Google Says New AI Technology Can Predict If Patients Will Die With 95% Accuracy
Google has designed an artificial intelligence system that it believes can predict with extreme accuracy whether patients will die within 24 hours of being admitted to the hospital.
According to the tech giant, its new algorithm is 95% accurate at predicting patient mortality, compared to 86% accuracy with traditional methods.
The system works by sifting through data points from a patient’s electronic medical records, including handwritten notes, to make predictions about the likelihood of death, discharge and readmission.
Dr. Mike Varshavski, a board certified family medicine physician, said he sees positive and negative aspects to this new technology.
"Data in hospitals has been tremendously underutilized. We're literally in the dark ages because half of our information is on paper, half of it is on systems that don't talk to each other," Varshavski said on "Fox & Friends." "There's a lot of bad mistakes happening simply because data's not connected."
He said he's glad that Google is making an effort to connect some of that data, but he has concerns about what happens to that data and who owns it.
"I hope, as a doctor, that these companies use the information that they get to benefit patients and not just the companies themselves," Varshavski explained.
He added that when the system determines that a person has a significant chance of dying, he wants to know what data was used to come to that conclusion.
"And then that decision needs to be evaluated by a human being. Because machines make mistakes, and sometimes they make mistakes based on faulty data," Varshavski said. "There needs to be human oversight of what these machines do."
Watch the discussion above.