CHICAGO (GenomeWeb) – Informaticians at the University of Nebraska Medical Center have developed a method for automating the coding and mining of molecular pathology test results to help clinicians assess the efficacy of targeted cancer treatments.
"We prepared the data so it can be consumed by the electronic health record and readily used," explained Scott Campbell, director of informatics for the public health and pathology laboratories at Omaha-based UNMC. "We've effectively made a molecular pathology report as readily presentable as a CBC — a complete blood count — or any sort of lab count."
This, according to Campbell, promises to speed the introduction of precision medicine into cancer treatment.
"People are expecting precision medicine now, but if we can't make our systems manage the data, then it's going to be pretty hard for our physicians to practice precision medicine with broad effect," he said.
"We looked at the entire sequencing pipeline and came to the conclusion that the information that we're getting out of our pipeline and into our molecular pathology reports is very rich in the sense that, in the report, you can understand it as a physician and use it once," Campbell explained. "But once that report has been filed away in an electronic health record system, the actionable data items are hard to discover and they're not computable because they're not stored in a computable way."
The co-lead on this effort, biomedical informatician and practicing internist James Campbell, said that clinicians approached the informatics department as the medical center was rolling out its genomic research database in 2013.
"The request was that they get better access to structured genomic data in the diagnosis of their cancer cases," James Campbell said. The reports existed on paper, but they were not easily accessible in clinician workflows, he noted.
To address this problem, the Nebraska team — funded in part by a National Institutes of Health grant — pulled next-generation sequencing data on 444 patients with Stage IV colorectal cancer from the UNMC's Informatics for Integrating Biology and the Bedside (i2b2) data warehouse. The University of Iowa also participated.
They developed a method to analyze and classify genetic variants based on the SNOMED CT, Variant Normalization Toolkit, or the Human Genome Variation Society nomenclatures to represent findings of molecular pathology tests. That result then is structured into a Health Level Seven International version 2 clinical message for export into an EHR.
"When the molecular pathologist signs out a report, the written document and PDF version is sent to the clinician, but simultaneously, we create an HL7 lab results message that serializes and indexes every gene that was assessed, all variants or lack thereof that were detected," Scott Campbell said.
"We've pushed this really complex data into a very well-used, well-understood form using standards — HL7 messaging, SNOMED CT, and some [Logical Observation Identifiers Names and Codes (LOINC)] for representing questions and HGVS for representing answers — and it's something that can be passed to clinical data warehouses," he explained. "It also prepares it for ready consumption into the electronic health record. The data then becomes very discrete and actionable," by oncologists and researchers alike.
"Scott and I reviewed the relevant standards and concluded that SNOMED CT and LOINC were the appropriate standard terminologies, but that the content really didn't exist in order to properly describe these observations," James Campbell said. "We have worked with both SNOMED and LOINC to extend the terminologies. Once we developed the extension for the necessary codes and code definitions, Scott went ahead and actually deployed them in our anatomic and molecular pathology departments."
James Campbell presented some of his work in March at the American Medical Informatics Association's annual Informatics Summit in San Francisco. Both men subsequently spoke to GenomeWeb by phone.
"What we found out is that we can answer very complex questions much more quickly and without chart review when the data is structured as we have done it," Scott Campbell explained. "When we encode our molecular data in this very straightforward way, it does not lose information. It compresses very complex data into a readily computable form, given today's frontline technologies," namely EHRs and clinical data warehouses.
The intention is to use the technology for research, clinical care, public health, and quality control of lab systems. Scott Campbell noted that the College of American Pathologists has told labs to be ready to self-assess capabilities and to be able to retrieve patient data based on certain mutations or when new research finds additional therapies or contraindications. He said that current methods make it "extraordinarily hard to get at" such data.
"The approach that Jim and I developed, in part with NIH funding, really makes those very tractable problems to assess and address, so it's very quick to identify all our patients back to 2013 when we started our sequencing operation," Scott Campbell said. "Who has these variants? Do we need to bring [patients] back? Do we not need to bring them back? Are they getting therapies that are not contraindicated by other variants?"
This SNOMED and LOINC extension makes molecular data fit within the constraints of today's health information management ecosystem, he said.
"People are expecting precision medicine now, but if we can't make our systems manage the data, then it's going to be pretty hard for our physicians to practice precision medicine with broad effect," Scott Campbell said.
"Now that we've shown that it can be done, now we have to do the necessary work with our Epic build teams and the clinical operation to make sure that it fits what the physicians need and want to see and is repeatable and reliable — all of the things that are necessary to take something into production. In my opinion, it's ready for regular operational consideration and hardening work for that purpose," Scott Campbell said.
Their work is part of the International Collaboration on Cancer Reporting, so the Nebraska-led team is developing a global rollout strategy for the extension and also working to include other types of cancer in the future, according to James Campbell. He said that UNMC currently is updating the terminology every six months, and once it becomes international, updates will follow the SNOMED release cycle.
ecause SNOMED-CT is freely available for noncommercial use in the US, the extension is available for free download by clinicians and academic researchers. "If a vendor wants to come along and develop a product that does it all behind the scenes and they're using it to sell their product, they can talk to our tech transfer people," Scott Campbell added.