CHICAGO (GenomeWeb) – Electronic health records giant Epic Systems is adapting to the age of precision medicine by building a dedicated, cloud-based sequence server to store large genomics files.
Peter DeVault, Epic's vice president of genomics and interoperability, did not divulge specifics of the technology, including whether or not Epic is partnering with a major cloud services company, nor did he offer a target release date. "We are engaging several of our customer organizations in the development of it and when it's ready for particular use cases, we will release it," DeVault said.
"We have aspirations at Epic of trying to become a clearinghouse for some of that knowledge, and make it available to all of our customers," DeVault said from Epic's sprawling complex of themed campuses — including farm life, a big-city experience, and the "Harry Potter"-inspired Wizard's Campus — in the Madison suburb of Verona, Wisconsin.
When the time comes, do not expect a big splash. Epic historically has not made product announcements and by rule does not issue press releases.
DeVault did, however, say that the sequence server will offer flexibility to Epic customers, which tend to be large, integrated healthcare delivery networks. "The idea is that it will be an off-the-shelf solution for organizations that want to use our complete solution or be able to plug into their existing bioinformatics pipeline to the degree that they might have one already," he said.
"You can use the phenotype information that's in the EHR to help interpret those sequences," DeVault continued, adding that interpretation needs change over time.
The idea behind the dedicated server is that Epic does not want to put large sequencing files into patient records, but wants to make this information available as needed.
"For one thing, it would be cognitive overload for most clinicians," DeVault said of tightly integrating clinical records and massive sequencing files. "And secondly, most of that information is not actionable today." It might not be relevant for the case at hand, either, since clinicians tend to be looking for one specific thing in the sequence at any given moment, such as a sign of a rare disease or a potential treatment pathway.
"A year from now, you may want to go back to that sequence and re-interrogate it, either because you know more about the patient or your questions have changed or, just as likely, we know more about the human genome. A variant that was previously of uncertain significance we now know to be pathogenic or to have some combinatorial effect with other genes," he added.
DeVault said that Epic's genomics team wants to make genomics data seem similar to any other element in an EHR.
"Part of what that means is making that data available for algorithms, for predictive modeling, for clinical decision support, or for analytics, just like any other data," he explained. "Part of our goal is to help elevate genomics and related data to the same level as other types of data in the patient's record, to be able to use genomic data for analytics purposes, decision support, and research."
Clinical decision support rules now can be written to take into account genetic variants as well as traditional phenotypic factors, DeVault noted.
Some of Epic's more advanced customers have been incorporating genetic and genomic data to some extent into patient records for close to a decade, DeVault reported. The EHR vendor formed its genomics team a little more than two years ago.
"It has been interesting in the last year or two to see both knowledge and awareness of genomically informed medicine across the board, as well as the desire to incorporate some of those findings into clinical practice," DeVault said. Hospitals have been particularly active in pharmacogenomics, he noted.
After noticing that many of its customers had been "kind of forging their own paths ahead," according to DeVault, the company decided to help them tackle all the new types of data showing up in clinical practice.
"Rather than just making do with the data structures and workflows for traditional laboratory results and observations, we wanted to make a purpose-built home for that data within the patient record. That's what we've been doing for the last few years," DeVault said.
Epic long has made laboratory information systems and other ancillary pieces of technology to support the core EHR. "A lot of genetic testing today is for panels of genetic traits … that often in size are not all that different from traditional laboratory observations," DeVault said.
In general, though, genomics data requires a different approach than earlier additions to clinical workflows. Notably, genetic observations are not simple numerical values, like, for example, a test for hemoglobin A1c.
"They can be differences in sequence. They can be differences in the proteins that are encoded by a gene. The data structures to hold that information need to be equally sophisticated, and that's what we have been doing," DeVault said.
"Another big difference, of course, is that with a traditional laboratory result like a blood chemistry result, the value of that result might change from day to day or week to week, but the way we interpret it changes very slowly," DeVault said. "With genetics data, that's stood on its head."
With the notable exception of somatic results of cancerous tumors, genetic observations in a specific patient do not change much over time, but the way in which results are interpreted do as researchers discover new markers and pathways.
To compensate, Epic has built a "home" within the EHR for genetic information, including what DeVault called "actionable variants," meaning variants with thresholds backed by clinical evidence, including pharmacogenomics and risk factors for cardiovascular disease and cancer.
He noted that the company has made some of its tools more robust in terms of changing variant interpretation in patient records as new knowledge emerges.
This movement to integrate genomic data into traditional medical records and workflows is happening in conjunction with an industry-wide push to create more interoperability of health information between healthcare organizations.
Epic has been the frequent target of critics in recent years for allegedly engaging in what federal officials in both the Obama and Trump administrations have dubbed "information blocking," but the EHR vendor continues to tout its efforts to promote interoperability of health data.
Epic and its competitors have had their hands forced to a degree not only by customer demand for easier flow of EHR data, but by the 21st Century Cures Act, a law that authorizes the US Department of Health and Human Services to impose fines as high as $1 million per episode of information blocking. The rule to implement that portion of the 21st Century Cures Act reportedly is under final regulatory review at the White House's Office of Management and Budget.
DeVault suggested that demand for data will break the impasse more than the threat of hefty monetary penalties, but stated that data liquidity is an ongoing issue.
"Interoperability is never a completely solved problem," DeVault said. "As new types of data come out, standards have to be adapted."
One standard that Epic and its competitors have embraced is Fast Healthcare Interoperability Resources, or FHIR, a Health Level Seven International specification that has an offshoot under development called FHIR Genomics. "When we can use that new technology for new types of data like we're doing with the FHIR Genomics resource in HL7, so much the better," DeVault said.
To DeVault, interoperability in genomics is more about transferring discrete, structured, machine-readable data from one place to another than it is about moving large FASTQ, BAM, or VCF files. "That requires some careful design of those data structures. We've modeled our data structures in Epic around the industry standards that are emerging for being able to transport those data structures," DeVault said.
Genomic data has required a shift in strategy.
"We had to step back from the traditional laboratory-results data structures where … if you think about a hemoglobin A1c result, it's a pretty basic, key-value pair. HbA1c is the key and the value might be 3.9," DeVault explained. "For a simple genetic variant, you have the name of the variant, its location, the predicted protein change, the transcript change, links to various knowledgebases that can tell you what we know about that variant, so it's a much more multidimensional data structure."
DeVault said that his team talks just about weekly with laboratory technology vendors about how Epic customers would like to integrate with their information systems. "We're at kind of a difficult time right now in late 2018 with the state of both science and the ability to get that discrete data, but I think we are going to see that resolved in the next couple of years," he predicted.