Supporting outbreak response through real-time epidemic forecasting

Our solution provides user-friendly, AI guided epidemic risk assessments in real-time to support outbreak response efforts and humanitarian health emergency activities across the globe

Dr. Britta Lassmann, Program Director, International Society for Infectious Diseases and Program for Monitoring Emerging Diseases (ProMED)

Over the past twenty years, at least 30 new human infectious diseases have emerged including HIV/AIDS, Ebola, SARS, MERS and COVID-19; The COVID-19 outbreak which began in Wuhan, China, rapidly developed into a pandemic with more than 2.7 million deaths since January 1, 2020. Together, emerging infectious diseases threaten hundreds of millions of people. Six of the ten “global health threats the world faces” highlighted by the WHO relate to infectious diseases. Globalization has exacerbated the threat of epidemics through easy, rapid, and relatively inexpensive air travel and the export of produce and other agricultural products, which facilitate the spread of pathogens. 

The ongoing COVID-19 pandemic and other recent infectious disease outbreaks have highlighted the global need for reliable data sources and forecasting algorithms that can assist in defining key risk factors associated with outbreaks, as well as their potential for further spread. 

Global health data, modeling frameworks, and tools to analyze and visualize the spread of infectious diseases all exist, but not in a single platform to monitor outbreaks, project case numbers, and predict geographic risk of disease spread in real-time. 

Our solution represents a significant advance over current forecasting methods where disease risk estimates are published in static publication forms. Making probability estimates readily available online through easy-to-interpret visuals, the proposed tool aims to ensure information on ongoing outbreaks is accurate, actionable, and timely. 

Our solution informs health decision makers at national and regional levels of the risks of an outbreak spreading in real-time, and aides government and non-governmental decision makers globally in allocating resources and preparing for the possible disease importation into their country or region.

The prototype application was informed by extensive user research in West Africa to make the data and forecasting outputs as user-friendly, accessible and actionable as possible. Through a collaboration with the Dalberg Design Impact Group and continuous user engagement, we developed solution design and product vision addressing urgent needs. 

Teams involved in this project are committed to making data generated through this solution available open access – as ProMED has successfully done with its outbreak reporting data for more than 20 years while following ProMED's guiding principles:

• Transparency and a commitment to the unfettered flow of outbreak information
• Freedom from political constraints
• Availability to all without cost
• Service to the global health community

For the pilot project and through a collaboration with the Dalberg Design Impact Group we refined the problem statement, user focus, user needs and opportunity areas. We identified and reached out to core users (national and regional public health decision makers and health managers) in West Africa to make the prototype, open access data streams and model outputs as user-friendly, accessible and actionable as possible Their input informed the product (www.mriids.org) schematic, navigation, UI elements and product use flow (59 interviewees across 45 organizations and teams in 6 stakeholder categories). List of participants and user research summary can be provided upon request. 

We plan to continue to engage with these users during the next stages of the project, specifically during the implementation phase. We will harness synergies with other partner organizations such as WHO, CDC Africa, and FAO and seek integration of the proposed tool into existing electronic health management systems, e.g. DHIS2. 

“Red sky in the morning, shepherd’s warning” - this used to be our ability to forecast weather. Today, meteorological forecasting plays a crucial role in our day-to-day lives, with great implications for our economy, health and general well-being. 

Our ability to forecast the spread of infectious diseases in real-time is barely existent and mostly relies on expert opinions. After working on this problem for two years, our team has made significant progress in:

1. Delivering a pilot forecasting tool, restricted to Ebola, and disease spread across Africa. 

2. Expanding the forecasting tool and visual outputs to COVID-19 on a global level.

In one year, we envision to be closer to meteorological forecasting through automation of data ingestion and further refinement of the forecasting algorithms. 

In three years, we envision to reach universal epidemic forecasting capabilities with readily available risk assessments for emerging pathogens of global concern. 

The proposed tool will be a valuable information resource for decision makers, fostering prevention and control resource allocation tailored to the risk level experienced by a country/region. 

Although we all hope that there will be no infectious disease outbreaks, we know that they are inevitable. As the WHO has stated, "The world will face another pandemic—the only thing we don't know is when it will hit and how severe it will be." Long-term, we will measure success of our solution by the impact on people’s health. Over the intermediate-term, project partners will measure the project’s success by the tool’s capacity to provide actionable data that informs public health decisions on the allocation of human and financial resources to forestall infectious disease spread. To do this, we will track how and where the tool is used and implemented and the outbreaks it is used to analyze, and monitor and evaluate the model’s efficacy in accurately predicting case/death counts and disease spread and the extent to which the information it generates is used to curtail or prevent infectious disease spread. 

Financial: Funds are needed for project development to reach universal epidemic forecasting capabilities. Once the tool is fully developed and automated, maintenance costs will be significantly lower. Teams involved in this project are committed to making data generated through this solution available open access. 

To foster scale up, the tool is designed in a modular, flexible manner with emphasis on unit testing to ensure robustness and sustainability over time. The model is developed in the open-source statistical software R; code is made publicly available via github.com to promote collaborative contributions. We expect the international infectious disease community will contribute to this activity with new research involving additional data sources/methodologies. 

Methodology: Despite the high sensitivity of digital disease surveillance data from ProMED and HealthMap, concern remains regarding the under-reporting of cases, especially early on during an outbreak. Our statistical framework was built to account for potential under-reporting and to correct estimates. As an outbreak unfolds, surveillance efforts are usually intensified and this is when surveillance data may become biased. Subsetting the analysis to specific temporal periods where we believe the bias is minimal mitigates the impact. The use of AI aids by adjusting corrections in real-time based on past observed patterns. 

Team members are affiliated with the following organizations:

• International Society for Infectious Diseases/ProMED, Brookline, MA (US)
• Imperial College London (UK)
• HealthMap at Boston Children's Hospital/Harvard Medical School, Boston, MA (US)
• Sussex University (UK)
• University of California, Davis, CA (US)
• Northeastern University, Boston, MA (US)

Partnerships and funding through the Trinity Challenge would be catalytic to allow project partners to move towards universal, real-time outbreak forecasting by expanding the forecasting algorithms to additional groups of pathogens and automating the data sharing and forecasting frameworks for rapid scalability. 

Google, facebook or other organization who are willing to share global mobility data for real-time integration into forecasting algorithms.