Episode 46: Inside Informatics with John D’Amore

In Episode 46 of the Informonster Podcast, Charlie Harp sits down with John D’Amore a leading voice in healthcare informatics, data standards, and interoperability.

From his early days in decision support to building and selling a data normalization company, John shares the experiences that shaped his career and the lessons he’s learned along the way. Together, Charlie and John dig into the evolution of interoperability, why data quality matters more than ever, and how standards like IPS and CCDA continue to influence care delivery across the globe.

They also explore the intersection of AI and data quality, the realities of modern EHRs, and why the next generation of informatics professionals will shape the future of healthcare. John also weighs in on the growing momentum behind the Patient Information Quality Improvement (PIQI) Framework and why establishing a shared, consistent approach to measuring data quality is critical for the industry. 

Contact Clinical Architecture

• Tweet us at @ClinicalArch
• Follow us on LinkedIn and Facebook
• Email us at informonster@clinicalarchitecture.com

Thanks for listening!

Transcript

00:02.20

Charlie Harp

Hi, I'm Charlie Harp, and this is the Informonster Podcast. Now, today on the Informonster Podcast, I'm joined by the one and only John D'Amore. John, thank you for joining the Informonster Podcast today. And usually the way I start this whole thing is for the listeners out there, in case I happen to have a couple of stragglers that have never heard or or know you, do you mind giving us a little bit of your background and your journey into healthcare.

00:33.89

John D'Amore

Well, thanks so much, Charlie, for having me here today. I'm really excited to be here. And, you know, I mentioned beforehand, I don't listen to a lot of podcasts, but I had a chance to catch up on the last six or seven from Informonster. And I'm just really honored to be here. This is a great audience, great group that you brought together, and I think incredible content. And I'll start with just my background, how I got into this whole informatics domain, which leads to a lot of Informonster type problems. I you know always want to be in healthcare. My entire career was focused around being able to get in healthcare and solve problems. I thought I was going a doctor. That's what I really was pre-med and thought I was going to go off to, but I got some really good advice early on from some actual doctors. They said, why don't you just see if you can do something in healthcare care and find it fulfilling? Because if you can do anything but be a doctor, that'd be a lot easier.

01:30.08

John D'Amore

And I listened to that advice, fortunately, and i went ahead and got into healthcare consulting, a little bit into healthcare investing through venture capital, and found that really working with data and and figuring out how to be able to use information to improve care was really the thing that I was most excited about. And it was something where there was a lot of need. In the 2000s, as I was getting going, we were transitioning from things that were largely paper or maybe a hybrid of paper with some billing systems or maybe some ancillary systems into what we'll call now fully fledged electronic health records or EHRs. I really found my calling for informatics when I was at Memorial Hermann. It was the largest healthcare care system in Texas at the time. And we had 12 hospitals. They got more hospitals now. They had lot of ambulatory facilities. And I ended up bouncing around a little bit, but ran their system wide decision support program, which wasn't clinical decision support. I'm not a doctor. I always get a preface that I play one on TV sometimes, not a doctor, not a clinician for things.

02:44.03

John D'Amore

But I ended up running that department and we did a lot of reporting for both the CFOs, but also for the CMOs, the chief medical officers and the CEOs as well to come up with clinical and operational data. And one of the projects that we ended up running was around being able to reduce the time between when results were shared for microbiology and when the antibiotic selection was changed. And we called it the drug bug matching project because we had to basically take back these microbiology results and say, all right, what antibiotics are going to work for what's been cultured on this patient and then what actually antibiotics are being administered today, which often isn't an exact match. Sometimes you have some patients that are on some things that kind of work. Sometimes they're just fine. Most of the time it's really just fine. And, but then a small fraction, they really need a change because they're being just administered antibiotics that that bacteria or that pathogen is be resistant to. And the consequence of this, that we ended up implementing was it used to sit on the fax machine of hospital pharmacists. And they had to, before the end of their shift, review all these results. And it would could take two hours. It could take four, five, four or five six hours if things were really busy in the hospital pharmacy. And we turned that into an automated process, which they just reviewed the results and released them essentially immediately once they released from the labs.

And patients got this treatment sooner. And the consequence of getting your and the right antibiotic sooner is that there were fewer septic sepsis cases. The patients got better sooner. And lower length of stay pharmacists had to do less work.

04:21.37

John D'Amore

Everybody was happy. It was just a win-win all around. It was just through the intelligent use of information. That was my calling for informatics. I was like, we have got to be able to use our information most effectively. And I took the app that we developed and then, this is really before we even talk about apps like on your phone, but it was an app, an analytic rule set. How do you want to talk about it? I took it to our EHR vendor at the time. And I won't mention who they are. i don't want to anyone under the bus. Um, but they basically said, John, this is incredible what you and the team have done. and I want to give credit. It was mostly a team. I, I was fortunate to be overseeing it. Um, it's incredible what you've done, but we can't take that app, that analytic anywhere else because the data models for how the microbiology results are stored and how the medications are stored is so different.

