Mark at Health 2.0 Conference: Insights into Healthcare Interoperability

apple spotify

Recorded at the Health 2.0 Conference, this episode of HealthTech with Purpose features Mark Scrimshire, Chief Interoperability Officer at Onyx Health, in conversation with Mindbowser CEO Ayush Jain. Mark tells the story of Blue Button 2.0: how an entrepreneur-in-residence stint at the HHS IDEA Lab turned a 1,700-page Medicare claims printout into a FHIR API, why he helped define the Explanation of Benefit resource with the HL7 community, and how launching at HIMSS in 2018 gave CMS the proof it needed to write interoperability rules for health plans. He then maps the US landscape as he saw it in 2023: ONC certification moving EMRs from DSTU2 to FHIR R4 and US Core, patient access and bulk APIs, and CMS pointing to community implementation guides as a recommended path rather than a mandate. Mark explains how HL7 connectathons and the Da Vinci Project produce the guides regulators later reference, including the Payer Data Exchange guide he leads. The conversation closes on the next hard problem: discovering FHIR endpoints and registering a million doctors to a thousand health plans without every connection being set up by hand.

Highlights

From a 1,700-page printout to a FHIR API

Blue Button 1.0 let Medicare beneficiaries download three years of their claims, but Mark points out that a typical record printed to about 1,700 pages of plain text. As an entrepreneur in residence at the HHS IDEA Lab, his brief was to turn that into structured data people could reuse. Rather than build something custom, he joined the HL7 FHIR community, helped define the Explanation of Benefit resource for complex claims data, and loaded Medicare’s Chronic Conditions Warehouse data into a FHIR server with APIs in front of it.

Blue Button 2.0 launched at HIMSS in 2018 after being opened to developers as a trial first, which Mark describes as a deliberate move so CMS could not easily close the door on it. By his count it now covers 60 million Medicare beneficiaries, and because CMS had already agreed to release three years of data, he “forgot to tell them” the team wasn’t going to throw older data away, so the window grew to seven years.

Why a shared standard changes how health tech scales

Mark’s clearest argument for FHIR is the pilot that works at one hospital and then stalls. A diabetes program shows great results, the team tries to scale it, and the next hospital runs a different EMR, so the interface gets rebuilt from scratch. With ONC certification now requiring EMRs to expose around 28 resources in the same format through FHIR R4 and US Core, he argues teams can spend their time on what to do with the data instead of on getting it. The certification shift he describes, the move from DSTU2 to FHIR R4, is what makes that baseline possible.

That per-EHR rebuild is still the problem most digital health teams hit first. Mindbowser works on it both ways: FHIR and HL7 integration built for your product, or ConnectHealth, which connects a product to 20+ EHRs through one integration model and runs inside your own cloud.

How implementation guides become regulation

Mark explains a sequence most people outside HL7 never see. Legislation bubbles up in Congress, payers and providers start drafting specifications, and when regulators get their deadline they look for implementation guides the industry has already tested so they can point to them. CMS supports accelerators like the Da Vinci Project because, as he puts it, it cannot write regulation if it cannot point to something people can already do. Connectathons, where around 200 engineers spend a weekend testing specifications, are where those guides get proven. The Da Vinci work he describes is the same foundation behind today’s FHIR prior authorization APIs.

He leads the Payer Data Exchange work group, which answers a simple consumer question: when you switch health insurers at the end of the year, why shouldn’t your data go with you? He also stresses that guides need to keep evolving, or the industry ends up with “islands of interoperability.” Kristen Valdes of b.well made a similar case for open standards and implementation guides in an earlier episode.

From checkbox compliance to business change

Asked for the biggest challenge, Mark says adoption through regulation is going well, especially in Medicare Advantage, but many plans have treated the APIs as a compliance box. He expects the next round of regulation to push plans toward seeing how the data can change their business, and cites California requiring plans operating there to implement the interoperability rule across their commercial plans. He contrasts FHIR’s pull model, where you ask for a specific piece of data with the right permissions, with X12 transactions and C-CDA documents that you wait for someone to send.

