An artificial intelligence tool can help doctors when deciding which COVID-19 patients can be sent home, according to a study that started in May.
Researchers at New York University used the AI tool to analyze thousands of COVID-19 cases. The team used data from Asia and Europe to develop the model.
It uses lab results, vital signs and oxygen requirements to determine how the patient will do over the next few days.
The study has found the AI tool can identify patients that will do well with 90% precision, which could help doctors prioritize care and make discharging plans for others.
Doctors always make the final call, but the AI tool can help reaffirm their decisions.
“It's augmenting, but it's also teaching a little bit, because you know when the physicians see a patient that looks well, but the models say they may not be well, they can look at the, the features right? The vital signs and labs and other things that are saying that this patient isn't well and then they can start to understand or create a mental model,” said Yin Aphinyanaphongs, Director of Translational Clinical Informatics for DataCore, NYU.
The researchers are still running randomized trials for about another two months, looking at the average time patients spend in the hospital.
The hope is the AI tool could help reduce the length of stay.
“That's extremely compelling because now you don't just show oh I have a model that can predict a favorable outcome, but now you have a model, right, that actually affects the outcome,” said Aphinyanaphongs.
Another team at NYU is now conducting a survey among doctors that have used the tool. One facility has already implemented it.
The researchers wanted to make sure it's easy for other organizations to start -- so the software can be up and running in about half a day.