05:08.49

John D'Amore

And I said, this can't be like, you can't ask 6,000 hospitals all to be inventing the best way to do this. We have to find ways to share best practices. And then I found out that really what we needed to work on was the data that was underlying that. So I went and got my master's in informatics. This was back in 2008, 2009 time for me getting started. And I was at the University of Texas School of Biomedical Informatics, which is now the McWilliams School of Biomedical Informatics, one of the first schools for informatics in the nation. And I ended up studying interoperability standards and how people were sending out this data because I saw there was a really big need to be able to improve data and data quality, the data syntax, as well as the semantics for that information. And that led into what I did afterwards, some which I ended up working at an EHR company. I worked at Allscripts for a couple years, and then I ended up founding my own software company to do data normalization and a lot of magic in between. It was called Diameter Health, and that was a good ride. And that we we sold that in 2022. I've been consulting, I do teaching. I teach at Boston University now for informatics.

06:19.79

Charlie Harp

That's an awesome story.

06:21.02

John D'Amore

Yeah, I know. it's you’ve said this on the prior podcast. People are in health care for the right reasons. and Not everyone, but the vast majority. People are here because we want to improve care. We know our healthcare system really wasn't engineered like a system. It wasn't designed to work together the way we need it to, to get patients the best care possible. People are here for the right reasons, and we're all trying to do that in our own way that we can help push that boulder up the hill.

06:49.00

Charlie Harp

What I find is that people that use the term to describe some part of themselves as health informatics are lifers. I mean, if somebody, I'm a developer and if a developer can easily go from one market to another market, one industry to another industry, um I think if you're a salesperson, you know, you can spend time in healthcare. But if that those that phrase, healthcare care informatics, is somehow woven into your DNA, you're youre you're likely not to go and and work for, I don't know, Snickers or something like that.

07:27.01

John D'Amore

I could never be a tire salesman. It just wouldn't work. Not that I... couldn't figure it out. It's just that I wouldn't be passionate about it.

07:35.60

Charlie Harp

There's a lot of cool data in tires, though, John.

07:35.57

John D'Amore

Yeah, I mean add I mean, you know I worked in consulting for a little bit and people always want to ploy on these projects that nothing to do with healthcare. care And i I just said, no, but it's not because they're not fascinating for things. you know I'll give a side story for it. When I was at the University of Texas School bio Medical Informatics, Ted Shortliff, um president of AMIA, well known from his work with artificial intelligence way back in the day with mice and antibiotics. He was still lecturing there. He wasn't teaching a full class for it, but I get to sit down a lecture for him. And in some of the work that he did on uncertainty and certainty factors and reasoning, they were able to use that in like, how does Folgers make the right coffee mixture all the time for it And frankly, you can make a lot more money doing that type of stuff um because like other industries need a lot of these things and healthcare is got the most, I'd say perverse incentive streams for how we pay for things and how we do things. The most well-intentioned people often enough. But the system, again, it's not a system. We call it a health care system, but it's really just a whole bunch of parts that we put together. And it works, kind of, most of the time, but it could be a lot better.

08:44.24

Charlie Harp

Yeah, absolutely. I think that it's got the it's got the issue that it's complex, it's personal, and it's critical.

08:51.57

John D'Amore

Mm-hmm.

08:51.63

Charlie Harp

And I think when you look at those factors, it's it's hard to disrupt. It's hard to figure out how to incentivize change other than passing laws, because that's an easy way to incentivize change.

09:02.77

John D'Amore

Mm-hmm. That's the easy, ah but we're not going to get lot more of those in the next couple of years, probably. like So we got to find other ways to do it. You know, one other thing that you talked about with informatics and lifers, which I think is so true is, um, part of what I actually lectured in just this last lecture in informatics, um, at Boston university. You know we really get into the last lectures talking about what is this domain of informatics. And and I think one of the key things that draws people to it is tremendously interdisciplinary too, which is you can't just wear one hat. You can be an engineer and come into it, but you can't just say, well, I don't need to know anything about the use case here for how the heart works or how a sepsis works. you know You have to understand some of the clinical nature. And same same way, if you're a clinician, you can be the best doctor and the best nurse in the world may have horrible when you get into informatics because You got deal with data and you got to deal with data structures and you got to think about things like standards and data quality and terminology for this information. and And I think the thing, the kind of the secret sauce to that is you also have to understand the system, which is unfortunately our system isn't just about doing the right thing for the patient. There has to be the right incentive structure for things to get adopted. and for things to be permanent. Now, that can often align with doing the right thing for the patient, but you just have to realize that doing the right thing for the patient doesn't necessarily mean that it's going to be easy to do. You to kind of find a way to move the needle for incentives, for quality measures, for population health, for risk adjustment, all those things that kind of need to move in the right direction.

10:31.36

Charlie Harp

And also kind of push against the the mentality that you're not just trying to check the box. It's the spirit of the law, not necessarily the letter of the law. But yeah, I think that when you when you talk about the interdisciplinary nature of health informatics, it's kind of a mixture of, you know, smart, curious, passionate, because as an engineer, I've worked in other industries. I've worked in manufacturing. I've worked in finance. But when I got into health care, don't know, was like a switch was flipped in me. And when you get into, you know, how does a ah micro turbidity scanner work? What does it mean when it's looking for growth or no growth? And when it's analyzing sensitivity versus susceptibility, yeah and it's, it's what I always tell the developers at work for me is it's not just about building something based upon a set of requirements.

11:17.15

John D'Amore

You've got to be curious. You've got to be curious.

11:25.51

Charlie Harp

You gotta kind of rock it.

11:27.07

John D'Amore

Yep.

