The FDA has agreed to a special authorization allowing a vendor to deploy its AI tool designed to predict adverse events in hospital ICUs.
The agency issued an Emergency Use Authorization for an ICU solution developed by CLEW Medical, whose CLEW-ICU uses predictive screening to identify patients with a higher likelihood of being diagnosed with serious conditions such as respiratory failure or hemodynamic instability. These conditions are frequently associated with COVID-19.
CLEW trained its AI using data on nearly 100,000 patients in ICUs. The AI-based algorithms are machine-learning models trained to predict such problems hours in advance. The product, CLEW-ICU, integrates care delivered by local and remote teams. According to the vendor, this clearance is the FDA’s first for technology in this category.
The FDA authorization follows a steadily increasing deployment of predictive tools in the emergency department, ICU and other hospital areas where patients might be at risk of sudden decline. These tools have been particularly popular as the pandemic has raged on, bringing in patients in particularly sensitive and volatile conditions.
For example, the University of Virginia Health System has rolled out an AI-based predictive analytics tool that monitors COVID-19 patients continuously and predicts whether they are likely to decline. The system, known as CoMET, delivers a graphic summarizing a given patient’s status, drawing on data from patients’ EHRs such as EKGs, lab results and vital signs.
A number of other organizations are working with tools in Epic. One example comes from Stanford University, where researchers have built out a tool known as the Deterioration Index. While Epic trained the prediction model based on hospital patients who didn’t have COVID-19, it still seems to offer some predictive value. Still, almost 50 health systems and hundreds of hospitals are using the Epic technology to create their own predictive analytics engines.
It appears that COVID-19 has sped up a trend that was already underway. For the last few years, hospitals have been taking steps to leverage their big data stores to predict some outcomes. In more recent times, they’ve gone from analyzing stores of past data to handling critical predictions on the fly.
It seems likely that after being put through the COVID-19 wringer, hospitals will be much better equipped to use AI for predicting health events. While it’s most unfortunate that they had to develop more sophisticated technology in a time of crisis, there will at least be benefits from the work they did to keep pandemic patients alive.