Skip to main content
Version: v0.1.10

Edit a document

You have a C-CDA document, and you need to correct one section (add a Problem, swap out the Medications) and send the fix, without disturbing every other section or losing the audit trail that this is a new version of the original.

editCcda(doc, options) is the read→edit→write loop. It takes a document from parseCcda, rebuilds only the section you target (through the same emitters buildCcda uses), and carries every other section through byte-for-byte. By default it stamps a CDA R2 revision: a new document id, the same version-series setId, an incremented versionNumber, and a relatedDocument typeCode="RPLC" naming the version it replaces.

import { buildCcda, editCcda, parseCcda } from "@cosyte/ccda";

// Start from a document the parser produced (here, one we built and re-parsed).
const original = parseCcda(
  buildCcda({
    patient: { mrn: "MRN-00042", given: ["Jane"], family: "Doe" },
    documentId: "DOC-1",
    problems: [{ problem: { code: "38341003", displayName: "Hypertension" }, status: "active" }],
    medications: [{ drug: { code: "197361", displayName: "Amlodipine 5 MG" } }],
  }).toString(),
);

const revised = editCcda(original, {
  sections: [
    // Replace the Medications section wholesale…
    {
      kind: "medications",
      mode: "replace",
      content: [{ drug: { code: "314076", displayName: "Lisinopril 10 MG" } }],
    },
    // …and add a Family History section the source did not have.
    {
      kind: "familyHistory",
      content: [
        {
          relative: { relationship: { code: "72705000", displayName: "Mother" } },
          observations: [{ condition: { code: "73211009", displayName: "Diabetes mellitus" } }],
        },
      ],
    },
  ],
});

revised.getMedications()[0]?.drug?.code; // => "314076"
revised.getFamilyHistory().length; // => 1
// The untouched Problems section is preserved:
revised.getProblems()[0]?.problems[0]?.value?.code; // => "38341003"

// A CDA R2 revision that replaces DOC-1 (version 1):
revised.header.versionNumber; // => 2
revised.header.relatedDocuments[0]?.typeCode; // => "RPLC"
revised.header.relatedDocuments[0]?.parentDocument.ids[0]?.extension; // => "DOC-1"

// Re-parsing the edited XML is a fixed point.
parseCcda(revised.toString()).toString() === revised.toString(); // => true

How a revision is stamped​

By default an edit stamps a CDA R2 revision: a new ClinicalDocument.id, the same version-series setId (minted when absent), an incremented versionNumber, and a relatedDocument typeCode="RPLC" naming the prior version, inserted at their CDA R2 XSD sequence positions, and surfaced back on the parsed header (setId / versionNumber / relatedDocuments). Pass revision: false to edit in place. A source with no ClinicalDocument.id cannot be revised: the RPLC link's parentDocument.id is a CDA R2 SHALL (1..*) and there is no prior-version id to name, so editCcda throws CcdaEditError (SOURCE_MISSING_ID) rather than fabricate one; use revision: false to edit it in place.

A minted setId is labeled as synthetic​

CDA R2 requires a replacement and its parentDocument to share a version-series setId, so one is minted when the source has none; minting it invents an identifier, and the invention is made obvious rather than hidden. A minted id is SYNTHETIC-SETID-… under a synthetic assigning-authority root (2.16.840.1.113883.19.5.99999, in HL7's example arc), and isSyntheticSetId(doc.header.setId) is the check. Both halves are required, because either alone is something a real document could carry. A setId the source already asserted, or one you pass as revision.setId, is never relabelled.

Nothing forces a receiving system to read the label. A receiver that ignores the prefix and the root treats a minted setId exactly as it treats a real one, and this library cannot make it do otherwise. A false from isSyntheticSetId is likewise not a promise the id is real, only that this library did not mint it under this scheme.

An edit reports what it wrote​

An edit reports what it wrote, over the sections it actually grafted. A grafted Plan of Treatment can carry a Planned Medication Activity short the effectiveTime its template SHALLs, so the returned document raises MISSING_PLANNED_MEDICATION_EFFECTIVE_TIME, appended after the re-parse's warnings, with the emitted XML unchanged. The scope is deliberately narrow in both directions and neither half is an optimization. It reads what survived into the emitted document, so an offending edit that a later edit in the same call discarded says nothing (sections is an ordered list, and reading the list rather than the result reports a violation against a document that does not have one). And it covers only what this call grafted, so an offending act the source brought with it is never re-reported: an edit is not a validator of a document its caller did not write.

Modes, and what an edit refuses​

An edit is fail-safe: an unedited section is carried by reference (never dropped), an empty content list emits a spec-clean nullFlavor="NI" shell (never fabricated entries), and an edit that would drop a SHALL required section throws a typed CcdaEditError. The required set is the one the document was parsed under: an R1.1-origin document (no R2.1 @extension stamp) is edited under the reduced set, and a document stamped for a release we do not model is not evaluated. mode is "add" (require absent), "replace" (require present), or "upsert" (default: replace-or-add). editCcda(doc, { terminology }) forwards an optional adapter to the final re-parse, so an adapter-rejected code in a grafted or untouched section is flagged; the edit still emits every code verbatim.

What editing does not cover​

The twelve editable section kinds are problems, allergies, medications, results, vitalSigns, immunizations, procedures, encounters, socialHistory, pastMedicalHistory, planOfTreatment, and familyHistory. Functional Status and Mental Status are buildable but not editable: each is assembled from three separate content lists, which the single-list edit shape does not fit. The Referral Note's narrative-only Assessment and Reason for Referral sections are likewise not editable. There is no entry-level append: adding one problem to an existing Problems section means a replace carrying the full entry set, which rebuilds that section from your typed input, so anything in the original section you do not carry over, including detail this library does not model, is absent from the result (every section you did not target is still carried through byte-for-byte). There is no way to remove a section at all, and the APND / XFRM document relationships are not implemented: an edit stamps RPLC only.

Reference​