AI Doctor

CHAPTER 7Clinical Decision Support

IF THERE'S ONE AREA that artificial intelligence (AI) won’t be a luxury for in the future of medicine, it’s decision support. But how do we define decision support? Well, it has many definitions. If an algorithm helps a radiologist to read a CT scan, that’s a form of decision support. If an algorithm helps to interpret a home urine sample to diagnose a urinary tract infection, that’s also a form of decision support. Both of these involve an algorithm that uses a single data file to arrive at its output. The CT scan is a single data file that feeds into the algorithm to diagnose tumors or hemorrhages. This is an important point: the algorithm can do its job of timely decision support because it has all the data it needs to do its job. That’s why these types of applications have formed the bulk of the Food and Drug Administration (FDA)‐approved algorithms so far.

There is significant overlap between what we discussed in the Diagnostics, Therapeutics and the discussion in this chapter. One can argue that some of these topics belong in the other chapters, or vice versa. Since decision support has always been one of the ambitions for analytics and AI, I think it’s worth deovting a chapter to this topic. Also, decision support that involves analyzing data from different sources and arriving at the best decision goes beyond analyzing one data file to make a diagnosis.

In the course of clinical workflows, a clinician often makes decisions based on ...