A groundbreaking artificial intelligence tool developed by researchers at Johns Hopkins and Duke universities is setting new standards in predicting the spread of infectious diseases. Supported by federal funding, this innovative tool outstrips existing forecasting methods, offering a new approach to managing outbreaks of diseases such as flu and COVID-19.
The Challenge of Prediction
The COVID-19 pandemic underscored the complexities involved in disease prediction. Lauren Gardner, a renowned modeling expert from Johns Hopkins, highlighted the difficulties faced during the pandemic. ‘COVID-19 elucidated the challenge of predicting disease spread due to the interplay of complex factors that were constantly changing,’ Gardner explained. ‘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 esteemed journal Nature Computational Science, the AI tool named PandemicLLM utilizes large language models, a type of generative AI famously used in ChatGPT, to predict disease spread. Unlike traditional methods that treat prediction as a mere mathematical problem, PandemicLLM incorporates reasoning capabilities, assimilating inputs such as infection spikes, new variants, and policy changes.
Advanced Data Integration
The team behind PandemicLLM introduced unprecedented data streams into the model, allowing it to accurately forecast disease patterns and hospitalization trends weeks in advance. The model consistently outperformed other methods, including those on the CDC’s CovidHub. ‘A pressing challenge in disease prediction is trying to figure out what drives surges in infections and hospitalizations,’ Gardner noted, ‘and to build these new information streams into the modeling.’
PandemicLLM relies on four primary data sources:
– State-level spatial data: This includes demographic information, details about the healthcare system, and political affiliations.
– Epidemiological time series data: This encompasses reported cases, hospitalizations, and vaccination rates.
– Public health policy data: This involves the stringency and types of government policies enacted.
– Genomic surveillance data: This includes characteristics of disease variants and their prevalence.
Predictive Testing and Future Applications
To validate the model, researchers retrospectively applied it to the COVID-19 pandemic, analyzing data from each U.S. state over 19 months. The tool demonstrated remarkable success, particularly during times of rapid outbreak changes. ‘Traditionally, we use the past to predict the future,’ said Hao ‘Frank’ Yang, an assistant professor at Johns Hopkins. ‘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.’
With the extensive data available, PandemicLLM can be adapted to forecast any infectious disease, including bird flu, monkeypox, and RSV. The research team is also exploring the potential of large language models to emulate how individuals make health-related decisions, which could assist officials in crafting safer and more effective public health policies.
‘We know from COVID-19 that we need better tools so that we can inform more effective policies,’ Gardner emphasized. ‘There will be another pandemic, and these types of frameworks will be crucial for supporting public health response.’
The collaborative effort included contributions from Johns Hopkins PhD student Hongru Du, graduate student Yang Zhao, Jianan Zhao from the University of Montreal, Johns Hopkins PhD student Shaochong Xu, Xihong Lin from Harvard University, and Duke University Professor Yiran Chen.
For more articles about advancements in AI technology and infectious disease forecasting, follow us at aitechtrend.com.
Note: This article is inspired by content from https://hub.jhu.edu/2025/06/06/artificial-intelligence-infectious-disease-forecasting/. It has been rephrased for originality. Images are credited to the original source.