A decade ago, Isaac Kohane, M.D., Ph.D., Marion V. Nelson Professor & Chair, Biomedical Informatics, Harvard Medical School, decided to tackle a major problem: each Harvard-affiliated hospital operated its own unique, siloed system of electronic storage and data management.
Kohane, a physician and data information engineer, wanted to link the disparate systems to improve clinical research. It wouldn’t be easy, requiring a massive amount of funding, technical expertise and an epic collaboration between sometimes competitive hospitals, but he hoped that if his team could pull it off, the Harvard model might pave the way for the future of electronic health data sharing in the U.S.
In May 2008, he and his colleagues received a Clinical and Translational Science Award from the NIH to launch the Shared Health Research Information Network, or SHRINE, which built upon an existing software system called i2b2. Today, SHRINE is among the largest efforts to streamline and share data across hospital institutions, and an example of the so-called “federated” model that many experts believe represents the immediate future of electronic health records (EHR) integration in the U.S.
Over the past decade, leading researchers, clinicians and policymakers have proposed various hypothetical models to address the country’s woefully disconnected EHR systems, which, if unified, could become a treasure trove for new research and fix gaps in clinical care. One “utopian” vision for the future imagines that all medical information will be stored in the cloud, “with each patient in control and accessible to his or her own information via online portals,” explained Richard V. Milani, M.D., chief clinical transformation officer, Ochsner Health System.
Another vision, borrowed from smaller countries like Scotland, imagines a future in which the government manages one central EHR database, explained John Halamka, M.D., MS, is chief information officer of the Beth Israel Deaconess Medical Center.
In the U.S., however, with its diverse population and complex health system, many experts agree that the immediate next steps will be to build a network of federated models similar to SHRINE. Under a federated system, each hospital will continue to manage and store data locally. However, layered on top of each local EHR will be a sharing architecture by which other institutions can ask for, and receive, data from their peers, delivered in a HIPAA-compliant and de-identified way.
“Think of it as a network of networks,” said Robert Califf, M.D., vice chancellor for Clinical and Translational Research, Duke Clinical Research Institute, and former FDA Commissioner. “I think there’s pretty good agreement on the future world, but the tricky part is how long it will take,” Califf said.
Fueled by the Affordable Care Act’s (ACA) mandates for EHRs, today there are countless siloed EHR systems across the country, with the FDA and other federal agencies taking a leading role to encourage various hospital networks, life science companies and healthcare IT companies to begin building interoperable systems.
A fully interoperable EHR that spans across multiple health networks would be a boon to clinical research, potentially saving millions for each new study because of such a system’s ability to identify patient demographics, determine feasibility of recruitment and quickly test a research hypothesis. All of this could take place in a real-world clinical research setting, without the need to build a separate research ecosystem.
This vision is fundamentally distinct from the way clinical trials are run today, said Califf. “Particularly with industry-funded trials, you build a neighborhood and, after living in it for a year, tear it down again,” he said. “Later, if someone else has a similar question, it has to be built up again.”
Although SHRINE began as a pact between three major Harvard health centers—Beth Israel Deaconess Medical Center, Children’s Hospital Boston and Partners HealthCare System—it serves as a useful model for efforts to build even larger federated systems across multiple institutions. In addition, the challenges that Kohane faced—both technical and cultural—are likely to be echoed in future efforts. Eventually, experts hope that multiple federated health systems will integrate into a nation-wide collaborative to which nearly all hospital and healthcare systems belong.
Much of the activity around EHR integration began under President Barack Obama’s administration, which created financial and legislative incentives to build research collaborations across organizations, explained Halamka.
“President Obama was very influenced by Silicon Valley,” said Califf. “As well, Vice President [Joe] Biden, whose son died of a brain tumor, was shocked when he went to the major cancer centers and each had its own insular source of data,” Califf said.
Establishing EHRs became a mandate of the 2010 Affordable Care Act (ACA), which also established Meaningful Use incentives to encourage health networks to leverage EHRs to reduce waste in the healthcare system.
“Healthcare is a $3.2 trillion industry, and we might be able to reduce up to a third of that by having information readily accessible when patients and doctors need it,” Milani said.
The ACA also established the non-governmental organization Patient-Centered Outcomes Research Institute (PCORI), which curates a network of hubs that use EHRs to conduct observational studies and clinical trials embedded within health networks. Today, PCORnet is comprised of 33 partner hubs, at least one of which, called SCILHS (Scalable Collaborative Infrastructure for a Learning Health System), makes use of the SHRINE system.
