Battling Counterfeit ePrescribing | Anthony Brooke, DrFirst, Inc.

Speech to text:

Mike Koelzer: Anthony, for those that haven't come across you online, introduce yourself. And tell our listeners what we're talking about today.

Anthony Brooke: My name is Anthony Brooke, and I'm the President of Provider Solutions and Technology at Dr.First. And Dr. First is one of the major contributors of prescriptions, or e prescribing software, that end up at a pharmacy. And so today, I want to talk about one of our newer inventions.

that we think will provide pharmacy one of the first ever prescription monitoring systems to have a standardized way to determine if there's an outlier. Is there something that they should maybe take a look at that's potentially [00:02:00] fraudulent?

Mike Koelzer: It's so fascinating talking to companies like yours because pharmacists, if you ask the average pharmacist, they'd complain about a few things, but they'd say, yeah, we pretty much have everything we need. We're doing it. Okay. And it's kind of like asking Somebody who had a old, phonograph vinyl record.

How are things going? Well, people aren't smart enough or have enough ingenuity to say, well, we wish we had a little thing we held in our hand that had 1 million songs on it. People aren't desiring these things because they don't. No, they need it even.

So I put this in there. It's like, someone said to me, what do you need? I'd be like, just, and they'd say, well, what if you could get a pretty good snapshot of who's coming in your door before you know anything else? Practically. 

Anthony Brooke: Yeah I think you're right. And it's something that has bit important for all the participants or stakeholders [00:03:00] along prescription management value chain. But I believe that this has really been in the realm of security officers and fraud officers and controlled substance officers. on that compliance level, trying to minimize risk.

And what we recognize that each one of us, where we do handoffs along that prescription chain, we're kind of blinded about what happens with the next step. And that limits our ability to make very informed decisions. And so while each one of us acts in the best interest as possible, as fraud, as bad actors become more sophisticated, Typically employing pretty advanced AI to game the system and figure out how they can make a couple extra billion.

From that, that's created enough incentive, but also a challenge for us that we decided maybe it was time. not to [00:04:00] necessarily earn more, but rather be a good citizen in that value chain and try and play nice with our pharmacy counterparts and show them what was happening up at the e prescribing level.

At the same time, it was kind of complex and how do you go through the volume and speed that pharmacists are challenged to do to get through all the scripts that they're presented every day and do their daily work. And so from that, we tried to consolidate down to have A simple score that they could know, huh, it's a score coming from Doctor First, a TRX score, and if it's this value, I should probably pay attention when it's that one of those drug therapies I worry myself with, and let them make a decision.

Mike Koelzer: My mother, God rest her soul, but It was easy in the day to be a child of my mother because she would just label anybody that she had reservations about. She [00:05:00] would just label them a hippie, and so we were kids and she'd say, Oh, that's so and so he's a hippie. Well, that meant everything to us. It meant they're not trustworthy. It meant a ton of stuff to us. Well, now. Everybody looks like a hippie. My sons have hair halfway down their back and my daughters have all the tattoos and all the fun stuff people have.

And I'm joking, of course, cause you can't put everybody into that category, but that's a lot of time. What pharmacists are left with is some kind of a a judgment maybe a stereotype, but

In our profession, you can get in. More trouble by not having those views than you can by having those views, because you're not going to get in as much trouble from the DEA if you don't feel something as if you do feel too many things. Unfortunately though, it's not. Always [00:06:00] fair because believe it or not, some people really do need pain medicine.

 And so sounds like your thing is using AI. you're not going to come out and say someone's a hippie. There's gotta be better ways to do it.

Anthony Brooke: Yeah, so, right now, I think that pharmacists do a pretty good job at assessing the patient or the patron that's coming to pick up a script. Because they have that information available, there's a person there they can actually look at, evaluate, ask questions of and do an investigation. What they have less access to is the prescriber that wrote that.

And better still, the concept that 1 in 10 prescribers has had their identity stolen. And so there's a legitimate, good, hard working prescriber out there who has someone doing the transaction out there that it's not [00:07:00] them. It's not in their practice. It's not even in the same software they tend to use.

And yet, they know enough that they've been able to steal their identity, create another account, and start prescribing on their behalf. And that's what we really focus on, because that's the part upstream from the pharmacy that we can really work with. That's something that we happen to, as a software that provides e prescribing in 270 plus medical record systems, not to mention all of our direct folks that use our software themselves.

With that, we get to see, where are they? What are they prescribing on? For whom? Is this outside of the pattern? We call that a digital fingerprint. And that's what our AI uses to identify Is this the same pattern or is this slightly different on this prescription? 

And what's unique is not doing it at the NPI level, but rather doing it at the prescription level.

