Linking Research and Patient Care: the University of Florida’s Bold Internet2 Adventure

I basically told you my end; my background is nuclear physics and computational chemistry. I’ve been working for many years to build infrastructure for chemists, physicists, and engineers, and now I’m responsible for building these high-performing networks and storage systems. In the last year, we’ve made a push forward to get to the top 10 in the nation as far as infrastructure of research computing is concerned. And now that it’s built, people want to use it to work with protected health information.

And Gigi has been working for multiple years on building infrastructure and protocols, for protected health information.

Gigi Lipori: So we’ve been working for a long time on warehousing data and providing it in a way that researchers can have direct access to it in HIPAA-compliant fashion. So we’ve gone along slowly and steadily to reassure people that we’re doing it in a HIPAA-compliant way. So as much as it’s a technical exercise to bring together research data, hospital data, anesthesia data, etc., in one place and work with it, the other part of it is sharing the data with researchers so they can form a hypothesis, and subsequently, that they can go back to the IRB (Institutional Review Board for the Protection of Human Subjects) for approval to link research to patient care.

What we do is NIH-funded [funded by the National Institutes of Health], and is called I2B2 (Integrating Biology to the Bedside), so they do their queries there, find a cohort and say to the IRB, I want to do this study on this set of patients, say 300 or 500 of patients, and then the IRB determines whether it’s appropriate, and once they do, we can broker the data to them. So we gather data and put it into this integrated data repository; we allow the researchers to query this using an I2B2 tool, in a way that’s protected and blinded. When they have a good hypothesis, they go back to the IRB and ask for permission to get protected data; then, upon receiving permission, we have an honest-broker process—there are five people in the university certified to give data back to them, and they’re all in my group. We broker that data back to them in a virtual environment that’s also safe, so they can’t drag that protected data anywhere else. But the beauty of it is that, before, they couldn’t explore these hypotheses safely, so they would just blindly go in and inspect medical records, which is dangerous, because you could end up seeing things you weren’t supposed to. This is safer for the human subjects, and it’s better for institution, because we can provide the data more efficiently.

Erik: This is very complex; it’s not something any institution can do in just a few months. And at Internet2, we were having these discussions; and everyone knows that it’s very, very hard, and we’ve been working on this for years. But we feel we’ve made some really good progress; and we hope that in the next couple of years, we will be able to scale this to the size that many of our health researchers are dreaming of doing. But you have to build all these details, make sure they all fit, and that everybody’s in agreement; because otherwise, if someone disagrees with it, the whole thing crashes again.

And this kind of development work could crash your EHR, correct?

Gigi: Yes, it could crash your EHR; and when you’re trying to look across a million patients to try to find a particular phenotype, there’s definitely an art to engineering all of this.

When will the average academic medical center be able to do all this?

Erik: I think it will be sooner rather than later, because there’s incredible pressure to get this done. There are several other universities doing something similar to what we’re doing. And the creation of an Internet2 data path will allow institutions to talk to each other; and they’ll quickly figure out ways to share information and expertise; we will gladly do that. This could turn into a watershed moment. And that is of course what Internet2 as an organization is hoping, that organizations can learn from each other and make progress. And people will write white papers and documents and share software that you can download. So I think it could be within the next couple of years; definitely by 2020, this is going to be standard.

Do you see genomic data use as being accelerated soon?

Gigi: You have the holy grail of phenotype-genotype matching, which is what people are after. And we pull in data from personalized medicine, so that we can have that genomic data available to do some genotype-to-phenotype matching, particularly around Clopidogrel (Plavix), so they’re able to do that; and if they’re looking for patients and trying to find all the breast cancer patients with a particular phenotype and who have allowed their specimens to be used, we can address that as well. It’s all good when you’ve got a cohort of 500 patients, but you need 50 samples in storage.