Clean medication lists
from messy data.
Your patient's meds are scattered across EHR orders, pharmacy dispenses, and self-reports. MedListIQ reconciles them into a deduplicated, clinically-classified list with a full provenance trail.
Many resources in. One clean med list out.
Reconcile across orders, dispenses, administrations, and patient-reported history using semantic matching — not just code equality. We handle the full terminology landscape (RxNorm, NDC, ATC, UMLS, free text) and derive status from auditable rules. Every medication ships with a provenance trail: inputs consumed, rules fired, fields enriched.
Many PDFs in. Same clean med list out.
Submit a referral packet, an H&P, a discharge summary, or progress notes. The same reconciliation engine runs across the document text — collapsing brand and generic mentions, surfacing conflicts when sources disagree, and preserving the page, section, and exact text-snippet evidence for every match.
Four phases of our pipeline
We designed our pipeline to handle the complexity of real-world medication data. Every phase was designed to be flexible to a wide variety of inputs and use cases, and native to modern healthcare and interoperability standards.
Extract
Find every medication in your data. Doesn't matter if it's structured records, scanned PDFs, or free-text notes.
Normalize
Map every extracted medication to standardized healthcare codes.
Reconcile
Deduplicate drugs across all sources, sort out conflicts, and keep the evidence trail intact.
Enrich
Expand every field with defacto standards across healthcare to promote semantic interoperability.
Typically three hires. One API.
Production medication reconciliation sits at the intersection of three specialties. Get the equivalent of three specialist hires as one API call — at a fraction of the cost:
Clinical pharmacy
What 'active' actually means when a stopped order and a recent dispense both exist. When to trust patient-reported meds. How to weight conflicting signals.
Health informatics
RxNorm TTY ranking (SCD > IN). SNOMED + NCIT enrichment for routes and methods. ICD-10 / SNOMED indications. Free-text to discrete.
FHIR integration
MedicationRequest / Dispense / Statement dedup. Inline codeableConcept vs. external medicationReference. US Core alignment. Bundle vs. flat-array.
Built for production use.
Every decision is auditable
Every inferred medication ships with its sources (which FHIR resources contributed), evidence (which rules fired), and enrichments (what we derived from the input). No black boxes.
Pick your payload
Three verbosity levels. `minimal` for a med-list UI alert. `standard` for EHR integration with RxNorm codes and dosage. `full` for clinical decision support with the complete provenance trail.
Normalized to the right codes
RxNorm SCD-preferred TTY ranking. SNOMED + NCIT mapping for routes and methods. ICD-10 / SNOMED for indications. Discrete data you can feed to downstream systems — or render human-readable.
Deterministic + versioned
Every response carries a ruleset_version (YYYY-MM-DD.vN). Pin to a version for stable integrations. No ML drift, no hallucinations, no unexpected behavior changes.
Stateless by design — no PHI persisted
Your FHIR payloads are processed in memory and discarded with the response. We never write patient data to disk. Only request metadata (timestamps, counts, status codes) is retained — for billing and observability, never the payload.
Who's building with this.
Care coordination
Platforms stitching records across multiple providers. MedListIQ dedups across EHR + pharmacy + patient input.
EHR integrations
Apps pulling meds from Epic, Cerner, Athena. MedListIQ normalizes the per-vendor quirks into one shape.
Telehealth + virtual care
Quick med-list ingest from a patient's uploaded records or HIE bundle. Ready for clinical review in seconds.
Clinical decision support
Rule engines need a clean active-medication input. Feed our output in; skip the reconciliation step.
“A medication inference engine built on substantive knowledge of prescriptions by an actual pharmacist — with zero introduced error from AI? That’s the dream.”
“Wow. This is something we’d probably want to use.”
Try it in 60 seconds.
Sign up, mint a key, interact with our preset scenarios in the playground or integrate with your own application.