An 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 recommendations 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 demonstrated 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 going into septic shock," study author Aldo Faisal, PhD, an associate 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 recommendations, 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.