Smart Solutions Battling Malaria in Liberia with AI
June 27, 2024
Want to try out demo Malaria Prediction?
Leave your email by filling out the form via the “Test it out” button
Introduction
On May 15, 2024, the Omdena community proudly presented their innovative project: “Developing an AI-powered App for Predictive Modeling and Forecasting of Malaria Prevention in Liberia”
This ambitious initiative brought together data enthusiasts from across the globe, including countries like Sri Lanka, India, the United States, Bolivia, Portugal, Brazil, and several West African nations, including Liberia. For many participants, this project provided an invaluable opportunity to apply their university-acquired skills in Data Analytics, Data Sciences, and Machine Learning Engineering to a real-world problem.
This project brought together different types of data and used advanced prediction methods to report the patterns in malaria spread and deaths. Its goal was to help prevent malaria in Liberia by making it easier to plan ahead and take action early. The project aimed to lower malaria cases and deaths, especially among children and pregnant women who are at higher risk.
Malaria remains a major threat, particularly affecting children under five and pregnant women. The project aimed to leverage technology and data to improve prevention efforts. By integrating various data sources and using advanced predictive models, the app helps health officials predict where and when outbreaks might occur.
The project was spearheaded by Liberian Data Scientist Daikukai Bindah, who guided the team through the complex process of developing and implementing this AI-driven solution. This article explores the significance of addressing malaria in Liberia and the innovative methodologies employed to overcome data scarcity and build a predictive model.
The Problem
Child receiving malaria testing and treatment in Sierra Leone. Photo by PMI Impact Malaria
Malaria remains one of the deadliest diseases globally, particularly affecting children under five and pregnant women. In 2020, Africa accounted for 95% of malaria cases and 96% of malaria deaths, with 80% of the fatalities being children under five, according to the report “Mathematical Modelling and Optimal Control of Malaria Using Awareness-Based Interventions” by FAHAD AL BASIR and Teklebirhan Abraha. The World Health Organization (WHO) highlights that malaria during pregnancy can lead to severe complications, including anemia, premature birth, and low birth weight.
Efforts to control and eliminate malaria have faced significant challenges. Anopheles mosquitoes, the primary vectors of malaria, have developed resistance to medications and insecticides. Inadequate health infrastructure, limited financial resources, and climatic factors further exacerbate the spread of malaria. The mobilization of people within countries also contributes to the disease’s proliferation.
Malaria-causing parasites, carried by Anopheles mosquitoes, include over 465 species, with 70 capable of transmitting malaria and 41 identified as major health threats.
Despite these challenges, there is hope. Advances in molecular identification, as highlighted in the study “Revolutionizing Malaria Vector Control: The Importance of Accurate Species Identification through Enhanced Molecular Capacity,” and innovative technologies, such as those reviewed in “Leveraging innovation technologies to respond to malaria: a systematized literature review of emerging technologies,” offer new avenues for combating malaria. However, these efforts require precise and continuous data collection to be effective.
Project Goals
The AI-powered Anti-Malaria App was supplemented with data to design better interventions for malaria. Some of the goals decided were:
- Malaria Transmission Risk Prediction: Identify high-risk areas and populations by using historical data, weather patterns, and human behavior to create accurate predictive models.
- Malaria Outbreak Forecast: Anticipate the timing and severity of future outbreaks through real-time data analysis of weather patterns, mosquito populations, and human mobility.
- Identification of Environmental and Social Determinants: Unravel the factors contributing to malaria transmission and vulnerability by examining large datasets to inform and optimize intervention strategies.
Methodology and Sources Of Data
The project faced significant challenges due to data scarcity. Collecting continuous and comprehensive data is essential for developing accurate predictive models. However, not all regions have the resources or infrastructure to gather such data consistently.
To tackle this issue, the team decided to focus on integrating data from various international sources:
- The DHS Program: Provided malaria indicators from The Demographic and Health Surveys (DHS) Program.
- World Bank Data Catalog: Offered valuable economic and social data.
- The Humanitarian Data Exchange: Supplied rainfall indicators at the subnational level.
- Malaria Atlas Project: Provided comprehensive data on malaria prevalence and control efforts.
By leveraging these sources, the team was able to gather sufficient data for the counties of Liberia, which was crucial for developing the predictive model. As Thomas James, a North American Data Scientist, explained during the project presentation:
What We Found From Analyzing The Data
Exploratory Data Analysis (EDA) was conducted to examine the intercorrelations of various rainfall measurements, identifying distribution patterns and their correlation with malaria deaths. The data showed significant variations in malaria cases and deaths across different counties in Liberia.
It can be seen how the Average Cases Value and the Average Deaths Value increase for each year in Liberia.
The data show that both cases are not reduced.
The country is organized by 15 counties and it is discovered that each one has different experiences from others with malaria. This can be seen with the number of cases and deaths in Greater Kru and River Gee counties.
Choosing the Model: After extensive testing, the team selected the Random Forest Regressor (RFR) as the best-performing model. Key performance metrics included:
- Mean Absolute Error (MAE): 5.305979429918038e-13
- Mean Squared Error (MSE): 6.930678541261898e-25
- Root Mean Squared Error (RMSE): 8.325069694159862e-13
These metrics demonstrated the model’s high accuracy in predicting malaria cases and deaths. The model also provided valuable insights into the influence of factors such as rainfall and health interventions (IRS, ITN, medical treatments) on malaria prevalence.
Application and Future Directions
The AI-powered app offers users a profound understanding of how environmental and social factors influence malaria transmission in Liberia. By precisely identifying high-risk areas and accurately predicting outbreaks, the app empowers health officials and policymakers to devise more effective prevention strategies.
This targeted approach ensures that resources are allocated efficiently, significantly enhancing the impact of malaria control efforts. Consequently, the app plays a crucial role in reducing malaria cases and deaths, contributing to improved public health outcomes and fostering a healthier, more resilient community.
Benefits of Using these Methodologies and Impact
- Enhanced Prevention Strategies: By predicting high-risk areas and future outbreaks, the app enables more effective allocation of resources and targeted interventions, improving overall malaria prevention efforts.
- Data-Driven Decision Making: The integration of various data sources provides a comprehensive view of malaria transmission, helping policymakers and health organizations make informed decisions.
- Improved Public Health Outcomes: By reducing malaria cases and deaths, the app contributes to better health outcomes for communities, particularly vulnerable groups such as children and pregnant women.
- Economic Benefits: Reducing the prevalence of malaria can lead to economic improvements by decreasing healthcare costs and increasing productivity.
Future Possibilities
- Incorporate Additional Features: Including human mobility, temperature, humidity, vegetation cover, and socio-economic factors.
- Real-time Data Integration: Enabling near real-time predictions by integrating real-time weather and healthcare intervention data.
- Explore Advanced Models: Utilizing Convolutional Neural Networks (CNN) or Long Short-Term Memory (LSTM) networks to capture more complex relationships.
Conclusion
The development of the AI-powered Anti-Malaria App in Liberia demonstrates the transformative potential of data-driven approaches in tackling global health challenges. By combining the expertise of a diverse, international team with advanced machine learning techniques, this project offers a promising solution to one of the world’s most persistent health threats.
Continued innovation and collaboration are essential to further enhance the app’s capabilities and expand its impact. By leveraging AI and comprehensive data analysis, we can move closer to a future where malaria is no longer a threat to millions of lives.
