal and postnatal journey and for the mother and infant to be supported with tailored interventions.  

Our evidence-based Model has been identified globally as a scalable solution that strengthens networks of care and directly addresses local health system challenges that prevent women and children from receiving safe, high-quality, care and support (see Why our Team for more information). m2m’s footprint in SSA, and extensive track record of collecting, monitoring, and analyzing a broad range of data, including antenatal data and adverse birth outcomes, will be used as a foundation to predict adverse birth outcomes among our clients, including pre-eclampsia, pre-term births, and miscarriages. Using our data to predict adverse birth outcomes through the categorization of pregnancy phenotypes could help identify women at risk of such outcomes and ultimately reduce them. We will use AI algorithms to assess pregnancy risk and predict which women are most likely to experience adverse birth outcomes, to improve pregnancy care and outcomes in sub-Saharan Africa in order to deliver targeted services.

The health of the global population will not improve in this century without improvements to health in Africa. SSA remains the only region globally where infectious diseases are the leading cause of death, and is characterized by high prevalence of HIV/AIDS, neglected tropical diseases (NTDs) and NCDs), and pervasive mental health and substance abuse disorders that remain neglected. The leading challenges in the healthcare sector in sub–Saharan Africa are: inadequate human resources for health, inadequate budgetary allocations, and poor leadership and management. WHO estimates that over 4 million health workers are needed to meet the region’s needs.

m2m provides services to pregnant and postnatal adolescents and women aimed at eliminating mother-to-child transmission of HIV, and ensuring that our clients can access the health care they need, and are retained in care. Through our integrated approach, we ensure our clients receive access to quality health services during pregnancy, childbirth and postpartum continuum, regardless of where they seek care. Since 2001, we have reached more than 14.5 million people across SSA with critical health services.

The risk of adverse pregnancy outcomes is distributed highly unevenly across and within population groups, with women living in poverty, adolescent girls, women with advanced gestational age, rural women, and women living with certain health conditions (such as high blood pressure, hypertension, diabetes, or sexually transmitted infections, including HIV) disproportionately impacted.

Notably, with 44% of its population below the age of 15 years, SSA is the youngest region in the world. With a large proportion of the population either having entered adolescence or due to enter adolescence within the next decade, SRH is one of the most significant health challenges facing adolescents in SSA. Pregnant adolescents do not benefit equally from routine antenatal care, so a more targeted approach is necessary to address their specific needs.

Impact: Our Model works. We currently reach more than 1.6 million clients annually. In 2021, m2m met or exceeded the UNAIDS 95-95-95 targets for ending HIV and achieved virtual elimination of mother-to-child transmission of HIV among our enrolled clients for the eighth consecutive year.

Trust: We are a trusted implementer. Our Model is cited as best practice for sustainable, people-centered approaches in the UNAIDS Global AIDS Update 2020 and has been profiled in UNICEF and UNAIDS flagship reports.

Knowledge: We are experts. We combine cutting-edge global maternal and child health knowledge—working directly with UN Agencies and the WHO—with Mentor Mothers’ on-the-ground community understanding and lived experience. m2m Head Office in Cape Town houses the Department of Technical Programming and Support, with technical and strategic information experts who work closely with country teams from design of programming, through implementation, monitoring and evaluation, and QA/QI for performance management and improvement.

Core to our organization is the involvement of the people in the communities we serve in all aspects of designing and implementing our projects. m2m trains and employs local women living with HIV as Mentor Mothers—community health workers who increase access to healthcare for women and their families. m2m Mentors operate in homes and health facilities at the heart of communities across ten countries in SSA. Since 2001, they have provided life-saving services to over 14.5 million clients, predominantly pregnant women, new mothers, and their infants.

m2m is committed to the principle of “nothing about us without us.” To ensure that the authentic needs and voices of our clients shape our programming, m2m employs local staff (the overwhelming majority are former clients), with deep roots in their communities and personal experience of the challenges facing clients and service providers. All programming is anchored in a community engagement model, involving input and feedback from community members and leaders. Programming will be introduced and promoted through quarterly community dialogues and other existing community meetings, networks, and local traditional leaders. m2m will work closely with the health service and relevant Ministries at local and national levels to create awareness and build skills to promote the delivery of a coordinated approach. Moreover, we will provide feedback loops on program results and impact to key stakeholders and decision-makers. Through community engagement, we incorporate the expressed needs, recommendations, and leadership of women and girls who are most at risk of maternal mortality and help to advance more equitable maternal and child health outcomes.

m2m is applying to the MIT Solve Challenge as winning would enable m2m to include AI as a component of our strategic information and digital systems, and enhance our ability to provide vulnerable pregnant and postnatal mothers and their newborns with targeted services to reduce maternal and neonatal morbidity and mortality. Both mentorship and partnership will accelerate m2m’s use of AI to optimize care for pregnant and postnatal women and their newborns to reduce maternal and neonatal morbidity and mortality in eight countries of m2m operation in SSA.