Despite this progress, most systems are still siloed. “Following the ACA, many grants were given [to develop] electronic medical record systems, but there is a lack of standardization,” said Rachel E. Sherman, M.D., MPH, FDA’s associate deputy commissioner for Medical Products and Tobacco. “In a sense, the field is still immature,” she said.
Back in 2008, before Harvard’s Kohane received the Clinical Translation and Science Award, he knew he would first need to iron out the technical challenges of sharing data between Harvard hospitals. “We went into this fast ‘skunk works’ plan to show it was technically possible,” he said. “Only then would be to get credibility from the [Harvard] institutions.”
Even within the Harvard network, there were many obstacles standing in the way of data sharing. “When we talk about interoperability,” said Halamka, “you need a common vocabulary, a structure and data model and some mechanism of sending data from place to place with integrity, so it’s not modified along the way.”
Complicating those technical requirements are the layers of legislation designed to protect patients involved in research projects. “Because of the history of human experimentation and the problems that have come with it, there are a set of [regulations] that create inefficiencies [when it comes to data sharing],” Califf explained.
With SHRINE, Kohane and his colleagues were able to address these concerns by de-identifying patient information, establishing electronic audit trials, requiring each user to complete ethics trainings and a number of other safety safeguards.
The next major roadblock was inconsistent medical terminology. “For a variety of reasons in health and medicine, we never agreed on the same names for the same terms,” explained Califf. “If you say, ‘I have someone with chest pain,’ that means different things to different people.”
The terminology problem can be solved by establishing industry guidelines and using data science to bridge the gap between similar but non-identical terms. For example, Califf pointed out, the FDA established the Clinical Data Interchange Standards Consortium (CDISC), which created an effective system to collect clinical trial information in a standardized way.
In 2008, the terminology problem created a considerable headache for the engineers behind SHRINE. However, in recent years the Centers for Medicare and Medicaid Services (CMS) and the Office of the National Coordinator (ONC) addressed this issue with the Stage 3 Meaningful Use criteria. This criteria, which physicians and hospitals are required to adopt by 2018, outlines a unified system called the Fast Healthcare Interoperability Resources, commonly known as the FHIR standards.
The FHIR standards “arose out of a general dissatisfaction with the previous standards, and arose at a convenient time in which policy mandates like the 21st Century Cures Act are shifting incentives” toward common terminology, said Kohane. “I do believe there’s an urgency to stimulate interoperability, and that is why these FHIR standards have been developed,” Milani said.
While “terminology is critical [to build interoperability,] it is not sufficient,” said Sherman. Kohane agreed that the terminology roadblock is “not trivial” but is “manageable. While building SHRINE, he added, “the technical stuff was groundbreaking but in my mind the least of the challenges.”
The biggest roadblock standing in the way of an interoperable healthcare system is institutional commitment, Kohane said. Again, this challenge was echoed in the early days of SHRINE, in which Kohane convened dozens of meetings to persuade Harvard hospital executives to participate.
Halamka, a key member of the SHRINE team, credits Kohane’s tenacity and “totally apolitical” nature for the ultimate success of the network. “It was very hard to convince people to centralize their data in a location they couldn’t control,” explained Halamka, which is exactly why SHRINE proposed a federated system that would allow each hub to retain local data control, while simultaneously submitting to an architecture that would map data to a “common ontology.”
“One of the things that makes the change so hard is that there is not enough in it for any big entity to make big changes,” said Califf. Outside of hospital networks, Califf added, large healthcare IT vendors have created proprietary software systems and have little incentive to collaborate.
“Right now, Epic systems are very interoperable, but the real problem is moving from one vendor to another,” Milani said. In 2010, Kohane and his colleague Joshua C Mandel built another piece of software employing federated principles called SMART, which made use of the FHIR standard and allowed participating institutions to download interoperability software in a marketplace.
Although an App Store-like architecture seemed to be at odds with the highly propriety health IT systems developed by EHR vendors like Epic and Cerner, some healthcare IT vendors have begun to adopt the SMART-like systems, Kohane said.
For example, Kohane believes one of the reasons Cerner won a recent data hosting contract from the Department of Defense was because the company promised to allow innovation to flourish through an add-on app system akin to SMART. Ultimately, customer demand will encourage large companies like Cerner and Epic to adopt interoperable systems, said Califf.
