WONDER maternal mortality reduction

Improving antenatal care for women offers a unique opportunity to impact maternal mortality, maternal morbidity, infant mortality and poor infant health outcomes. In the ten countries of southern Africa (Angola, Botswana, Lesotho, Madagascar, Malawi, Mozambique, Namibia, South Africa, Zambia, Zimbabwe), there are roughly 533 maternal deaths /100,000 live births or 200,000 maternal deaths per year.  In addition, approximately 10X as many women suffer maternal morbidity.

The WONDER telehealth solution:

1. Rapidly identifies high risk pregnancies and developing emergencies during pregnancy – using AI and low-cost telehealth devices.  

2. Integrates easily to existing public health infrastructure and protocols.

3. Has been successfully piloted in two states in India (Tamil Nadu – 15,000 mothers, Bihar – 55,000 mothers).

This proposal seeks funding to partially cover the cost of (a) innovation and converting the WONDER algorithm into an AI software to the WONDER telehealth model and (b) scaling in southern Africa starting with Zambia. 

Maternal mortality takes the lives of tens of thousands of women every year in southern Africa. In addition, maternal mortality has a lifelong devastating effect on children and the entire family. For each case of maternal mortality, there are between 10 and 25 cases of maternal morbidity. Published literature shows a direct relationship between maternal mortality, morbidity and perinatal mortality, including stillbirths and early neonatal deaths.  Interventions that save the lives of mothers have been documented to improve all other mortalities. 

Poor maternal health is the start of a lifelong journey of ill health and often leads to severe physical, psychological and social consequences for millions of women in southern Africa. Maternal health is a point of leverage and has been found to positively impact family economic, education and health outcomes.

A number of systemic factors impact maternal mortality include early childbirth (< 18 years) and delivery outside a medical institution. However, once a woman is pregnant, the single most critical factor that leads to maternal mortality is the delay in identifying a developing risk and taking timely action. This is our focus.

Women who die during childbirth rarely go directly from being in a healthy state to death. Typically, there is a crucial transition period, during which care-providers can intervene and institute appropriate treatment in time to prevent serious complications and death.

The WONDER System is a cloud based comprehensive digitized telehealth solution using maternal early warning criteria and an electronic birthing unit modeled after eICU. The core of our solution is an algorithm based on maternal early warning criteria recommended by the RCOG & NICE of UK and ACOG. The system creates alerts through an internal algorithm based on clinical history, signs & symptoms along with lab results and vital signs in real time to warn the providers of the dangerous clinical status of the patient and provides treatment guidelines put forth by various leading organizations around the world along with references.  

 Vital signs for women during pregnancy, labor and post-partum are monitored using inexpensive technology that include mobile technology, computing hardware, telehealth platforms and two-way communications. The algorithm continuously scores women on potential risk factors and initiates alarms for intervention. The system is integrated into existing health-systems and protocols easily.

The primary beneficiary of the project are

Phase 1: pregnant women in Lusaka, Zambia;

Phase 2: pregnant women in Zambia;

Phase 3: pregnant women in rest of southern Africa.

In addition to women in urban locations, pregnant women in rural and hard to reach areas with limited access to maternity care will be a key target population for this initiative.

 Multiple studies across the world has demonstrated that only institutional deliveries and skilled obstetric care can save women’s lives.  By educating pregnant women in local vernacular by familiar community health workers about the importance of identifying warning signs of pregnancy complications care can be instituted rapidly.

The proposed project is focused on reducing the maternal mortality rate (MMR). In southern Africa MMR is 533/100,000 live births. For these 10 countries to meet goal #3 of SDG by 2030, they will need to improve at >15% year, a rate never achieved before in any comparable time period in any of these countries. Based on the gains we witnessed in the pilot in Tamil Nadu (25% in 2 years), we are setting out to reduce MMR by a cumulative 50% in first 5 full years of implementation in these countries.

While sustained efforts by global organizations and national health systems have improved maternal health indicators across the world, much remains to be done. 

