Skip to main content
Premium Trial:

Request an Annual Quote

Lack of Adequate Genomic Data Standards Presents Major Hurdle to Precision Medicine


CHICAGO — Electronic health records vendors and leaders of precision medicine programs alike have been decrying a lack of adequate standards to support efficient use of genomic data in clinical care. Key standards-making bodies are aware of the issue and say they are doing what they can to expand and modernize their specifications.

However, such changes take time, due in part to the bureaucratic nature of standards-setting processes and also due to the general slow pace of change in healthcare.

"I am eager to see the disparate players support the kind of standards that vendors can implement," Charles Jaffe, CEO of standards development organization Health Level Seven International (HL7), said, referring to vendors of technology as well as genomic data. He said that in the clinical realm, the issue is matching genotypes with phenotypes.

A working group from the Association for Molecular Pathology published a paper in the Journal of Molecular Diagnostics in October suggesting that EHRs fall short when it comes to handling genetic and genomic data, and that they will continue to be inadequate until appropriate standards are in place. Like others, the AMP team said that genomic data is far more complex than any other kind of laboratory information an EHR might handle.

That falls in line with what Nephi Walton, associate medical director of precision genomics at Intermountain Healthcare in Salt Lake City, said when recently asked about common complaints leveled against major EHR vendors.

Walton said that it can be quite difficult to get genomic data from the molecular laboratory to the EHR, though he cited lack of standards more than anything that any vendor is doing. Intermountain uses Cerner's EHR, and Walton previously was a clinical genetics specialist at Geisinger Health System in Pennsylvania, a customer of Epic Systems.

A key problem, according to Walton, is that there is no standardization of genetic phenotypes. This makes it difficult to maintain variant classification across patients and reports, resulting in patients with the same variants receiving different care.

"Part of the scaling problem is how you get the data in there," Walton said. "You have to build an interface for every lab you use, and then you have to build a genetic phenotype for every lab result you can get in."

HL7 has taken on genomics by including an add-on to its Fast Healthcare Interoperability Resources, or FHIR, standard since 2019, but it has not yet been widely or uniformly adopted.

Walton said that the FHIR Genomics module component still is not mature and has "significant error in it." He took issue with how the standard classifies genomic phenotypes and its inability to handle compound heterozygotes and de novo variants.

Compounding that problem is that Epic adopted a variation of FHIR that HL7 later abandoned and therefore is one that nobody else supports. A useful standard, Walton said, must be able to keep up as genomics evolves.

"Genomics is a rapidly changing field, and at some point you have to say these are the granular elements of genomics and this is what we need to capture and build on," Walton said. "Otherwise, there's nothing that anybody can grab onto, because if you want to model everything that's going to happen in the future in genetics, you're never going to be done."

He said that a weakness of the HL7 Clinical Genomics Work Group is that it takes up so much time that clinical geneticists cannot be fully involved. "Clinical geneticists, we understand the data the best, but it's hard for us to engage as much as is required," Walton said. His strategy is to get involved when he can and to make sure industry partners like vendors understand what clinical geneticists need.

Another issue is that FHIR is not yet widely used among genomic laboratories. There, the preferred standards tend to come from the Global Alliance for Genomics and Health (GA4GH).

However, Peter Goodhand, Toronto-based CEO of GA4GH, estimated that there are more than 200 genomic data-sharing initiatives worldwide, though most are research-focused. Some working more on the practice side include the Global Genomic Medicine Collaborative, the Medical Genome Initiative, and the Mobilizing Computable Biomedical Knowledge project.

HL7's Jaffe said that everyone in the standards world wants to create their own standards "because they believe theirs is superior." However, that hinders interoperability.

Goodhand said that GA4GH has "not a lot of direct connection with the EHR vendors." However, some vendor representatives participate in working groups.

Goodhand, a past president of the Ontario Institute for Cancer Research, noted that most of GA4GH's work comes out of the research world, though an increasing share is applied, translational research that originated within health systems. "What we haven't reached yet is clinical utility and value from genomics in routine healthcare," he said.

Before the COVID-19 pandemic, Goodhand would go to healthcare industry conferences and hear from oncologists that they have just a few minutes with each patient. They would love to read through complex genomics reports to find the best course of treatment, but they don't have the time, so they need clinical decision support that considers genomic factors. Clinical decision support is usually delivered through EHRs and laboratory information systems.

HL7's Jaffe said that some purveyors of genomic data ask users to look up information themselves. "It really isn't the way healthcare works," he said. "Clinical workflow doesn't work that way."

Goodhand said that he would prefer to tailor standards for clinical practice in concert with the health IT community, noting that "we've got to fit in and map to what's already in the system, as opposed to thinking we can create some standalone thing that only works in genomics."

He said that GA4GH recently started a cancer community to identify the genomics elements most important to cancer researchers, clinicians, and patients. "There's been some great progress, but I think there's a lot more that can be done there," he said. He said he would "definitely welcome" more direct engagement with vendors.

"There's still a long way between the research-driven healthcare professionals in teaching hospitals looking for these complex, nuanced situations that can be addressed outside a health record and what people need in community care and routine healthcare," Goodhand said. Community-based clinicians don't necessarily need genomics itself, but they need the knowledge that genomics can uncover.

"From a GA4GH point of view, what we're really interested in is a learning health system. How does clinical practice come back and inform research?" Goodhand said. "I think that's still ahead of us in most cases."

