NEW YORK (GenomeWeb News) – The informatics subgroup of the Clinical Pharmacogenetics Implementation Consortium is involved in an ongoing effort to enable hospitals to implement pharmacogenomics-based clinical decision support in their electronic health record systems.
Highlights of the initiative and the resources the subgroup is developing were presented at last month's Summit on Translational Bioinformatics meeting in San Francisco.
CPIC was formed in 2009 as a collaborative project between PharmGKB and the Pharmacogenomics Research Network. Its goal was to establish guidelines that would help clinicians use the information provided by genetic test results to make prescribing decisions for specific drugs and select optimal treatments for patients. These guidelines, though, do not provide direction regarding whether or not tests should be ordered.
"The key assumption here is that ... high-throughput preemptive phenotyping will become more and more widespread, so you are going to have this scenario where the clinician will be faced with having the patient's phenotype available even if they didn't order the test," James Hoffman, a medication outcomes and safety officer for pharmaceutical services at St Jude Children's Research Hospital and one of the members of the CPIC informatics group, told GenomeWeb after the conference. The guidelines, he said, are designed to help clinicians in such cases understand how to incorporate the test information, where available, into decisions about appropriate therapies and treatments for their patients.
CPIC has published several guidelines since 2011 for genes such as TPMT, CYP2C19, and CYP2C9 and associated drugs such as thiopurines, clopidogrel, and warfarin, and made these available in PharmGKB. It established the informatics working group in 2013 to identify and resolve informatics challenges associated with implementing the guidelines within EHR environments and enabling pharmacogenetic clinical decision support (CDS) at the point of care.
One of the key challenges, Hoffman told GenomeWeb, is that unlike in other areas such as drug interactions, which already have established knowledgebases for decision support, the pharmacogenetic space is still maturing and as yet does not have similarly in-depth resources to work with and to provide support at the point of care. The hope and intent of the CPIC informatics group is to come up with resources and tools that make it possible to combine clinical information from the EHR with the information from the CPIC guidelines — and do it in a vendor-agnostic fashion so that different hospitals can adopt and implement the tools regardless of which EHR system they use, he said.
One of the group's activities has been to create comprehensive translation tables that help hospitals implement the CPIC guidelines within their EHRs and use them for clinical decision support. These tables match genotype/diplotype to phenotype to clinical recommendations based on the guidelines. They currently contain over 700 rows worth of information linking, in some cases, multiple genotype test results for a single gene to associated phenotypes and assigned risk levels. These tables, which are publicly available in PharmGKB, help hospitals design and implement workflows within their EHRs that match patients' test results to expected phenotype and recommendations about therapies or dosing. They could also set up alerts to warn physicians about treatments that would be harmful to patients based on their test results.
Sites that choose to use the tables in the EHRs are free to implement them in whatever way they choose. During his talk, Hoffman gave an example from St Jude's own system to illustrate how these tables would be implemented and used in actual practice. If a physician enters an order for the drug simvastatin in the EHR, the system checks for the results for an SLCO1B1 genetic test, which provides useful information about the patient's ability to metabolize statins. If the test results indicate that the patient is at risk for developing simvastatin-associated myopathy, the system then sends up an alert to that effect and suggests prescribing a lower dose or trying an alternate treatment.
In addition to creating the translation tables, the informatics subgroup along with members of the broader CPIC organization are also working on standardizing terms that are used in reporting on pharmacogenomics data across clinical laboratories. Currently, "if you look at what clinical laboratories report, it's all over the board," he said. For example, "for an individual carrying two non-functional TPMT alleles, we identified 11 different ways laboratories may report that result," and "that's a challenge for the field and for electronic health records where you need that standardization to make all this work and work efficiently," said Hoffman.
To that end, CPIC launched the CPIC term standardization project, which, Hoffman explained in his talk, seeks to standardize allele function and phenotype terms in the CPIC guidelines and harmonize that with terminologies used by other groups such as ClinGen. So far the group has defined terms that need to be evaluated and standardized and these are being evaluated by experts within the domain.
Another area of interest to the informatics group is coming up with vendor-agnostic mechanisms for sharing knowledge and information about clinical decision support across sites, Hoffman told GenomeWeb. Currently, it's pretty easy for organizations that use the same EHR vendor to share information with each other; sharing information across systems is a much harder prospect. "There are some standards and ways to do it but not a lot ... [that's] kind of a next step for us."