The next problem: a million doctors, a thousand health plans

Mark describes his spec-writing role, semi-jokingly, as creating the next generation of problems to solve. Once enough health plans support these APIs, manually setting up every connection stops being workable, and that pressure forces the next standard. The open questions he names are how an app discovers FHIR endpoints, how it gets permission to access them, and how to register a million doctors to a thousand health plans while keeping access secure. His view is that piece solutions won’t get there; it takes the community acting together.

  • Full Transcript
  • Ayush Jain [00:26]

    [Opening portion of the introduction is not captioned.] …the next 30 minutes of his time with us, especially talking about his experiences with interoperability, because he’s a man who has seen it all, being part of interoperability right from when it all started. He’s been one of the people who were part of setting it out and then implementing it at various places, and today he kind of leads his life as a mission to bring interoperability closer and closer to our lives. So welcome, Mark. Thank you so much for joining us. How are you doing today at the Health 2.0 conference?

    Mark Scrimshire [01:03]

    I’m doing great. It’s been a really interesting conference, with lots of interesting people from around the world. So you get these different perspectives, because the US is very different from the UK, from Europe, from India and the Far East. It’s always fascinating to meet people and see those differences. So it’s been good today.

    Ayush Jain [01:29]

    So Mark, one of the things that pops out when I look at your profile is your invention of the Blue Button. So talk to us more about that experience, especially the time when you invented it and the context around it.

    Mark Scrimshire [01:45]

    Okay, so yeah, it was interesting. Blue Button originated back in 2009 as Blue Button [inaudible]. The whole idea there was that it came out of meetings of tech folks. I was not actually involved in that, but the idea was: shouldn’t I be able to log into the portal, whether it’s for a government service or for the VA, and shouldn’t I be able to actually retrieve my data and be able to do what I want with it? So that was Blue Button 1.0, and that was deployed by the VA, the Veterans Administration, and also by the Centers for Medicare & Medicaid Services (CMS). For CMS, it basically enabled three years of a member’s health data, their claims, to be downloaded. So if you were a Medicare beneficiary you could do that, but if you tried to print that out, a typical record was 1,700 pages of just text. How do you make sense of that?

    I had just come out of working with 3M on taking their products and systems to the cloud in healthcare, and I was urged to go to the US Department of Health and Human Services (HHS). They had an organization called the IDEA Lab, and in there they had entrepreneurs in residence. They basically brought in people from private industry and applied them to a task, to solve a problem in the government. So you were effectively paid to go and cause trouble inside the government, using entrepreneurial skills to really try and crack a problem.

    The problem I was asked to address, which I had been very much a supporter of for a long time, was Blue Button. The idea was: wouldn’t it be better if we could actually turn that into more structured data, so that it could be much more easily reused?

    So I went in and failed at the task. I was supposed to do this in a year, but what we ended up doing was a proof of concept, and then going to find sufficient funding to actually take this through to implementation. It was very much a case of standing on the shoulders of giants. At that time, this was about 2015, the Health Level Seven (HL7) Fast Healthcare Interoperability Resources (FHIR) specification was starting to move. You were starting to see electronic medical record (EMR) vendors, because of various moves by the Office of the National Coordinator for Health IT (ONC) that were pushing, through meaningful use, to say you need to have some sort of application programming interface (API). Well, rather than everybody have their own APIs and just perpetuate the problems we’ve had in healthcare, there were meetings convened where basically the EMR vendors said, okay, we’ll support FHIR as a common specification.

    So with that happening, I basically got involved with the HL7 community, and FHIR is very much a community. I started going to the meetings and said, I really need a profile, a data structure, that can express all of this complex claims information. So I basically worked with the FHIR community to define what’s called the Explanation of Benefit profile in the FHIR specification, a completely new resource.

    Then we took the data that Medicare had. We basically built a platform to be able to take that data from their Chronic Conditions Warehouse, which is a massive database covering data for every Medicare beneficiary going back 20 years. We basically took that data, extracted it, converted it into FHIR format, loaded it into a FHIR server and built the APIs on the front of that, very much following the emerging specifications that were going on in the community. So rather than do something custom, we tried to embrace the emerging specs.

