A new AI tool, PandemicLLM, developed by Johns Hopkins and Duke universities, offers unprecedented accuracy in forecasting infectious disease patterns, potentially transforming public health responses to future pandemics.
Researchers at Johns Hopkins and Duke universities have developed a cutting-edge AI tool that significantly outperforms current models in predicting the spread of infectious diseases, a breakthrough that could transform how public health officials manage outbreaks.
The tool, named PandemicLLM, leverages advanced generative AI techniques, similar to those used in ChatGPT, to enhance prediction accuracy. Its development comes in the wake of challenges highlighted by the COVID-19 pandemic.
“COVID-19 elucidated the challenge of predicting disease spread due to the interplay of complex factors that were constantly changing,” author Lauren Gardner, a modeling expert at Johns Hopkins who created the globally utilized COVID-19 dashboard, said in a news release. “When conditions were stable the models were fine. However, when new variants emerged or policies changed, we were terrible at predicting the outcomes because we didn’t have the modeling capabilities to include critical types of information. The new tool fills this gap.”
Published in the journal Nature Computational Science, the research introduces PandemicLLM as a game changer in the field. Unlike traditional methods, this model incorporates a variety of data streams, including state-level spatial data, epidemiological time series data, public health policy data and genomic surveillance data.
Pioneering Application and Testing
The researchers applied PandemicLLM retroactively to data from the COVID-19 pandemic across each U.S. state over 19 months. The tool consistently outperformed existing models, particularly during volatile phases of the outbreak.
“A pressing challenge in disease prediction is trying to figure out what drives surges in infections and hospitalizations, and to build these new information streams into the modeling,” Gardner added.
PandemicLLM processes information such as recent infection spikes, the emergence of new variants and the implementation of public health measures like mask mandates. This allows it to reason rather than merely calculate, providing forecasts that are both robust and flexible.
Implications for the Future
“Traditionally we use the past to predict the future,” added author Hao “Frank” Yang, a Johns Hopkins assistant professor of Civil and Systems Engineering. “But that doesn’t give the model sufficient information to understand and predict what’s happening. Instead, this framework uses new types of real-time information.”
This adaptability means the model is not restricted to COVID-19 alone but can be adapted for predicting other infectious diseases, such as bird flu, monkeypox and RSV.
The team also plans to explore how large language models (LLMs) can simulate individual health decision-making processes to inform public health policies better.
“We know from COVID-19 that we need better tools so that we can inform more effective policies,” Gardner added. “There will be another pandemic, and these types of frameworks will be crucial for supporting public health response.”
Source: Johns Hopkins University