Forecasting Health Futures - Precision Disease Prediction Tool

Enable detection and response to outbreaks through precise prediction of malaria incidence, using advanced meteorological, epidemiological, and GIS data modeling.

Dr. Kaushik Sarkar

Mosquito-borne disease creates a daunting cloud cover between global surveillance efforts and ground-level disease patterns. There are more than 220 million malaria infections each year alone, and millions more of dengue, yellow fever, West Nile and others. COVID-19 shares 7 of 10 primary symptoms with malaria, a disease accounting for the majority of inpatient and outpatient visits in many countries. Even as data collection improves, limited diagnostic capacity hinders the ability of health systems to distinguish malaria from other febrile illness among facility-level surveillance data.  

The combination of high malaria burden, natural volatility due to weather-related factors, and weak diagnosis and reporting systems serve to obscure the emergence and outbreaks of other novel febrile illnesses, including COVID-19 and likely the next pandemic disease.  

Climate change is exacerbating this problem as new geographies become malaria endemic, and changing/extreme weather events drive unusual incidence patterns. Our solution provides health systems the ability to predict shifting mosquito-borne disease incidence sufficiently to distinguish it from other disease threats, equipping countries to find and fight them more quickly and effectively.   

Hot spots for malaria become blind spots for the rest of the world, as we work to contain COVID-19 and prepare for the next pandemic. 

FHF-PDP supports decision-making at all levels of health systems in countries where the model is adopted, primarily those countries where malaria is, or may soon be endemic, and where climate change is impacting the disease’s national footprint and patterns.  FHF-PDP helps health leadership use data for detection and response purposes, even as it also supports local decision-makers in routine planning around the control of malaria and other mosquito-borne disease.

To better understand the needs of our primary users, Malaria No More has established a Strategic Support Unit within the Government of Odisha’s Ministry of Health & Family Welfare. Our team is embedded within the State Vector Borne Disease Control Program and has worked closely alongside staff for more than two years. Together we work to strengthen data collection, to test, prove, and operationalize the FHF-PDP dashboard, and optimize its utility to the local/state program’s staff.  

At the same time, we’re working to understand how needs vary from one state and country to the next. We recently convened an Expert Committee on Climate & Malaria in India to better understand dynamics of scaling throughout India, and plan workshops globally in advance of our co-pilot in Senegal and eventual scale throughout Africa.

While components of the health system data integrated in FHF-PDP’s India pilot will require approval before making it broadly available to the public, many other sources of local data included in the model may be applied to benefit other public health programs more immediately.  The OpenStreetMap information, for example, will be readily available to researchers and public health stakeholders. India is a large country with many hard to reach or completely inaccessible areas which have inadequate access to healthcare. By improving knowledge of where people are, the routes to get them, and optimizing around weather, India’s health system can bring better care to people who need it.

We anticipate multiple peer-reviewed papers resulting from the groundbreaking work behind FHF-PDP’s design, implementation and impact evaluation. Our own project staff and members of the Expert Committee on Climate & Malaria in India aim to publish research pieces and will have full access to all FHF data.

Furthermore, FHF-PDP itself is designed to be freely implemented in other settings, and its use will benefit the population of any government who adopts it. The model itself is a public good.

FHF-PDP is being proposed as a Trinity Challenge solution because its predictive capacity will improve the ability of health systems to respond to emerging disease outbreaks, by controlling for likely levels of malaria-cause fever and more quickly revealing other worrisome disease patterns. That enhanced detection ability will enable quicker response times, earlier intervention, and appropriate allocation of resources to benefit communities where outbreaks of any disease occur - offering disproportionate benefit to those vulnerable populations living in difficult-to-access areas. FHF-PDP currently covers a population of 1.3 million, and will soon be scaled to cover 43 million.

Its disease detection capacity is secondary to FHF-PDP’s primary purpose, however. In routine environments, precise geographical malaria prediction will first benefit covered populations by enabling more efficient use of limited state health system resources; Odisha has India’s highest malaria burden. FHF-PDP will enable more targeted malaria prevention campaigns, the pre-positioning of needed commodities, and the precision deployment of community health workers. We expect these actions to reduce the total malaria burden and improve associated health outcomes materially.  

FHF-PDP is approaching completion of its validation in two high-burden districts of Odisha state, and is poised to scale to the remaining 28 districts in 2021.  

The formation of FHF’s Expert Committee on Climate & Malaria, and our partnership with the India Meteorological Department, presents a compelling pathway to scale in India. FHF and IMD teams will hold a national workshop in late 2021 to introduce FHF-PDP to other states, prioritize states that would most benefit from the enhanced dashboard, and reach national scale in 2022. 