Mike Koelzer: [00:08:00] I think as you're talking here, I think to the the little captcha things, with check on the box. And this, I just heard it. Secondhand, but you're looking for non perfection here. Computers can just go and tap on the thing. You're looking for an old fart like me, who, can't find the pointer on the mouse for a while, it's tucked away.

Somewhere on your screen and you move it around and then you slow down and do this and you miss it and you come back and all that kind of stuff. And you alluded to it there with maybe the IP address, maybe what time of day it is, has the doctor used in the past?

What has been their style and putting all that together, especially with AI and just saying, this is outside of the norm.

Anthony Brooke: Absolutely. And so, like many others, and even pharmacies, we started with statistical analysis. With a smarter than average human looking at things, trying to find something that didn't seem to line up. With pretty straight logic. What's unique about the approach that we've taken is that by using [00:09:00] a a new type of AI that came out in about 2022, we applied it to the same problem we had been working on for a while.

What we found was it was better at being able to come up with, they call them dimensions or edges. You can think of it as the relationships between things. So not only that time of day or the type of device or. Things that you would think of, but also the relationship between them. I'll give a simplistic one.

if you're a prescriber registered in State 1, but you happen to be in State 2, eh, maybe you're at a conference. But then you're prescribing for a patient that's in State 3, and you've never interacted with that, and the prescription's being delivered in State 4, those become, each one of them by themselves, perfectly legit.

But the relationship between each one of those becomes something that starts to identify what we call a digital fingerprint that's unusual to way that [00:10:00] prescriber historically has prescribed. And that's how we start to use, and one of the big pieces we're looking for is something being an outlier.

Something that's different than usual and probably merits a little bit more inspection.

Mike Koelzer: well, I Pride myself on this one, Anthony. And, I've never received an award or anything for it, but I pride myself on being able to tell when people have bad toupees on, and I think like I'm a hundred percent in my mind, but I'm not because the people with good toupees.

I don't know. I might be 50 percent for all I know. And I think we pharmacists, we think that maybe we're catching these things. We can't hold a candle to AI and to the things your company's doing, because there's things that even if we had a board of people looking at these things, they wouldn't come up with the ideas, even that [00:11:00] AI has, going through your system and so on.

Anthony Brooke: Absolutely. Funny that you make the toupee one, considering that I'm bald.

Mike Koelzer: Well,

Anthony Brooke: Luckily, it's not a sensitive piece for me at

all.

Mike Koelzer: Anthony, you don't know if I am or not, because I might be in that just good to pay category.

Anthony Brooke: That's that's right. That's right. So, with that, you are correct. And that's one of the areas that computers in general have been good at for a long time, is they can take in millions of parameters, where Even clever people have a very hard time going there. And that's what we found is that while we had standard fraud programs that helped our law enforcement do investigations and figure out whether or not there were bad actors on the network, what we found was they couldn't hold a candle to what AI could detect.

And by taking in so many more parameters and doing so on the fly and giving us something that in [00:12:00] milliseconds. It would take us hours of pulling data, correlating. We found that it worked for us, and so hence now we're at the beginning of partnering with Pharmacy to say, prove us right, prove us wrong, have a look at this value because maybe we need to tweak our algorithms further, maybe we need to train them further, but we need your feedback because we do think it's pretty consistent, so we've been running this for about a year now.

Mike Koelzer: Anthony, you and I had talked and I had brought up the state programs to track opioid use and so on, and they're probably different names for different states and When you and I first talked, I'm like, well, what do you do more than that? But as I talk now, I'm thinking of being in a bunker and the grenade is sitting next to me.

And that's kind of like my state program it would be a lot better if I was able to have somebody point out the first hundred steps [00:13:00] before that, before the grenade is sitting at the bottom of my. Feet. Ahem.

Anthony Brooke: it is one of the things that we are really trying to do with this is catch as far upstream as possible. Hey, we think there's probably an issue. We can't say for certain. Why don't you have a look, see if it matches your risk profiles and your certified and educated opinion as a trained pharmacist to decide is this right or wrong.

So we look at it as another. measurement, not a replacement to any of the existing programs, even the ones that internal pharmacies have for fraud prevention.

Mike Koelzer: Anthony, when I think of this, I get excited because think of nefarious things. We're going to catch people and things like that. Is that the focus, or is there anything in here that might be Whenever this drug is written , doctors seem to miss that it's actually this problem, kind of like a medical [00:14:00] look back to the beginning, or right now is the focus on the bad guys.

