Nephi Walton, M.D., has experience with precision genomics programs at two pioneers in the space: Geisinger Health in Pennsylvania and Intermountain Health in Utah. At this week’s meeting of the American Medical Informatics Association (AMIA), Walton explained why it is important to have a place in the electronic health record for genomic data.
Walton, associate medical director of precision genomics at Intermountain, noted that genomics has become a national and international priority, but only recently have EHR vendors created ways for clinicians to store discrete genomic data. “There remain many challenges around adoption and implementation of putting genomic data into the electronic health record,” he said.
He described some lessons learned in this space and challenges still to be overcome. Intermountain Precision Genomics is working with partners to sequence 500,000 patients; it currently has over 80,000 enrolled. They also have a major initiative for pediatric patients. Walton said it should become the largest repository of pediatric whole genome sequences. “We are doing whole genome sequencing in that initiative and linking all that to healthcare data.”
When he was at Geisinger, he worked on the MyCode program. They are doing whole exome sequencing in collaboration with partners and have over 250,000 people enrolled and over 200,000 sequenced currently.
He said that at Geisinger they constructed clinical decision support and patient- and provider-facing information for all the variants they tracked as well as for pharmacogenomics data, and Intermountain has deployed “passive” decision support.
One of the challenges identified at both Geisinger and Intermountain in seeking to provide clinical decision support around pharmacogenomics is that physicians aren't looking for this kind of data when they prescribe.
“It's not something that they're thinking about as part of their process,” Walton said. “We need to have a home for genomic data in the EHR. Physicians don't really think about or take the time to look for a pharmacogenomics report in the EHR before they prescribe something. This can be potentially dangerous if you're really ramping up a pharmacogenomics program,” he said. “We have 10,000 pharmacogenomic reports; if a physician prescribes something and doesn't look at that because they're not used to it, there's a potential liability there, because we had information that it could harm the patient if you prescribe certain medications, or that might not be as effective. But the problem is there's not a central place to look for it.”
One reason people don't look for this data is that so few patients have it, but that is going to change in the future, especially as genomics expands exponentially,” Walton stressed. “There's so much more that we can do in terms of managing healthcare based on genomics, and that's just going to increase.”
Another challenge to overcome is the need for adoption of standards to pass genomic data from the laboratory into the EHR. “This is extremely complicated,” he said, “and I've worked with a number of different laboratories and EHR vendors on this. The laboratory wants to wait until the EHR adopts something, and the EHR vendor wants to wait until the laboratory adopts something and nothing really happens,” Walton said. “As we add more laboratories, or get more laboratory results in, it takes a lot of work to get those results to go in the right place in the right way so they can actually be used.”
They sought to figure out the role of the laboratory information system (LIS) in this process. “The laboratory information system seems like something that really should hold genetic data, but what genetic data should it hold? And how does it pass it to the EHR? And how does it store it?”
Both Geisinger and Intermountain actually bypassed the LIS, because there wasn't a defined way for the LIS to interact with the EHR, he said, “but that's an important thing to think about as we look at the LIS of the future: how are they going to handle genomic data?"
“Another thing we ran into is inadequate and discordant standards for genomics in HL7 genomic reports and the FHIR molecular sequence resource,” he said. “This is still an area that is in flux. The implementation for Geisinger was Epic, and the implementation for Intermountain was Cerner. Neither one has got everything figured out.”
Still another challenge in genomics is maintenance of clinical decision support in a rapidly changing field. “In genomics, we learn new things every day. There are lots of different guidelines. Trying to decide what to put in and how to manage all that is really difficult, particularly when you have to involve resources from the EHR team in order to make some of the changes,” Walton said.
Another problem is that there are not a lot of standard resources for patient/provider information that can be accessed with something like info buttons, Walton said. “That's really important, particularly if you're a smaller healthcare system that might not have the resources to develop all that internally, or don't have the genetics expertise, because there is a shortage of genetics providers. And that's something that I think really needs to be built in order for this to scale. Because only the institutions that can afford to have those resources internally can do it currently.”
There is a different perception of discrete genomic data put into the EHR vs. having genetic data in a PDF report, Walton added. With the PDF, people think that only clinical geneticists are going to look at it, whereas if you put it in the patient's chart, front and center, the physician might actually do something with this, and they might not understand genetics. “So that's a real perception challenge. I have seen bad things happen when people have taken variants out of context.”
Finally, varying workflows need to be taken into consideration, Walton said, because there are very different needs for clinical genetics and population health genetics. “Clinical geneticists want variants of unknown significance that they can look at in the context of the patient and maybe redefine trying to determine what the genetic cause of a patient's condition is, whereas primary care providers don't want that. They want certain variants, with certain guidelines with certain actions that you will take based on the variants,” he said. “Those workflows are so different that we had to completely abandon the clinical genetics workflow in order to implement genetics for population health for the larger population. That's something that's important to take into account. Interestingly, Epic has taken more of a population health approach to how they implemented genomics, while Cerner is taking a more clinical geneticist approach. It would be nice for them to meet in the middle and both build something that worked for both worlds, but it'll be interesting to see how those things play out.”
