Search for: All records

Sepsis is a dysregulated host response to infection with high mortality and morbidity. Early detection and intervention have been shown to improve patient outcomes, but existing computational models relying on structured electronic health record data often miss contextual information from unstructured clinical notes. This study introduces COMPOSER-LLM, an open-source large language model (LLM) integrated with the COMPOSER model to enhance early sepsis prediction. For high-uncertainty predictions, the LLM extracts additional context to assess sepsis-mimics, improving accuracy. Evaluated on 2500 patient encounters, COMPOSER-LLM achieved a sensitivity of 72.1%, positive predictive value of 52.9%, F-1 score of 61.0%, and 0.0087 false alarms per patient hour, outperforming the standalone COMPOSER model. Prospective validation yielded similar results. Manual chart review found 62% of false positives had bacterial infections, demonstrating potential clinical utility. Our findings suggest that integrating LLMs with traditional models can enhance predictive performance by leveraging unstructured data, representing a significant advance in healthcare analytics. 

Evaluating public health interventions during disease outbreaks requires an understanding of the spatial patterns underlying epidemiological processes. In this study, we explore how Large Language Models (LLMs) can leverage spatial understanding and contextual reasoning to support spatially-disaggregated epidemiological simulations. We present an approach in which a system dynamics model queries an LLM at key decision points to determine appropriate mitigation strategies, informed by local profiles and the current outbreak status, and incorporates these strategies into the simulations. Through a series of experiments with COVID-19 data from San Diego County, we show how different LLMs perform in tasks requiring spatial adaptation of mitigation strategies, and how incorporating connectivity information through Retrieval-Augmented Generation (RAG) enhances the performance of these customizations. The results reveal significant differences among LLMs in their ability to account for spatial structure and optimize mitigation strategies accordingly. This highlights the importance of selecting the right model and enhancing it with relevant contextual information for effective public health interventions. 

Evaluating public health interventions during disease outbreaks requires an understanding of the spatial patterns underlying epidemiological processes. In this study, we explore how Large Language Models (LLMs) can leverage spatial understanding and contextual reasoning to support spatially-disaggregated epidemiological simulations. We present an approach in which a system dynamics model queries an LLM at key decision points to determine appropriate mitigation strategies, informed by local profiles and the current outbreak status, and incorporates these strategies into the simulations. Through a series of experiments with COVID-19 data from San Diego County, we show how different LLMs perform in tasks requiring spatial adaptation of mitigation strategies, and how incorporating connectivity information through Retrieval-Augmented Generation (RAG) enhances the performance of these customizations. The results reveal significant differences among LLMs in their ability to account for spatial structure and optimize mitigation strategies accordingly. This highlights the importance of selecting the right model and enhancing it with relevant contextual information for effective public health interventions.