Anthony Brooke: So it's really interesting that you asked that question. I think you intuited a little bit of what we're doing roadmap wise. We started this to catch bad guys. Now we're starting to talk and have early conversations to say, can we use this for better decision support? Can, and whether it's on a single prescription or maybe at a health system level, where we let them understand what the variance is between the various prescribers who have the same job mission.

Where we help maybe at real time, where within our own prescribing solutions. And our platform there of, should we be showing you that this one looks as you've written it before you've signed and send this one looks like it's a little bit of out of parameter of what others or even yourself have done.

[00:15:00] We have not instituted that because we need a lot more validation from clinicians. But that is definitely something that we've made pretty pictures of. And maybe another time I'm happy to show you some of those plottings of those clusters that very clearly describe this is what a digital fingerprint looks like.

And you could name 150 NPIs. You could name a specialty. You could name a U insert, your slice that you would like to see. And we can show you a fingerprint of how they break out. Into the various clusters and the types of things that they typically are doing. Where that'll give you some false positive is maybe they're leading edge and they're doing a new drug therapy that most haven't started that's been approved, but most haven't started using, you'll also catch that.

So it's not just bad actors, it's just simply saying this is outside of the norm.

Mike Koelzer: I'm not smart enough [00:16:00] for this, but I've talked to somebody and they were just saying that you can tell what decade a physician graduated by what they're writing I'm not saying they're all stuck in their decade, but you can tell if they're writing this much of this, it means they probably graduated here and so on.

And that doesn't mean new is always better, but it would be kind of cool to see what the different levels are doing and have some communication there.

Anthony Brooke: It could be. This is where we walk a careful path because we don't want to be influencing what is prescribed. That's the job of the prescriber. But our software needs to 

provide

enough 

information for them. 

Mike Koelzer: Anthony, that's the job of the PBM.

Anthony Brooke: Exactly. 

Mike Koelzer: Alright, let's be honest

here. 

Anthony Brooke: our job is to truly surface up and provide them with the right information to [00:17:00] make an informed decision. And that's a lot of what our previous AIs have done, where we fill in the blanks or normalize information for them to lower that cognitive burden for a prescriber. This one brings in a new dimension to where we think we can say, of what you're suggesting, it's this far away from what you or others like you have been doing.

You can almost think of it as an inverse of when you've been to a good e commerce website. And it says other people like you are also interested in this and you're like, wow, actually I am interested in that. When it works, it's great. When it's poorly implemented, then you're frustrated by it. Right? And that's, we're in that evaluation period from potentially adding it as additional information for prescriber on deciding what's the right drug therapy.

We're still in the evaluation to see is it helpful or is it annoying? But on the fraudster side, that part we've [00:18:00] been able to really be able to detect. This one is definitely an outlier.

Mike Koelzer: Well, of course we have to Focus on the fraud stuff. That's the fun stuff. Now when we're seeing them earlier, there's probably intervention points. Because as I think about it now if I knew that the doctor was I would be more apt to question it too. In other words, It helps me to know when I call a doctor up and say, Hey, this looks a little fishy, it helps me to know that doctor might've seen the low score instead of this being his brother in law or something like that.

And now you've, I'm just making this up, there's some emotion there so that would help. That would help to see, to know that everybody's seeing this number as it goes along.

Anthony Brooke: I think you're right. I think that we're starting with [00:19:00] pharmacists in pharmacies to validate that what we're putting out there for an outlier score that they agree and not agree by the onesie twosies, but here are the last 15 million scripts. Yep. We agree. Because that starts to be pretty good consensus, that this is working not just because we said so, but because pharmacists have validated this is working for them.

Then I think we go upstream, the opposite direction, to that prescriber and let them see that too, and let them react accordingly.

Mike Koelzer: Anthony, what you were saying there, is that in your research for this model, or is that in real time where the pharmacist sees it first, because I thought you were saying maybe even this is going to be shown earlier, or maybe it's just shown to the pharmacist, but. You've looked earlier

Anthony Brooke: Correct. So, to make the score [00:20:00] on a 1 to 10 scale, very sophisticated that way we, of your outlier score, we have looked backwards to what this prescriber has done historically. We also are looking at, for the formation, because remember, it's not about the person who's prescribing, it's around the specific prescription, this one prescribing event.

And so we've looked at all the relationships, the edges. that are involved in this particular prescribing event. And from that, we assess a score once that script is complete. And that's where we're right now putting it. So that means that it's out of the hands of the prescriber

and on its way to the pharmacist.

So right now, only the pharmacist would see that.

Mike Koelzer: Sure. We're looking back at the history and the outlier, but the score is not influencing that can influence us, but we're seeing all this stuff that happens. 