In addition, it would increase m2m’s visibility and create opportunities for additional partnerships with other organizations towards achieving improved health outcomes and the global goal of Universal Health Coverage. The opportunity to collaborate through MIT’s ecosystem and Gates WFO Innovation Accelerator, learning from a cohort of experts will strengthen m2m’s ability to further develop and strengthen our solution, providing access to digital technology experts and cutting-edge developments in its use in health care. 

The earlier we identify individuals most at-risk, preventable at-risk birth outcomes will decrease. Through our community health workers, m2m can take predictions from the AI solution a step further by using predictions to directly inform care. We can communicate to clients why they have been classified in a certain risk profile, and identify and share with them how they can lower their risk. Accurately labeled and representative data are vital to machine learning.  Most antenatal risk stratification models are built using data from medical records in high-income countries which may not be suited to pregnant women in sub–Saharan Africa where women face additional health challenges, such as the high prevalence of HIV. Although we collect a range of data from our clients, we will supplement this with new detailed information on their experiences during pregnancy and birth. To build a custom machine learning model for antenatal risk stratification we will use a combination of existing detailed data and the additional data we collect to predict the risk of adverse birth outcomes among antenatal clients. 

Our impact goal over the next five years is to contribute to reduced maternal and infant mortality across sub–Saharan Africa. We currently work in six of the eight counties in the region that contribute 74% to these rates.

To ensure a clear pathway to impact, m2m will measure the impact of our activities in contributing to health and wellbeing, behavioral change, and care/treatment outcomes, including reducing maternal and neonatal mortality and other adverse birth outcomes, and improving reproductive, maternal, and newborn health. We will employ a robust Monitoring and Evaluation approach to track key indicators through our signature performance management dashboards that routinely identify quality assurance and quality improvement (QA/QI) needs, and support data-informed decision making. m2m will conduct an Acceptability, Feasibility and early Learning Study (AFeL), to identify early lessons and opportunities to improve/pivot. We will use our digital client management tools—to track priority appointments and individual progress against key indicators and measure aggregated progress against program indicators, including attendance of services. To establish pre-implementation benchmarks, m2m will conduct a Knowledge, Attitudes, and Practices (KAP) survey among targeted clients that will be complemented with baseline data collated from existing sources and current reach. The project will also conduct an end-line evaluation at project completion to assess progress made against project objectives and goal(s). m2m's QA/QI approach brings implementation science and effectiveness to the table, and our evidence- and science-based frameworks will bring deep data analysis and applied research to MIT Solve and the world. During the project, we will rapidly produce and disseminate data and reports, create knowledge forums, and put information in the hands of local and national decision-makers to create a pool of data for improving AI models stratifying antenatal risk.

Contribute to reduced maternal and infant mortality (aligned with SDG3 goal of less than 70/100,000 births by 2030)

Improve reproductive, maternal, and newborn health

Number of clients flagged by risk and referred to maternal health services. 

Number of clients successfully referred and receiving advanced maternal care. 

Number of high-risk women identified and linked to a Peer Mentor for intensive care and support. 

Number of emergency systems set-up to support maternal health emergencies (by type in country). 

Number of safe motherhood groups created at facility and community. 

Number of safe motherhood groups facilitated by trained health workers. 

To reduce maternal and neonatal morbidity and mortality and adverse birth events, we will develop and deploy AI to integrate the identification- and addressing- of maternal and newborn health risks and their preventable causes into our evidence-based (see Why Your Team section) Mentor Mother Model. The AI model will predict the risk of adverse events in antenatal and postnatal women and their newborns and enable Mentor Mothers to prioritise those most at risk. We will enhance understanding of safe motherhood, obstetric, and newborn care among pregnant women, mothers, and parents, including improved knowledge and decision-making abilities needed to seek care; will increase access to client-centered differentiated care during pregnancy, childbirth, and postpartum continuum; and will strengthen health system capacity to provide obstetric emergency and newborn care. In the short and intermediate-term, this will lead to increased access to safe, high-quality, and coordinated care. In the long term, the use of models to identify and link those in need of obstetric and neonatal care will become part of sustainable health system efforts to reduce maternal and newborn mortality and morbidity. Our peer-to-peer Model has a successful track record across maternal, newborn, and child health, non-communicable diseases, and HIV (see Why Your Team). Adding AI will enable m2m to better identify those most at risk and link them to care. To measure our causal links and the efficacy of the AI model between inputs, outcomes, and overall impact, we will employ research methodologies including comparing baseline data with midline and endline data.   

Our evidence-based Model has been identified globally as a scalable solution that strengthens networks of care and directly addresses local health system challenges that prevent women and children from receiving safe, high-quality, care and support (see Why our Team for more information). Through our integrated approach, we ensure our clients receive access to quality health services during pregnancy, childbirth and postpartum continuum, regardless of where they seek care.

m2m’s digital health platforms have successfully used technology to maintain strong human relationships and drive behavior change for disease prevention and the promotion of health seeking practices. We understand that digital solutions are complementary to face-to-face service delivery. We bring a rigorous focus to monitoring and evaluation, shaping our programs based on robust data captured through a suite of innovative digital and mobile health tools, including:

• Commcare App 1: A custom-developed electronic client appointment diary used for comprehensive client management and follow-up.
• Commcare App 2: A custom-developed electronic family folder used for case management at community level.
• DHIS2: A centralized database used as a standard data repository for programmatic data.

