CDS algorithm that predicts COVID-19 complications receives Emergency Use Authorization from FDA

Dascena's COViage system uses demographic and vital-sign data pulled from a COVID-19 patients' EHR to calculate their risk of hemodynamic instability or respiratory failure.
By Dave Muoio
03:05 pm
Share

The FDA has issued an Emergency Use Authorization (EUA) to a clinical decision support (CDS) tool that predicts when an adult patient admitted to the hospital with COVID-19 is at risk of further complications.

Developed by Oakland, California-based diagnostic algorithm developer Dascena, the COViage Hemodynamic Instability and Respiratory Decompensation Prediction System reviews demographic and vital-sign data stored in a patients' EHR, including age, gender, heart rate, respiratory rate temperature and blood pressure.

Using machine learning models, the CDS system calculates the patient's risk of experiencing either hemodynamic instability (unstable blood pressure) or respiratory failure. If these outcomes are determined to be likely, the system sends a one-time advance notification to the healthcare provider.

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

The COViage system integrates with a hospital's EHR system, and displays its notifications within the system's standard user interface. It should note be solely used to guide care, but interpreted by a clinician alongside other relevant data.

The EUA was first issued on September 24, according to the FDA's website.

WHAT'S THE IMPACT?

Although it's not a full-fledged regulatory approval or clearance, Dascena's EUA allows the company to deploy its algorithm during the extent of the public health emergency.

According to peer-reviewed study data cited by Dascena and the FDA, the system has demonstrated sensitivity of 0.708 for hemodynamic instability and 0.720 for respiratory decompression. Dascena also noted in a release that its algorithm's demonstrated area under the receiver operator characteristic curve was 36% higher than the standard of care Modified Early Warning Score (87% versus 64%).

Automated early notifications can help providers better identify patients who may be in need of proactive intervention, or who simply could benefit from increased clinical attention during the course of their illness. What's more, understanding which patients require more resources can improve efficiency within the hospital – a particular concern when mounting cases introduce logistical strains on a facility and staff.

The FDA and Dascena note that there is still a chance that the COViage system could miss a case or generate a false positive, and stressed the need for appropriate interpretation by a trained clinician.

“COViage demonstrated the ability to help diagnose respiratory decompensation and hemodynamic instability earlier and more accurately than the standard of care," Ritankar Das, president and CEO of Dascena, said in a statement. "We are excited to bring this machine learning algorithm to the bedside, which may enable the preservation of many lives and improve allocation of hospital personnel."

THE LARGER TREND

Since March, the FDA has exercised its emergency capabilities to permit a number of early or in-development health products to market, most notably with COVID-19 diagnostic testing platforms.

However, this effort has extended to a range of different medical devices, including those that use connectivity or algorithms to improve care inside and out of the hospitals. Some noteworthy examples here are VitalPatch's EUA for monitoring COVID-19 patients' QT intervals, and Eko's EUA for an ECG low ejection fraction tool.

Similarly, the agency has also made allowances for the distribution of low-risk behavioral therapy devices during the public health emergency, leading to the early rollouts of Akili's EndeavorRx, Pear Therapeutics' Pear-004, and Orexo's various digital therapeutics for depression, alcohol use and opioid use disorder.

MobiHealthNews:

The latest news in digital health delivered daily to your inbox.

Thank you for subscribing!
Error! Something went wrong!