Holden Karau, a trans woman from Canada, was surprised — and not in a good way — when she first had to navigate the U.S. health care system.
She says she received a string of insurance denials as she sought care. At first, she thought it was just because she was seeking gender affirming care. But, then she got hit by a car, and realized this is a problem just about anyone can face.
So, Karau did something about it. The software developer used her programming skills to help other people do something she had become pretty good at: Appeal health insurance denials. And she did using AI.
The result is a website called FightHealthInsurance.com, where anyone can go to get AI to help them write their health insurance appeal. Karau joined The Show to talk more about it, beginning with her first introduction to the American health care world.
Full conversation
HOLDEN KARAU: So, it started off very confusing and, in fact, the first time I went to a doctor's office in America. I remember being asked for my visa and just being super confused and thinking they wanted to know my immigration status and, and then things just kind of got, got worse from there.
You get lots of denials. Initially, I thought it was just because I was trying to access gender affirming care and at the time, that was maybe, like, not a super common thing for, for people to do. And so I figured, you know, maybe this is just, you know, a corner case of the American health care system and it works normally for other people. But later on that, that very clearly became not the case,
LAUREN GILGER: Right And then you get hit by a car, right? Like and things got even crazier.
KARAU: They did, they did. Yeah. So in 2019, I got hit by a car, you know, there there's of course the initial shock like, “ok, I broke my bones like everything hurts. Oh dear God, how do I survive?”
But then I'd say like within like, you know, several hours of being hit by a car, it was like, ok, I remember reading an article about the hospital that the ambulance brought me to. They're not in network with anyone, and this is the only tier one trauma center in, in San Francisco. And it's just like, “ok, well, probably where I need to go, but like this is also going to be super annoying to deal with.”
I still remember being high on the painkillers that they were giving me trying to figure out health insurance.
GILGER: So it sounds like you spent a lot of time after all of this happened, like fighting insurance claims that were denied. Did you win? How to go?
KARAU: Yeah, so I won most of them, which is super exciting and to be clear, also not normal, right? And like I would also say that I frequently would go to, like, unreasonable lengths.
So I remember like I was at one company and they didn't want to pay for certain stuff for me or some, some other folks who I, who I talked to and I was like, OK, so like this is a company that clearly cares about their public image and they care about being seen as like, you know, a queer friendly employer, right?
Like, they put in all this effort, but on the other hand, like also this is at the second level of denial, the next level of appeal for this kind of health insurance plan is you sue them. And it's like, so if I sue you, this will be public and I'm fine doing that.
Most people are not particularly willing to take their employer to court over, you know, not huge amounts of money, but I was kind of just like, “this is really wrong and I'm really annoyed with you. So like, I will sue you because I'm, you know, not a lawyer, but it looks like I will probably only be responsible for my costs if I lose. So I really don't have a lot to lose here whereas you have a lot to lose.”
So, like, “would you want to consider, like, maybe, you know, changing things?” And they actually, like, I'm really happy about that outcome eventually because not only did I get them to, like cover it, I actually got them to change some of how the policy worked for people going forward. And that's, that's where I was like, “oh man, I can actually like, you know, by being completely unreasonable,” as the saying goes, “I can force the world to change in small little bits.”
GILGER: That's amazing though, that's amazing. So, OK, so, right, you're a software developer, also, like in your day job. So you have this, this issue that you've been through, you understand that you can make a change here and you decide to kind of do it and do it on a level that is accessible to everybody. And you find this really great use for AI, you trained AI to help people write their health insurance appeals. So tell us about Fight Health Insurance here.
KARAU: Totally. So, I think a lot of the uses of AI are, perhaps not super fantastic, and that's as much as I'll say because I want to continue to be employable in the Bay area.
But like, yeah, I think this is a pretty good use case, because, like, insurance letters are something that most people have very limited experience writing and you have to write them in certain ways for them to like, take you seriously at all, right?
But then the problem is for most AI Right, like what you, what you really want is you want a lot of examples of it going well, right? So you want a lot of insurance denials and then you want a lot of successful insurance appeals. And the problem is the people who have that are the insurance companies and the insurance companies weren't exactly excited to share information with me.
So it was like, “OK, well, that's a challenge. How do I, how do I find information so that I can make this model?” And so, I looked around a bunch, and then the State of California, we have an independent medical review board and our independent medical review board actually publishes their decisions. And that is fantastic because it means that you've got all of this information about like, “hey, this is a thing that an insurance company denied. This is why the person appealed it and this is why we either approved or rejected that appeal.”
That is the information that you want. And then we have this data set, we train a model. And so the first model that I train, on this data set, isn't great, get back to my day job for a while. And then another health insurance company, technically pet health insurance, denies my pet's health insurance claim. And I'm like, “no, this is a bridge too far. I am going back to my project and I'm finishing it.”
GILGER: Wow. So it was your dog's health insurance that really set you over the edge here? OK, OK.
KARAU: That was the final push. That was the final push.
GILGER: So you train this A I to do this very specific thing by finding this very specific data set, right? And working with that because health insurance appeal is very specific. Are you essentially using their own language here to help people write their appeals?
KARAU: Very much so. So, it's very much using the language of they're called “second” or “third level appeals,” depending on the plan. But using that kind of language to produce the right kind of things fundamentally like most people, I think most people know why they need the medicine, but they don't know how to say it, right?
Like, “I know that I need to take my hormones,” that is apparently like not sufficient medical justification. So instead you have to say like, “according to the standards of care” like, “this revision,” blah, blah, blah, blah blah.
GILGER: Right, OK. So this is pretty new. But how is this going, like have people been successful so far in getting these denials reversed?
KARAU: So the majority of the people that I've talked to so far are like, hey, I submitted my appeal but the insurance company has not gotten back to me yet. And there was one person who was like, yeah, no dice. They got back to me and said no.
But then there was another person who got back to me and was like, “yeah, they said yes.” And actually the person who said yes was not their insurance company. It was the Department of Managed Health Care, and so that's like a second level external appeal. And so their insurance company had said no, but then they wrote the appeal to the State and the State said yes.
And that was, I don't know, that was the happiest email I got that day. It's not going to change the world yet. But I like I have this, like, dream of we make it so that it costs the insurance companies more to deny care. So, they do it less often and that might not be a reasonable dream, but it is my dream.