Maternal mortality offers a uniquely strategic point of leverage to impact the lives of women, children and their families in significant ways. It is a global problem and any success in reducing maternal mortality will deliver outsize returns. 

Our solution is based on inexpensive technology and is designed for resource-constrained locations. It seeks to build an AI component to make the software more adaptive and accurate. It aligns with the Maternal and Newborn Health challenge section.

Three things make the proposed solution unique:

1. The solution relies on inexpensive, off-the-shelf medical devices and computer equipment that are available across the world 

2. The solution integrates easily and quickly into existing hospital and public health protocols without major infrastructure requirements or integration costs

3. To our knowledge there is no other maternal mortality reduction solution targeted for deployment in resource-limited countries that leverages the power of AI

In its present form the core of the WONDER system is a static algorithm hard-coded into the system. This algorithm has already shown strong results in the India pilots.

This proposal aims to convert this algorithm to a learning model and make it more accurate, flexible and more adapted to local (southern Africa) conditions.

The core technology that powers WONDER is a comprehensive Electronic Health Record system (compliant to HIPAA and ICD10), an algorithm based on Maternal Early Warning Criteria and a decision support database that is populated with the latest ACOG (American College of Obstetrics and Gynecology), RCOG (Royal College of Obstetrics and Gynecology), WHO (World Health Organization), and Surviving Sepsis Campaign guidelines for conditions that are common to the antenatal, labor and postpartum phases of pregnancy. The entire platform is set up on Microsoft Azure Cloud with enterprise level security. The care providers of staff in hospitals and clinics use the provider web portal to manage patients and monitor alerts.

Through the two pilots completed in India (Tamil Nadu - 15,000 patients, Bihar - 56,000 patients) we have assembled a large database of patient data, outcomes and interventions. This MIT Solve submission seeks funding to build and test alternative ML models with this data to extend the WONDER solution into a learning model. This will vastly expand the versatility and accuracy of the WONDER algorithm and its effectiveness.

The team currently has data for over 70,000 women from the two pilots. The proposal seeks funding to convert the WONDER algorithm to an AI software with machine learning capabilities. Multiple ML models could be tested with a portion of the existing data and fine-tuned and tested with the remaining data.

The current WONDER technology has been proven in the two pilots to work effectively to reduce maternal mortality and in-hospital eclampsia. In Tamil Nadu, India, the WONDER system was tested in a two-year pilot study starting in July 2017. A total of 15,184 patients were monitored during labor and the postpartum period. Within limitations of the study, the incidence of in-hospital eclampsia was reduced by 88% and in 92% of cases, timely treatment was started within an hour of identifying the abnormality, a goal recommended by The Council on Patient Safety/Women and Mother's Health Care.  Maternal mortality was reduced by 25% over local benchmark figures. The WONDER system identified at-risk patients, directed skilled care to patients at risk for complications, and helped to institute treatments on time, demonstrating a potential solution for women in underserved locations.

In Bihar (Darbhanga), the software is currently in pilot with 56,000 pregnant women.

See attached image from Hindustan Times, a national newspaper in India  

The WONDER solution proposed here is based on the following published findings.  Approximately:

a) 85% of all maternal mortality is preventable;

b) 25% of maternal deaths occur during the antenatal period

c)  50% of maternal deaths occur during labor.

d) 20-25% of maternal deaths occur during postpartum.

Key to our proposal is gaining timely visibility to the emergence of risk factors in a pregnant woman. In the current system, pregnant women interact with the public health system sporadically.  

The proposed solution

Activity: 

·       WONDER electronic health record (EHR)with alert system based on maternal early warning criteria

·       WONDER App for remote patient Monitoring – using low level providers with embedded alert system

·   WONDER App for hard to reach areas with alert system to alert providers that patient’s clinical status is serious and requires immediate evaluation, treatment and the need to transport patient to a tertiary care facility

·       Telehealth system with ability to consult through video or audio call with the skilled provider at anytime

·       Robust patient education in their own language regarding warning signs of pregnancy

Output:

 ·       Provide quality maternity care to patients in poor, resources restricted Low and middle income countries.