11:27.14

Charlie Harp

You gotta understand. And you almost, like I was at First Data Bank for 10 years and people say, I'll talk to people and say, are you a pharmacist? I'm like, no, but like you said, I play one on TV. I know enough about pharmacy to be to be to be interesting and to be dangerous and to know that albumin and bilirubin aren't two guys I went to college with.

11:37.90

John D'Amore

Yep. That's a good one.

11:51.23

Charlie Harp

So what are you working on today? What are you excited about? I know we're going to talk about data quality. And I know that you're teaching and you see, I see your posts on LinkedIn and you seem to really love doing that. And that's awesome to see.

12:02.85

John D'Amore

right. I do, I do because it's all about, you know, I think about how healthcare is not going to be solved by you and me, Charlie, unfortunately, we're going to leave a lot of problems to that information.

12:11.56

Charlie Harp

but come on.

12:14.26

Charlie Harp

It's not too late for us, John.

12:15.63

John D'Amore

Well, no, we don't know what to do. I'm not saying we aren't going to lot great stuff for things, but I think the next generation is we're going to need more informatics professionals in the future.

12:18.42

Charlie Harp

Okay. oh yeah All right.

12:26.34

John D'Amore

So teaching is an inspiration for that. But, you know, some of the things I've worked on the last, you know, three, four years, um you know, since the acquisition of my software company is getting back to the standards. It's really funny. When I first found my way, when I was doing research back in grad school to the HL7 community and the standard stuff for it, I said, oh, dear God, i do not want to write these things. I do not want to do it. I just, I would literally say in meetings that I don't want to write your standards. I just want to use them. Because at the end of the day, and coming up with the right data model or a slightly different data model, if we all just do it the same way we'll be 10 times better than we are today even if we don't agree entirely that that was the right way to do it and i just wanted to to implement data standards is what i really said and to have questions answered like well how do you represent this data in that CCDA standard the HL7 V2 standard or now that FHIR standard for things and in the last three four years I have actually now become the thing I didn't think I wanted to be which is, I started writing a few of them. i you know I got pulled in along the way to become an editor um onto the CCDA standard while was still out when I was at the software company. So that's a that's a big standard. It's how we share clinical documents. It was forced on us by meaningful use and by EHR adoption, but it's got an incredible traction. I've talked to people and so sometimes they don't know just how much CCDA is exchanged in the United States. Some of the recent estimates are nine billion a year.

13:58.55

John D'Amore

You have to kind of put your head around that, nine billion a year. Well, nine billion what? Nine billion medical summaries, which could mean, know, could be hundreds of billions of lab results and allergies and medications that are associated with that. And you'd say, well, 9 billion, how much is that? but Well, there's only about a billion care interactions in the ambulatory setting per year. There's like 40, 50 million hospitalizations, 100, 200 million ER visits. We're exchanging CCDA multiple times for every actual care encounter we're doing, which is just kind of like blows your mind when you actually think about that. So like interoperability today is not the same problem that we have when I got started, which is just that the data was stuck and it did not flow.

And like it flew, but maybe on all on ah on a fax machine, right?

14:44.30

Charlie Harp

Yeah, maybe my RS-232 connector is but

14:48.39

John D'Amore

We have broken down the data flow problem, but now we've got an ocean of data and we got some variable quality. Some of this is potable water. We can drink it, we can use it. And a lot of this stuff is just questionable in terms of, is it complete? you We get to ask questions about that dimension. You could ask, you know is it actually representing what was there? Is it plausible? offer they say so Things that you really picked up, I think, in the PIQI Framework is just incredibly exciting. And I've been all obviously a big fan of following that for the past couple of years. So I get to writing the standards. Because you asked about what I was doing recently. I'm writing some standards now. So i be I've continued to do work on the CCDA to FHIR project from HL7, so how to do transformations between CCDA and FHIR. I was an editor. I helped that one get published. And then most recently, we have the International Patient Summary. Which a lot of people say, oh, International Patient Summit, isn't that just for international travel? for And I say, no, it's really actually much more about to be the successor to what we used to call the CCD or the continuity care document. It is one of those CCDAs that we exchange all the time in the U.S. It really is meant to be the FHIR-based format for how we will share and data as patients move from one provider to another provider. But they don't have a real-time FHIR API connection where they can just query for all this data magically we have to send a package of data it's a bundle in FHIR people kind of accustomed to that um its first thing is a composition so it's a type of document um and FHIR and you put together the same sections you put together in a ccd which is you put together here's your allergies and here's your meds and here's your procedures and problems and immunizations lab results vital signs etc But now the nice thing is that the CCD, many people might not know this, was a US standard.

16:31.00

John D'Amore

And because now we're starting with the IPS, the International Patient Summary, that's an international standard. So now we can actually get on board and have the rest of the world be talking the same language with us, which is which is a nice, it's been a really fun thing to work on in the past four years. And I've been very proud to be part of that project. And again, I've not done the, I'm an editor, but I'm not, I haven't done the happy lifting. There's been people who've working on this for a decade to get this off the ground.

16:55.00

Charlie Harp

I think having that least common denominator set of information that we can exchange across borders is really important for a lot of reasons. When you look at the the IPS community, which which countries are like super active um in that community?

