AI Used to Predict Acute Kidney Injury 48 Hours Earlier

DeepMind, the Alphabet-owned AI company created an AI based technology that can predict acute kidney injury 48 hours faster.
Donna Fuscaldo
Human kidney HYWARDS/iStock

You can now add predicting acute kidney injury to the list of artificial intelligence's accomplishments. 

DeepMind, the artificial intelligence company owned by Alphabet, revealed in a new research report that it developed technology that can predict acute kidney injury in patients up to two days earlier than possible today.


Working with the U.S. Department of Veteran Affairs, DeepMind applied AI to a deidentified electronic health record database that was collected from VA medical facilities. The database was comprised of 703,782 adult patients across 172 inpatient and 1,062 outpatient sites. 

Acute Kidney Injury Could be Predicted 48 Hours Earlier 

The research, which was published in Nature, showed the system could accurately predict acute kidney injury 48 hours quicker and predict nine out of ten patients who would require treatment such as kidney dialysis because of severe deterioration. The model predicted 55.8% of all inpatient episodes of acute kidney injury and 90.2% of all acute kidney injuries that required further invasive treatment.

DeepMind also revealed the results of a peer review of its mobile medical assistant it called Streams. Its been in use at the Free London NHS Foundation Trust since the early part of 2017. The app uses an AKI algorithm to spot patient deterioration, can access the medical information at the bedside and communicate instantly with clinical teams.

The Royal Free said Streams saved it up to two hours each day with specialists reviewing urgent cases within 15 minutes or less. The process would usually take several hours. Fewer cases of acute kidney injury were missed with the rate at 3.3% instead of 12.4%. The average cost of admission per patient was also reduced by 17%, DeepMind said.

DeepMind's AI Tech Could be Used as a Preventative Treatment 

Mustafa Suleyman, co-founder & head of applied AI and Dominic King, health lead, at DeepMind said in a blog post the technology could be used as a preventative treatment and to avoid invasive procedures. It was also developed in a way that the researchers said could be used to predict other diseases and deterioration including sepsis, which is an infection that can kill. 

Most Popular

Acute kidney injury occurs when a kidney stops functioning properly all of a sudden. It impacts one in five hospital patients in the U.S. and UK. It known for being hard to spot with kidney deterioration happening at a fast rate. The condition kills half a million people in the U.S. each year and 100,000 in the U.K., typically because it is not detected soon enough. With early intervention on the part of doctors, 30% of the cases could be reduced, DeepMind said in the blog. 

"Over the last few years, our team at DeepMind has focused on finding an answer to the complex problem of avoidable patient harm, building digital tools that can spot serious conditions earlier and helping doctors and nurses deliver faster, better care to patients in need," wrote Suleyman and King. "This is our team’s biggest healthcare research breakthrough to date, demonstrating the ability to not only spot deterioration more effectively but actually predict it before it happens."

message circleSHOW COMMENT (1)chevron