March 19, 2021

Next level data delivery

Illustration by Keith Negley

“I want to change the world, not just at Vanderbilt.”

In October 2019, in the course of delivering an online seminar sponsored by the National Institutes of Health (NIH), Paul Harris, PhD, casually included this aspiration. His talk was part of a series of NIH “webinars” examining how to involve more hospitals and health systems in clinical trials. Harris was describing an opportunity that arose in 2017 to integrate research software from Vanderbilt University Medical Center with electronic health record (EHR) systems used around the world.

Given the specialized context of his talk, when Harris spoke of changing the world, he wasn’t overstating matters. His audience of clinical trialists and informatics insiders would have implicitly grasped that the tools Harris was describing could save untold research dollars and perhaps hasten the future of medicine. Investigators have come to see that, despite how dirty, sparse, variable and suspect EHR data can be from a research perspective, the road to precision medicine does in fact go through the electronic health record. Handled properly, patient data, in the aggregate, is seen as providing a window into disease risk and what makes patient outcomes better or worse — all the more if it can be joined with lifestyle, genomic and biomarker data.

And then there’s the continuous spate of clinical trial data flowing through the EHR. According to one market research firm, global spending on clinical trials is estimated to reach $70 billion by 2027, and a sizeable portion of this expense goes for manual data extraction from the EHR.

To an outside observer it might seem simple and obvious, this notion of a user-friendly, all-purpose clinical research application hooked to the EHR, siphoning patient data, the instant it arises, into an institution’s myriad clinical and translational research projects. There were, however, substantial barriers to bringing it about, having only partly to do with what society has come to recognize as the inherent sensitivity of personal health information (enshrined in federal statute effective since 2001).

This story begins where many stories about biomedical research and informatics begin, with acronyms and abbreviations — REDCap and FHIR being two. Conceived and developed by Harris at VUMC and launched in 2004, REDCap, or Research Electronic Data Capture, is a secure web application geared to support online and offline data capture for clinical and translational research. In 2007, Rob Taylor became REDCap’s lead developer and in 2015 became its manager of application development.

Starting in 2006, Vanderbilt made REDCap freely available for non-commercial use, and today it’s used in 141 countries at more than 4,800 organizations comprising the REDCap Consortium.

Paul Harris, PhD, conceived and developed REDCap in 2004. It’s now used in 141 countries. Photo by Joe Howell.

“If you go way back to the very beginning of REDCap,” Harris said, “one of the most common questions all along the way at Vanderbilt has been, could we get data out of the electronic health record into REDCap faster?” The solution at VUMC and several like-minded centers was to bring EHR data into REDCap in nightly batches from a research data warehouse, via a middleware application built for this purpose. But not all health care organizations have a research data warehouse. While this worked well at VUMC, most REDCap Consortium organizations took a pass on this solution or lacked the wherewithal to implement it.

FHIR (“fire” to its friends) stands for Fast Healthcare Interoperability Resources. It’s an electronic data exchange standard introduced in 2014 and widely published beginning in 2017. It allows different systems to talk to each other and share structured data — discrete, predefined data, as opposed to text or images. The EHR is, of course, rife with such data: demographics, lab results, vital signs, medications, allergies, patient problems. EHR vendors have adopted FHIR en masse and scientists have seized upon it as an accelerant for clinical research.

On Nov. 2, 2017, VUMC switched from a homegrown EHR to software from health information technology behemoth Epic Systems Corp., based in Verona, Wisconsin. This change at VUMC, coupled with FHIR’s rapid ascendance, prompted Harris and the REDCap team to modify the application: data would now flow directly into research projects from an institution’s FHIR-compliant EHR servers, based, as always, on privacy protections and well-defined permissions granted to each researcher-user.

This innovation is called REDCap Clinical Data Interoperability Services (CDIS), or simply REDCap on FHIR. It opens the way for organizations across the REDCap Consortium to put EHR data to secondary uses far more efficiently, not only for research, but for quality, safety and operations improvement.

 

New freedoms

Now that CDIS is built and has begun delivering the goods — securely and relatively effortlessly, with research teams able to use it themselves and no longer reduced to pleading for data from beleaguered health IT departments, and with patient privacy all the better because research data coordinators no longer need to roam through charts for data — it has some experts thinking differently about what’s possible in the realm of clinical research.

“What we’ve created is a solution where we get real-time data out of the EHR and deliver it to real-time screens that we, as researchers, control. And now you’re in that space, the if-you-think-it-you-can-do-it space,” said Harris, professor of Biomedical Informatics and director of the Office of Research Informatics.

Chris Lindsell, PhD, is working with Harris to chart a future for REDCap on FHIR technology. Photo by Susan Urmy.

New freedoms presented by this data interoperability are paramount for Chris Lindsell, PhD, professor of Biostatistics at VUMC. “If you’re working inside the EHR environment, you’re constrained by health IT, by governance, by all the rules and regulations and documentation,” Lindsell said. “The moment you as a researcher can interoperate with the EHR from outside the system proper, you no longer have to live within the health IT model, and you become totally unconstrained.”

