Infrastructure for Pharmacogenomic Decision Support

Pharmacogenomic decision support requires formally codified gene-drug interaction rules: IF (Patient.CYP2C9 = *2/*3 AND Medication = warfarin) THEN recommend 25-50% dose reduction WITH supporting CPIC guideline reference. CPIC and PharmGKB guidelines must be translated into machine-executable decision logic, not referenced as PDF documents. Phenotype classifications (poor/intermediate/extensive/ultrarapid metabolizer) must be formally defined with associated dosing implications per drug class for the AI to generate specific, actionable dosing recommendations without pharmacist re-derivation of each alert.

Pharmacogenomic decision support requires systematic capture of PGx test results in structured form linked to patient records, with consistent fields for gene, variant, and phenotype classification. When a PGx panel is ordered, results must flow through a defined capture pathway—not as free-text lab notes—ensuring the AI can query genotype data at medication ordering time. Template-driven capture of PGx ordering rationale and clinical response to recommendations supports ongoing improvement of the decision support logic.

Pharmacogenomic decision support demands formal ontology: Gene entities (CYP2C9, CYP2C19, VKORC1) with Variant attributes (*1, *2, *3), Phenotype classifications mapped to Diplotype combinations, and Drug-gene interaction rules with Dosing recommendations. PharmGKB and CPIC provide the foundational knowledge graph that must be implemented as machine-readable schema. Without formal entity-relationship mapping from Patient.Genotype → Phenotype → Drug.DoseAdjustment, the system can't traverse the gene-to-recommendation pathway autonomously.

Pharmacogenomic decision support requires unified API access to the genomics/lab system (PGx test results), CPOE (active and new medication orders), EHR (allergies, phenotype documentation), drug-gene interaction databases (CPIC/PharmGKB APIs), and pharmacy dispensing system. At medication order entry, the system must simultaneously query patient genotype, active medications for interaction checking, and current dosing recommendations—requiring a unified access layer that assembles this multi-source context in real-time without pharmacist data retrieval.

Pharmacogenomic guidelines evolve as CPIC publishes new or updated guideline versions and as PharmGKB adds newly characterized gene-drug interactions. Event-triggered maintenance—new CPIC guideline publication triggers review and integration of updated dosing recommendations—is appropriate for this capability. While the pace of PGx evidence evolution requires attention, same-day updates aren't typically required; a policy-triggered update process within days of guideline release is sufficient.

Pharmacogenomic decision support requires API-based connections between the genomics laboratory system, CPOE, EHR clinical documentation, pharmacy dispensing, and external PGx knowledge databases (CPIC, PharmGKB). These connections enable real-time gene-drug interaction checking at medication ordering. The existing EHR-to-pharmacy-to-lab integration infrastructure supports most connections; the addition of genomics laboratory API access and CPIC knowledge database feeds completes the required integration set.