State lawmakers have approved and sent to the governor a bill that would generally preclude health insurance companies from relying on artificial intelligence to decide whether to deny a claim. The measure had near unanimous support in both the state House and Senate and would require a medical director to individually review any denials.
The bill had the support of, among others, the Arizona Medical Association.
Dr. Michael Dunn, a board member of the group and a family physician working in urgent care in the East Valley, joined The Show to discuss how prevalent is it that insurance companies are using AI in decision making.
Full conversation
MARK BRODIE: How prevalent is it that insurance companies are using AI in decision making, at least based on your experience?
DR. MICHAEL DUNN: Yes, well, over the last few years it had actually become a lot more prevalent and the insurance companies for the most part are denying that they're doing this, but we know that, you know, through various court actions that are that are taking place that they are in fact doing it.
Yeah, I know in my own practice we are seeing a market increase in the number of denials and and downcoding denials that we are getting in our billing, and it has the the fingerprints of AI all over it, mainly because the the denials just don't make any sense and they're wrong most of the time.
BRODIE: When you say the denials don't make sense and they're wrong, is that a matter of opinion, or is it like sort of a poorly written ChatGPT essay?
DUNN: You know, it's a combination of both actually. You know, there is some subjectiveness in the billing process. And so, and that that is part of the problem, but that subjectiveness, you know, it, it works on both ways, you know, physicians could potentially overbill, but what we actually see more of it are the insurance companies underpaying, you know, for the services that, you know, that we're billing for.
And then the other part of that is that when they either deny a claim or down code a claim, they're supposed to give you a reason for it and usually it's just about a half a sentence why they are doing it and that blurb makes no sense whatsoever. I mean, we, you know, we, we look at the guidelines and, you know, the billing and the documentation that we've submitted, and they just don't make any sense and and and that's where we believe that, you know, that AI is, you know, is the culprit behind, you know, these problems that we're having.
Because we didn't, we didn't see it like that before, you know, you could always tell that there was some a human behind it and, and there was, you know, some, some thought behind it.That didn't mean we agreed with it, but you could see that, you know, that they were trying to make a case for their position.
BRODIE: So are you seeing more decisions with which you disagree now than you did before? Because I think, you know, if you talk to a lot of physicians, you talk to a lot of patients, it's not like they always agreed with the decisions to approve or deny, especially deny claims that insurance companies were making, you know, before.
DUNN: Sure. Yeah, yes, I would say that we're, we're definitely seeing more, more denials and, you know, in, in my own practice, you know, we're, we're kind of the on the low end of the totem pole in terms of reimbursements, you know, obviously the, the surgical specialties, you know, where their, their billing is, is much higher than ours would be, they're, they're seeing that on a, on a much grander scale. You know, we see it and, you know, we lose, you know, $15 or $20 you know, when they see it, they might lose, you know, thousands of dollars on a case. And, it's definitely is more prevalent than it has been, in times past.
BRODIE: How might this impact how doctors like you do your jobs and treat your patients if insurance companies are using AI to determine whether or not a particular claim is going to be approved or not?
DUNN: Well, I mean, at the end of the day it's, you know, we're a business and you know, to keep our business moving forward and staying healthy, you know, we have to get paid. We've got lots of overhead that seems to go up every year and staff members that need to get paid and you know, they would like raises.
For the most part, insurance reimbursements have been stagnant for the last 20 years, and in some cases they're lower than they were 20 years ago. I don't think the general public recognizes that, but as a physician, you know, with single digit profit margins, it just doesn't take much or too many claims being denied that were in the red.
BRODIE: And presumably that would make you think a little bit maybe more differently about how you treat your patients, maybe tests that you order or diagnoses that you give them. Is that the kind of thing you're talking about?
DUNN: Well, absolutely it all goes into it. I mean, you know, certainly, you know, in the last few years, a new technology, new to us anyway, in our practices is PCR testing. And we can now do that in the office versus sending it out to the labs but it's expensive technology. And there are much better tests, yet now the insurance companies are starting to see these claims come in from, you know, private practice offices, and now their expenses are going up. Well, now we're starting to see them deny those or reduce the reimbursements on them.
And so it's, it does make us think twice, you know, do we want to innovate? Do we want to have these new modalities in our office that are expensive to us but we may not get paid for.
BRODIE: Do you think that there is any role for AI in this space with health insurance companies and maybe even an initial screening of denying versus accepting a claim like do you see any place for AI here?
DUNN: Oh, absolutely, and in fact this bill, HB 2175, does not keep the insurance companies from using AI. And between you and me, I think they'd be crazy for not using it as a screening tool, as you suggested, but they need to be accountable to their patients and the AI cannot be the end all be all. A human, preferably a human that is experienced in the area of, you know, where they're making the denial, you know, needs to review the case and make that, that final.
And that's in fact what HB 2175 does is it, it holds the insurance companies accountable and makes sure that somebody reviews the case prior to a denial or, you know, either a denial of a claim or a prior authorization goes into effect.
BRODIE: So maybe the kind of thing where the AI does sort of an initial screening, it separates into the yeses and the no's, and then a human goes through and looks at the nose and says, OK, these really should be denied, these maybe we can let through.
DUNN: Yeah, I mean, I think that that makes sense from a business standpoint. AI has a lot of potential. It has potential in my office as well. It can really alleviate some of those administrative burdens that we have and in fact, we've, you know, I've started using it as, you know, as a business owner of my practice. It's been useful, but I can't 100% depend on it.I have to look at the work that's been generated and make sure that I agree with it and refine it, but it does make my life a lot easier and I think it will only get better with time.
