AI-powered coding and CDI solutions augment human beings, but do not (and cannot) replace them

By Brian Murphy

Machines are a human augmenter—not a human replacement.

I love the way my recent Off the Record podcast guest Rhoda Chism frames technology in her CDI practice. As an amplifier, extending her reach and allowing her to touch more patients, but not a substitute for her unique skillset.

Two examples from the episode—one illustrating the incredible potential of AI-powered CDI, the other its limitations—are worth listening to. I’ve included a couple excerpts below, edited for clarity.

Augmenter—and life saver

“Iodine has also, to my mind, saved lives. This particular patient was a pre-op patient that was having some sort of surgery done, and she got a pre-op chest X ray. Well anesthesia. If you’re familiar with the pre-op they want to make sure that the patient has a heart and that they have lungs, but this chest X ray, when you read the report, said something didn’t look right, and they recommended a CT scan. Well, they cleared her for surgery because, I mean, she had a heart and she had lungs and she was breathing. Well, when this chart prioritized because there was this finding on the chest X ray that hadn’t been addressed, it allowed us to send a query. The surgeon saw it and ordered a CT scan of her chest. And it looked like a speculated mass. So she came back a few months later and had a lung cancer removed for cure. Now, if nobody had seen that, would somebody else pick up on it? I mean, I don’t know.”

Limited clinical context

“Let’s say perhaps that Iodine prioritizes the case for possible heart failure. But when you open the record and you see that the patient got four units of blood, and the lasix was given because it’s something we do when somebody gets too much volume. A lot of elderly people, if you think about each unit of blood having about 350 milliliters, while they might need the red blood cells, they don’t need all the extra volume. So the computer system is not going to know that, that Lasix was because it’s a routine order for somebody that’s getting two units of packed cells or four units of packed cells. So, you’ve got to have that clinical knowledge to discern what is truly an opportunity or a mismatch versus what is an integral finding. If you’re looking through a computer system, you could be, “oh, this patient might have heart failure.” When in reality, as a nurse, having seen thousands and thousands of these patients come back from surgery, I know that all of that is integral. So while it does a wonderful job, it’s still not a clinician.”

Human and machine: Healthcare needs both, working in harmony.