Issue link: https://beckershealthcare.uberflip.com/i/1058489
36 DATA INFORMATICS & ANALYTICS Duke University Hospital rolls out AI system for sepsis By Harrison Cook D urham, N.C.-based Duke University Hospital in Novem- ber launched Sepsis Watch, a system that uses artificial intelligence to help identify patients in the early stages of sepsis, according to IEEE Spectrum. Duke University Hospital deployed the system in its emergency department. e hospital plans to eventually extend the tool to the general hospital floor and intensive care unit. "e most important thing is to catch cases early, before they get to the ICU," Suresh Balu, project lead and director of the Duke Institute for Health Innovation, told IEEE Spectrum. e Sepsis Watch system can identify cases based on numerous variables, including vital signs, lab test results and medical histo- ries. e AI was trained on information from 50,000 patient re- cords and more than 32 million data points. While operating, the system pulls information from medical records every five minutes to evaluate patients' conditions, offering real-time analytics that physicians can't provide. When the AI system detects a patient who may be in the early stages of sepsis, it alerts a nurse on the hospital's rapid-response team who will either dismiss the alert, place the patient on a watch list or contact a physician about starting treatment. e system will also walk staff through a sepsis treatment checklist using protocols outlined by the Surviving Sepsis Campaign. "e model detects sepsis," Mark Sendak, MD, physician and data scientist, told IEEE Spectrum. "But most of the application is focused on completing treatment." n 'AI Clinician' outperforms physicians in treating sepsis patients By Mackenzie Bean A n artificial intelligence software system trained to recommend sepsis treatments achieved better patient outcomes than physician-recommended treatments, according to a study published in Nature Medicine. Researchers used data on 17,000 cases from intensive care units nationwide to train the system — called the AI Clinician — to issue recommendations on fluid and vasopressor administration. Researchers then had the AI system offer treatment recommen- dations for 79,000 cases it had not seen before. The system made treatment decisions every four hours, according to IEEE Spectrum. The AI Clinician recommended lower doses of IV fluids and higher doses of vasopressors than what patients received from human physicians. Patients who received treatment doses similar to those suggested by the AI system demon- strated the lowest mortality rates. "It's not mimicking the perceptual ability of the doctor, where the doctor sees certain symptoms and says the patient is go- ing into septic shock," study author Aldo Faisal, PhD, an asso- ciate professor of bioengineering and computing at Imperial College London, told IEEE Spectrum. "It's really cognition that is captured here. We're not just making the AI see like a doctor, we're making it act like a doctor." The researchers plan to test the AI Clinician in an actual hospital setting, using real-time EHR data to issue recom- mendations, although physicians won't know or act on them, according to IEEE Spectrum. If the system proves effective, researchers plan to commercialize the software for hospitals to use across the globe. n Machine learning model helps scientists identify source of deadly viruses By Mackenzie Bean R esearchers created a machine learning software that analyzes virus' genetic information to predict which groups of animals the virus will likely spread to, according to a paper published Nov. 1 in Science. For the study, researchers at the University of Glasgow in the U.K. collected epidemiological and genetic data on several hundred viruses with known animal hosts that can spread to humans. Using machine learning, researchers created a computer model to predict which animal groups would most likely host a virus, based on its RNA genome. The model can predict one of 11 likely animal groups, such as rodents or primates, but cannot identify a specific species as the likely virus carrier. "We would love to know the species," said Daniel Streick- er, PhD, a disease ecologist at the University of Glasgow and lead study author. "But this is a way to hopefully get to the species faster." When used on known viruses, the system identified the general vector type 90.8 percent of the time and the host reservoir type 71.9 percent of the time, according to STAT. At present, scientists mostly rely on circumstantial ev- idence to link an emerging human virus to its animal reservoir. The machine learning model could help scien- tists prevent future outbreaks from occurring in humans, according to Dr. Streicker. "Until you know what the reservoir is, it's difficult to gauge risk, and it's difficult to do anything to stop a disease from emerging," he said. n