Scientists have developed an artificial intelligence system that uses the same technology behind Chatgpt to predict outbreaks of infectious diseases, marking the transition from traditional mathematical models to inference-based predictions.
The tool, called PandemicLLM, outperforms existing latest approaches in predicting hospitalizations for Covid-19 in 50 U.S. states in 19-month tests.
Exceeded digital processing
Unlike conventional prediction models that rely purely on mathematical calculations, the new system treats disease prediction as a reasoning problem. It processes multiple types of information simultaneously, including infection data, government policies, demographic information and genetic monitoring of viral variants.
“Covid-19 sheds light on the challenges of disease spreading due to the interaction of changing complex factors,” said Lauren Gardner of Johns Hopkins, who leads the research team. “When conditions are stable, the model is good. But when new variants appear or policy changes, we are very bad at predicting outcomes because we don’t have the ability to model the information types to include the types of key information types. The new tools fill that gap.”
Adapt to new threats in real time
One of the most promising features of the system is its ability to incorporate new information about emerging virus variants without the need for complete retraining. During the testing process, the researchers provided model information for the dominant SARS-COV-2 BQ.1 variant at the end of 2022. The system’s performance is 28% higher when it can access this real-time genetic information.
What makes this particularly important is the speed of response. Traditional predictive models struggle when biological or social conditions change rapidly. However, AI systems can process textual descriptions of new variant features and immediately break them down into predictions.
Confidence level guides decision making
Rather than producing the usually unreliable precise numerical predictions, the system divides future trends into five levels: large decline, medium decrease, stable, medium increase and large increase. Importantly, it also provides confidence for each prediction.
Tests show that when the system expresses high confidence (85% or higher), its accuracy reaches a one-week forecast, a one-week forecast of 75% and a three-week forecast of 77%. This reliability measure may be crucial for public health officials who make resource allocation decisions.
Learn from pandemic restrictions
The COVID-19 pandemic has exposed serious gaps in existing predictive capabilities. Traditional models performed quite well in periods of stability, but when new variants emerged or policies changed, traditional models experienced huge failures. The researchers specifically designed their systems to address these shortcomings.
“The pressing challenge of disease prediction is trying to figure out what drives the trend of infection and hospitalization and build these new information flows into modeling,” Gardner explained.
The system handles four different data streams:
- National demographics, health care capacity and political characteristics
- Disease surveillance data, including case, hospitalization and vaccination rates
- Government policy information, such as masking authorization and collection restrictions
- Genetic surveillance data tracks virus variants and their characteristics
Freezing parameters improve efficiency
To make the system more practical to the real world deployment, the researchers found that they could “freeze” most basic language model parameters during the training process. This approach maintains prediction accuracy while significantly reducing computational requirements, a key consideration for health sectors with limited technical resources.
This study, supported by the National Science Foundation and the Centers for Disease Control and Prevention, also shows that different types of information become more or less important at different stages of an outbreak. For example, policy and demographic factors have shown to have greater influence during increased infections, while epidemiological trends dominate during periods of stable or phase decline.
Future disease preparations
Assistant Professor Johns Hopkins, who specializes in reliable AI, “Traditionally, we use the past to predict the future.” “But this doesn’t provide enough information for the model to understand and predict what is happening. Instead, this framework uses new types of real-time information.”
Researchers envision adapting to other infectious diseases, including influenza, avian flu and respiratory tract comprehensive viruses. With the proper data flow, the system may provide early warnings for various outbreaks.
Gardner highlighted the broader implication: “We know from Covid-19 that we need better tools so that we can inform more effective policies. There will be another pandemic and these types of frameworks are essential to support public health responses.”
The study emerges in natural computing science in nature and represents a new approach to infectious disease monitoring that combines artificial intelligence reasoning with traditional epidemiological data to create more robust and adaptive prediction tools.
Related
If our report has been informed or inspired, please consider donating. No matter how big or small, every contribution allows us to continue to deliver accurate, engaging and trustworthy scientific and medical news. Independent news takes time, energy and resources – your support ensures that we can continue to reveal the stories that matter most to you.
Join us to make knowledge accessible and impactful. Thank you for standing with us!
