Pneumonia remains a significant health challenge worldwide, accounting for 20% of hospital admissions in the United States. Researchers at Northwestern University have developed a machine-learning approach to better predict outcomes for pneumonia patients by analyzing electronic health records (EHRs). This new method identifies five distinct clinical states, three of which are strongly linked to disease outcomes.

The study highlights that one specific state correlates with a 7.5% mortality rate within 24 hours. "Other approaches to classifying the state of pneumonia patients are not as discriminatory," said Luís Amaral, the lead author and Erastus Otis Haven Professor at Northwestern's McCormick School of Engineering. He emphasized that current classification methods provide limited insight into patient recovery chances.

Historically, pneumonia has been categorized based on acquisition—community-acquired, hospital-acquired, or ventilator-acquired—but these classifications do not effectively predict individual prognoses. The research team used data from Northwestern's SCRIPT project and another clinical dataset to identify clusters that more accurately forecast mortality risks.

The paper detailing this approach will be published in the Proceedings of the National Academy of Sciences (PNAS). Feihong Xu, a graduate student involved in the study, mentioned ongoing efforts to apply these techniques to sepsis models.

The study received support from various institutes including the National Heart, Lung, and Blood Institute and the National Institute of Allergy and Infectious Diseases.