Oracle doesn't need FDA approval. Most EHRs are excluded from the definition of a medical device by the 21st Century…
HIStalk Interviews David Lareau, CEO, Medicomp Systems
David Lareau is CEO of Medicomp Systems of Chantilly, VA.
Tell me about yourself and the company.
I’ve been with Medicomp about 20 years and CEO for 10. Medicomp’s core business is connecting all of the clinical information and data that is in an EHR, whether as terminology codes or free text, and making diagnostic sense of it, either for the providers at the point of care or for people reviewing the record for diagnostic relevancy, which is important now with Medicare Advantage and value-based care. We’ve been building this for 45 years. It seems that the industry is moving in the direction of not just trying to paid for transactions and coding them, but getting paid for caring for patients effectively and proving that it was done. It’s an exciting time for us.
Doctors say they are burned out from keystroke overload and entering data that doesn’t contribute to patient care. How might that situation improve?
There are a couple of ways to go about it. If you approach the EHR like a burden, as most of them are now, you’re just trying to isolate the clinician from the EHR. Ambient AI is a play in that space, saying that we’ll just listen to what’s going on in the room, and then maybe at the end, we’ll tell you or a reviewer that you have to meet these quality measures and your documentation might not be sufficient to pass a Medicare audit. Capture stuff at the point of care, get it coded as best you can, but don’t really use the EHR as a tool for the clinician. Just try to isolate the clinician from the usability of the system.
Or, you try to put the information in front of the clinician, at the time they need it, for the patient and the condition or multiple conditions that they are dealing with. Here are the clinical quality measures that apply. You’ve met the documentation requirements. You have all the information properly done. Then present them what they need, when they need it, so that the EHR becomes a data repository, not a repository of text and other stuff that has to be dealt with after the fact. If you can’t do that, if you just keep popping up stuff that’s not clinically important when they are thinking clinically, they are going to get burned out and they will be frustrated.
Value-based care, 21st Century Cures, and TEFCA have increased the need for what we think of as diagnostic interoperability. Either diagnostic interoperability between systems or diagnostic interoperability between the clinician and the system itself, saying, I’m dealing with this patient. They have these conditions. Show me what I need. Show me what reflects the way I think and work. Also, let me complete my work here and get on to the next patient.
One of the big things that people contact me about is Medicare Advantage. Medicare Advantage is not saving the government the money that they thought it would. There’s more and more people of the baby boom generation retiring and they are living longer with chronic conditions. We have to bend the cost curve down. How do we do that? One solution they’ve come up with is to take better care of the patients and their chronic conditions. CMS has said, we’re going to come and look at your records, and we want to see evidence that you have managed, evaluated, assessed, and treated every one of these conditions for which you’re claiming risk adjustment and risk adjustment solutions have been on. Make sure we get these things coded so that we get a higher risk adjustment factor for each patient. Fine, but are they really taking care of the patient? Does their documentation prove it?
That’s where we are seeing the most interest in what we do. At HIMSS, we will be promoting whether you have the processes and technology in place to protect yourself against Medicare Advantage fraud audits, because that’s about managing patient conditions, not just getting the diagnosis code right. We’re getting lots of interest on that from people who haven’t talked to us before.
It’s an exciting time to be in our industry. Some people get excited about AI, and other people poo-poo it. There’s a great place for it if you have good, clean, high fidelity data. Then it can empower these learning models and algorithms. The industry is in such a state of flux because of all that. W are just glad we are in the space we’re in.
Microsoft and Oracle are now deep into the healthcare application area via acquisition, and both companies have placed big bets on cloud and speech recognition. What changes do you excpect and how will they affect other companies?
There’s a great place for speech, text, and the technology. Natural language processing, NLP, which a lot of these approaches rely on, provides at best 75% to 85% data fidelity. Most of those systems are trying to find codes in text – SNOMED, ICD, CPT – through language models. They work pretty well, but when you are trying to get a full clinical picture of the patient, you need to turn all of that into computable data that you can filter diagnostically. That’s what enterprises are being asked to do. Manage these patients, especially under Medicare Advantage and value-based care, manage their chronic conditions, and show that you did it. A lot of the models are relying on reviewing that stuff after the fact to make sure we did it.
We were pretty excited a few years ago when we got approached by Emtelligent, which has a natural language processing engine. They wanted to add our concepts that are in our engine to the roster of vocabularies they looked at. I told them that we weren’t really interested in that, but if they could do a version of their engine that targets our vocabularies, then we can filter that stuff diagnostically. We can take the text record and say, show me what in this record applies to chronic renal failure versus diabetes and then pass that to algorithms that say if it looks like it’s documented adequately to pass a Medicare Advantage audit or not. There’s a real exciting mix of voice navigation and voice capture of information, but that still needs to be turned into data that is computable. We sit in the middle of all that.
How does the growth of ChatGPT and other AI tools impact company strategy?
The Gartner Hype Cycle says that it takes a while for hype to build, but I’ve never seen such an upward thrust in the hype cycle when ChatGPT came out from OpenAI. But there are valuable uses for this, because that kind of technology at its core does statistical analysis of data and pattern recognition. If the data is good and the information that you’re trying to process is best processed as data like they’re seeing now — images, MRIs, and mammograms in a consistent format – there’s an opportunity to get high fidelity data out of that and apply AI to it.
Machine learning is valuable for remote patient monitoring, for patients who are willing to do it at home, for monitoring their hallmark findings for chronic conditions. Trying to support the clinician at the point of care is problematic, unless you just say that we’re going to use this stuff to capture all the information. We’re going to use voice, speech, and sound and turn it into something and then process after the fact to figure out if we have gaps in care. That whole framework for where this stuff is and where it fits now versus in 10 years, we are constantly looking at that.
