Over the years the Chesapeake Regional Information System for our Patients (CRISP), the regional health information exchange serving Maryland and the District of Columbia, has gradually added new services to help with quality reporting and care coordination. Now it is launching a research initiative to let participants use its data to help with clinical research.
“Early on in CRISP’s evolution there was a recognition that an HIE would be a natural part of a learning health system — that there are useful things to find out from data that is aggregated from multiple institutions,” said Ross Martin, M.D., program director for research and transformation at CRISP. He is responsible for establishing CRISP's new clinical and operational research capabilities, in support of participating organizations' research efforts. In an interview with Healthcare Informatics, Martin described some of the steps involved in setting up the research initiative.
CRISP is a successful HIE. It sends out millions of notifications each month, and an increasing number of providers are signing up for its clinical query portal. A few years ago, the HIE began working with the state to get the regulatory authority to have research as a permitted purpose, with several parameters around it. “Last summer we also did an amendment to our CRISP participation agreement so that providers can say,’yes, we know that the data we are sharing with CRISP can, under certain circumstances, also be used for research,’” he said. Since then, Martin helped establish a research subcommittee, whose members include:
• Dr. Christopher Chute (Chair) – Bloomberg Distinguished Professor of Health Informatics at Johns Hopkins University
• Ms. Del Price – Patient representative
• Dr. Robert Rudin – Research Scientist, RAND
• Dr. Kate Tracy – Associate Professor and Director of Clinical Translational Research and Informatics Center at the University of Maryland School of Medicine
• Dr. Neil Weissman – President of the MedStar Health Research Institute
The research subcommittee creates use cases and says, “Here is a type of study we want to be able to approve;” the clinical committee blesses that and says if somebody asks for data and it fits this use, you can go ahead without having to run it up the flagpole yet again, Martin explained. “That empowers the research subcommittee to approve data requests. If a data request doesn’t fit under a use case, we can’t approve it. But we might say we are ready to look at creating a new use case and approve it.”
In the first approved use case, the only way you can get access to data is for the patients to be part of a patient-consented research study. “You have to get consent at the individual level,” Martin explained. “Right now if your data is flowing through CRISP, no researcher can look at it, unless you sign an explicit document saying ‘I want to participate in this study’. If you think of what people want from our data, this fulfills one slice of it, but the larger opportunity is for de-identified data. However, we are not prepared yet to approve a use case like that.”
The holdup for doing anything beyond the single current use case is that CRISP has not yet started educating patients about their opportunity to have their data used in research. “Right now, if you look at what every doctor and hospital that participates has to tell their patients about CRISP, it says your data could be used for your care, for public health reporting, and operations. But there is nothing in there that says we could use it for research,” he said. “Because we are an opt-out state, until we give patients the opportunity to say, ‘I’m sorry, but I don’t want my data used for research,’ we really can’t go forward with broader use cases. So we are taking very incremental baby steps and only doing studies where the patients whose data could be used are actually signing up for it.”
The researchers are insisting on this approach. “They do not want patients realizing that researchers are using their data and they never even knew about it. They want to be good communicators about opportunities,” Martin said. CRISP also is working on a “data router” capability to allow people to more granularly control their opt-outs. They could say they want it used for clinical purposes with certain providers, but they don’t want to participate in research.
The one study that has been approved so far has actually been going on at Johns Hopkins University for more than 30 years. They have been looking at IV drug users in the Baltimore area since the 1980s. Martin explained that the study has a staff that goes to a lot of trouble to track down medical records from other facilities outside Hopkins. They will see these people every six months, and ask if they have been hospitalized. And the subject might say they were at Holy Cross on Oct. 13. If the researchers do a data request to Holy Cross, and the patient had the date wrong, Holy Cross will say that person is not in our records and they won’t give them any data. But because researchers will now have access to CRISP’s encounter notification service, they can know exactly when these enrolled patients show up in somebody else’s emergency room. And they can look in the query portal and get information on that patient. And if they still don’t have enough information from the query portal, they can request data from the hospital. “Their efficiency is going to be much greater,” Martin said. “Our ability to give them a lot of data that now takes them an incredible amount of effort to get is going to save them a lot of time and money and potentially show them more data than they would have gotten otherwise.”
Beyond patient consent and awareness, another obstacle is that CRISP is essentially plumbing, Martin said. “We are not designed to be a data lake of the sort that most researchers would use. We haven’t set up our data infrastructure to accumulate vast quantities of data, to be able to share and look at in aggregate. We show people data on a patient. We are set up technically to be more like pipes than a repository. It is not a trivial thing for us to serve up a big set of data. And frankly we are very busy with our prime directive of supporting clinical care.”
CRISP is working to change its data architecture to make it more streamlined, so that every time someone asks for a new type of data it is not struggling to create that one-off solution. “We have a data architect working on it, but it is going to take a year or a year and a half for a lot of this to materialize,” Martin said. The driver is better clinical care, he added, but a byproduct of that will be the ability to support research in a more fundamental way.
Eventually CRISP will be able to deliver a de-identified data set of all post-prostatectomy patients within a geographic area over a span of time. “We have a ways to go before we can technically and logistically support that, and we don’t want this to be a distraction from our prime directive,” he said. “The nice thing about the uses we are approving now is that we are giving them access to the same tools we give to clinicians and filtering it to their research subject patient panel.”
CRISP has set a goal of supporting 10 research studies by June 30, and Martin believes the demand is there to make it happen, even given the current parameters.
Although offering new services can be a way for HIEs to sustain themselves, Martin noted that the research initiative is not a money-maker for CRISP. “One of the parameters is that the only people who can make data requests to us are participants in CRISP,” he said. “If we tried to turn this into a sustainability center, we would basically be taxing our own users. We didn’t think that was a great approach. We are charging and doing it as a cost-recovery model. Once we have it working, we do want to make sure the incremental cost of adding a research unit and servicing it is covered.”