This 2014 photo shows the rundown corridors of the general operating wing at the Mulago National Referral Hospital in Kampala, Uganda.

Google’s AI tool can predict the probability of a hospital patient’s death

19.9%

Google’s Brain Team is using AI to predict the future of patients after admission to a hospital, including length of stay, chance of return, and even the probability of death.

Published   |  Photo by AP Photo/Rebecca Vassie
This 2014 photo shows the rundown corridors of the general operating wing at the Mulago National Referral Hospital in Kampala, Uganda.
19.9%

Google says its AI tool does a better job than the hospital’s own prediction tool because it can process more types of data, including handwritten notes.

This 2014 photo shows the rundown corridors of the general operating wing at the Mulago National Referral Hospital in Kampala, Uganda.
19.9%

In a recent research report, the team shared the story of a woman with late-stage breast cancer who sought care at a hospital.

This 2014 photo shows the rundown corridors of the general operating wing at the Mulago National Referral Hospital in Kampala, Uganda.
19.9%

The hospital’s computers gave her a 9.3% chance of dying during her stay at the hospital, while Google’s AI tool gave her a 19.9% chance. The woman died days later.

This 2014 photo shows the rundown corridors of the general operating wing at the Mulago National Referral Hospital in Kampala, Uganda.
19.9%

But Google still has challenges ahead, including getting the data it needs to train its AI. The current version processed 46 billion data points from the health records of 114,000 patients.

home our picks popular latest obsessions search