Curate reason for medication discontinuation or switch

This recipe is part of Advanced Data Curation: capturing facts that do not exist in OMOP. It uses Data Curation with a custom registry to pull clinically decisive facts that live only in free-text notes and have no native home in OMOP's coded domains.

User goal
Capture why a medication was stopped or switched, which is context coded medication exposure alone usually cannot represent.

When to use this
Use for real-world tolerability, adherence, switching-pattern, or pharmacovigilance questions where the rationale behind a stop or switch matters.

Required inputs/data

Dataset built from a cohort or ingestion results (see the "Build a dataset" recipe reference in Scenario: S3 to cohort to dataset to Data Curation) for patients with the relevant medication exposure
Note types likely to hold the rationale, such as progress notes, medication reconciliation, clinic notes, oncology visit notes, or discharge summaries
A custom registry defining the discontinuation fields

Custom registry fields (illustrative)

discontinued_drug
discontinuation_reason
adverse_event_detail
switched_to
discontinuation_date

Patient Journey Intelligence features involved

Dataset Builder / Explorer
Data Curation Studio
Custom Registries
Curation Results and evidence review
Result export

Step-by-step workflow

Build a dataset scoped to patients and note types where the drug history appears.
Define a custom registry with the discontinuation fields.
Start a curation job over the dataset.
Review each extracted reason against its evidence span.
Correct and validate values; resolve notes that list multiple reasons.
Export the structured table and join it back to the patient cohort on patient, drug, and date.

Expected output

Structured discontinuation reason record per drug episode
Evidence-linked, reviewed values
Exportable table that augments coded medication history

Edge cases / limitations

A single note may list several reasons; capture the primary reason and retain evidence for the rest.
Distinguish "held" or "on hold" from true discontinuation.
Free-text drug names must be normalized before joining.
Absence of a stated reason is not the same as no reason.

User value

Enables tolerability, adherence, and switching analyses that coded data alone cannot support.

Next step​

Next step