COVID-19 severity prediction tools limited by inadequate datasets

Dozens of AI systems aiming to predict severe disease or death in COVID-19 patients may not be reliable, according to a recent article in STAT News. Because the models were developed in isolation, with many tested and trained on local data, it’s unclear if they are generalizable to broader populations. The datasets used by many also lacked diverse sets of patients, leading one early study of the models to conclude they were highly susceptible to bias. A separate study found a model, which was successfully deployed in Wuhan, China, yielded results “no better than a roll of the dice when applied to a sample of patients in New York.” (STAT News article, 11/18/20)