·       Reduce maternal and neonatal mortality and morbidity

·       Preserve skilled care to the sickest patients

·       Address problem of lack of skilled obstetric care

·       Hub and spoke model enables one or a few providers to deliver care to multiple centers and different areas regardless of distance

·       Better Obstetric outcome- (eclampsia, postpartum hemorrhage, sepsis)

·       Improve patient education and encourage patients to use contraceptive options provided by the Government 

Outcome if Scaled

·       Improved access to skilled obstetric care across resources restricted areas

·       Reduce global maternal and neonatal mortality and morbidity and ability to meet Sustainable Developmental goals for all countries (to reduce the maternal mortality below 70/100000 live births and neonatal mortality rate to 12 per 1000 live births by 2030)

·       Generational impact on families, as well as social, cultural and economic impact of communities.

Current number serving: 56000+ pregnant mothers 

In one year planning to serve: 500,000 pregnant women.

In 5 years: 2 Million

Our primary goal over the next year is to transform WONDER to an AI software and scale within Bihar, India and the 10 countries in southern Africa.

Bihar – in one district we presently cover 56,000 women. We are looking to expand to all 38 districts starting next year and this will allow us an opportunity to reach 7m women. However we are taking a conservative view that we will only reach 500,000 women in Bihar in one year.

Southern Africa and Bihar – southern Africa has approx. 6.5m women who are pregnant each year. Along with 7m million in Bihar this gives a total of approximately 13.5m women who we could reach. We are taking a conservative view that we will scale to 2m women in 5 years.

Medical and legal

Buy-in from doctors

Reluctance to use technology

Adds to physician workload 

Cost:

Cost of devices to monitor patients, hardware and training, storage and maintenance, Clinical and 24/7 technical support

Infrastructure

Internet availability, Government support and possible reluctance  to change policies

Medical and Legal

A significant hurdle is to secure buy-in from doctors. In our two pilots we encountered some initial resistance from doctors who felt some loss of power and control by bringing the WONDER system into their existing ward operations. However, when these doctors realized the advantage of identifying the at-risk patients early, we observed a turnaround in physicians attitudes.

Additionally, getting the National and local Obstetric Societies to see the value of the system is critical. Our approach has been to work in full transparency and engage in proactive conversations with all such entities.

Costs

At this time both our pilots were funded by our funders and the local authorities did not have to commit funds. As we begin scaling we expect costs and any diversion of local funds to this initiative to be a significant barrier. We are hoping to demonstrate successful pilots in Lusaka, Zambia and use that to influence the local ministries of health and finance.

Infrastructure

We hope to work with the local telecom providers to ensure internet access to transmit data in real/near-real time. However, the current implementation of the WONDER system also has capability to work with no data capability and do a store and forward approach of patient information.

6 Full time staff and 10 part time staff

This proposal comes from a unique group of committed professionals with a passion to save women and childrens lives

The team is led by the following 4 individuals:

Dr. Narmadha Kuppuswami, MD is a US Board Certified obstetrician and gynecologist with nearly 40 years of clinical experience in the Chicago area.  Her expertise is both clinical and technical, as Dr. Kuppuswami began investigating software and emerging technologies as a means to improving outcomes for pregnant women early on in her career. After witnessing a maternal death in her teen years, she has remained focused on working to save lives of pregnant women her entire career.

Dr. Suresh Subramanian, an MIT alum has spent 15+ years working in the high-tech sector for Fortune 200 companies, during which time he became intimately familiar with both healthcare technologies and effective project management in resource-limited settings. Separately, in 2003 he founded the Power of Love Foundation in the US and the Matero Care Center (in 2004) in Lusaka, Zambia to deliver innovative pediatric HIV care to infants and children.

