Implementing organisation D-Tree International
Period of implementation 12/2019 – 11/2021
Digital tools Mobile App, Machine Learning, Artificial Intelligence
Country Tanzania
Sector Health
Budget 349.887 €
Contribution to SDGs sdg 3


What is the project aiming to achieve?

Zanzibar, like many sub-Saharan countries faces high levels of neonatal mortality, it is estimated at 29 per 1000 live births. The primary challenges contributing to this are limited access to emergency obstetric care, delays or inability to seek care, and biological risk factors that go undetected due to lack of contact with providers.

But predictive analytics as an innovative approach to big data has the potential to improve effectiveness and efficiency of service delivery of maternal, new-born and child health in Low and Middle-Income Countries (LMICs). However, significant challenges still exist for building these models on traditional clinical datasets, which are often incomplete or inaccurate in setting such as Zanzibar.

The project aims to personalise and improve maternal, new-born and child health in Zanzibar by integrating predictive analytics into the national digital community health system by using machine learning. This innovation will enable community health workers to pre-identify women with high-risk pregnancies and target prenatal and postnatal services to mitigate risk and improve outcomes.

This project supports the Zanzibar Ministry of Health to scale up a nationwide digital community health program which will equip over 2,000 community health workers with mobile apps and support them as they provide home-based services to pregnant women and children under 5.

The app provides step-by-step guidance to the community health workers to support assist with antenatal and postnatal care, including screening for danger signs, birth planning, counselling and education and referral coordination.

Digital component

D-tree will further develop a predictive model for perinatal risk which has been partially developed based on data collected from over 40,000 pregnant women. This model takes into account already-captured client data from the program as well as secondary data streams including geospatial data from drone imagery which can help identify differences in geography, (quality of roads, etc.). It also includes anonymised telecommunications data and mobile money transactions to indicate vulnerability of communities based on types of phone-based transactions, and GPS coordinates for health facilities and villages to indicate vulnerability based on distance to health facilities.

D-tree will also work with public health and medical experts, including the Zanzibar Ministry of Health to design a modified version of the community health app which will include enhanced decision support protocols and tailored care pathways based on the predictive model developed in phase -1. For example, based on the predictive model, an algorithm will be developed within the app to continuously assess a pregnant woman’s risk of perinatal mortality. If she is identified as high risk, community health workers will be prompted to engage in a tailored care plan specifically designed to reduce the risk of perinatal mortality based on the risk factors identified. This may include increased frequency of visits, more tailored counselling and education, and accompaniment to a health facility to receive appropriate and timely medical interventions.

The modified mobile app will be deployed over a period of 12 months in 1 district with approximately 200 community health workers. The project will analyse the results of the implementation to determine relative performance of the modified app and will then work with the Zanzibar MOH to develop a scaling plan if effective.