dischargers
An Abridge-powered agent that does the tedious prep for discharge medication reconciliation: it pre-populates a draft reconciled list from every source (home meds, inpatient orders, and the ambient transcript), then surf
Core trick
Builds a draft reconciled medication list from home meds, inpatient orders, and transcript, then shows source conflicts side by side with quotes.
Tags
Description
An Abridge-powered agent that does the tedious prep for discharge medication reconciliation: it pre-populates a draft reconciled list from every source (home meds, inpatient orders, and the ambient transcript), then surfaces exactly where the sources disagree — each conflict shown side-by-side with the transcript quote and prescriber as evidence. The clinician still makes every call; we just turn "assemble the picture from scratch, then decide" into "the picture is already assembled and the disagreements are highlighted — decide."
README excerpt
# discharge-agent An Abridge-powered agent that does the tedious **prep** for discharge medication reconciliation: it **pre-populates** a draft reconciled list from every source (home meds, inpatient orders, and the ambient **transcript**), then **surfaces exactly where the sources disagree** — each conflict shown side-by-side with the transcript quote and prescriber as evidence. The clinician still makes every call; we just turn *"assemble the picture from scratch, then decide"* into *"the picture is already assembled and the disagreements are highlighted — decide."* **Decision support, not decision replacement.** We make the decision easy, not the decision. > Every existing tool has either the **conversation** or the **structured record**. > We reconcile the two against each other — three-way (transcript ↔ FHIR ↔ > prescriptions), across **multiple prescribers**, with the **transcript quote as the > evidence** for each flag. See [`PROJECT.md`](./PROJECT.md) for the full brief: the validated problem, the pharmacist-absence gap, the multi-specialist hero scenario, the Epic (SMART on FHIR) integration story, and the competitive landscape. ## Repo layout - `PROJECT.md` — project brief / pitch - `docs/` — ideation, data-pipeline notes - `data/` — our own 100-record labeled discrepancy eval set + generators - `synthetic-ambient-fhir-25/` — provided Abridge dataset (25 synthetic ambient + FHIR encounters) used as the data substrate - `backend/discharge_agent/` — FastAPI se