Hospital Discharge Planning Cockpit (Next.js) — Workflow Coordination + Safe, Explainable AI Summaries

Align on target users (RN case manager, hospitalist, pharmacist, social worker, utilization review, unit clerk), key entities (patient, encounter, discharge plan, tasks, meds, insurance issues, notes, documents), and the end-to-end discharge workflow states (Admit → Inpatient → Discharge Planning → Ready Pending Items → Discharged). Produce a lightweight spec with acceptance criteria for cockpit views: readiness checklist, follow-up tasks, medication review, insurance blockers, care-team notes, and discharge summary generation.

Design TypeScript domain models and JSON fixtures for synthetic patients/encounters, including demographics, diagnoses/problems, labs/vitals (synthetic), medication list (home/inpatient/discharge), follow-up tasks, insurance prior-auth blocks, notes, and audit events. Establish a synthetic data policy: no real PHI, generate deterministic synthetic IDs, and include a clear banner in-app indicating synthetic/demo mode.

Implement a privacy posture checklist for the prototype: synthetic-only guardrails, no external telemetry by default, redact/avoid free-text PHI patterns, and add in-app disclaimers that outputs are assistive and require clinician review. Add an audit log model for key actions (task completion, med reconciliation changes, summary generation) to support traceability.

Set up route structure and layout for cockpit: /dashboard, /patients/[id], /patients/[id]/timeline, /patients/[id]/discharge. Add shared components (patient header, status chips, task list, note composer, blockers panel). Ensure accessibility basics (keyboard navigation, ARIA for key controls) and consistent visual language for safety-critical alerts.

Implement local persistence (IndexedDB or localStorage via a thin repository layer) for synthetic data edits: tasks, notes, med review decisions, blocker resolutions, and generated documents. Provide migration/versioning for stored schema, and add reset-to-fixtures capability for demos/testing.

Build the primary discharge cockpit panel: readiness checklist (clinical stability, pending tests, mobility/OT, patient education, transportation), follow-up tasks (assigned owner + due date), medication review status, insurance blockers, and care-team notes. Include clear “not ready” reasons, required next actions, and a timeline snippet of recent events.

Implement task creation and routing with owners, due times, status, and dependencies (e.g., prior auth must resolve before discharge). Add escalation rules (overdue/high-risk), quick handoff between roles, and a “today view” for care-team workload. Ensure tasks are explainable: show why a task exists and what evidence triggered it.

Create medication review surface: compare home vs inpatient vs proposed discharge meds, highlight deltas, duplications, dose changes, and allergy/interaction flags (rule-based on synthetic data). Add structured decisions (continue/stop/change) with rationale and required co-sign prompts. Gate discharge readiness on unresolved high-severity medication issues.