MIT Solve

Basic Information

Our tagline:

Offline-first predictive algorithms to identify at-risk households and increase health worker efficiency and effectiveness in the world’s hardest-to-reach communities.

Our pitch:

Medic Mobile combines decision support for doorstep care, prioritization for home visits and follow-ups, messaging, and actionable analytics for managers. The tools we build are free, open-source, and deployed at scale in the hardest-to-reach areas. Evidence-based workflows come together in the software to support health workers and families – helping to ensure safe deliveries, track outbreaks, screen and treat illnesses door-to-door, communicate about emergencies, and more. Our application is used by over 20,000 health workers supporting hundreds of thousands of households in 14 countries across the world.

Medic Predict is our first implementation of machine learning algorithms within the Medic Mobile application. Our aim is to put data to work for those who are most marginalized, helping health workers and health systems deliver high-quality, equitable, targeted, and timely, healthcare for everyone.

The Medic Predict tool will use key individual, household, and community variables to predict which individuals are at highest risk of experiencing negative health outcomes, for example the risk of a child not receiving treatment in a timely manner after becoming ill. The tool will combine static predictors with dynamic views of community level risks, as well as patient and family medical history, to produce individualized risk scores for each patient in a health worker’s community. The Medic application will then use this information to direct health workers to the families who need their assistance most, and to offer targeted interventions to address the factors that may have caused these individuals to be at risk. Our offline-first implementation will ensure that our analytical tools are available to even those in the hardest to reach communities.

At the core of our approach is a focus on Human Centered Data Science - the integration of data science with human centered design - to produce quantitative data products that are supported by the qualitative needs of our product’s users. Our aim is to go beyond simple “black box” models and to create predictions that both resonate with our user’s understanding of their communities, while also offering new insights that help them perform their roles more effectively and efficiently.

To accomplish this, we have completed extensive interviews with frontline community health workers as well as their supervisors, to understand how they view risk in their communities, and how they might use this information to plan their daily activities. We want to ensure that any risk information we present to our users translates into their local context, and that it offers them additional insights beyond the knowledge that they’ve gained from their years of work in their communities.

We firmly believe that with careful design, including a focus on iterative rounds of human feedback, the Medic Predict tool can help make the health workers using our application more efficient and effective, and can ensure that the families that they work with on a daily basis receive the best care that can be delivered.

Where our solution team is headquartered or located:

San Francisco, CA, USA

The dimensions of the Challenge our solution addresses:

Effective and affordable healthcare services

About Your Solution

What makes our solution innovative:

While there have been significant efforts to improve community-based diagnostic tools using machine learning, our approach is focused on using operational data from existing community health programs to measure an individual’s risk level. Our tool will provide timely, accurate predictions in an offline-first setting where computing resources on user’s devices are limited.

We are also focusing our design efforts on leveraging the wealth of expertise that health workers possess about their work and the communities they serve. We want to ensure that we’re giving health workers information that complements the wealth of knowledge they’ve gained through their years of experience.

How technology is integral to our solution:

The Medic Mobile app is a critical tool for patient management and decision support for the health workers who use it. The Medic Predict tool will leverage the data that they’re collecting to create machine learning models which will offer insights on which households in their community may need additional care or attention. Our aim is to use machine learning to supplement health workers’ knowledge, and to identify patterns in our data that are not evident at first glance.

Our solution goals over the next 12 months:

Over the next 12 months our aims are: 1) to pilot the Medic Predict tool with a small group of health workers in Kenya; 2) to evaluate the success of the pilot and document key learnings from the implementation of the predictive algorithms within the app; 3) to expand usage of the Medic Predict tool to an additional 1,500 health workers in Kenya; and 4) to prioritize additional use cases and partners for future expansion of the Medic Predict tool.

Our vision over the next three to five years to grow and scale our solution to affect the lives of more people:

A key priority for the next three to five years is to build our tools to the point where we can start applying cross-project learnings, building tools that allow us to predict outcomes in one context or region based on the collective data from other partners’ implementations. By doing this we’ll be able to offer Medic Predict as an out-of-the-box tool that partners will be able to implement with less historical data than traditional machine learning tools would need to generate accurate predictions.

The key characteristics of the populations who will benefit from our solution in the next 12 months:

Pre-natal
Child
Adult
Lower

The regions where we will be operating in the next 12 months:

Sub-Saharan Africa
East and Southeast Asia
South Asia

How we will reach and retain our customers or beneficiaries:

Our solution will be deployed within the Medic Mobile app which is currently used by nearly 20,000 health workers in 14 countries. After an initial pilot with 30 users in Kenya, we will look to expand use of the predictive algorithms tool to a broader user base of 1,500 additional users within Kenya. We envision Medic Predict becoming part of our core application, which is projected to be in the hands of 200,000 community health workers by the end of 2021.

How many people we are currently serving with our solution:

Over 2017, Medic’s tools were used by nearly 20,000 health workers to support:

4 million healthcare services;
Screening and registering women for 225,621 new pregnancies;
91,327 deliveries with skilled care at a health facility;
1,087,167 doorstep assessments for children including 822,857 diagnoses and 296,665 clinical referrals.

Currently, health workers send in records representing over 700,000 actions at patient’s doorsteps each month. Partners using Medic’s tools to facilitate proactive health care have seen substantial improvements in key health outcomes; one partner in Uganda documented a 27% reduction in child mortality; while another in Mali documented an astounding 10x reduction over three years.

How many people we will be serving with our solution in the 12 months and the next 3 years:

Our pilot project will utilize 30 community health workers who each serve roughly 110 households. Following a successful pilot in early 2019, we will expand use of the tool to an additional 1,500 users in Kenya. Our aim over the next several years is to implement the Medic Predict module as a core tool in the Medic Mobile application, allowing our global base of mobile app users to have access to these methods.

About Your Team

Partnership Potential