Nevertheless there is concern about any data that goes into an EMR. There is always concern about what happens in the worst case scenario where something very bad happens to a patient and their whole record is examined in a courtroom after the fact. If, say, a machine learning model predicts that thing X will happen to the patient and it wasn't acted on, a hypothetical lawsuit may go worse for the provider than if the predicted risk of thing X weren't part of the record.

This doesn't mean that the doctors bury their heads in the sand, but it is a part of the reason providers are resistant to adopting these tools. Relying on their own knowledge and intuition may not be better but it won't hurt them as evidence in a hypothetical malpractice suit. The decision support tool has to be sufficiently better to outweigh this.

I hold no opinion on the validity of this line of thinking, but I can say that it is part of the calculus, at least for some.