Skip to main content
Version: v0.1.6

Find RxNorm concepts by RXCUI, NDC or name

These calls run over a graph you loaded with loadRxNormGraph (see navigate the RxNorm graph). The rows below are synthetic, with made-up RXCUIs and NDCs.

By RXCUI​

import { loadRxNormGraph, getConcept } from "@cosyte/terminology";

const graph = loadRxNormGraph({
  conso: "3|ENG||||||||||RXNORM|SCD||lisinopril 10 MG Oral Tablet||N||",
  rel: "",
});

getConcept(graph, "3"); // => { rxcui: "3", tty: "SCD", name: "lisinopril 10 MG Oral Tablet", suppressed: false }
getConcept(graph, "4"); // => undefined

A concept takes its term type (tty) from its defining atom: the normalized name, such as IN, SCD or BN. Synonym atoms (PSN, SY, TMSY) name a concept but never type it. A concept whose atoms are all synonyms, or all of term types the engine does not model, is left out and reported on graph.warnings as TERM_RXNORM_UNTYPED_CONCEPT. If a concept you expected is missing, look there first.

import { loadRxNormGraph, getConcept } from "@cosyte/terminology";

const graph = loadRxNormGraph({
  conso: "8|ENG||||||||||RXNORM|SY||a synonym only||N||",
  rel: "",
});

getConcept(graph, "8"); // => undefined
graph.warnings[0]?.code; // => "TERM_RXNORM_UNTYPED_CONCEPT"
graph.warnings[0]?.line; // => 1

suppressed is true when the defining atom's SUPPRESS column is anything but N or empty.

By NDC​

Pass RXNSAT.RRF when you load the graph, and resolveNdc maps an NDC to the RXCUI it is attached to in that release.

import { loadRxNormGraph, resolveNdc } from "@cosyte/terminology";

const graph = loadRxNormGraph({
  conso: "3|ENG||||||||||RXNORM|SCD||lisinopril 10 MG Oral Tablet||N||",
  rel: "",
  sat: "3||||||||NDC|RXNORM|00000000001|N||",
  version: "RXNORM_2026AA",
});

resolveNdc(graph, "00000000001"); // => { resolved: true, ndc: "00000000001", rxcui: "3", status: "active", asOf: "RXNORM_2026AA" }
resolveNdc(graph, "0000-0000-01"); // => { resolved: false, code: "TERM_RXNORM_NDC_UNMAPPED", ndc: "0000-0000-01" }
  • The NDC is matched exactly as it appears in RXNSAT.RRF, the 11-digit form without hyphens. Normalize a 10-digit or hyphenated NDC before you call.
  • An NDC in the release is active as of that release, and asOf carries the version you passed when loading. The mapping changes between releases, so keep asOf with the answer.
  • This release reads only RXNSAT.RRF, so a resolved NDC is always active. The status type also names obsolete, alien and unknown, from RxNav's NDC history, which the engine does not read yet; it never reports them without that data.
  • An NDC not in the release is TERM_RXNORM_NDC_UNMAPPED, never a guessed RXCUI.

By name, approximately​

approximateMatch is the one search that is not exact, and it is labeled that way: every candidate has approximate: true and a score from 0 to 1: the number of words the concept's name and your query share, divided by the number of distinct words in the two together. No other call uses it.

import { loadRxNormGraph, approximateMatch } from "@cosyte/terminology";

const graph = loadRxNormGraph({
  conso: [
    "1|ENG||||||||||RXNORM|IN||lisinopril||N||",
    "3|ENG||||||||||RXNORM|SCD||lisinopril 10 MG Oral Tablet||N||",
  ].join("\n"),
  rel: "",
});

const candidates = approximateMatch(graph, "lisinopril tablet");
candidates.map((c) => [c.concept.rxcui, c.score]); // => [["1", 0.5], ["3", 0.4]]
candidates.every((c) => c.approximate); // => true

approximateMatch(graph, "no such drug"); // => []
approximateMatch(graph, "lisinopril", { minScore: 0.9, limit: 1 }).length; // => 1

Candidates come best first. The defaults are a minScore of 0.34 and a limit of 20. No match is an empty list, never a nearest guess. Show approximate candidates to a person to choose from; do not store one as if it were the code you were sent.

Reference​