Patient Triage Artificial Intelligence May Boost Future Outbreak Response
A patient triage AI model can accurately predict disease severity and length of hospitalization during a viral outbreak using clinical and metabolomics data.
Source: Getty Images
By - Researchers from Yale School of Public Health (YSPH) have developed an artificial intelligence (AI)-driven patient triage platform capable of forecasting length of hospitalization and disease severity during a viral outbreak, according to a recent study published in Human Genomics.
The tool is designed to improve patient management and resource allocation using machine learning (ML) and metabolomics data.
“Being able to predict which patients can be sent home and those possibly needing intensive care unit admission is critical for health officials seeking to optimize patient health outcomes and use hospital resources most efficiently during an outbreak,” said senior author Vasilis Vasiliou, PhD, a professor of epidemiology at Yale School of Public Health, in a press release.
The platform uses COVID-19 as its disease model, utilizing a combination of untargeted plasma metabolomics data, patient comorbidity, and routine clinical data to generate predictions.
To develop the tool, the research team leveraged data from 111 COVID-19 patients admitted to Yale New Haven Hospital and 342 healthy, COVID-negative individuals between March and May 2020. Patients were then placed into categories based on treatment needs, such as not requiring external oxygen, requiring positive airway pressure, or requiring intubation.
Doing so allowed the researchers to identify a panel of multiple metabolites correlated with COVID severity and disease progression, including kynurenine, hydroxytryptophan, picolinic acid, glucuronic acid, and allantoin. The analysis further highlighted elevated blood eosinophil as a novel potential biomarker for disease severity.
Patients who required positive airway pressure or intubation were also found to have decreased plasma serotonin levels.
The findings demonstrate the potential utility of predictive AI for public health, the research team concluded.
“Our model platform provides a personalized approach for managing COVID-19 patients, but it also lays the groundwork for future viral outbreaks,” Vasiliou noted. “As the world continues to grapple with COVID-19 and we remain vigilant against potential future outbreaks, our AI-powered platform represents a promising step towards a more effective and data-driven public health response.”
Despite these promising results, the study had multiple limitations. First, data were collected before the advent of most COVID-19 treatments and vaccines, which could impact metabolite changes observed in patients.
The possible role of race and ethnicity in the findings also cannot be excluded, as the cohort of COVID patients contained a higher proportion of Black patients, while the healthy cohort was mostly white.
This research reflects a growing interest in leveraging advanced technologies to support outbreak response.
In 2021, researchers from the University of Pittsburg School of Medicine and Carnegie Mellon University showed that a method using ML and whole genome sequencing could improve infectious disease outbreak detection.
The tool, known as the Enhanced Detection System for Healthcare-Associated Transmission (EDS-HAT), clusters patients with similar infections and identifies potential transmission routes.
Over the two years from which data were sourced, the research team estimated that EDS-HAT could have prevented dozens of infectious disease transmissions and saved UPMC Presbyterian Hospital hundreds of thousands of dollars.