17:14.65

John D'Amore

Well, the interesting thing is because we have so much CCDA in the United States, the United States is not leading. We're going to be following. There's a project that actually just got announced in September that the U.S. will be doing its own version of this. But we're following in this respect because we actually are exchanging CCDAs. And lot of other nations never did that. um So in Canada, they still got PDFs and faxes flying around and that's really suboptimal for things. So Canada is was one of the first nations to really pick up and move this forward about four or five years ago. They really started the process of using this for domestic exchange. And now several provinces have rolled that out as kind of their standard. They don't call it the IPS because they're using it domestically. It's called the PSCA, our Patient Summary Canadian Edition. But it follows the format of the IPS and can be used internationally as well. The IPS use case and use cases are important because it keeps you scoped. it keeps you focused on what you're trying to do. Our use case is for unscheduled cross-border care. Now cross-border, clearly nations are a border, but cross-border isn't just national borders.

18:17.83

John D'Amore

It can be regional, provincial, state borders. It can be organizational borders. It can be just, you're going down the street to another hospital and they don't use the same EHR instance. That's a border. Graham Greve, kind of godfather for FHIR, loves to say your front door is a border. It is, you know, when you leave the house for it. So anytime you have to communicate in information across border, um you know, for unscheduled care and to have that kind of baseline of information, as you said, the IPS is a is a good fit. And and we're seeing lots of nations. So other nations that are doing besides Canada, there have been several places in Europe that have been picked up. Brazil has had a pretty active usage for it. Australia and New Zealand um are going along this path ah pretty well. um And I'm sure I'm missing some nations in there as well, too. Actually, one of the ones that is a cross-national project is the Hajj pilgrimage. Actually, for the last two years, when pilgrims would come, they actually have to get preclearance they have to collect data. So Malaysia, Oman, Saudi Arabia, and Indonesia are all using the IPS to provide that baseline of information. So if you're on pilgrimage to Mecca and you get ill, the first thing that they can look at is your IPS from your care providers. So they have a little bit of baseline for where you are.

19:36.38

Charlie Harp

That's very cool. And I'm not trying to turn this into gotcha journalism with some of these questions. So I'm going to put you on the spot.

19:44.68

John D'Amore

Oh, go for it.

19:45.94

Charlie Harp

I'm going to go into the the nerd weeds a little bit. So if I'm creating an IPS and I'm exchanging data from one region to another, let's say in Europe, for my time back in my some of my previous lives, one of the things that I was acutely aware of is for example, different countries have different national drug terminologies, variants like DM&D and ARG storm in the us is there when when you're looking at producing IPS, are there semantic or terminology binding norms, or do you really get what you get and you make the best of it based upon the the country of origin?

20:24.87

John D'Amore

Well, I'm going to give a, for that's a great question. I'm going to give a multi-part answer to this.

20:28.94

Charlie Harp

sure

20:29.15

John D'Amore

You know, the first thing is that we have textual fallbacks and that, you know, a little bit in FHIR, we actually did a lot of work on text and text representation in IPS because a lot of FHIR is like, oh, mean, now you get the code, you don't need to worry about the text. You know, well, we're moving beyond text for it. And then this cross-border scenario, you actually kind of need that as a baseline because there can be some times where you don't understand, you if you're coming from the US, you might be sending an RX norm code or a medication. I'm not going to mean a hoot to people in Netherlands they're trying to carry you for things. So the first thing you'd say is you want to have a textual representation for it. And that also helps when you might be doing translation. Remember that you're going to get an IPS produced in the U.S. It'd be in English, but it's not going be in Dutch when you land in the Netherlands for things. So text fallback is the first baseline for this. But we actually have been working because one there are five standards development organizations that work. through the Joint Initiatives Council on IPS. They're part of the what's called IPS Coordinating Committee, IPSCC. And one of them is SNOMED. So SNOMED has made available a subset of their ontology, SNOMED CT, available worldwide for free because not every nation licenses SNOMED yet. In the U.S., we license it, so we're national affiliate. A lot of other nations do it, but not everywhere. And the nice thing is that, you know, while you don't have the millions of terms you' have available in SNOMED, you've got tens of thousands, But those would try to pick common ones that end up in the SNOMED IPS terminology, and they're freely available worldwide so people can use them. So you can refer to your medications. So how did this actually work on the wire, Charlie, which is probably where you want to get? You send that RXNorm code because that's what makes sense locally. But if you can translate that to a SNOMED code or to a WHOATC code for it, those are freely available worldwide medications terminology. SNOMED is not ah quite as good as medications as it is a lot of other places for it. If you can send those globally available terms, um please do that as well, because that makes it more likely that recipients are going to be able to understand that information on the other side of the ocean.

22:32.12

Charlie Harp

But to your point, the nice thing about it is even if you don't have the means to ingest that data and process the codes natively, you can still render that so a human being can see that you're on these maintenance and you're doing this and you're doing this. So even as just a representation, i think that that that's hugely valuable for that kind of thing. It's very cool work that you guys do on the IPS.

22:55.91

John D'Amore

Well, thanks so much. It's been, and again, it's been a real pleasure and it's, you know, it's been fun because a lot of my software work was all based on the U.S. U.S. health information exchanges, U.S. payers for it. And we're weird in the U.S. You get this refreshing perspective when you go where deal with health care overseas because it's a lot different for how they how they approach things.

