The leader of CORHIO, one of the most progressive health information exchange (HIE) organizations in the country, continue to innovate forward across a broad range of areas. The Denver-based CORHIO already connects 65 hospitals across Colorado—virtually all of the inpatient community and academic facilities in the state—and connects around 5,000 physicians statewide as well.
As the organization’s website notes, “CORHIO empowers people, providers and communities by providing the information they need to improve health. Our advanced health information exchange (HIE) technology, data analytics tools and expert consulting help healthcare providers access information that saves lives, streamlines care coordination, reduces costs, and improves clinical outcomes for millions of people.”
Recently, CORHIO’s leaders, including Morgan Honea, the HIE’s president and CEO, and Mark Carlson, its director of product management, have been pushing ahead to connect providers across the state both more broadly and more deeply—extending out into the behavioral health sphere as well as facilitating the sharing of more granular data across Colorado, through data normalization work. Honea and Carlson spoke recently with Healthcare Informatics Editor-in-Chief Mark Hagland regarding their current initiatives. Below are excerpts from that interview.
You’ve been expanding some of your core data-sharing capabilities of late, correct?
Mark Carlson: Yes; we certainly do have some activity and infrastructure that we’ve been building out. I’ll focus in on clinical and population health first. One area we had identified a couple of years ago in terms of being able to generate information for population health, at the state level, or in partnership with ACOs, came about as the state pushed forward an initiative called “regional care collaboratives” with ACOs. As part of that initiative, we did work on packaging and bundling notifications around ED visits and hospitalizations and discharges, for providers, as well as helping smaller physician practices in that area. And we’ve been looking at expanding out that concept around clinical indicators, initially focusing on labs.
We have 65 hospitals sending data into CORHIO, and we had 30-plus representations as to how a hemoglobin a1c might be represented, in terms of vocabulary and coding. So we used NLP [natural language processing] to help us with that, to help move forward in disease management in areas like diabetes. We’ve also focused on another use case with our Department of Health in Colorado, around an influenza use case, where we’re able to flag a positive influenza use test and track for an inpatient admit that occurs within 48 hours, to map the cost of care as well as the ability to access supporting resources that hopefully would avert an inpatient admission.
That’s what we’re working on—normalization across general labs and clinical metrics; and as we expand our data types, we’re expanding towards social determinants, as well as labs that extend beyond the general labs.
Morgan Honea: I agree with everything that Mark said. I would just add that this is, really, in my opinion, kind of a second evolution around the interoperability question. We’ve got a tremendous HIE with tremendous participation here in Colorado. The important fact is that, after laying the infrastructure for a statewide HIE, it next becomes imperative move into normalization across data sources, so that you’re not changing vocabularies or nomenclature.
That sounds like “HIE 2.0,” in terms of the advanced work, doesn’t it?
Carlson: That’ll work.
What’s next or top priority now for providers?
Honea: Our top priority is to continue to expand the type of data available in the HIE. In that context, we’re facing up to the incredible challenge of continuing to integrate behavioral data into the system. We’re also working with state agencies, to make sure that folks are getting the best care coordination for the best outcomes possible. And probably the highest demand from our clients is fewer queries and more push notifications and types of functionality, greater integration into EHRs [electronic health records] and other population health-type tools, with really clean, neatly packaged data, which is where this conversation becomes more important, because as Mark said, with hemoglobin a1c, things get very messy as the volume of the data grows, if you’re constantly having to clean up the data. So providing the data in interoperable, easily usable ways, is a top priority.
Carlson: And you have to follow the money in terms of reimbursable events and other value-based areas. So as we improve our inbound CCDA-type activities, we want to improve the quality of submission at the level of formatting as well as presence of charted measures, as being able to format and report those out, from practices, including around broader performance measures.
With regard to the capture and sharing of data, are you making any use of artificial intelligence? And where is that going?
Carlson: One of our core initiatives is, how do we become more situationally aware? I’ve looked at FHIR as a path forward, in that context Whereas the CCDA is a blunt-force instrument, FHIR provides the opportunity to be a lot more precise in packaging and bundling data. For example, we’ve been working on a use case for an anesthesiology group. They want to see problems, meds, last treatments, discharge summary, they don’t want to see everything. FHIR helps us to bundle and package data, and then via an API connection, they can receive more precise information that meets their needs, rather than via all-encompassing data. More targeted, based on clinical needs.
Honea: The ability to get down the discrete data level, understand the data points and bundle and share them, is where I think things are going. A CCDA is a big, narrative summary of an encounter, and doesn’t get down to that level of granularity.
Carlson: In the media right now, there’s been a lot of discussion around where the next steps of IBM Watson should be. And we’ve had this discussion with a lot of vendors in the past, where we’ve been introduced to some very compelling functionality; but then some wonderfully designed tools absolutely choke on some of the variability of the nomenclature in the data. And that prevents us from getting to advancing the Quadruple Aim. Those learnings and market information that we’ve gathered over time, indicated our absolute need to partner with organizations that have a foundation for creating mappings that are clinically valid and reliable and backed with the expertise behind it. That can help us get to the population health insights that you’re referencing when you mention AI.
Do you think you’ll be able to incorporate some social determinants of health data into what you’re sharing?
Honea: That’s an area where I’m spending some of my time now. I have no doubt that we’ll run into the same challenges with local code sets and varying terminologies with that type of data that we’ve had with clinical data. I don’t see that that process is strictly limited to hospital and clinic data; I think it will go across all sorts of systems; and when we share from one program to another and one type of data system to another, we’ll face the same types of challenges and requirements for data standardization. So we’ll probably rinse and repeat every time we go out and get another data point.
Carlson: We are working with United Way 211, understanding how their community resources and curated content and partnerships are working, and getting insights from diabetic prevention programs and food banks—the data quality is as variable as some of the source organizations involved. I think this opens up a whole new opportunity for whole-person care, but it will pose some of the types of data normalization and use challenges as clinical data.
How do you see the next few years evolving forward at CORHIO?
Carlson: We’ve touched on a lot of priorities—ECQM work… our learnings in various areas. It’s a big lift to ingest the CCDA documents and get consistency at the data level. Our partner organizations continue to work with us and with Wolters Kluwer, to work on various types of data together. When we spoke at the HIMSS Conference earlier this year, Morgan and I talked a lot about data normalization work and about data visualization, and about being able to visualize risk across counties and the state, to identify pockets of need. And in that context, the social determinants data will help us understand where the food deserts are, and where high levels of chronic diabetics live. We have a number of mountain and rural communities that are fairly isolated, so our opportunities to impact that, are large, but so are the needs, and thus, we need to address data quality issues.
Honea: I agree with everything that Mark said. We’ve got this never-ending effort to include programmatic elements, site-specific elements, into the HIE, every kind of element—that work will never end. But I’m also continually focused on the question of how we as a state, with only 5 million people, can identify ways to leverage the infrastructure built with significant investment at the federal, state, and local levels, to advance our overall HIE efforts as a state, and minimize the risk of continually building new silos of data that will just require new efforts in the same fashion? How do we improve coordination when folks are moving across different geographies or service areas, without rebuilding existing infrastructure? How do we partner with communities to get the biggest bang for our buck? That requires a lot of planning and coordination and collaboration.
Carlson: For HIEs to provide value, Morgan and I often say, it’s data versus documents. Document exchange has a very valuable place in the broader landscape, but where the HIEs are differentiating themselves is at local-level attention and relationships and meeting community needs, and where we can operate at the data level to provide the insights to drive patient care quality forward.