    The net result was that after two and a half, three years, in 2018 we were able to launch Blue Button 2.0 at the Healthcare Information and Management Systems Society (HIMSS) conference. The CMS Administrator announced it. Prior to that, we had actually released it to the developer community as a trial, an important step in opening the door so that CMS didn’t feel they could close the door on it. We rolled this out, we had a number of consumer apps wanting to connect, and that grew and grew. The net result is that there’s a resource there now covering 60 million Medicare beneficiaries.

    When we designed the API, I was rather naughty, given that I was there to cause trouble. The original Blue Button 1.0 released three years’ worth of data. So you’d already made this decision: you’re releasing three years’ worth of data. I just forgot to tell them that we weren’t going to throw the data away. So as we went forward, it was four years’ worth of data, five years, and it’s now seven years’ worth of data. And that’s how Blue Button 2.0 came to be.

    I think that ultimately led to CMS turning around to the industry. One of my reasons for going to CMS was that they’re a third of the healthcare industry. They spend something like a trillion and a half dollars a year supporting various plans: Medicare Advantage, the Children’s Health Insurance Program, Medicaid and so forth. So if you can get the major healthcare payer in the country to do something, a lot of people tend to follow. But CMS also has the advantage that they write regulations, so those government-funded plans, like Medicare Advantage, have to do what CMS tells them to do.

    At Onyx, the team that I was involved with, I was the first person on the project, but we built a team behind it, and we basically built Blue Button. We took those lessons, because we could see that CMS was then going to write regulations that basically said: look, we’ve done this, we proved we can do it at scale, your insurance plan can do the same. And that resulted in regulations that went into effect in 2021. So it all started in solving a smaller problem.

    It was an interesting problem and it was very challenging. Looking back, it seems like, oh, it’s just a bunch of data. But actually designing the project so that you don’t hit those government roadblocks that come every step of the way was the challenge, because it’s an incredible group of people in government, in places like CMS, really dedicated to what they do and really trying to do the right thing. They’re very careful in doing that, because they have this responsibility for all this data. So it’s really trying to help them and prove that this has been done before in different ways, and just really trying to get them over that final step.

    Ayush Jain [10:30]

    No, makes sense. I think a lot of things had to come into place.

    Mark Scrimshire [10:39]

    Yeah, I mean, this is Blue Button. I think it goes back to the original Blue Button 1.0. One of the chain stores that does office supplies has the Easy Button, and it’s really that: why can’t I just click a button and this happens? It was blue because a lot of what goes on in healthcare is blue, whereas recycling and energy efficiency is green. So it became the Blue Button. When we did the API, we simply [inaudible], so it just made sense. You want that recognition. People were already starting to know: I can get my data using the Blue Button.

    Ayush Jain [11:31]

    It seems so obvious. Yeah, great. So Mark, tell us more about the state of interoperability today: one, how it has evolved, and second, what the future holds. Because again, you’re someone who is in this space very, very intimately, and [inaudible] the best eyes on everything that’s going on.

    Mark Scrimshire [12:06]

    It’s actually important to realize that there are two thrusts going on here. I’ll talk about the US more than internationally. In the US, you’ve got the Office of the National Coordinator for Health IT, where they’re setting specifications and requirements that tend to deal with what we call certified products, so basically the EMRs. They were really leading the charge on what was then meaningful use, so they’ve laid down certain specifications.

    Then you’ve got CMS. You’ve also got the Veterans Administration as well, so again another large payer. The VA is more about delivering overall care, and they want to follow similar sorts of standards. CMS has the power of the purse across a large portion of the industry, because they’re funding things like Medicare Advantage, they do Medicare fee-for-service, and they’re supporting Medicaid, which is administered at a state level. But they can provide funding for the states. They have the 90/10, where they’ll provide 90% of the funding for the states to do something, provided they do the right thing. So you’ve got these combined things going on.

    So what’s happened recently? If you look back to last year, the ONC has been working through regulations around information blocking, trying to encourage the free flow of data, securely. Part of that, which comes out of legislation from Congress, is that they have updated their requirements on certified products.