23:13.82

Charlie Harp

Yeah, when I was... spent time in China and in Australia and the UK. It is very different because the way we look at the data we produce in the US is it's very, I mean, it's it's kind of native to how we do health care in the US. It's very billing oriented. And I always I often think that the next generation you know, EMR, the next generation truly clinical EMR is is unlikely to to evolve natively in the U.S. And i don't I don't love saying that. But I think that as long as as long as we're treating EMRs like cash registers, is that a terrible thing to say? Then they're not gonna be clinical first. And I think you need to have something, and maybe we'll come up with something that we bolt onto our current processes, but and maybe it'll evolve out of ambient care or something. Maybe it'll it will Maybe AI will just synthesize it out of nowhere. 

24:16.15

John D'Amore

No, it's not gonna be out of nowhere.

24:17.48

Charlie Harp

It'll ride in on digital wings. No.

24:20.54

John D'Amore

We're gonna need some high data quality to feed AI if we're gonna actually get work, Charlie.

24:25.27

John D'Amore

And that's, that's, you know, always thinking about ah data quality before I came on here for it.

24:26.18

Charlie Harp

That's absolutely right.

24:31.03

John D'Amore

And it's so funny to mention AI, because I think that a lot of people I've, I've actually been at board meetings. I won't say where, but I was at a board meeting and I got a question when I was giving a board presentation. I wasn't a board member of that organization. They said, well, our health information exchange doesn't need to worry about investing in data quality because AI is just going to fix it all.

24:49.90

John D'Amore

And I just, I could not grasp that question in terms of how insane it sounds. It's like when people started the internet and says, you know what the internet, we don't need money anymore.

25:01.80

Charlie Harp

Yeah.

25:01.97

John D'Amore

Money is like an input to making the internet work, right? If you can't buy things, you can't transact, you can't do anything for it. The internet didn't obviate money, it made money more important. AI data is the input to making all of these things happen that are so immensely valuable. And yes, you know, generalized pre-trained transformers, GPTs, you know, they when they transform that data, they have a deep representation that, um you know, is a is something that can be common across many input representations of that. So it is like a normalization aspect for it, but that that actual layer change, first it's not exposed, it's hidden, it's a hidden layer, right? That representation for it, so humans can't interact with it. And it changes every time they update the software. So that's not ahs not terminology, that's not standards.

It can help us in so many immense ways for it, but when you can feed better data into it to get started, you're just gonna get better outputs for these things. Data quality is not going to be less important in the age of AI. It's going to become more important. I know people are going to disagree with that, but I don't mind saying it.

26:03.82

Charlie Harp

I'm not going to disagree with that. I think that, one of the things that scares me is that we're doing all these things around AI and, and you know, where it's providing care, where it's summarizing care. And i always try to remind people that AI language models are a rear facing technology.

26:21.75

John D'Amore

Yes.

26:22.03

Charlie Harp

It takes content that was created by us historically, and it's using that to to do essentially a weighted probabilistic decision process. And one of the things that I learned when I was with Zinx is that every six years, the evidence changes and we decide we're totally wrong. Those things we're doing, oh that was crazy. That was actually hurting people. We're going to do it better now. work

26:47.82

John D'Amore

Margarine's not good for you. Let's have butter instead. I mean...

26:50.04

Charlie Harp

yeah That's right. That's right. Eat bread. Now I'm not allowed to eat bread. Make up your mind. I don't know. what's gonna i'm so i'm supposed to so Next week, I'm going to start smoking cigarettes. I think that the problem we have is we're relying on history.

27:03.59

John D'Amore

Mm-hmm. Mm-hmm.

27:03.99

Charlie Harp

And we're relying on the quality of that data to drive models, to help us make decisions going forward. So the models are built on questionable, uncertain data. We're prompting it with questionable, uncertain data. And then we're taking actions on that. And I think that, I think AI is, I think GPT, chatbots, generative AI, large language models, whatever you want to call it, I think it can be a huge accelerant because of its capacity for dealing with large amounts of data and identifying patterns and helping us distill things out. But I'm still, as someone who used to build decision support for a living, I'm still a believer in creating deterministic guardrails so that I can point to that and say, the reason why I decided to do that is because of these inputs, this process chain, and that output. As opposed to I threw it into a machine and it told me to do it.

28:01.67

John D'Amore

I actually talk about this when we go over LLMs and a AI and and a lot of these things, which is they're they're great tools. They're going to change the world. Like, let's start with that as a baseline. It's not that you can ignore these things for it. But when you can do a logistic regression or you can do a rule set and come up with the same outputs, do that first. One, it's computationally so much cheaper. But two, you can explain it. Like, this is why your risk of readmission or your risk of heart failure or death is this percentage. It's because of these five or 10 or 15 factors that go into it. It's so much more understandable to the average person rather than saying, well, it's a black box. We don't really know why they said that this was a high risk type of thing for it. If you can get the same or often better outcomes by doing those things, start there. When you're not getting the results you want through deterministic, through rule sets, through the more basic math, that's when you start looking to AI to start working with big data.

28:57.12

Charlie Harp

Or combine the two together.

28:58.85

John D'Amore

Oh, yeah.