Anthony Brooke: You're seeing the [00:21:00] history without having to have that cognitive load, and it's where, quite frankly, not all have responded well to my analogy on this, so I've refrained a little bit, but I do think of it as like a FICO score,

Just not to a person, but for a single prescription. Hey, do you want to give this loan based on this additional value that's pretty standardized and consistent?

Yes or no, that's up to your institution. What are your rules, parameters? All the things that come to play there, plus their own judgment of whatever it is they're selling in the marketplace. We see it in a very similar way, where for this prescribing event, we're empowering them to have a little bit of a quantified view of how that prescription came to be, and then let them make their own informed determination as to whether or not that's one of the scripts that they want to fill or not.

Mike Koelzer: It's an interesting take. I was talking to one of my guests a year ago, and he wrote a book on the opioid [00:22:00] crisis and some of the chains down in Florida that are making it difficult for the rest of us up here now. We're kind of preaching to the choir in my mind, but, On the other side, though, part of me was saying, well, how the hell do pharmacists know? We're trusting the doctor. We don't really know these people. We don't know exactly if they're in pain or not. And so on. And he was saying, well, that's your job. You've got to know the person enough.

And I think even the DA law says we have to know the doctor and the patient, whatever no means he said, that's your job. You have to know enough about them. And so on. It's like, I see that, but I've got distant family member that has a bad neck. He's had a bad neck from his job for 30 years.

I know what kind of job he does. And I can picture him holding something onto his neck like that. And we fill his pain medicine and I have. No doubt what this is for. If I didn't know him and if I didn't know his history, I might throw him in the same batch as all the other people I think are just drug addicts.

I don't [00:23:00] know! I don't know enough about it, Anthony. I just don't know.

Anthony Brooke: I think you're right that pharmacists have a hard role to play in this, where they're the last gateway before a drug therapy is released to the wild. And so they have quite a bit of work. And we've focused a lot on the patient and particularly with things like opiates where we know there's abuses.

What about non controlled substances? What about cancer drugs? We're finding that fraud isn't just happening with recreational drug use, but rather also it's happening on areas where there's aftermarkets or profit to be made in non illegal ways on not just opiates, not just codeine, promethazine to create purple drink, right?

And so when that comes to play, very often [00:24:00] it's Yes, you could identify it with half the equation around the patient, the person coming to pick up. But what about the prescriber?

We've implicitly trusted that if you had a script pad, and it was the proper print, then it must be legit from a legit document.

And then we found out that script pads could be stolen easily, so we shifted to digital, and that secured us for a while. Until our bad actors started to become more sophisticated, and now They can emulate your digital identity sufficiently enough to get privileges to prescribe. And how do you tell?

And our own investigations seem to point to this is about 1 in 10 problem. That's a pretty large number of providers that have had their identity compromised.

 I just watched a movie last night. Nicholas Cage was in it. It was Dream Something or Other [00:25:00] and I don't want to give it away to everybody but basically it was the new system of fame but also cancel culture and through no fault of his own, he was canceled. And the doctors are the same way. If you start getting that, or you start seeing these Fake ones coming through. We're only human. We heard it was stolen. We know that it was a forgery and that kind of stuff.

Mike Koelzer: But as humans, a few months later, it's like, what the hell happened there? I don't think that doctor was bad, but something was goofy there and here he's a victim or maybe wasn't even his pad. The person got it off the black market or something.

He might not have. Anything to do with it, but that connection in our head is maybe there.

Anthony Brooke: that's right. That's right. And again, our company is not one to decide who's. Good. Who's bad? We really stick in the middle to help facilitate outliers, but we often do work [00:26:00] with authorities to provide investigations. We're seeing an uptick in Retired physicians being one that are particularly targeted because they're not as active and they don't quite know as much and their information has been out there longer and they are of that slightly older generation and they, it's not restricted to them but that just happens to be a class of providers that we've seen used often that a brand new account gets signed up somewhere else and as far as anyone can tell it's a legit doctor.

That's because they have your entire identity enough to get through. All of the current state IDP.

It can even be that they're an active, legitimate prescriber, and they received prescribing through their electronic medical record system, which came from their employer. They just come in and work all day. And their identity is compromised. And someone creates a different account, in a different state, in a different place, and they'll never know.[00:27:00] 

Because when the pharmacist does a request for change, for instance, that message is going to go back to what? The office you prescribed from, right? So it's not going to go to you, the person, it's going to go back to that source said they just don't know until the authorities are like, Hey, why are you prescribing these things?

Let's talk about this. And I'm like, I'm not even treating patients in that state.

Mike Koelzer: When you were talking about maybe that being from a doctor who's older, you're saying that they might be active, but maybe they're just not picking up on it. They're not watching it as closely as somebody else might, or maybe they're a little bit further on in the life of technology and so on.