GA4GH is in the process of refreshing its roadmap. The current one contains three strategic imperatives: make the standards work better together, make them easier to implement, and get closer to the front lines of healthcare.

GA4GH now has eight workstreams including one dedicated to clinical and phenotypic information.

The organization supports 24 "driver" projects, including national-level genomics programs such as All of Us in the US, Genomics England, and the Genome Medical Alliance of Japan. Other projects are international in scope, such as the Human Cell Atlas, Variant Interpretation for Cancer Consortium, and the European Genome-Phenome Archive.

There are some small projects working close to the realm of patient care, including one GA4GH driver project, the BRCA Exchange. That project has created a database of 70,000 BRCA1 and BRCA2 variants that can be looked up from a mobile device, but Goodhand said that the information still does not easily feed into a clinical decision support system.

"What healthcare needs from the genomic community is better translation of the knowledge, better packaging of the knowledge, and [the ability to] feed it in in a way that can be digested in the front line of healthcare," he said.

"As we develop the genomic knowledge, it doesn't always fall into clean, crisp decision criteria, but it's incredibly valuable and informative," Goodhand added. "But because it's not black and white, it's more difficult to bring over into clinical."

Culture is another factor, since old habits die hard in healthcare. "Many frontline [healthcare] systems and providers have not grown up in the genomic era. This is new knowledge to them," Goodhand said.

Jaffe mentioned the phenomenon called alert fatigue, in which clinicians start tuning out excessive alerts from clinical decision support systems, potentially missing critical information. Another HL7 specification, Clinical Decision Support Hooks, aims to help technology developers "tier" alerts according to importance.

CDS Hooks is in place for things other than genomics right now. In the future, Jaffe expects it to be able to tell practitioners that a diagnosis can be confirmed with a genome sequence. "That's the promise, and that's what I hope someday to have realized," Jaffe said.

Before that can be realized, Jaffe said, genetic and genomic data have to be put into a machine-readable format, not a PDF file typical of test reports. An EHR system needs to alert physicians that genetic data even exists.

Framing it in the context of the VHS-Beta battle for superiority in home video decades ago, Jaffe suggested that health IT vendors, like pharmaceutical manufacturers, sometimes win out based on marketing or lack of customer feedback. That may be happening with regard to EHRs and genomics.

"We, the standards community, haven't given them the rationale for changing their input to something that is more readily consumable," Jaffe said.

"We haven't gotten to the point where the people who buy the EHRs demand the genomic data," Jaffe added. "That's our fault."

There also is the question of cost versus benefit in implementing new technologies in clinical practice. HL7 finally has some evidence that there is tangible financial benefit to interoperability.

Health insurer Cambia Health Solutions, parent of Blue Cross Blue Shield licensee Regence, sponsors a healthcare innovation hub called Cambia Grove. Each year, Cambia Grove selects a class of research fellows. In 2021, the fellowship program focused on measuring return on investment of data interoperability in healthcare, collaborating with HL7 to calculate the value of widespread FHIR adoption.

Specifically, Cambia Grove looked at value metrics of the Da Vinci FHIR Accelerator Project, a private-sector initiative led by HL7 to help payors and providers improve healthcare outcomes within value-based reimbursement frameworks by applying the FHIR standard.

"By automating data into downstream systems, an organization can not only have a more complete and real-time picture of the patient’s experience, but also gain that picture in a cost-effective manner. These advantages promote improved collaborative decision making between providers and patients that results in better clinical care, faster care gap closures, and lower bottom lines," one report from the research fellows said.

"I think of it as a scorecard for implementers, whether they are payors, providers, etc., to be able to measure the value of what a particular use case is delivering for them," said Kirk Anderson, Cambia's VP and chief technology officer, and current Da Vinci Project board chairman.

"In the first few years of Da Vinci, when we were trying to get this idea off the ground, we had mainly been measuring success in terms of how many use cases are being implemented," Anderson said. That worked fine for a while, but now that FHIR is specified in federal healthcare interoperability rules, it is an inadequate metric, he said.

US interoperability regulations under the 21st Century Cures Act that call for broader use of FHIR has accelerated uptake of the FHIR standard, but the regulations do not explicitly say EHRs have to support genomics.

David Degandi, Cambia's senior strategist in interoperability, said that Cambia Grove is now looking for a custodian to govern the value-metric framework it spelled out in its report, and also to run proof-of-concept programs. He said that the framework will likely end up within HL7, which will then make the content available globally. There is no timetable for a handoff.

Degandi urged patience in adopting the Cambia Grove recommendations. "We're going to start small, and it'll mature and grow as people become more familiar with it and learn how to really make best use of it also. Standards bodies go slow," he said.

From HL7's perspective, the organization now has a way to measure the value of interoperability. "We need to be able to articulate that to [vendors] … so that you'll get far better care for much less money if you incorporate genomic data into the decision process."

Standards organizations are also still looking for real-world examples to spur interest in interoperability of genomic data for precision medicine.

"We've standardized a lot of what happens in VCF [files], but from a clinical perspective, people are saying that we need more consistency, we need less ambiguity because it's for clinical purposes," Goodhand said. GA4GH would like to be able to benchmark informatics systems against each other to improve translation of knowledge for use at the front lines of healthcare, but Goodhand said that the organization cannot do it alone. It will take collaboration with other standards bodies.

"I think healthcare needs to build some successes with genomics so that it can become more comfortable with bringing more data into it," Goodhand said.