    The original EMR certifications were on an older version of FHIR. In the community we call it DSTU2, Draft Standard for Trial Use, basically. So the EMR vendors had done a selection of types of records in standard format using that early version of FHIR.

    With the work going on in the CMS interoperability rule that went live in July 2021, CMS basically pointed to FHIR version 4. This is really the widely deployed version, and it has been the base standard. They pointed to implementation guides the community had created to address certain areas, like formulary data, provider directory data for your network providers, and clinical data, which uses the terms US Core or the United States Core Data for Interoperability (USCDI). Then the claims data, which had basically been taken on from the CMS work that we did: an organization called the CARIN Alliance, which is focused on patient engagement and involvement, basically took that and built an implementation guide based on the new version of FHIR, Release 4.

    Those were pointed to by CMS. So: we’re not saying you have to, but if you use these guides, you will meet the requirements of the rule. And that leaves scope for us as an industry to make improvements.

    So what happened last year is that the ONC, on October 6th, basically expanded the definition of electronic health information (EHI), basically saying there’s not much electronic information that you as a patient cannot get your hands on. So we’re going to see more of that gradually getting structured.

    At the same time, they said that by the end of last year, EMR vendors, if they wanted to be certified, and if I’m a health system and I go to buy a new EMR, I want it to be certified, would have to be on the latest version of FHIR, Release 4 at the time, using US Core version 3.1.1 for the clinical records.

    So now we’re sitting here in 2023, and the EMR vendors’ certification means that they’re doing, I think it’s 28 resources: encounters, conditions, procedures, practitioners, the whole gamut, all in the same format. Plus they have to make that available as a patient access API, so you can go to your doctor, connect an app and be able to download your data. That builds on the work CMS had done, and basically requires the same standards.

    The other part of the ONC certification is bulk access, so you can do a population health type query into that data. That’s really important for the industry, for example for a plan to be able to go knocking on a provider’s door and say: we have a business relationship, I’m trying to manage my health plan members, here’s my list of health plan members, can I have their data? That’s just starting to roll out.

    So now there are standard ways. Because if you look at what’s happened traditionally, you might have a pilot addressing some radical improvements in diabetes. What happens is you do the research, you do a pilot, you roll it out at one hospital, and then you want to say: it’s great, we get great results, let’s scale it. Now you go to the next hospital and it’s a different EMR, so you’ve got to reinvent.

    Now, with FHIR as a standard, saying here’s a baseline, all the EMRs, or at least all of those that matter, will talk the same way. You now don’t have to keep redesigning the interfaces. You can spend more time working out what you need to do with the data than you spend getting the data. But we’re still in the early days, and I think we’re going to see a lot evolve over the next few years. I think it will start to snowball, as people realize they can get that data and then put it to use really quickly.

    Ayush Jain [19:21]

    So Mark, the one thing that was, in fact, leading on your profile was your [inaudible]. So talk to us.

    Mark Scrimshire [19:44]

    I think it’s the fact that, starting with Blue Button, because we built it on FHIR, I’ve been involved with one of the work groups within HL7. At least three times a year we have what we call connectathons. That’s where you get 200 geeks in a room for a weekend, hammering on the specifications to see that they work and where they can be improved. I’ve lost count; I’ve attended 20 or 30 of these connectathons.

    And so the Da Vinci accelerator. What HL7 did is, in the EMR world they had the Argonaut Project, and that proved successful at coordinating the activities, using agile methods to keep pushing the ball forward in the EMR world. Well, a bunch of folks got together and said there were some real problems in the US between payers and providers, lots of manual processes and so on. So the Da Vinci Project was an accelerator where a bunch of EMR vendors, provider organizations and payers got together and said: we need to solve certain problems.

    In that project, the various key players fund development of new implementation guides, because at the end of the day, interoperability only works if you’re all doing the same thing.

    It’s been interesting, because what you’ve got is organizations like CMS supporting those accelerators, because they can’t write regulation if they can’t point to something that people can do. They need the specifications. So what’s happening now is you may see legislation bubbling up in Congress, and payers and providers may get together and say, well, we really need to address this issue, and they’ll start to work on the specifications, saying this is how we could do it. Then, when the legislation hits, the regulators have a certain amount of time to go and write regulations to actually see that the legislation works. They’ll be looking to see that those specifications, those implementation guides, have been built, so that they can point to them or recommend them in their regulations. So you’ve got this constant evolution of these implementation guides happening.