Anthony Brooke: I'm saying that we have noticed a slightly higher trend towards older doctors, just of accounts because they are more dormant or less But there isn't anything for them to notice. [00:28:00] That's what's scary about this. When your credit card's stolen You have to go and look at your balance or your transactions to see like, I didn't just rent a cab in Tokyo, I'm in New York.

How does that work? But there isn't that for prescriptions. There isn't a lookup for a provider to say, show me the last 90 prescriptions in the whole world that this particular MPI did.

That's not available. And so because we're distributed in health care, they wouldn't know. Whereas, when you're, let's call ourselves an aggregator, Dr.

Furst, where we're across many different health systems, many different medical record systems, we can start to see that cross. We're like, why is there some weird traffic over here that seems to be an outlier to what this prescriber normally does.

It can be an active provider just as much. The fact is that you would [00:29:00] have no idea. If there were a separate physical person who set up a clinic under your credentials in a different state, how would you know?

Mike Koelzer: So Anthony, let me think here. All right. So you've got a doctor who's current and then they might set something up a couple states away. And I'm thinking from a uh, novice on this I'm thinking, well, the database would check that you can't have.

Two different things going on. One, people are probably not looking as closely as they should, but also the hackers can do about whatever the hell they want to do these days.

Anthony Brooke: Enough so that they're able to positively come through with your identity. And as if they're a legit person. Take this person who was a neurosurgeon. And set them up in a dentistry practice and start prescribing. Even though those two things typically are unrelated, may even be geographically separated.

And [00:30:00] those are the types when I spoke to you earlier about dimensions or edges, really the relationships between things that our AI models are looking at. To try and say, this is funny, something's off. And there has to be enough of those off, which is how you and I try and decide whether something's right or not.

Is we take a one measurement, two measurement, three measurement, and we keep saying that's one strike against you, two strikes against you, three, and that's why We happen to use a geeky thing called ECOD, and it's, the C is cumulative. It's like, there's enough things off with this that, I don't know, let's call it an outlier.

I can't say it's bad. It could be I just set up a new practice. That too would trigger a lot. But I can say that this is not like the others. Within that digital fingerprint, or the standard behavior in which This provider and other providers like them typically prescribed.

 When you're looking at our [00:31:00] prescription level scoring for Outlier, it's transactional where each time we're recalculating, looking at the past and what we understand. Then part of the behind the curtain is that we do create an additional score for all the stakeholders that we've seen before.

Have I seen this patient? Have I seen this provider? Have I seen this pharmacy location? Right? And then how do those things, this payer, this particular drug therapy, Each one of those, we have our own markers that we're like, at right now, we're at this comfort level for each one of those stakeholders,

and then those typically are only updated on about a weekly basis, whereas the actual prescribing event is done in real time, like you said, in like chat, GBT, it's calculated on the fly, and then it's sort of forgotten.

Mike Koelzer: Anthony, My dad years ago When you knew you had a bad player in this [00:32:00] maybe the patient they stole the scripts or something and my dad one time kind of set up a sting You know where he was gonna have the police out there waiting for this guy.

And so and this is 30 years ago And he said, Mike, that's the last time I did that. He said, went down there. I sat for, a whole day in court, the person didn't show up and the judge did this. He said, they just didn't do anything about it. And I think that's where pharmacists are a lot. Now too, it's hard to follow through. And so you kind of push these people down the road and hopefully everybody pushes them out of the way. With your system here, I imagine you have the same because you don't have a secret button that's going to open up the trap door and put them in, the paddy wagon or anything.

Anthony Brooke: That's right. We're in no way some kind of centralized authority. We simply noticed that we had enough view of unique data that wasn't available to pharmacies. [00:33:00] That we would be able to tell, particularly at a prescription level. This one's different than others. Like it. I still can't confirm whether that's good or bad, because it's human nature to sometimes handle things.

Even if you're just doing something most providers, they'll do friends and family. So they'll be at the Thanksgiving dinner table, someone's got a problem, and here, let me write you a Z Pack. That's a very normal thing to do. Even if that is a dentist who never writes a ZPAC because they wouldn't use that for any of the procedural things they do, right?

And so, with that could be flagged as well, and that's okay. 

We think that's perfectly acceptable. What we want to do is just give that indication, give that additional her script, her prescribing event metric that a pharmacist can help [00:34:00] identify and say, do I want to actually fill this?

Do I want to give it to the person who comes to pick it up?

Mike Koelzer: It's more information. It doesn't automate putting someone in the slammer, but I guess we don't want that. I guess we want, innocent till proven guilty and all that kind of stuff. And I'm not sure how well a machine would do that.

