Imagine a world where predicting disease outbreaks could be as routine as forecasting the weather. This is the vision of researchers at Carnegie Mellon University, who are developing advanced methods to predict infectious disease trends, hoping to provide early warnings to public health officials.
According to Roni Rosenfeld, a University Professor at Carnegie Mellon, outbreaks can occur at any time, making it crucial to have predictive systems in place. Rosenfeld, who co-founded the Delphi Research Group in 2012, aims to create forecasts that help control the spread of diseases like influenza, RSV, and COVID-19.
Forecasting Epidemics with Data
The Delphi Research Group uses an array of data sources to detect trends and generate forecasts. Rosenfeld explained, “Delphi tries to provide early warning to public health authorities by scanning our indicators for unexplained upward trends.” These indicators offer a detailed view of disease dynamics and can estimate risks for the coming weeks.
Understanding the risks associated with respiratory illnesses is essential, especially since vulnerable groups can face a risk up to 40 times higher during peak flu season compared to off-season, Rosenfeld noted. The group strives to make information accessible so individuals can make informed decisions based on the current and predicted levels of disease circulation in their areas.
The Delphi Research Group has expanded its team to include experts like Will Townes and Bryan Wilder, among others, to enhance the accuracy of their forecasts by aggregating and analyzing real-time data.
Building a Robust Data Infrastructure
Delphi’s data repository, Delphi EpiData, includes over 5 billion de-identified records from various indicators. This extensive database supports accurate epidemic forecasting by confirming trends and distinguishing them from random fluctuations.
“We learned that perhaps the biggest obstacle to useful forecasts is the lack of data,” Rosenfeld said. He emphasized the importance of situational awareness to improve future projections. Partnerships with organizations like Sleep Cycle provide additional data on symptoms like coughing and congestion, which can indicate early signs of outbreaks.
Partnerships and Real-Time Data
Delphi’s success largely depends on partnerships that provide valuable data. Rosenfeld highlighted the importance of collaborations with healthcare companies for data access. Roughly half of Delphi’s indicators come from such partnerships, ensuring all data is de-identified.
Zico Kolter, Director of the Machine Learning Department at Carnegie Mellon, praised Delphi for its approach to applying machine learning to real-world problems, stating, “Delphi reflects what MLD is about: combining strong statistical foundations with modern machine learning to tackle urgent, real-world problems.”
Delphi maintains each version of data as it was reported, allowing them to test forecasts under the same conditions faced by decision-makers. This approach enhances the accuracy of their models and forecasts, which are vital for public health decisions.
Innovations and Future Preparedness
During the COVID-19 pandemic, Delphi rapidly scaled its operations, involving volunteers and collaborations with tech companies like Google. Their database and forecasts supported public health officials and informed the public about pandemic trends.
In 2023, Delphi was recognized as one of the national Centers for Outbreak Analytics and Disease Modeling by the CDC, which allowed them to enhance their data systems and expand their infrastructure significantly.
Adam Johns, Delphi’s engineering manager, emphasized the importance of the CDC’s support, which enabled them to develop a more robust and scalable data platform to accommodate new data sources.
Enhancing Public Health Data Accessibility
Delphi’s Epidata platform offers users the ability to access and analyze health data without advanced programming skills. The platform supports public health officials, researchers, and individuals in understanding disease risks and making informed decisions.
Peter Jhon, Delphi’s executive director, stated, “One of our core values is to make our data as maximally accessible as possible.” The platform also guides users to other data sources, encouraging the use of data for research and education.
As Delphi continues to innovate, their goal remains to provide timely and actionable information to mitigate the impact of infectious diseases and prepare for future public health challenges.
Read More Here