28:58.91

Charlie Harp

I mean, I was, you know, like you guys, like your history, we do a lot of the semantic normalization stuff and we have a bunch of algorithms. We do all kinds of cognitive processing. And I've had people say, well, Charlie, I could just use AI to do my semantic normalization. And I'm like, you know, I can get you like 95% of the way there with deterministic logic in a repeatable way. Are you sure you want to throw that away and go to the pachinko machine and you're fit? Bet everything on the fact that that's going to come up with something equal. You can you can use it for the 5%.And if you can pick up the phone and call the source and say, what the heck is X1755327, it might be able to do something that that that a deterministic algorithm can't do.

29:44.41

John D'Amore

Well, I'll tell you a little one about data quality, which will relate to your kind of pharmacy background for it, which is when I was at my software company, we actually developed an aspect for medication sick person. So to be able to understand any way that those instructions, you know, take one tab by mouth twice daily or one tab POB ID, you know, and understanding that those are the same. And people immediately said, well, we should just use NLP for this. But remember how many CCDA as I talked about processing? Um, if you're processing a billion CCDAs a year and each CCDA has, know, let's say 20 medications on average, you'd say, oh, what? Nobody has 20 medications. Actually, some people have a lot more when you start looking at all stuff for it. Let's say 20. So now you get 20 billion medication entries for it. If how many milliseconds can you spend parsing each one of those? If you actually want to get through all the data and if you're throwing out each one of those out to a LLM or an NLP engine port, you're talking about hundreds of milliseconds before you get a response per medication. Wow, use deterministic things to get you 95% of the way there. And then, yeah, maybe for that other portion, you can you can spend a little bit more time and spend a little bit more money on that portion. But most, and we found that actually it's not 95%, but for was like of them could just be deterministic.

31:00.68

Charlie Harp

Yeah, they're they're, they have to be human readable. They're predictable. They have to be unambiguous because a provider or somebody has to read and interpret it. So you're absolutely right. I think that in those types of scenarios, the deterministic algorithmic logic should be used. Now, when it comes to things like language models, the data prep, creating data, saying, what are all the ways I can say this? I think there's a lot of potential there for AI, and we've we've been looking at how we can leverage that. And I also think assessing when you have a bunch of options, what's my best option here? So once again, assisting a human, catalyzing them, helping them make a decision faster. I think there are some really great opportunities for AI there. And who knows how AI is going to change?

31:45.20

John D'Amore

Well, one thing I definitely known because I've known a few being in informatics, you get to realize that there's a lot of terminologists out there too, for an terminologist, like some your MVPs. People who can understand terminologies and understand how to work with a transitive closure of SNOMED. I'm like, I still have trouble with all this stuff. But with AI, you can turn those terminologists into like, they can vibe code problems solutions to problems that they couldn't do previously. Because they're not like not coders, but like it's amazing how AI as a tool is going to empower you, but it's not it's not magic. It's not going to solve all your problems. And now, hey, we don't need to worry about under interoperability or normalized data or we're going to need to to continue to worry about those things, work on those things. But now we've got some great tools to help us do do these fixes a lot faster.

32:31.61

Charlie Harp

The trick is for people not to be lazy with the technology. I’ve done some code and I've leveraged different, different LLMs. I've used Claude, I've used GPT, and I've had them write code snippets and blocks of processing logic. And what's funny is I never, one of the differences when I do that, I always go back and look at the code.Bbecause sometimes I learn things, sometimes I'll like I'll learn some new way of doing it that wasn't around when I was doing living. Right. And sometimes I'll look in and I'll say, why on earth did you make that decision? And and it's and I think that what's important is, you know, it's great if I can can write code that would have taken you as a human 40 minutes to write it and it can do it just like that. But I think you still have to have the wherewithal and the knowledge and the experience to go back and look at it and say, well, that's not what I asked you to do wrong.

33:26.30

John D'Amore

Oh, absolutely. But, but it makes you much more productive because you're an editor of code rather than a full content creator and the other thing by the way i know you talked about your love of documentation you're probably better than me for at this. I was horrible at commenting my code and i gotta say i've used cursor and and copilot and a few other ones were oh gosh they're so they force you because every function is commented you're like oh my gosh this is like how I should have been coding the entire time um for a lot of it so it gets you it can put you into some good practices but I agree it does not eliminate the need but it can make you more productive.

34:01.85

Charlie Harp

Yeah, I agree. So data quality. I appreciate when when you look at what we're doing with PIQI, I appreciate your kind words about PIQI. And honestly, for me, it's been, it's been a great experience. It's been a lot of fun intersecting. I mean, with the folks in the standards, it's really been my first standards foray. And there are aspects of it that are, um,

34:27.86

John D'Amore

Infuriating?

34:28.90

Charlie Harp

No, I don't get mad. I just, it's one of those things where I get, it's almost like when somebody asks a question, I know a lot of people that would get frustrated with that, but I appreciate the fact that some person in their own time took the time to read a document to ask me an insightful question about something. I mean, I always, I kind of step back and go, the fact that they took the time, they showed me the respect, they they went and looked at it. I can, I might say, okay, now I got to explain what this word means or why I chose this word. And a lot of times I just say, yeah I'm just, I just chose the wrong word. Tell me what you, what you prefer and and we fix it. But I temper any frustration I have when I'm going through that process with a deep appreciation that somebody else took the time because we live in a world where a lot of people don't, they don't read emails. They don't, they don't look at things before they come to meetings. And when people do that, I just feel a deep sense of appreciation because, you know, it, it makes everything work. And, and I'll honestly going through this process with HL7 and going through the Alliance work groups, sometimes I have to explain myself. Sometimes it makes me think about why I said what I said. And sometimes, more often than not, I come away with, with something that's better than what I had before.