Harris is working with Lindsell and others to chart a future for this technology. To facilitate clinical trial recruitment, when a patient who happens to appear to meet a trial’s inclusion criteria enters the hospital or clinic or logs into the institution’s patient portal, the EHR can be made to sense this and immediately initiate the recruitment process via REDCap. In the case of pragmatic trials that don’t require patient consent — because you’re simply comparing real-world outcomes for accepted but contrasting treatments that are already in wide use — researchers are in a position to go straight from automated screening in the EHR to randomization and automated data collection in REDCap without passing go. To facilitate multisite trials, de-identified EHR data could be made to flow from various far flung instances of REDCap to an umbrella REDCap database at a national coordinating center. These scenarios have drawn interest from the NIH, and some of them are already in testing at multiple sites.

 

Expanding in the U.S.

But what are most researchers doing today with REDCap on FHIR?

Some 82 projects at VUMC have used the new tools. Working with Epic, Harris and the REDCap team created a plug-in that makes integration of REDCap with the Epic EHR especially simple. (It’s now the most popular app in the research category in the Epic App Orchard.) To date, 70 institutions have initiated adoption of CDIS, including 62 Epic customers and eight customers of another large EHR vendor, Cerner Corporation, based in North Kansas City, Missouri.

First CDIS use case: a team needs selected EHR data for a clinical trial or health care improvement project. Having mapped the appropriate EHR fields to a case report template created in REDCap, a data coordinator imports data into the project straight from EHR servers with a few clicks, one patient at a time, based on medical record numbers and dates of service. Data exchange is immediate, and the connection can be left open to automatically gather data from future dates of service. This REDCap service is called Clinical Data Pull (CDP).

Second CDIS use case: for a specialized research platform, patient registry or retrospective study, a team has been given authority to gather a wide range of EHR data for all dates of service from hundreds or thousands of patients. They can have their data in minutes, and, again, the connection can be left open to gather future data on the cohort. Using a pick list, a data coordinator outlines the needed data, then gives REDCap the list of medical record numbers for the research cohort (as easily furnished, perhaps, by the EHR’s search filter). This REDCap service is called Clinical Data Mart (CDM).

“There’s no coding or programming knowledge needed,” said one enthusiastic CDP user, Seibert Tregoning, senior clinical research data specialist with the Vanderbilt Coordinating Center. “If there are, say, 20 or 30 data points per person per study visit, all those data can be pulled over with just a few clicks. Before, a research coordinator would need to click into participants’ health records one by one and transcribe or copy and paste the data. All that navigation in the EHR has been removed.”

The University of Texas Southwestern Medical Center in Dallas was an early adopter of these tools.

“Prior to this, for researchers needing clinical data, we had to hand those requests to a team of programmers,” said Teresa Bosler, an information resources manager at UT Southwestern. “The team at Vanderbilt is phenomenal; they have really changed clinical research.” In all, 10 active projects at UT Southwestern use CDIS.

As UT Southwestern language translators work with patients and care teams, they pull patient demographics from the EHR into a REDCap data base, the easier to log case-by-case details on their services.

An oral surgery team at UT Southwestern sought to survey several hundred patients who’d received a particular procedure. They used the data mart feature to pull all the contact information into REDCap in one go. (REDCap also powered the eventual survey.)

At Weill Cornell Medicine in New York City, HIV researcher Marshall Glesby, MD, PhD, has created a specialized research platform linking subjects’ survey responses to their EHR data. He says REDCap on FHIR, in combination with algorithms that identify comorbidities, is speeding discovery for his team, who no longer need to look in the EHR at all. “These things together really saved, undoubtedly, hundreds of hours of having to review patient charts to extract the information,” Glesby said, “and we probably also have more accurate information.”

Tom Campion, PhD, directs research informatics at Weill Cornell. “We saw these new tools as a great opportunity to continue to bring together clinical and research workflows,” said Campion (who earned his doctorate in biomedical informatics at Vanderbilt). “Getting data into our studies from our EHR in real time was huge for our investigators.”

Weill Cornell has 13 projects underway that use REDCap on FHIR (and 26 that use REDCap’s earlier, non-FHIR data pull method).

In the cardiology clinic at Yale Medicine in New Haven, Connecticut, a cardiologist uses clinical data pull to simplify research subject recruitment, opening REDCap from within the EHR during the patient visit and pulling the study data over with a couple of clicks. This is the first project supported by the Yale Center for Clinical Investigation using CDIS.

“It’s seen here as a way to cut costs and reduce errors,” said Andrew Poppe, PhD, a research data analyst at Yale. He adds, “It eliminates the ‘swivel chair problem,’” a reference to the common scenario of research data coordinators swiveling between two computer screens, transcribing or copying and pasting data. “Somewhere in that swivel is the opportunity for errors.”

Alex Cheng, PhD, research assistant professor of Biomedical Informatics, is helping to guide continued implementation of REDCap on FHIR at VUMC and its dissemination to other institutions.

“We’re seeing a steady uptake of CDIS among REDCap consortium members in the U.S. and some international interest as well,” Cheng said. “Many VUMC researchers have realized the improvements to efficiency, timeliness and accuracy of capturing EHR data through CDIS. As word spreads, we believe implementations at other institutions will accelerate.”

The REDCap team is housed in the Vanderbilt Institute for Clinical and Translational Research.