The Indiana Health Information Exchange, the U.S.'s largest HIE and one of the longest running, serves more than 19,000 doctors and 70 hospitals. It has data on more than 7 million patients. Its CEO, Dr. Marc Overhage, recently talked with InformationWeek senior writer Marianne Kolbasuk McGee about the tech obstacles HIEs face.
InformationWeek: What are the biggest technology issues for HIEs?
Overhage: One of the biggest technological challenges is creating interoperable interfaces with the source systems--for instance, the laboratory system that generates the clinical test results or the electronic medical record in the physician's office. While there are standards and the way to do it is pretty well understood, there is a limited workforce in which to find the people.
InformationWeek: What's the bottom line on developing interfaces?
Overhage: It comes down to the skilled workforce. There are different ways to create the interfaces, but you don't always pick the easiest path because there are a lot of unknowns and moving parts. For example, we've had the situation where, after we've decided how to construct the interfaces, they've changed the lab system. So you end up starting over. It's not unknown how to do the work. It's just a lot of work to make these systems interoperable ... because there are a lot of systems, and they all have work that has to be done to integrate them.
InformationWeek: What are the biggest tech challenges for more mature HIEs?
Overhage: We're beginning to feel the strains of scale. We crossed the mark not long ago of having 7 million patients' data. Using the federated repository model, we've started having to work harder on performance issues and things like that than we used to. We're doing some collaboration with Children's Hospital in Boston on database scaling ... On the other end of the spectrum, where people are going to hit the wall as they grow and expand, is that it just gets big.
InformationWeek: How "real time" is data about a patient in an HIE?
Overhage: The vast majority of data that comes from local laboratories or hospitals is essentially real time, though that's not a computer science definition. Within seconds of results being available from a laboratory system, that data is available to the exchange. There are examples of data being delivered in batches. For instance, a lab in Texas delivers results in batches monthly.
InformationWeek: What do the physician offices contribute?
Overhage: The data that flows from physician offices varies widely. We have practices where everything essentially flows in real time. For instance, if they write an order for a complete blood count for a patient, it's available within seconds in the repository. That's one extreme. ... There are a large number of practices that get data from the exchange, but we get zero data from them because we don't have interfaces to their appointment scheduling system or their electronic health record, if they have one.
InformationWeek: What about the security challenge?
Overhage: Increasingly, questions like patient matching or patient identity management aren't technological challenges anymore. ... Things like managing security are pretty well known now. When I think about technological challenges, I think more of the areas where there's still a bit of the Wild West, where we don't know how or where to do it in an efficient or effective fashion. That helps me crystallize the problem when it comes to interfaces: You can do it, but it's not efficient or effective to scale across the 4,300 practices that participate in our exchange. How do you scale across 70-something hospitals and labs? And how do you get interfaces to those 5,000 organizations in an affordable, efficient fashion? That's the key technological challenge.
InformationWeek: Who's responsible for building those interfaces? Is that the burden of the participants or the HIE organization?
Overhage: The answer is both. Work has to be done typically at the source system--say, the lab system. They may have to license or purchase that interface tool and capabilities from their vendor. They have to implement the interfaces. Then they have to work with the exchange to create secure exchange capabilities. In our case, that's a VPN. They have to work with the exchange to make sure the messages--the results that the vendors' products produce--are interpretable.
The exchange then does a lot of work to standardize or normalize the data to make sure that salt or sodium is called the same thing in each hospital or laboratory. That requires work from the source system to validate that data has been correctly translated and identified.
InformationWeek: How long will it take to solve the interface challenge?
Overhage: It's a matter of time until we see the national standards like RxNorm for medications and LOINC for test results trickle down and get embedded in the source systems. But it will be a significant challenge--a 15-year process. These systems have a long shelf life. When you invest in a radiology system, those are big investments, and you don't likely replace them. So it will take time.