Evidence & methodology
How PolySafe's dataset is built, where it comes from, and what it doesn't cover yet.
Primary evidence sources
-
Product labeling
Current prescribing information / package labeling is the primary source for contraindications, warnings, and documented pharmacokinetic interactions (e.g. CYP3A4/CYP2C9 inhibition, P-glycoprotein effects).
-
RxNorm (NLM)
Used as the reference point for standardized, normalized drug naming — so "the same drug" is represented consistently even when different brand or generic names are used.
-
DailyMed (NLM)
Current FDA-submitted labeling, used to support interaction, contraindication, warning and clinical-pharmacology claims.
-
Published clinical literature and case reports
Used to supplement labeling where clinically important — particularly for quantifying real-world effect sizes (e.g. specific INR changes, AUC/Cmax increases) rather than relying on mechanism alone.
PolySafe does not copy proprietary interaction content from commercial databases like DrugBank, Micromedex or Lexicomp. Content is drafted from public labeling and literature, and licensing terms are checked before any future redistribution.
How severity is assigned
Each record gets one of four severity levels — 🚫 Contraindicated (C1), 🔴 Major (C2), 🟠 Moderate (C3), or 🟡 Mild (C4) — based on what the labeling or literature actually supports, not on how alarming the mechanism sounds. Two rules guide this specifically:
- C1 is reserved for genuine contraindications. A drug that should generally be avoided (like grapefruit juice with cyclosporine) is still C2/Major with an "avoid" action — not automatically C1 — unless labeling states an outright contraindication.
- Route and formulation matter. The same drug can carry different risk by route — e.g. IV ceftriaxone with IV calcium-containing solutions is a real neonatal contraindication, while oral calcium has no established interaction with ceftriaxone at all. Collapsing these into one record would create false alarms.