    Because of my involvement with Blue Button, I was asked to take part and actually author one of the guides. So I run one of the work groups, around Payer Data Exchange. It’s really dealing with how payers are going to exchange data between payers. When you take your health insurance, and at the end of the year you decide to move to a new health insurer, why shouldn’t your data go with you? Well, how are you going to make that happen? Payer Data Exchange also covers what happens when a provider comes knocking on the payer’s door.

    We published a first version of the guide a couple of years ago. Then, as we saw the new interoperability rule, the follow-on on prior authorization, bubbling up, we could see the likely things that were going to be pointed to. At the end of the last administration there was a rule they put out for comments, and then it was withdrawn, so we had some ideas of what was going to be required. So we built a new version of the guide, which has to go through the balloting process. I’m just in the process of resolving everything that came out of those ballots, to be able to publish that and then go to the next stage of updates, which will actually try and address what’s going to be in the new rule that has been put out for comments and that we’re now expecting to drop. It’s just an ongoing drive to really try to improve the specification and get people to adopt it.

    Ayush Jain [24:43]

    [inaudible] As the Chief Interoperability Officer, what does your day-to-day look like, apart from coming to podcasts and speaking at conferences?

    Mark Scrimshire [24:56]

    At Onyx, some of us are the core team that founded Onyx within our parent company. We were spun off three years ago as a startup. Whereas our parent company is very much in the [inaudible] space, we’re addressing health plans in the commercial marketplace.

    So my day-to-day role is a combination of spending a lot of time talking to chief medical information officers (CMIOs), chief information officers (CIOs) and the like about what this interoperability means and what new things we see coming that they need to be prepared for. At the same time, I’m basically paid to help write these specifications, so I spend some of my week actually working on things like the Da Vinci Project and supporting the CARIN Alliance and so forth.

    And then we have a whole suite of products, all built around FHIR interoperability. We saw the writing on the wall for the health plans, so we went out and partnered with Microsoft and built a cloud-based compliance solution for patient access. But what we actually delivered was a complete FHIR interoperability platform, so new things like clinical data acquisition [uncertain], cost transparency and prior authorization can all be built on top of this platform.

    We’ve spent the last three years deploying this for commercial health plans, like divisions of Humana, a couple of Blue Cross Blue Shield affiliates and a number of other smaller health plans, but we’re also currently in three different Medicaid state agencies, basically allowing patients to get access to their data. Now we’re gearing up around what payers are going to have to do to interact with providers to streamline prior authorization, and how payers could take advantage of the bulk APIs being made available via the EMRs to get more data to improve population health management.

    So that’s my day: driving the overall product development for Onyx and all of the different features we’re adding to our roadmap, supporting the specifications, and explaining what we’re doing to potential customers.

    Ayush Jain [27:35]

    Good. So Mark, tell us more about [inaudible]. The question is: what is the biggest challenge that you see in healthcare interoperability? [inaudible] Of course, a lot has been discussed and talked about, so I would rather ask you: what’s the one major thing that, according to you, needs to get started, because it poses most of the problems in terms of adoption?

    Mark Scrimshire [28:08]

    I think we’re seeing pretty good adoption by means of regulation. A large part of the industry, at least for certain product lines, primarily Medicare Advantage and such like, have had to implement this. The question is how they’ve gone about it. Is it just ticking the box for compliance, or are they really seeing that this can change the way they do things?

    This next round of regulation, I think, is really going to switch the industry from this checkbox compliance to “I see how this can transform my business.” We’re already starting to see that in California. At the end of last year, there were regulations that said any health plan operating in California has to implement the interoperability rule. So they’ve got to basically do what CMS did, but for all of their commercial employer plans, et cetera.

    And the question is, [inaudible] being a bit ahead [inaudible], in a different way.

