PillHarmonics: An Orchestrated Medication Clinical Decision Support Service - Project Summary
Medication related adverse events account for over 2 million hospital stays and 3.5 million physician
office visits per year. Medication decision support, when implemented correctly, can have a significant impact
on these numbers, enhancing patient safety and improving drug efficacy. But while drug decision support is
now commonplace in Electronic Health Records (EHRs), many issues remain, and clinicians are generally
unsatisfied with the lack of patient specificity and inappropriate context of medication alerts.
Add to this the fact that drug-gene alerts are becoming increasingly important. Studies show that over
half of all primary care patients are exposed to pharmacogenomics (PGx) relevant drugs; that 7% of FDA-
approved medications and 18% of the 4 billion prescriptions written in the United States per year are affected
by PGx interactions; and that nearly 98% of individuals have at least one actionable variant by current
guidelines.
PGx findings are most commonly integrated into the EHR as non-actionable PDF reports. Structured
EHR-specific solutions are emerging, and several groups are experimenting with HL7 FHIR and CDS Hooks
standards. A common theme across these efforts is that PGx is implemented apart from other types of
medication decision support, leading to disjointedness of alerts. For many years, groups have suggested the
need to integrate PGx with other types of identified medication interactions. Evidence suggests that such a
holistic approach can address patient safety issues (e.g., by juxtaposing conflicting drug recommendations)
and alert fatigue (e.g., through greater alert precision).
However, merging PGx into an environment that already has many usability challenges risks obscuring
the benefits of such alerts. In response, this project aims to develop a medication decision support service,
‘PillHarmonics’, that seamlessly integrates drug-gene interaction checking with other types of medication
alerting (such as drug-drug, drug-allergy, and drug-condition), thereby enhancing patient safety through
minimization of adverse drug events and decreasing alert fatigue via more precise surfacing of relevant alerts.
In the planned prototype, PillHarmonics will gather FHIR-formatted clinical data from an EHR, simulated
by a HAPI FHIR server; FHIR-formatted genomic data, in this case from Elimu’s genomic data server; and drug
knowledge, in this case from First DataBank and PharmGKB. The service translates identified interactions into
normalized data elements which are exposed as structured FHIR DetectedIssues, one DetectedIssue per
interaction. The PillHarmonics service will be demonstrated via a CDS Hooks application that generates
integrated alerts in response to the addition of tacrolimus or clopidogrel to a patient's existing medication
regimen. AIM 2 evaluation will entail a ‘perceived usefulness’ assessment of the PillHarmonics algorithm using
an established evaluation instrument.