We’ve decided for now that our place is to make it possible to take in all this information — whether it’s text or codes from these various terminologies and code sets — diagnostically organize it, and present it back to the user. Eventually that kind of information will be valuable for ChatGPT or other AI algorithms that then apply machine learning to detect patterns that would otherwise not be detectable. We are constantly looking at that.
People used to call our stuff AI back in the 1980s, not the same way that people do now because we built it using physicians who determined what’s appropriate when you’re thinking of one diagnosis versus others. That’s valuable data. Getting data acquisition and being able to diagnostically filter it is important. We do that pretty well. If people can start applying AI and machine learning to the data to our data points, it will be valuable. We’re pretty excited about it.
ChatGPT provides a chatbot-like response to user input as an ongoing conversation. Will that affect the usual software design paradigm of static screens full of data entry fields and submit button at the bottom?
The chat paradigm is an evolving target. As a conversation proceeds, different things seem to become relevant. The challenge is that clinicians, not just doctors, are pretty highly trained users. They’re not like me going out on the internet and typing in a few searches to put together an itinerary for a three-day visit to Phoenix, I don’t know anything about Phoenix, never been there, so it’s a good tool for that.
When you’re dealing with a highly trained clinical user, and when you think about physicians — medical school, internship, residency, their experience – they are already pretty good at clinical pattern recognition. They would like systems to present to them what they know they need to do their work. That’s what we try to do.
ChatGPT does that by searching the internet to find things that the user is not familiar with and and constructs information for presentation. Our engine does that from a diagnostic framework, pulling all this stuff together. But the technology inside things like ChatGPT will be more useful to the clinician when they’re dealing with conditions that they are not familiar with. For example rare diseases. The National Organization of Rare Diseases has a list of 1,100 to 1,500, depending on how you count them. Rare diseases that in some cases, if detected early, will lead to much better outcomes. If missed, there’s not much you can do about it. You can’t really prompt every clinician to consider the symptoms, history, physical exams, and tests that are relevant for every diagnosis the patient might have.
But with artificial intelligence running in the background, you can present the things to the clinician that make it usable for 98 or 99% of all patients. An algorithm runs in the background that says, this patient might have this condition. If you want to see the hallmark findings of it, click here. If not, go about your work. They tell doctors in medical school that if they hear hoof beats, think horses, not zebras. For things like the zebras in medicine, AI and machine learning could be valuable.
Medicomp has made few announcements of executive changes, acquisitions, or funding, which usually dominates the headlines of other companies. How does that position the company in a challenging economic environment?
People have a tendency to chase the latest hot thing. If you guess right, great. But if you guess wrong and you give away equity or control, you can no longer focus on what the core business is or the core value that you bring. We’ve been clear from the beginning that we wanted to focus on providing a tool that presented information to clinicians, the way they were trained and the way they need it.
To do that, you need patient capital. You can’t chase quarterly results. You have to approach your people as the most powerful, valuable, and non-replaceable resource in the company, because when you’re creating software and intellectual property, turnover kills you. Change of focus changes or ownership kills you. People say, we’re such and such and it’s in our DNA. I always say to them, yes, until you get a new CEO, and then who knows what’s going to happen?
We’ve been consistent in what we’re trying to do. We’ve never gone into debt. We don’t chase the latest thing. We’ve always thought it was going to be critical at some point in this industry to move away from tracking transactions to get paid to tracking conditions to get better outcomes. Our engine was built to do that. We’ve been able to retain that focus and get enough people interested in using our stuff so that we had the revenue to stay on track and we had the opportunity to continue to our core engine and all the mappings as the industry changed. Then, adapt for what we needed to for our core mission, which is diagnostically connecting data and presenting it and tools for documenting it, if people want to use our documentation tools.
Changes are fine if you really need to change your focus, vision, or mission. Peter Goltra set one out for us a long time ago and we’ve been able to stay with it. We’ve been pretty happy with that. It has also allowed us to keep the people we need to adapt to things like Meaningful Use, 21st Century Cures, ECQMs, quality control measures, and TEFCA interoperability. Figure out what you’re doing, get really good at it, and stay at it until something tells you you’re doing the wrong thing. So far, we’ve been fortunate that we seem to have made the right big choices whenever we needed to.
What elements will be important in the company’s strategy over the next few years?
We think the healthcare IT industry is on a path to realizing that the clinical record of a patient, regardless of where it resides, should be computable data that will power analytics, AI, and machine learning. The challenge is going to be filtering that data and presenting it to the various people who need it and meeting all the requirements that are being forced down on the providers by all kinds of things. Home healthcare has a set. Hospitals have a set. Ambulatory has a set.
We think that over the next three to four years, we will see an increasing move and realization that the important thing is caring for the patient using AI machine learning, and other techniques for identifying people in a population who are at risk. But you still have to somehow deliver the care for each of those conditions, one patient at a time. The industry is coming to the realization that it would be much better for these health information technology systems if we had data, not just a bunch of stuff electronically stored. We are excited because of the realization in the industry that data is paramount to everything.
One more C-suite statement wih nothing really to say that is not known already. Two pertinent examples in the above:
“Bending the cost curve” in re Medicare Advantagehas not saved money. Always the soltuion from the top is cut costs, when the real problem has been gaming on the part of the insurance players. No one of the parties with skin in the game, including patients but especially providers, pharma, third party payers and all the vendors– is alone in participating in the greed which drives our system more than any other factor.
“Capture stuff at the point of care” That horse left the barn along time ago as clinicians have become data entry clerks and it is particularly bad in primary care which is the most vulnerable and valuable point in the health care information universe.
Bottom line: Same old, same old.