    In some ways, I think in the US we started, certainly the health plans, with X12 for exchanging financial transactions. The Health Insurance Portability and Accountability Act (HIPAA) of 1996 requires financial transactions to go through that X12 standard. FHIR has really come along as a way where, instead of using things like Consolidated Clinical Document Architecture (C-CDA) document-based exchange, you can now go and ask for a specific piece of data and pull it yourself, if you’ve got the right permissions, rather than wait for somebody to send it to you. So we’re seeing that sort of evolution.

    If you’re not familiar with FHIR, it’s fully internet friendly. We would say it’s a RESTful interface: it’s JSON or XML. So there are lots of programmers out there that understand how to work with that, which means it’s much quicker and easier to develop for. So I think we’re going to see the use snowball.

    But it’s a case of really keeping on developing these implementation guides so that people have got a common way of doing things. Otherwise we’re just going to have islands of interoperability, and we really need it to be more pervasive.

    We still have issues to address in terms of how you discover those FHIR endpoints, so how do you know where to go, and then how do you get permission to access them? The tools we have in place today aren’t really scalable for that. That’s why part of my role in writing specs, I describe it semi-jokingly, is I’m creating the next generation of problems to solve. If we get a number of health plans supporting these APIs but they’re having to manually set up every one, then somebody’s going to say: this cannot go on, we have to find a better way. Which ultimately pushes us forward.

    So that’s really where this next phase of evolution is going to be: about discoverability and easier access. And I don’t necessarily mean more open access; we’ve still got to keep things secure. But how do I go and register a million doctors to a thousand health plans? That’s the sort of scale of challenge we’ve got to grapple with, and it requires us really acting as a community to get there. We can solve it with piece solutions, but we’ll still be in more or less the same problem as we were.

    Ayush Jain [32:15]

    Cool. So lastly, I’d like you to tell our audience a story, something that is close to your heart, where you had some experiences in health tech or in interoperability that you’re proud of and would like to share.

    Mark Scrimshire [32:37]

    I think just the achievement of CMS Blue Button has been far more than we ever expected. Folks like Niall Brennan, who was the chief data officer at CMS, originally put the pitch forward to say we need to bring someone in to do this. I remember back in 2014, we were in front of the executives at Health and Human Services saying this is what we were going to do. And then five years later, there’s regulation out there saying these health plans have to do it.

    At Onyx, we built a compliance tracker to see how many health plans have actually implemented, and there are literally hundreds, five or six hundred health plans in the US, that have implemented patient access. It’s just that they haven’t necessarily told their members that it’s there. So in many ways, that is the incredible achievement, [inaudible] built upon an incredible team that deployed this stuff and proved it really can work. That, ultimately, is really satisfying.

    Ayush Jain [34:03]

    Yeah, thank you so much, Mark. [inaudible] Thanks so much. Thanks a lot.

About Mark Scrimshire

At the time of recording (August 2023), Mark Scrimshire was Chief Interoperability Officer at Onyx Health, a company he describes as spun off from its parent (NewWave, which launched Onyx as a subsidiary in May 2020) three years earlier as a startup focused on health plans in the commercial market. Before that, he served as an entrepreneur in residence at the HHS IDEA Lab, where he worked on CMS Blue Button and helped launch Blue Button 2.0 in 2018, after working with 3M on taking its healthcare products and systems to the cloud. He helped define the Explanation of Benefit resource in the HL7 FHIR specification, has attended 20 to 30 HL7 connectathons, leads the HL7 work group on Payer Data Exchange, and contributes to the Da Vinci Project and the CARIN Alliance.

Building a product that has to talk to more than one EHR? Tell us what you’re connecting to, and we’ll walk you through whether a custom build or ConnectHealth gets you there faster.

Discover More Episodes

Let’s #Transform Healthcare,# Together.

Partner with us to design, build, and scale digital solutions that drive better outcomes.

Location

Global Tech Teams LLC, 525 Washington Blvd, Industrious at Newport Tower, Jersey City, NJ 07310, United States.

Contact

+1 408 786 5974
contact@mindbowser.com
BOOK A QUICK CONSULTATION

Have a Healthcare Project in Mind?

Let’s discuss your goals, workflows, and next steps in a focused consultation call.

Calendar icon Schedule a Call

Contact form