35:55.19

John D'Amore

You know, I think, mean, the community of HL7 is, is the most valuable asset. It's not any of the standards they produce for it is the people that are brought to this and they are thoughtful and they spend time. And a lot of them are people who cover time, like as part of their job, right? It's not their full-time job. They still to go back and code or go back and do oversight for projects for it, but they make time to show up, to read the standards, to be informed and then to give feedback for it. And it is, I use the word infuriating kind of in jest, but it's, it's hard because you do get pushback and you have questions and, and they're generally good questions, which means you got to think about you're like, huh, I hadn't thought about it that way. And, and they got a point. Sometimes it's easier when they don't have a point, you can just kind of, you know, that's not really where we want to go. Um, but they, they come back with good questions for it. And I think about it like that, that, Bounds, phrase that's always used democracy is you know it's not a great form of government it's just the least bad version of what we've tried to date for consensus based standards making is not the most efficient way to go forward but it's like the least bad way that we figured out bring to bring together a whole bunch of stakeholders across the Community give them an opportunity to to and i'm so proud and honored that you guys are balloting. You know, I know it's an informative guide, you know, to get started with PIQI for it, but I'm so proud that you're balloting it. You're putting it out there in the domain for public to comment on, for public to give feedback on, and it's making it better. I think I've already seen improvements since you guys get started, you know, six, nine, 12 months ago. And I think it's just going to be such a tremendous asset, you know, that I've been talking about this my entire career. I used to actually have this on my website before we had a marketing team, Charlie, who had said, why don't you not put quotes on your website? It looks cheesy. But you know it's Lord Kelvin, which you cannot manage that, which you cannot measure. or i don't Deming also said, I think it's a similar aspect for this.

37:45.76

Charlie Harp

He said, if you can measure something and explain it in numbers, then you've begun to understand it I love the Lord Kelvin quote. It's a great quote about data quality.

37:56.60

John D'Amore

And that's what we need because we jumped in healthcare care because we always know where we wanna go. wanna go to higher quality. We want to go to value-based care. wanna go to the say, hey, we know that we're improving colorectal cancer screening, diabetes management, all these things that are really important for patients for it. We jumped to the quality measurement without asking the intermediary question, which is, is the data or are the data feeding these quality measures of appropriate quality before we try to actually do the outcome measure for things. And i think PIQI is really filling a need and a gap for it. Now I'm excited with what other organizations are doing, right? NCQA is doing some work on it. I think there's a lot of coordination and movement in the community, but I think that you're pushing this, again, going back to what I said earlier, you're pushing this rock up the hill. It's not easy work to move this forward. and And frankly, it's harder. It's much easier just to say clinical architecture is going to some software, you know, buy our stuff. You're actually moving this forward in a consensus and standards are driven process which benefits everybody. I think that there's really good benefits for for Clinical Architecture too, but this is something that's great for the community. So I just, I really want to say thank you for that. And I think that it's it's it's exciting to be part of that. I you know I gave a few Jira tickets um um on the first ballot for things. And I think it's it's great progress to see what's moving.

39:11.82

Charlie Harp

And i I really appreciate your involvement in it and everybody's involvement. I think that with PIQI, what I have to keep explaining to people, because people say, Charlie, we check our quality. We have 8,000 quality checks in our data. And I'm like, well, yeah, there's a difference between a developer doing a check-in schema then the kind of quality checks we're talking about in PIQI and the biggest difference is it's not a replacement for the quality checks that people are doing when they land data in a repo and they're checking for referential integrity and all these other other things that they validate. The difference is, is that quality goes from being an adjective to a noun because we have a scorecard and we're saying, according to this rubric, is your data quality or not? Because one of the things and I don't know if you encountered this in your career, I'm sure you have, is that when you're selling things that improve the quality of data, the answer you get is like that money. Python's get where, you know, we're we're in search of the Holy Grail. Oh, he's already got one. You know, the data quality is fine.

40:09.71

John D'Amore

Yep. Yeah.

40:12.27

Charlie Harp

I told them we already got the one with like, oh, our data quality is fine. I'm like, are you, are you, are you sure? Have you looked at it? And the cool thing, the thing that i love about PIQI and the thing I said to my team is, because we were debating this early on that we can make a product or we can make a difference. And if we can have a scorecard that we can agree as a, as an industry, this is what quality is. These are the things we measure for this use case. And people put the data in that it's not a question of, I have great quality. The score will tell you if you have good quality. And the other thing that came up, it came up at the nncqa i did a presentation in NCQA, and the question was, what's a good quality score? I said, well, if the rubric articulates the minimum ah requirements to be usable, the minimum score is a hundred percent, right? Because if you build a rubric that says, this is what I need, and you don't give me that 100%. ah hundred percent Now you can have an arbitrary scale. And and for me, anything below 100%, the purpose of that is the second I in PIQI, which is improvement. What do I need to do to get to 100%? It's like being in college. you know what What do I need to do to get the A? And so I've been really humbled and excited about all the all the interest and activity around it And Levitt Partners has been great.

