Genomics pose 'daunting' test for EHRs

Terabytes of data mean unprecedented storage and interoperability issues
By Neil Versel
10:33 AM

Think parsing the growing amount of information in electronic health records is tough now? Just wait until genomic data starts showing up in EHRs.

"The number of individual genetic tests is daunting," Peter Tarczy-Hornoch, MD, chair of the University of Washington's Department of Biomedical Informatics and Medical Education, said this week at the American Medical Informatics Association's annual symposium. Each needs "structure and storage."

A fully sequenced and analyzed genome contains about a terabyte of information, Tarczy-Hornoch explained during a well-attended session on integrating genomic data into the EHR, creating unprecedented storage and interoperability issues.

Asked about how to move around such large files, Christopher Chute, MD, founder of Mayo Clinic's Division of Biomedical Informatics, recommended just recording variants in the EHR, not the entire genome. "It's a kind of lossless compression," he said.

Chute and Tarczy-Hornoch joined other top informatics researchers to review the October issue of the journal Genetics in Medicine, which was dedicated its to the integration of genomic data into electronic health records.

Christopher Chute, MD, founder of Mayo Clinic's Division of Biomedical Informatics, wrote about applying big data to translational research.

"The special issue is intended to provide guidance to the many sites that are embarking upon or planning for the integration and implementation of genomic medicine into clinical care," the co-editors of the issue, Marc S. Williams, MD, director of the Genomic Medicine Institute at Geisinger Health System in Danville, Pa., and Joseph L. Kannry, MD, lead technical informaticist of the EMR Clinical Transformation Group at Mount Sinai Medical Center in New York, wrote in their introduction.

"Authors of each article were asked to provide guidance in the form of a 'how to' and 'what to think about' for anyone tasked with launching a genomics program and integrating these data into the EHR at the point of care."

In other words, "Healthcare is incredibly complex," Chute said.

"Healthcare data is more than Google search," he told a packed AMIA meeting room. "Close enough is not good enough."

Most of the authors featured in the special issue are members of the EHR integration workgroup of the Electronic Medical Records and Genomics (eMERGE) Network, a federally funded, nine-organization consortium hosted at Vanderbilt University Medical Center in Nashville, Tenn. Several of the papers focused on the experiences of eMERGE institutions.

One implication, according to Williams, and co-editor of the special issue, is that clinicians and researchers soon might be able to build patient phenotypes from EHR data.

Williams was one of 18 listed authors of a paper examining stakeholder engagement strategies at eMERGE sites. "We're clearly in a movement to open more data to patients," he said, in explaining that article.

However, Williams cautioned that integration of genomic data into EHRs may increase the risk of compromising this highly personal information. "Current privacy and discrimination laws may not be adequate," he said.

"One of the big takeaways for me" was the length of the data cycle, Williams said, explaining a paper written by Andrew G. Ury, MD, chairman of Seattle-based ActX, a startup seeking to make genomic information clinically relevant to physicians. Genomic data has a much longer "shelf life" than other health data such as vitals or test results, Williams said.

It is a cycle instead of an arc with an endpoint, Williams explained, and genomic data should be constantly reused "as we gain more knowledge about the patient."

Genomic profiles need to be available for the life of the patient, travel with the patient, be structured and possibly resequenced as technology improves, Williams said.

Genetic knowledge has been changing over time, according to Tarczy-Hornoch, co-chair of the EHR working group of the Clinical Sequencing Exploratory Research (CSER) Project, an NIH-funded effort to examine practical, ethical and legal requirements for integrating genomic sequencing into clinical care. He is the corresponding author of a paper explaining CSER's survey of various efforts to include genomic reporting in EHRs.

Clement J. McDonald, MD, director of the National Library of Medicine's Lister Hill National Center for Biomedical Communications and former director of the Regenstrief Institute in Indianapolis, was not on the AMIA panel, but he spoke to Healthcare IT News after the session. "Things slip" during sequencing, potentially creating small errors, this longtime informatics expert said.

He said that even one error in 1 million is a lot when dealing with a genome containing 1 billion data points. "Depending on what they hit, they could be important," McDonald said.

"You've got a big problem with interpretation and it keeps evolving as you get more data," McDonald said.

Referring to a Journal of the American Medical Association commentary he co-authored earlier this year, Chute said there is no real coding system for genomic data.

However, McDonald told Healthcare IT News that Health Level Seven International already has a code for reporting how patient-specific data corresponds with the Reference Sequence Database, a collection of curated, publicly available DNA, RNA and protein sequences from the NLM's National Center for Biotechnology Information. McDonald recommended that genomic data include Ref Sequence codes.

Once data is in the EHR and is being used to diagnose or treat, "this is no longer genetics land," Kannry said. "It's routine care." And thus, clinical decision support is necessary. Kannry said CDS requires actionable, discrete data, genetic or otherwise.

Joshua C. Denny, MD, a biomedical informaticist at Vanderbilt, told AMIA attendees that that genomic information does not have to be stored within a core EHR, but it does need to be connected to clinical decision support. Vanderbilt only puts genetic information into the EHR "if there is CDS around it," Denny said.

As a result of eMERGE work, genetic results now are "visible passively" in the EHR at Vanderbilt, Denny said. "It gets people used to looking at it." However, it does not necessarily change prescribing habits, which is why Denny said CDS needs to be connected.

Genomics probably will lead to "versioning" in clinical decision support, Kannry predicted. "I actually think we're going to have to version these things" so physicians know what information led to a particular decision, he said.

He said he believes it will take 20-40 years to get to the point where a doctor can tell patients what they will die from, and when.

[See also: HIT unprepared for 'omics' onslaught ]

[See also: HIT unprepared for 'omics' onslaught ]