CHICAGO – Current electronic health records fall short when it comes to handling genetic and genomic data, and they will continue to be inadequate until appropriate standards are in place, according to a new report from the Association for Molecular Pathology.
The fundamental problem is that genomic data is far more complex than any other kind of laboratory information an EHR might handle, said Alexis Carter, an informaticist and molecular pathologist at Children's Healthcare of Atlanta. Carter chairs the EHR Interoperability for Clinical Genomics Data Working Group for AMP and is lead author of a new paper on the subject.
"We've been sending regular, simple lab data over to EHRs for a very long time. They don't have nearly the complexity that genetic and genomic data do," Carter said. "It is going to be difficult to do, otherwise somebody might have done it already."
The findings of the working group, formed in November 2019 to provide recommendations to the AMP board of directors, were published in the Journal of Molecular Diagnostics earlier this month.
This follows the July release of a series of guidelines from AMP to promote the sharing of genetic variant data via public repositories. The earlier recommendations, however, were not peer-reviewed.
According to Carter, the organization had heard numerous complaints from members about the difficulty pathologists had in reporting discrete genomic data into EHRs.
While other forms of healthcare information may not be as complex as genetic and genomic reports, there remains an interoperability problem across the EHR world, enough for the US government to threaten vendors, providers, payors, and other entities with fines for "information blocking." Data is largely siloed within institutions, and the electronic transfer of discrete, computable information often is difficult even between healthcare organizations that use the same EHR vendor.
Carter said that genomics presents a unique challenge. "You can't send discrete variants over and divorce them from the main overall interpretation for a specimen, because there's a lot that goes into that interpretation," she said.
Another thing the working group kept coming back to was the need for a uniform method of performing variant reanalysis, reclassification, and reinterpretation.
Like the paper itself, Carter did not single out anyone by name, but said that some vendors have "lost" the need to connect variants to other factors when handling genomic data. She said that some of the genomics modules or add-ons that EHR vendors have introduced in recent years fall short.
The AMP working group has not yet received feedback on its paper from the EHR vendor community, she noted.
Meditech, Epic Systems, and Cerner are among those that have updated their technology for clinical genomics, particularly pharmacogenomics. Allscripts Healthcare Solutions actually sold genomic data integrator 2bPrecise to AccessDx Holdings earlier this year.
Carter said that some members of the working group had direct experience with modules like these, while others looked at the technology and decided not to implement it. "Some people had implemented [these systems] and were reporting really serious difficulties with getting data across in a way that it was not going to be misinterpreted," she said.
A key reason for these experiences was a lack of standards for integrating genomic data into EHRs. Carter noted that it is difficult for vendors to build proper technology when customers are asking them to follow different methods and strategies.
She said that the working group hopes to work in a multiorganizational way to figure out what needs to be in industrywide standards for transferring genomic data. There needs to be a standard to describe variants in a way that is computer-readable, for example, even if an individual variant gets separated from others in a report. "That, to my knowledge, has not been done," Carter said.
It is also essential that the interpretation does not get lost if one variant is separated from others or several variants are pooled into a single report.
Patients being tested for a genetic cause of a disease often get multiple targeted panels if the first test does not turn up anything diagnostically significant. "What [clinicians] really want to be able to do is to take all of that information that comes back and amalgamate it into a single view over time," Carter said. "If you're talking about taking individual variants out of a particular report and putting it into a global view of all these variants from all these different laboratories, that's when it becomes really important to have a very structured standard to make sure that interpretations are not lost."
For example, the presence of a PIK3CA variant in a tumor does not necessarily mean that the patient has an inherited PIK3CA variant. "It's very important to make sure you're making that distinction for physicians when they're looking at things at the EHR," Carter said.
Communication is also important. If there is a change following a reanalysis, that information has to be communicated to the patient and his or her care team, the report said.
Carter said that the paper is meant to be informational more than a call to action, though the workgroup stated in the article that it would be actively involved in implementing its recommendations.
The next step will be a full-scale literature review "to see if we can find who has been able to solve what aspects and how," Carter said. AMP is also putting together a conference on variant reclassification and reinterpretation in order to help develop standards in those areas.
The report acknowledges that other organizations have attempted to address the genomic data question.
Health Level Seven International (HL7), for example, has a clinical genomics working group, though Carter said that that group's clinical messaging specification is too dependent on Logical Observation Identifiers Names and Codes, or LOINC, a coding system for lab testing and results.
HL7 has also taken on genomics by including an add-on to its Fast Healthcare Interoperability Resources standard, though FHIR is more about moving data than computing it. Plus, according to Carter, pathology labs tend to use an earlier, text-based version of HL7 messaging that does not include FHIR.
The working group's article also cited the US National Human Genome Research Institute's Electronic Medical Records and Genomics (eMERGE) network, Sync for Genes, ClinGen, and DIGITizE, a former National Academy of Medicine project that evolved into the HL7 clinical genomics working group.
The authors neglected to mention the Global Alliance for Genomics and Health, an oversight that Carter rectified in a webinar about the AMP report that will debut next week.
Once there are standards, AMP would like to make sure that variants are always interpreted in the context of the original specimen. "The specimen source, test details, variant allele fractions, and combinations of variants matter," Carter said in the webinar, which was prerecorded.
The absence of clinical decision support rules and standards for genetic and genomic tests can lead to unnecessary testing and even create patient safety hazards, for example, by neglecting pharmacogenomic clues when prescribing medication, she added.