You know, folks like you that have added your brains to what we're doing. um And, you know, for me, it's one of those things where when you can do something, you know, when I look at somebody like Graham Greed, when you can do something, that makes the industry better, that's something you you can take away you know as you as you ride off into the sunset. You can say, hey, I was part of that thing that made a difference. And so I'm very grateful to everybody for being involved in it.

42:07.82

John D'Amore

Well, I think you said it well. I think that also just the way that you've started to categorize this, you know one of the ways that I tried to use words to talk about this, which is like there's a syntax or syntactic normalization that needs to get done for it. Like, hey, when you have something in quotes, you can't have a space or in this code on this code feel for it. That is not the level that PIQI is really aimed at for, it which is not trying to get to that syntax, which is part of XML or JSON or whatever data format that you want to use. It's not something that fits nicely into schema. It's much more of those rules of like, does this make sense for it? You know, if you've got a vital sign for a patient and it's 12,080 milligrams of mercury, I mean, there's nothing wrong with that from a syntactical presentation aspect, but that's not within the range of a human being at least, i don't know, maybe there's something out there that can have that blood pressure, but you know what happened, which is the nurse, rather than writing 120 over 80, the systolic diastolic fields just get all crammed in that one field.

And you know, this is something where we need to have the rubrics to, ah to work on this. And I think having those different dimensions, like some of them are just going to be, you know, pass or fail, like a hundred percent, or it's a no go for it. But there are some things where like, you know, we're not expecting every medical diagnosis that's ever been done on every patient to always have an onset date. It's not always known. It's not always relevant to ask, um, for things. But if you want to say, Hey, are we doing good documentation? You need to have a way of saying, well, what was our rate of onset date documentation five years ago? are we better than that today? You know, one of the things I were like i shared it when we talked about previous at a previous meeting, which is like LOINC code normalization. people and Everyone would kind of agree we're probably using a lot more LOINC codes today than we did 10 years ago in healthcare. But at what rate? Were we 50%? we're like 70%? Or were we at 40% and now 42%? it where we at 40% and now at 42% percent I don't know. Nobody can answer these questions. And until we come up with the right frameworks like PIQI to be really put together SAMs, put together kind logic models of what we're going to measure, and then do that consistently over time, we're not going to be able to make the improvement that we want to.

44:15.10

Charlie Harp

Yeah, it goes back to Lord Kelvin. If we start to measure it, we can make we can start to change it and observe the change we're hoping to see, right?

44:22.22

John D'Amore

Absolutely.

44:23.35

Charlie Harp

Hey, I'd like to get your commitment that you're going to do this again with me. We're going to talk about, there's so many things that we could talk about. Like I want to talk about code systems. I want to talk about how FHIR represents code coding collections. I have like a million things. Can I get your pre-agreement?

44:38.76

John D'Amore

I will pre-agree right now.

44:41.22

Charlie Harp

All right.

44:43.55

John D'Amore

I will come back for it this is, again, I'm going I've told you I've listened to the last six or seven, four. I'm going to go back and listen to 40.

44:50.02

Charlie Harp

Oh, well, if you go back too far, they get a little rough. I was new.

44:53.83

John D'Amore

Well, that's okay because, again, you're doing this and and this is not easy easy to do and, you know, there's a marketing benefit to podcasts for it, but I think you're putting together an asset for all my informatics students. You know, this is something to listen to, to learn, to to have an engaged dialogue for it. And I think that you just do a great job, Charlie, interacting with people on a personal basis because we're all humans. We're all, we're all trying to find our way through this muddled night and figure out how we make things a little bit better in the world before we're at the end of it. 

45:20.06

Charlie Harp

And it's another thing about our industry. You know, i think I think at one point you and I would have been considered to be competitors, right?

45:27.12

John D'Amore

I think so.

45:27.89

Charlie Harp

And and i I think that I've had a whole bunch of people out there in the industry that have been in a competitive situation with me. And I don't know that I've met too many people that I don't genuinely really like.

And but I like the fact that in healthcare, care even though we might be in a competitive situation, that doesn't mean we can't agree, we can't have conversations, we can't get along, we can't grab a drink. And so that's another great thing about, about this industry in the community.

45:52.85

John D'Amore

I didn't I know I didn't come to healthcare care to win. The winning for me is making patient care better, not having a successful business.

i think I think the same of you, Charlie, like which is, i can there i've never even though I competed with you in the past, I never really said anything bad about Clinical Architecture because you guys were doing good work. We're all rowing the boat.

46:10.06

John D'Amore

you know We'll see which boat come moves forward at what pace for it, but um we're all trying to make healthcare just a little bit better.

46:16.31

Charlie Harp

Absolutely. All right. Anything you want to share before we sign off?

46:20.19

John D'Amore

No, thanks so much. I'm excited. I'll come back. You let me know when. Awesome.

46:23.33

Charlie Harp

Yeah. We'll make it a regular thing.

46:26.13

John D'Amore

The

46:26.69

Charlie Harp

All right. Thank you, John. um As always, thank you for listening. I'm Charlie Harp with,

46:34.05

John D'Amore

Informonster Podcast.

46:35.49

Charlie Harp

well, no, you say your name.

46:36.82

John D'Amore

Oh, and John Tull. Sorry.

46:39.99

Charlie Harp

on the Let's say it together. On the Informonster podcast.

46:43.00

John D'Amore

The Informonster.

46:44.82

Charlie Harp

Thanks a lot.

46:45.75

John D'Amore

Thank you.