In less than a year, the pandemic has spurred the development of countless artificial intelligence models designed to aid in the diagnosis of Covid-19 and spot early warning signs of severe illness among those infected. But so far, only a few have received emergency use authorization from the Food and Drug Administration. That scarcity is a sign of the newness of these tools — as well as the murkiness of the regulatory landscape at a time when unapproved algorithms are being widely tested and rolled out in the clinic.
The latest such authorization, which amounts to a conditional approval of a product, was granted to a system meant to predict whether hospitalized Covid-19 patients are at risk of needing intubation — a heads-up that could allow clinicians to take mitigating steps and plan accordingly. The model was developed by Dascena, a San Francisco Bay Area company working on clinical machine learning systems for conditions including sepsis and acute kidney injury.
When it comes to making predictions about the trajectory of hospitalized Covid-19 patients, “the need is very significant,” said Ritankar Das, Dascena’s president and chief executive officer. He pointed to data indicating that emergency intubations are associated with poor outcomes, including greater likelihood of death and brain damage, as evidence of the importance of “being able to do this in a planned way.”