The FDA has issued an Emergency Use Authorization (EUA) to a Clinical Decision Support System (CDSS) that provides when an adult patient hospitalized with COVID-19 is at risk of further complications.
The system, COViage Hemodynamic Instability and Respiratory Decompensation Prediction System, was developed by Dascena, an Oakland-based company.
COViage examines demographic data and vital parameters stored in a patient’s electronic medical records, including age, gender, heart rate, temperature, respiratory rate and blood pressure. Using machine learning models, it calculates the patient’s risk of hemodynamic instability (unstable blood pressure) and respiratory failure. If these results are deemed likely, the system sends a notification to the physicians.
Even if it is not a real regulatory approval or authorization, the Dascena EUA allows the company to use its system during the health emergency.
COViage was evaluated in a clinical trial that enrolled 197 patients in emergency rooms or hospitalized in five U.S. hospitals between March 24, 2020 and May 4, 2020. The patients had a confirmed diagnosis of COVID-19 and their vital signs were detected and laboratory tests were performed within two hours of their arrival in PS or hospitalization. The data were analyzed by COViage and the Modified Early Warning Score (MEWS) standard of care for comparison.
The result of respiratory decompensation leading to mechanical ventilation, defined as invasive ventilation requiring an endotracheal tube or mechanical ventilation that does not include BIPAP or CPAP, was evaluated 24 hours after the model’s predictions were made. The COViage algorithm obtained an area under the receiving operator characteristic curve of 87% compared to 64% for MEWS (an increase of 36%), demonstrating substantially higher sensitivity and specificity.
The data from this trial were published in an article entitled “Prediction of Respiratory Decompensation in COVID-19 Patients Using Machine Learning”: The READY Trial, in the peer-reviewed Computers in Biology and Medicine journal.
Automated early notifications can help physicians better identify patients who may need proactive intervention, or who may simply benefit from increased clinical attention during their illness.