Creating a Home for Genomic Data in the Electronic Health Record

Introduction

Genomic data generated from massively-parallel sequencing have the potential to improve health through targeted clinical management. The integration of genetic information into healthcare is considered a major priority with several national and international initiatives focused on this goal[1]. Only recently have EHR vendors allowed for the storage and access of discrete genomic data. However, many challenges remain related to adoption and implementation of such technologies. Aside from storage of discrete genomic data there are new clinical workflows that are introduced by large scale population sequencing that have not been addressed by EHR vendors. Having worked through the integration of genomics for the two largest EHR platforms and for two large scale sequencing programs, the MyCode® Community Health Initiative (MyCode)[2] at Geisinger, and the Heredigene® Population Study (Heredigene)[3] at Intermountain Healthcare, we have performed an analysis of the current state and future needs for true genomic data integration.

Methods

At Geisinger through MyCode and in collaboration with the Electronic Medical Records and Genomics (eMERGE) network[4]we imported discrete pathogenic and likely pathogenic genomic variants for the Centers for Disease Control and Prevention (CDC) Tier one conditions (Hereditary Breast and Ovarian Cancer syndrome, Lynch syndrome, and Familial Hypercholesterolemia), and variants from seven pharmacogenes. The C282Y variant for hereditary hemochromatosis were also included to assess for challenges related to implementation of autosomal recessive conditions. We constructed context sensitive maintenance guidelines in the EHR for each of the disease conditions and built best practice alerts for medications impacted by the pharmacogenomic variants. Patient and provider facing information was developed for all disease variants as these were accessible through both the physician chart and the patient portal[5]. Similarly, at Intermountain Healthcare through Heredigene and the RxMatch pharmacogenomics platform we constructed a passive decision support system for 17 pharmacogenes and patient and provider facing information for CDC Tier one conditions and hereditary hemochromatosis. At both institutions, we designed data infrastructures and interfaces to handle the clinical workflow of the return of results process as neither EHR vendor had the capability to manage this workflow. Weekly calls were held with pertinent institutional stakeholders to address challenges and barriers to implementation as they arose. A list of these challenges was maintained throughout the process. After each implementation, we analyzed the noted challenges and identified barriers. Metrics were also gathered on usage of CDS for pharmacogenomic (PGx) at both institutions, and changes in management where tracking was possible.

Results

We were able to successfully implement genomics in both systems with compromises. We encountered vendor specific informatics challenges on both EHR vendor platforms including the inability to store discrete variants for pharmacogenomics, necessitating use of star allele diplotypes, and having inadequate infrastructure for autosomal recessive conditions when the patient is a compound heterozygote. Although these were vendor specific issues the vendors data implementations were based on legacy standards, which had deficiencies that contributed to these problems. Other global challenges include: 1) Need for a “home” for genomic data 2) Need for standardization of genetic phenotype. 3) Difficulty in sending discrete data from the genetic testing laboratory directly to the EHR. 4) Addressing the role of the Laboratory Information System (LIS) in the genomics process. 5) Inadequate and discordant standards for genomics in the HL7 genomic report format and the Fast Healthcare Interoperability Resource (FHIR) molecular sequence resource. 6) Challenges related to maintenance of patient/provider information and decision support in a rapidly changing field. 7) Lack of standard resources for patient/provider information. 8) Disparate implementation needs for clinical geneticists compared to primary care providers. 9) Different perception of discrete genomic data versus scanned PDF of the genetic report. 10) Difficulty in maintaining variant classification across patients and reports.

A technical infrastructure was designed to manage the return of results process at each institution that enabled 1) tracking of notifications of patients of their results, 2) genetic counseling processes, 3) provider notifications, 4) specialist referrals, 5) result based clinical actions, 5) cascade testing, 6) variant classification and reclassification.

Metrics on CDS usage in pharmacogenomics demonstrated that 73% (40/55) of providers changed prescriptions based on active decision support for new prescriptions but none of the providers changed dosing or medication when the patient had been established on the medication. Metrics for passive decision support demonstrated that usage was primarily restricted to those who ordered a PGx test for a specific purpose and only used after the initial test result came back and rarely used outside of the ordering provider or by the ordering provider for another indication.

Discussion

While we were able to achieve some success in integration of genomic data into the EHR, there are many challenges that need to be addressed. One of the biggest challenges moving forward is the long-term maintenance of gene/variant specific information and decision support. Genomic medicine is a rapidly changing field and EHR systems require support from vendor specific technical personnel to implement changes to patient/provider facing information and decision support. Engaging this technical team can be challenging given their responsibility to maintain all aspects of the EHR and they are generally not skilled in genomics requiring an alignment of teams to make any changes. EHR vendors adhering to more open standards for genomics and decision support, including SMART on FHIR, and CDS Hooks would allow for integration of external applications where maintenance can be performed by staff with genetics training and less technical skill. Another important component to this, is the need for maintained resources such as PharmGKB that are Infobutton compliant and can allow to quick access to the most current recommendations and information on a specific gene or variant. To our knowledge no vendor has implemented Infobutton support for genomics, and outside of pharmacogenomics no such Infobutton compliant resources exist.

Difficulty in passing data from the laboratory to the EHR is a considerable challenge and one of the greatest components of this challenge is the standardization of the genetic phenotype (i.e., disease, metabolizer status, gene specific phenotype, modifier variant phenotype). We were able to capture discrete genomic data from reporting laboratories, however; laboratories used their own proprietary JSON structured format. Developing the ability to pass phenotypes required manually building a process for each phenotype and each lab, which is neither desirable nor sustainable as more laboratories and conditions are added. To effectively enable integration of genetic data in a scalable manner requires adequate standards for storing and transmitting genomic data and standard API’s that allow ancillary systems to interface with the EHR. While we have developed an external data architecture for managing the return of results process it is unclear at this time how or if this should be integrated into the EHR.

Active decision support was more effective in engaging providers and causing change, especially in situations of a new diagnosis or new medication. Passive decision support was less effective. We believe this was due to providers not seeking genomic information prior to prescribing even when it existed. Lack of training of providers about genomics likely contributed to this, as well as the relatively small number of patients that had genomic results making the information not common to seek. We believe as genomic information becomes more available and providers become better trained on its use that it will be sought after and must have an established place in the EHR similar to imaging or laboratory results. More research is needed to understand and optimize workflows for genomic medicine and to test novel methods of delivering genomic information to patients and providers.