Work in Senegal is slated to begin in mid to late 2021, with integrated weather data in a Senegal dashboard by early 2022. FHF is also already working with PMI on plans to disseminate FHF-PDP to decision-makers in their 27 focus countries via a targeted workshop, identify and prioritize those countries most likely to benefit from the enhanced predictive capacity, and develop a plan to include implementation in their work plans.

Within three years, we hope to have FHF-PDP scaled across India and Senegal, and beginning operations in 2 - 3 other African countries, with plans in place to also support the model’s adoption in another 7 – 10 countries in coming years. 

Over the course of FHF-PDP’s implementation and ongoing refinement, we have plans for evaluation via technical, quality, consultation and training indicators on data sources and integration, data models and forecasting, and data visualization. Some of these include:

• % of data sources captured 
• % of inputs with missing data 
• % error reduction following staff trainings
• # of stakeholders consulted for model feedback
• % improvement of accuracy on validation set 
• % pre-processed data accurately visualized
• # graphs interpretable to program managers

In addition to monitoring the progress of FHF-PDP’s roll-out, FHF will also measure the success of the pilot through operations, outcome, and impact indicators, including:

• % users trained 
• % users completing data input on time 
• % missing data 
• % incorrect/invalid data 
• % improvement in blood examination rate/surveillance 
• % reduction in stock out for diagnostics 
• % reduction in stock out of drugs 
• % improvement in reporting by frontline health workers 
• % improvement of reporting quality 
• %completion of timely campaign-style interventions  
• % reduction in malaria incidence 
• % reduction in local transmission rate 
• % reduction of malaria in under-5 children 
• % reduction of malaria in pregnant women 
• % reduction of malaria in tribal population
• % increase in accuracy of forecasting model

The main challenge facing FHF-PDP’s scale is the accessibility of high-quality weather data.

Once fully implemented, FHF-PDP’s ongoing operation and maintenance can be supported by host governments at existing staff levels, if sources of sufficiently granular weather data exist.

The cost of private sector weather data, however, is potentially prohibitive. While most countries do have a national meteorology institution, the quality and granularity of that data varies and may not be sufficiently nuanced to preserve the integrity and precision of results we observe in our Odisha pilot, where IBM’s The Weather Company has provided those data free of charge for the pilot.

Determining the granularity needed for accurate prediction will be a priority in the next stages of FHF-PDP validation and refinement. In India we are working with the India Meteorology Department, and anticipate their data will approach the level of TWC’s and be well suited to FHF-PDP.  

FHF will pursue similar arrangements with national institutions as we scale to other countries, and explore regional collaborations to reduce cost where private sector data is needed. We will also explore development-supported means of improving national meteorological capacity, through outreach to the National Oceanic and Atmospheric Administration (NOAA) and other potential partners.

Malaria No More leads the Forecasting Health Futures collaboration, which also includes PATH, Tableau Foundation, IHME, and Reaching the Last Mile, an initiative of the Abu Dhabi Crown Prince Court. FHF has funding and/or formal affiliation with IBM’s The Weather Company, the Government of Odisha, and the India Meteorological Department. 

Malaria No More admires the resolve of The Trinity Challenge's members to identify and support unique solutions to the world's most pressing health challenge. We feel that the Forecasting Healthy Futures initiative and its Precision Disease Prediction tool is an excellent example of a technology and data-driven innovation to better prepare the world for the next health emergency.  We seek the balance of start-up funding required to complete FHF’s pilot activities in Odisha state, scale to the rest of India, and replicate in Senegal, while we have solid plans in place to pursue financial support in other places for eventual expansion throughout Africa.  

We are also looking to secure additional partnerships with organizations who bring technical expertise and the ability to ensure sustained implementation and continued improvement of the model. Many of The Trinity Challenge members could help us ensure the success of this innovation through explicit collaboration and advocacy. For example, Palantir could likely assist with refinement of the model’s analytics and ongoing enhancement using alternative data sources, and the Bill and Melinda Gates Foundation could contribute to the model’s adoption in new geographies. Many of the academic institution members could assist with research and dissemination of our work.  

Many of The Trinity Challenge members could help us ensure the success of the FHF-PDP innovation through explicit collaboration and advocacy. For example, Palantir could likely assist with refinement of the model’s analytics and ongoing enhancement using alternative data sources, and the Bill and Melinda Gates Foundation could contribute to the model’s adoption in new geographies. Many of the academic institution members could assist with research and dissemination of our work.