Christopher Mulela, RN has led the work of the Matero Care Center in Lusaka, Zambia for 16 years. Mr. Mulela brings deep understanding of local conditions and partnerships with neighborhood leadership and local healthcare and technology organizations.

Karenna Groff has worked intimately with health technology and artificial intelligence as a student in the MIT Biological Engineering department, as well as through her work researching epilepsy at Boston Children's Hospital.

The team presently works with two partners. HP, Inc (earlier known as Hewlett Packard, Inc) and GC Infosys.

HP partnership is described here http://h20195.www2.hp.com/v2/getpdf.aspx/4AA7-3018ENW.pdf

HP Inc. is one of the biggest information technology companies in the world operating in over 85 countries. In 2016 HP was one of the earliest supporters of the WONDER software and provided all the hardware, computers and displays that were used in the pilot project in Tamil Nadu.

GC Infosys is an information technology implementation company headquartered in Malaysia and operating across Asia and Africa. GC Infosys has a track record of implementing large system change programs globally.

We anticipate GC Infosys to be part of our implementation and scaling in southern Africa through their offices in Mozambique.

Our current business model is that of an NGO operating through grants from foundations and donor entities. Our funding portfolio includes private donors, private and public foundations, and state governments.

It is important to note that all of the partners in this proposal are non-profit organizations and no portion of the funds requested will go toward any purpose other than directly paying for the development of the AI software and scaling within Zambia.

In addition, it is important to note that none of the partners or individuals within this project will be paid any remuneration (beyond student research stipend for currently enrolled students) from the amounts requested.

Our current plans are to continue to tap into our donor base to write grants, tap into our current donor base and raise necessary monies. Our funders include private donors, private and public foundations, and state governments.

For the proposed plan to scale in southern Africa, our funding will be used for:

a)    developing the AI software

b) tailoring the electronic health record software to the southern Africa context

c) implementation support

d) evaluation and communication.

The primary expenses of rollout and scaling will be the responsibility of the respective governments and their public health systems.

The WONDER solution and approach to reducing maternal mortality is truly a low-cost, easy to implement approach that has been piloted in highly resource-constrained settings in India. Prima-facie' it is ready to be scaled to southern Africa.

By converting the software to an AI solution, it vastly enhances the capabilities of an a product that is already supporting over 56,000 women in live usage. Adding machine learning capabilities will allow for the software to be finely tuned to the local needs and more accurate in its predictive power.

The four members developing this proposal bring decades of experience in obstetrics, information technology, southern Africa.

We are looking for support in taking the software to the next level – to add the AI capabilities. Funding is only one part of our needs. Our larger need is to be partnered with a team/department/organization with strong AI capabilities. This is the reason we are applying to Solve

We seek partnership in

a)    adding machine learning capabilities to the WONDER software

b)    converting a static algorithm WONDER software into an AI software

c)     tailoring the UI and app to the southern African context

We bring to the proposal:

a)    A stable software that is complete (in non-AI form) and is field tested.

b)    A team that has experience in obstetrics, information technology, southern Africa, implementing and managing large projects

c)     Deep connections in Zambia with local health and for profit and non profit corporations

We look forward to partnering with MIT teams working in AI and healthcare technology. Many of these leaders are based in the CSAIL at MIT. These teams can advance the WONDER solution by:

a)    Assigning students to work on creating and testing the various ML models

b)    Providing professor-level leadership to ensure that the project is technically sound and reliable

c)     Partnering with us in reaching funders to help support additional fundraising to complete the project and scale

We are excited to be applying for the Innovation for Women Prize.

 Maternal mortality continues to be highly preventable problem worldwide. While the birth of a child is a powerful symbolic validation for the continuation of the human species, the risk of death during pregnancy, labor, childbirth and postpartum continues to disproportionately impact women in many parts of the world. Worldwide, maternal mortality and maternal morbidity are high when access to health care is limited. Because of the shortage of healthcare infrastructure and skilled providers, the struggle to eliminate this highly preventable tragedy has remained unfulfilled. Maternal mortality is now, for the most part, concentrated in sub-Saharan Africa and south Asia. The countries of sub-Saharan Africa collectively account for 66% of all maternal deaths worldwide. We believe we have the opportunity to make a real and sustained impact on this issue through the WONDER software and approach.

