Ann: Positive Preliminary Results from SMARTCOUGH-C-2 Study, page-115

While the results aren't what we expected, they aren't too bad. Enough for us to go for FDA on a number of them (just looks like pneumonia isn't enough), and we still have croup results coming later.

What we can pull from this is the app is as accurate as a doctor, this shouldn't be disregarded since some applicable uses (beyond hospitals which seems less so at the moment) are still telehealth, on the field diagnosis, and consumer/self-diagnosis (the last I don't put much weight to).

Last thing to note is the algorithm can keep improving as long as they keep building their dataset (and assuming they haven't overfit the model). The model used in this trial will also not have been the most current one (seeing as Aus double blind was going at the same time).