Synthetic medication courses¶

openmed.training.synthetic.medication_courses materializes explicitly declared synthetic prescription, titration, adherence, discontinuation and outcome facts. It bundles no drug vocabulary, dose recommendations or real patient data.

MedicationCourseSpec uses controlled medication/indication/unit/route codes, relative start/stop days, a positive initial dose, and optional DoseChange, AdherenceEvent and MedicationOutcome collections. Dose changes lie strictly inside the active course; taken/missed events lie in its half-open active interval. Duplicate/conflicting same-day dose or adherence declarations reject. Outcomes may follow discontinuation only inside an explicitly declared followup window. Their gold relation is observed_after_start, never inferred causation.

generate_medication_course(spec, seed=...) returns stable opaque IDs, gold half-open regimen intervals, ordered events and relation labels. Seeds vary identities; doses and outcomes come from the caller's declared synthetic recipe, never randomly invented clinical facts. No model, network or restricted terminology download occurs. Each event family is capped at 1,000 entries.

Use validate_medication_course independently to reject regimen gaps, invalid doses, dangling relations, mismatched stop dates or adherence outside the linked regimen. Controlled error categories contain no private source values. These fixtures support extraction/relationship evaluation; they do not establish drug safety, clinical plausibility, treatment utility or medical-device readiness.