Meanwhile, the pharmaceutical, biotech and medical device industries are reluctant to deviate from the tried-and-true research and development process.
“With collaboration, there is some loss of autonomy,” said the FDA’s Sherman. “To change business practices, there have to be incentives. In the long run, you have to convince [the industry] that this will be more effective, more efficient, better for formulary positioning and better for patients.”
Another important group that needs to be brought to the table is the healthcare practitioner, said Sherman. “When we introduced electronic medical records, it did not help them improve their practices,” she said. To encourage physician support of interoperability, “we need to prove it will create an easier workflow and more time with patients.”
FDA and EvGen
Today, the FDA is taking a leading role in guiding the industry’s broad vision toward a “learning health system,” in which clinical research is integrated within day-to-day clinical practice and data can be queried immediately to answer research questions. To that end, the Agency recently accounted a multi-stakeholder collaborative called the National Medical Evidence Generation Collaborative, or EvGen.
“The cost of clinical trials has become so high that it is impeding the ability of companies to develop new therapies,” Califf said. The FDA’s EvGen collaboration was set up, in part, to encourage the life science industry to make use of real-world clinical care evidence in later-stage trials.
“We have a clinical care environment and we have a clinical research environment, and up until now they have been more or less separate,” said Sherman. “The heart of EvGen is to integrate clinical research and clinical care. Ultimately, we believe this will create a more efficient and lower cost system that brings better health outcomes to patients.”
Most clinical trials are set up in a “parallel universe,” and then pivot to the real world upon approval, said Sherman. “We would like to see that pivot happen much earlier,” she said, referencing the FDA’s willingness to look at the totality of evidence, including data from a real-world “environment where the evidence is richer.”
“We have to make sure the industry believes us, because old habits die hard,” Sherman said. When it comes to the EvGen collaboration, the FDA’s role is to build evidence that a learning health system can benefit all stakeholders, which it plans to do by building on current successes like the Sentinel Initiative, a post-market surveillance system launched in 2008.
EvGen is much bigger than the FDA, Sherman explained. “We cannot own and operate EvGen because we don’t own and operate healthcare in the U.S. We can be a catalyst, an enthusiastic participant, [but success will rely on] some sort of public-private partnership,” she said.
Success with SHRINE
In 2008, SHRINE was still just an idea. By 2012, it was operating independently of Kohane’s supervision. In fact, in 2012 Kohane stumbled across groundbreaking research that had made use of the SHRINE system—research he was not aware of at the time.
The paper, published in Nature, was a study of peripartum cardiomyopathy, a rare type of heart failure that can happen during pregnancy. “Apparently some investigators at Beth Israel Deaconess Hospital had developed an interesting hypothesis about how vessel growth was out-of-balance in these patients,” Kohane said. “They only had a handful of patients, but they had heard about SHRINE.”
Using the SHRINE system, the investigators built an inclusion criteria and identified enough women to validate their findings, with statistically significant results. “Without the SHRINE system,” Kohane said, “it would have taken them years to do this research. This was able to happen in weeks.”
In clinical research, the difference between research taking weeks and taking years can mean the difference between something getting done, or never getting done, Kohane said. SHRINE has also been used in studies of autism co-morbidity, juvenile idiopathic arthritis, peripartum cardiomyopathy, colorectal cancer, diabetes and likely others.
For SHRINE, the “next frontier,” Kohane said, is to set up a mechanism for researchers to re-contact patients for consent to enroll in clinical trials. Once key players start “seeing the value of shared data,” Kohane said, the cultural and technical challenges that still exist today will quickly resolve.
“At the heart here of this group of efforts is delivering better healthcare to Americans,” said Sherman. “As a patient myself, I can say that far too often we are left scratching our heads and making the best guess about care. Given the amount of data we have in this country, that is not acceptable, in my opinion.”
Sony Salzman is a freelance journalist reporting on healthcare and medicine. She earned her Master’s degree from Boston University’s Science Journalism program, and has won multiple awards for narrative writing and radio journalism. She can be reached at firstname.lastname@example.org, or on Twitter @sonysalz.
This article was reprinted from Volume 24, Issue 06, of The CenterWatch Monthly, an industry leading publication providing hard-hitting, authoritative business and financial coverage of the clinical research space. Subscribe >>