Hypotension can be fatal, but machine learning can detect it early

A drop in blood pressure can be hard to predict using standard clinical measurements such as heart rate or cardiac output – but experts say a new AI tool could give clinicians a valuable 15-minute lead time.
By Kat Jercich
02:45 PM

Hypotension, or low blood pressure, is extremely common in surgical patients – but it can have serious negative outcomes, including myocardial injury, acute kidney injury and death.

At the same time, said Feras Hatib, development director of algorithms and signal processing at Edwards Lifesciences Research, clinical parameters such as cardiac output, heart rate or respiration are often poor indicators of looming hypotension. 

Those parameters are extremely good for noticing when a patient is unstable, Hatib said during his recent HIMSS20 Digital presentation with Boston Strategic Partners Principal Sybil Munson, Building a Machine Learning Model to Drive Clinical Insights. But "once the [hypotension] event has happened, there is not much you can do."

Learn on-demand, earn credit, find products and solutions. Get Started >>

A retrospective study using the Cerner electronic health record database from 2010 through 2016 reiterated the importance of reducing hypotension in surgical patients. 

"We identified a significant association between hypotension and clinical outcomes within the septic critically ill population," said Munson, with hypotension associated with acute kidney injury, myocardial injury and increased mortality in this group.

Hatib and his team developed what they called Hypotension Precision Index technology. The tool consists of a sensor connected to a catheter measuring radial artery pressure and a monitor. 

The higher the HPI number, Hatib explained, the greater the chance of hypotension occurring in the next 10 to 15 minutes.

The tool also directs the clinician in terms of best steps to treat the incoming hypotension. 

By using machine learning techniques such as sequential feature selection and data from 1,334 patients, Hatib explained, the team was able to predict which measurable cardiovascular parameters can predict hypotension.

The HPI has shown a high degree of clinical success, Hatib said. He pointed to a German study in which researchers used an HPI-based clinical protocol to determine treatment for patients. 

"They were able to significantly, dramatically, and completely eliminate hypotension in these patients," he said.

HIMSS20 Digital

Experience the education, innovation and collaboration of the HIMSS Global Health Conference & Exhibition… virtually.

Kat Jercich is senior editor of Healthcare IT News.
Twitter: @kjercich
Healthcare IT News is a HIMSS Media publication.

Want to get more stories like this one? Get daily news updates from Healthcare IT News.
Your subscription has been saved.
Something went wrong. Please try again.