Anthony Brooke: Right. And this is where some long time ago I worked in finance for one of the large trading houses and we were charged to find insider trading. That's what we did. Right. And we tried to find it before the authorities did, and police it ourselves. And exactly that. We even used the first generation of AI within expert systems, neural networks to find anomalies.

And then from that, we put it into a queue for an expert person to review.

I can foresee a world where we do that on behalf of a pharmacy chain, where they say, you know what? When it's these conditions, please set it aside and I'll [00:35:00] have a team go review it before it touches the local facility pharmacy.

Okay, I could see that happening. It's not today. We first need to get through that validation part with our pharmacy partners to have them prove that this is working for them and that the, essentially the math is right. And that the AI is picking up insightful things for the downstream pharmacy.

We're pretty comfortable because we've been able to accurately identify bad actors that then FBI, DEA, Sheriff's Department come back and say, Hey, we're running an investigation on this. Can you give us data? And sure enough, our system's already been like, Yep, that looked kind of weird. 

It looked like an outlier.

And we've been doing that for a year. So we're pretty comfortable But now it's time to really partner with Pharmacy and have them come back and respond and say yes. We agree with you. And we agree with you by the tens of [00:36:00] millions, not a hundred.

Mike Koelzer: was commenting on a human judging like there was value in that I just read an article though a couple weeks ago It was saying that if you go to court for any reason like a traffic violation Go like A half hour after lunch and you get a lot better treatment from the judge than like at 11 45 or at like 4 it was remarkable how, here it is justice.

You think it's kind of black and white and they said, if a judge is hungry, look out, be careful. Anthony, here's something I hate now. I consider myself pretty good computer wise, technology wise, and this last double lock verification threw me off when they started talking about a passkey.

Because I'm like, what the hell is that? Physical? Is it not physical? Is it a thumbprint? What the hell is this? It was throwing me off. As I think about some of this stuff though [00:37:00] there's a rating of. The patients and so on. But as you bring up even the duplicate parallel universe of a good doctor and a fraudulent doctor and so on, are there any newfangled ideas about verification and so on.

Passwords and all that kind of stuff that would say, Hey, this person set this up, but it's not the real person because they didn't get the text code or something for this.

Anthony Brooke: Right, right. So, I think what you're asking around there is around second factor. And that's one of the core concepts between electronic controlled substance prescribing. And what we have for that class of drug therapies that fall under controlled substance. And we have a forced second factor exactly that.

The little code that's generated, it generally uses some type of either hard or soft token. They [00:38:00] say it generates a code, the code gets put in. Most recently our firm has been working on creating what we create our own authenticator, simply because we wanted as ease of a process as possible. No entering a code, just click yes, that's me on your phone.

Right. 

Mike Koelzer: mean doctors don't like tricky technology?

Anthony Brooke: I would say, not just doctors, there's not a user out there that really likes

those. 

Mike Koelzer: at me. Yeah.

Anthony Brooke: But we've been thinking through and actually have in some of our retail software that it's not just second factor, it can be third and fourth and fifth and sixth factors, such as this let's take this particular behavior of prescribing that we call a digital fingerprint scan. We could say in our own software, and we have for some opiate type prescriptions in our own retail software, to say look this is one that is a little bit tricky. We're gonna force you to do something a little bit extra. Not because the regulations say so, but [00:39:00] because we think it's the right thing to do.

Or because your outlier score is a little too high. We want an extra validation. Even if that validation's not something a user has to do as an action. Or rather, we do that real time lookup to see whether or not you're still active at the DEA. We do that real time lookup to see that your device is one that we've seen before.

That helps us understand whether or not you are going through this electronic experience with good intentions or bad. And try and nip it off a little bit earlier. So we have a limited degree of that in our direct to consumer. But in our greater platform We don't do that because it's up to our enterprise customers to make those types of decisions.

That led us to thinking through this specific solution of a prescription monitoring system to Rx, where there's a standardized value that they can make their own [00:40:00] decisions. 

And, 

should that be validated, I could see that an enterprise would tell us if it's above a certain score, Just stop it. Don't even let it go forward.

And we could program that in, but right now we don't have that programmed in. And I do think there's an opportunity, and probably one that'll end up eventually getting mandated, that we'll probably do EBCS style Second factor, bound tokens, some of those annoyances. For all prescriptions.

because right now the controlled substances are held to a slightly higher standard.

but that also means a higher hassle factor for our prescribers. And that's where our software, we're constantly trying to work out ways to save you just that one click,

much less Five. Because it means something when you're seeing 35, 40 patients a day to save you that one click each time.

