Games

CSIRO scientists develop machine learning tool to predict patient deterioration

A new machine learning tool that can predict patient deterioration has been developed by scientists from the Commonwealth Scientific and Industrial Research minyak atsiri esential oil Organisation, Australia’s national science agency.

WHAT IT DOES

In collaboration with Princess Alexandra Hospital and Metro South Health, the researchers harnessed EMR data to develop the said tool that can predict when a patient’s vital signs are likely to reach a danger zone, leading to their decline.

The new clinical decision support tool can alert staff about a patient’s risk of deterioration that could lead to possible death, cardiac arrest, or unplanned admission to ICU.

Additionally, the alert tool can set out the reasons for its warning and notify clinicians of the need for intervention.

Based on a test study involving over 18,500 patient records, the ML tool achieved full sensitivity for prediction windows two to eight hours in advance for patients with 95%, 85%, and 70% risk of deterioration.

WHY IT MATTERS

CSIRO scientist Dr Sankalp Khanna noted that some hospitals still cannot access patient data electronically. “Until now there hasn’t been a way to harness all the data in the EMR to predict patient health,” she said.

The CSIRO developers are now in talks with partners for a clinical trial to test how the ML tool can be best implemented into clinical workflows.

THE LARGER TREND

The creation of this ML tool follows the development of a similar ML tool for quickly detecting emerging dangerous COVID-19 variants. Called VariantSpark, the AI tool looks into the RNA of a whole variant, instead of just its spike protein.

In the Asia-Pacific region, Apollo Hospitals has also used AI/ML to harness a large set of patient data in developing a tool to predict disease risk. Last year, the hospital group introduced its AI-powered Cardiovascular Disease Risk tool which provides patients with their risk score.

Leave a Reply

Your email address will not be published. Required fields are marked *