During the COVID-19 pandemic, we rapidly adapted our services to reach and serve clients using virtual digital tools. Now, we offer eServices alongside our face-to-face services, providing clients with an innovative hybrid (in-person and virtual) delivery model. Our eServices include:

- Peer-via-Phone services: a combination of bespoke and scripted scheduled phone calls focusing on key services, client retention in care, adherence to treatment, COVID-19, and TB screening.

- m2m’s Virtual Mentor Mother Platform (VMMP): an interactive WhatsApp-based chatbot providing clients with critical health information, including on healthy pregnancy and motherhood, 24 hours a day

We will adapt our face-to-face services, eServices (Peer-via-Phone and VMMP), and digital health platforms (Commcare App1 and App2, and DHIS21) in SSA, using Artificial Intelligence (AI), to stratify clients’ antenatal risk/risk of adverse birth outcomes into low, medium, or high risk. Using data (biomedical, clinical history, diagnostic screening, fetal monitoring, genetic, demographic, and behavioral) will enable us to differentiate clients by risk profile for active follow-up, education/counselling, and linkage to services, promoting behavior change, improving health outcomes, and accelerating the decline of maternal mortality to less than 70/100,000 births by 2030 (SDG 3).

Building on our existing mHealth systems to include an AI component will enable client data to be analyzed and used to guide recommendations towards health services and reduce or eliminate risk of adverse maternal, neonatal and child outcomes.

We will equip community health workers with valuable transferable skills—digital literacy, online case management, client support, data use, and analysis for client care. The proposed activities build on our existing mHealth platforms, local strengths, and leverage pre-existing programs, they respond to community needs and advance more equitable maternal health outcomes. Advancing data science and AI modelling to develop an algorithm that predicts’ the risk of experiencing adverse outcomes in antenatal and postnatal women and their newborns will enable risk assessment to be carried out early during the antenatal and postnatal journey and for the mother and infant to be support with tailored interventions.  

The AI model will be scaled-up and adopted across m2m platforms in Lesotho, Malawi, South Africa, Ghana, Kenya, Zambia, Angola, and Uganda. Where there is evidence of the success of AI in stratifying antenatal risk and reducing adverse birth outcomes we will use it to advocate for key stakeholders (e.g. government at all levels) to adopt our approach. As a stand-alone model it can be integrated into government mHealth, eHealth (medical records), and even telemedicine platforms.

m2m employs women living with HIV as Mentor Mothers and Peer Mentors, community health workers, providing services at health facilities and households in their local communities. m2m Mentors are trained to use m2m’s mHealth tools to gather client information using mobile devices. App1 and App2 are run off CommCare, an open-source mobile platform. Mentor Mothers can capture data on their devices while offline and without internet connectivity, and later sync data to the server once an internet connection becomes available, uploading to DHIS2, a centralized database used as a standard data repository for programmatic data.

m2m is committed to the principles of diversity, equity, and inclusion, and opposed to all forms of discrimination on the grounds of race, gender, disability, nationality, religion, sexual orientation, or HIV status. Our model is based on female social and economic empowerment, and inclusive community engagement to address gender and health inequities. Our clients are among those most vulnerable and marginalized – pregnant and postnatal women living with HIV, their HIV-exposed or infected children, and their families. Equity is core to our approach and impact, positioning local women living with HIV as the agents of change at an individual, community, and systems level. Since 2001, m2m has employed over 11,700 Mentor Mothers, often their first formalized employment. Mentor Mothers become economically and socially empowered, and act as change agents reducing HIV-related stigma, and demonstrate that women should be at the heart of solutions to improve maternal and child health. Equity is a key consideration at all levels of the organization, with women constituting 89% of staff and all m2m Country Directors being local women or leaders from their region. m2m’s employment equity policy promotes inclusive and equal opportunities in the workplace, including accessibility for employees living with disability.

mothers2mothers (m2m) is a global nonprofit that unlocks the potential of women to eliminate paediatric AIDS and create healthy families across Africa. m2m trains. Employs and helps empower women living with HIV to work as community healthcare workers in understaffed health centres and underserved communities. Through a peer-to-peer approach, these Mentor Mothers deliver a range of health services, advice and support to

women and their families. m2m's partners include Governments, Multi and Bilaterals as well as Strategic Partnerhips with Foundations and Corporates.

m2m is pursuing alternative financing models to help ensure long term financial sustainability. These include raising more Unrestricted Funding that will allow the organisation to invest in innovative and strategic initiatives that provides transformational benefits to the organisation and the clients we serve.

The foundational cost for m2m’s digital health platform is already funded and the financial support being requested would be deployed on scaling up and enhancing m2m’s existing digital platform.