The solution outlined here that relies on inexpensive technology (computers, cell phones, blood pressure cuffs, thermometers, pulse oximeters) and a proven software that is now being proposed to move to AI is ideally suited for resource-constrained settings in Asia and Africa.

Maternal mortality is one of the most preventable of causes of mortality among women, particularly in resource-constrained countries. Reported as deaths per 100,000 live births, maternal mortality in 2017 ranged from a low of 2 in a handful of European countries to a high of over 1100 in South Sudan.  A total of 189 countries signed up to the goals of the Millennium Development Goals in the year 2000 to reduce maternal mortality to below 100 by 2015, a figure that was met by less than two thirds of the participants. As of 2017, a third of the countries in the world had not reached the MDG goal of 100. In 2015 and as a follow up to MDG goals, 193 countries came together to establish and ratify 17 Sustainable Development Goals (SDGs) with a target date of 2030. Goal number 3 of the SDG covers Maternal health and specifically lays out the objective of reducing worldwide MMR to 70 or below by 2030. 

Sub-Saharan Africa (that includes southern Africa) accounts for 66% of all maternal deaths during pregnancy, labor and post-partum. There is no path for the world to achieve its SDG goal #3 without impacting the maternal mortality rates in southern Africa significantly.

This proposal focuses on southern Africa and seeks partnership support and funding to scale a software and approach that has proven itself in large pilots (over 70,000 women).  

We are excited at the opportunity to apply for the 'AI for Humanity' Prize. We believe that the present proposal aligns with the guidelines of the AI for Humanity Prize for the following reasons:

1. The proposed solution targets a major global challenge – maternal mortality. Despite major strides in women’s health, maternal mortality continues to stubbornly remain high across much of sub-Saharan Africa and south Asia. The signatories to the Sustainable Development Goals have set an ambitious target for 2030 that will require significant improvement in maternal mortality rates in the coming decade.

2.  The proposed solution is a stable software that has proven itself in multiple large-scale pilots and is ready for conversion to an AI software.  While maternal health and causes for maternal mortality are similar across many countries, our experience has shown that the cohort groups in different countries have differing characteristics. For example, the American college of obstetrics lists 160 mm of Hg as the threshold for identifying a critical case of sever-eclampsia. Our experience in India showed that women with blood pressure in the 140 mm range were developing eclampsia. AI software trained on large data local to the region will be far better at predicting emergencies.  

3. The team has extensive experience in obstetrics, information technology, southern Africa and implementing large projects – increasing the probability of success of the project.

Reducing maternal mortality and morbidity offers a unique point of leverage to impact the health of women,newborns, family stability and economic conditions. Sustained efforts over the last many decades has brought significant improvement in maternal health globally, however, much remains to be done.

Sub-Saharan Africa accounts for approximately 66% of the worldwide maternal deaths. There is no path for the world to achieve our SDG goal #3 by 2030 without improving MMR in sub-Saharan Africa by over 15% per year between now and 2030 - a rate of improvement not seen before. 

The solution proposed here targets a point of strategic change – early visibility to risk factors in a pregnant woman through near real time monitoring of vital signs. The ubiquity of cheap digital technology and AI algorithms allow for this solution to be adopted and scaled without major investments by the government in infrastructure or deployment costs. In two pilot tests in India (in Tamil Nadu with 15,000 women and in Bihar with 56,000 women) the solution in its current form has shown success at scale in reducing maternal mortality and morbidity.

We are now proposing to vastly improve the current system by making it an AI system by incorporating machine learning models to continually improve the prediction. This will allow the system to be more flexible and better tailored to individual regions of the world.

The Bill and Melinda Gates Award would accelerate this work and make it a reality.