Mike Koelzer: I think I read it about Tesla or something or I forget where but they're saying that like One of these [00:41:00] cars doesn't have a reverse Button or something they know they can tell if you've got something eight inches in front of you and nothing behind you and it can tell that your right shoulder lifted up the seat Because you're looking behind you or something like that It's one less button to press and might not sound like a lot, but one becomes two becomes three becomes four and pretty soon you're doing things that are more human nature than having the machines get in the way.

Anthony Brooke: And I think that's an appropriate use, and we'll continue to see. There's lots of conversations about AI and all the destructive bad things that can be done for them. I think in a practical standpoint, and something that we've seen, because we've been actively using AI for about 8 years. at scale is we take steps away for providers by applying AI and workflow.

Do things like fill in the blank when there is one, because it could go do a lookup for you in the background, 

right? Clean up and [00:42:00] normalize information. These are problems that we've taken on for years. But I think as we go forward and as there's been additional excitement with the introduction of generative AI for about a year and a half or it's been really mainstream thanks to OpenAI.

That has brought a lot of workers to understand that it can help us do our daily tasks with that one less click,

With that one less interaction. And that's why, 

Early on when I'd be asked about generative AI I use a, for most people, it's a. new way of navigating digital information and software.

It's actually a new navigation or UX paradigm to where instead of doing a search on a body of information, you're asking a questions and having a conversation, you're not just looking for keywords and phrases, you're looking for concepts

 [00:43:00] and humans are very complicated. And so often we have ambiguous intent of what we want and what we're seeking and how we phrased that.

And that's a little bit of what we've been able to capitalize on our outlier detection there. Of 

Mike Koelzer: being able to say, Huh, the intent doesn't seem to be the same.

I was looking online of your guy's stuff And I know that part of that even plays out, of course, with human approval, but part of that plays out. I think he was even talking about Dr. Sigs, , The questions of

what did they mean by this? And. What are they getting at here? And maybe putting it into plain English again with the human intervention. We got to say that, but you know, that kind of stuff to clear up the whole damn thing.

Anthony Brooke: So that happens to be one that we probably do more than anything else, is normalize or fill in the blanks on SIG, the [00:44:00] instructions about a specific drug therapy. Take one daily with meal, right? 

Or with breakfast. There's over 800 ways that we've seen that stated, that same simple phrase of Take one daily.

And so whether we're normalizing the way it's written, something as simple as that, or whether we're actually filling in the blank because it didn't make it across because you went from one record system to another record system, and when the data got imported, they slightly viewed the data differently. For provider, that just means that they spend time filling in empty fields because that's what the record system has in front of them instead of practicing medicine. And so that's one that we often and most often use within our clinical AI solutions is that we're not making determinations of what the drug therapy should be.

We're trying to clean up the data. To be able to optimize it such that you can quickly [00:45:00] make your own professional decision accurately.

Mike Koelzer: Anthony, you would have kept my late father happy because he would tell me, he said, Mike, when you're writing a prescription out and the doctor says, take three times a day until gone, he said, don't put gone down because people are going to think they have to take it till they die. And he said, put until all taken.

And I'm thinking, nah, I'm like, Dan, what the hell? I'm thinking now, like, you think people, if I put till gone, they're actually going to think they have to take something until they die. I think it was just a way to like, try to share his wisdom with me, but he'd be happy about that because the AI would not let till gone go through.

It would massage it for regular humans to interact with it.

Anthony Brooke: Correct. And that's where, when we work with like our pharmacy specific products that pharmacy can tell us to say, I want gone, or I don't want gone. I want it codified this [00:46:00] way. Daily should be day. Daily, like, abbreviated. Sure, you tell us that, and that's what the AI will flip it to when it gets to your destination.

If it's upstream a health system, we also have the ability to help them choose that through. To say, how do we want to present our SIGs at a health system level? So we can do that even before the script's created.

Mike Koelzer: Yeah. And people think they have the answer. It's like, well, our system does that because we use this sig code. It's like, yeah. What if someone uses a different code or what if in the heat of the moment, you're not ready for it? I mean, we don't have those in our pharmacy, but I know that some people have certain codes again if you've got to do almost more than the computer's doing to get something to happen, it's not there yet.

It's gotta not be there. Period.

Anthony Brooke: Agreed. And so that's where we allow our customers to have choice, but we think that the [00:47:00] appropriate application of AI is put it in workflow, not a separate step that you have to go to the AI and do something, but just put it in the natural workflow of getting your job done. And do simple little thing 

that, that help. Lots of folks right now are working on ambient listening. And I think that's an area in healthcare that is drawing some interest and excitement. Some because we can relate to listening to a conversation and taking actions ourselves. So we can anthropomorphize it, we can make it human. But some it's because it actually really can pick out.

