Every clinical claim cites a public source.

Normalized names and identifiers (RxCUIs) for prescription and OTC drugs in the U.S. Maps brand names, generic names, salts, and dose forms to a unified concept.

Each parsed medication is sent to the RxNorm REST API to resolve a rxcui. All downstream queries use the RxCUI, never the raw name.

Structured Product Labeling (SPL) drug labels — dosage & administration, contraindications, drug interactions, clinical pharmacology, and more.

Fetches the indications_and_usage, dosage_and_administration, and drug_interactions sections by RxCUI. Cited at the section level (e.g., FDA §2.1).

Authoritative source of the same SPL labels in openFDA, with additional historical and version data. Used as a fallback and for label provenance.

Falls back to DailyMed when openFDA returns no record for a particular RxCUI. Same access pattern — RxCUI only.

Minimum list of drug-drug interactions recommended for clinical decision support. Originally published by the ONC, peer-reviewed. 302 paired rules, severity-graded.

Target core dataset — transcription from the published PDF into structured JSON is in progress. Each shipping rule cites ONCHigh #N. Quarterly review.

Drugs associated with QT prolongation and torsades de pointes, graded by risk: Known / Possible / Conditional / Avoid in congenital LQTS.

QT-prolonging drugs are flagged when multiple appear in the same regimen — combination raises risk multiplicatively. Cited as CredibleMeds [tier].

Specific timing rules derived from FDA labels — e.g., levothyroxine ↔ calcium 4h gap, tetracyclines ↔ dairy 2h gap, sucralfate ↔ ciprofloxacin 6h gap.

Each rule is a small JSON object with the two RxCUI patterns, a required gap, and a citation to the originating FDA label section. Subject to quarterly review.

Sewa ingests DDInter 2.0’s Major-severity drug-drug interactions (37,687 pairs across 1,492 drugs, RxCUI-keyed). Used as a secondary tier for pairs the curated set doesn’t cover. Curated rules always win on overlap.

When a DDInter-sourced flag appears in your schedule, Sewa reports the severity and points you to your pharmacist — the bulk DDInter download does not include mechanism or management text, so we’re not in a position to repeat clinical guidance ourselves. Moderate, Minor, and Unknown-severity records are present in the parsed dataset but not surfaced in v0.1 — they introduce more clinical noise than signal at the patient-to-pharmacist handoff. Your pharmacist has access to fuller interaction databases (Lexicomp, Micromedex) for comprehensive screening.