We've got one that it can listen to your conversation, if you choose, with a patient, and it will pick out all the drug therapies and queue up the scripts that were discussed that aren't currently on the past medical history for that provider as active. Not because you will prescribe that, but rather because, hey, these are the ones that you might need to do and we could save you a couple keystrokes.

By saying, yep, [00:48:00] take that one and now let me finish editing it the way I want that drug therapy and send it. Just cause it's a little bit faster. And those are the things that I think we'll see happen more and more all throughout healthcare. Not just for doctors and prescribers, but rather all through healthcare.

 I need agents, as an engineer might say, that help with this little thing, 

and it doesn't in any way replace the human. It augments the grind task so that you don't have to have that.

Mike Koelzer: I've talked to a lot of people about block change just out of interest and everybody I've talked to, and I'm sure you'd agree with this. It's like, those tools are kind of cool. And when they come out the. Techies like to look at 'em and things like that. For something to truly be an impact, it happens without really anybody knowing it.

And that's the same thing with ai. It's like early adopters think it's cool to do this or that, but when it comes down to it, it's [00:49:00] gonna help my mother-in-law she doesn't know what the hell the AI is, but it's gonna help her do something. And that's where the beauty of it comes in.

Anthony Brooke: That's right. That's right. And we've gone through multiple waves of this, and fortunately or unfortunately, I'm old enough to remember them. That is true. When the first wave of AI came through, consumers tended to not know. But yes, Ben Jerry's was mixed by AI. 

That was one of my earliest career software pieces, right?

Nabisco cookies were cooked by AI and baked by that. And that was amazing to see that shift. 

Then it kind of went to a little bit more of the forefront with boom we would start using these search engines, and some of them worked better than others. And the reason they worked better is 'cause they started to employ ai.

Not that you knew it, not that you had any concept, you just knew it would pull up those search results. You're like, thanks, ask chiefs

Thanks. 

Google. And that was where we [00:50:00] started using AI at mass scale.

Anthony Brooke: Our noise canceling that we use on headphones these days, all of that, they're all influenced by this impact.

And I think. The new wave that's creating the excitement of generative AI is because it truly emulates what we think we do as our own task. And so we're fascinated by that because it is a little bit more human. And I think that also creates some fear. And that's where in healthcare, I do think it's the right decision to let AI type technologies filter, sort, present information, and then let it Good.

Mike Koelzer: My wife I said, Margaret, I can't wait to get, the AI car my daughter lives 12 hours away from us. At nine o'clock, I can decide what movie I want to put in the popcorn, get in my driving pod. Fall asleep [00:51:00] at midnight and wake up at, nine in the morning in her driveway.

And my wife said, I don't trust that. And I'm like, are you kidding? Have you seen me or talk to lately? You think me having, 30 meters on my car that at the speed of light are making decisions based on. 8 billion people's decisions. You think that's better than this old fart?

I'm like, no way. So I'm ready. I'm ready for it, Anthony.

Anthony Brooke: I do think that there are guardrails that still need to be put in place because we aren't fully ready. And how do you deal with things that some of the cars are analog 

and some of the cars are 

not? And that's where I think it becomes difficult cause it creates enough variability that even well created AI still struggles 

to

anticipate that much chaos. 

Mike Koelzer: that's gonna be interesting where you've got a, 12 year old right now, whose dad is. Talking to him [00:52:00] about, stock cars and all that kind of stuff. This kid's going to be 92 some year that's 80 years of having a kid at heart wanting to drive his car.

And so that's going to be a real interesting crossover.

Anthony Brooke: It is I think it also gets to Social things that we tend not to think about. Why are most cars capable of hitting 100 miles an hour, but there's no place in our country where you can go legally 100 miles an hour? And so then if the machine is regulating it, Well, can you really still speed? And that's where we start to say, huh, I don't know if I like the 

assist because it's imposing on freedom.

Mike Koelzer: yeah, it's going to be interesting. Golly, Anthony, thank you. That was really fascinating. the interplay of true AI is really cool how it's affecting our daily life now and [00:53:00] not being, out there it's here. So thanks for your time and sharing that with us. That was really fascinating. 

Anthony Brooke: Absolutely, Mike. I'm pretty excited about what we're doing with our True RX program and the application of AI to solve what I believe is kind of an important problem that no one quite has a handle on.

 Really cool stuff. thanks again, Anthony.

Anthony Brooke: Thank you.

You've been listening to the Business of Pharmacy podcast with me, your host, Mike Kelser. Please subscribe for all future episodes.