Discharge Domain Model

Canonical discharge concepts

Verdicts:

• ready
• ready_with_caveats
• not_ready

Blocker categories:

• clinical_stability
• pending_diagnostics
• medication_reconciliation
• follow_up_and_referrals
• patient_education
• home_support_and_services
• equipment_and_transport
• administrative_and_documentation

Core laws

• Discharge Gatekeeper MCP owns the baseline structured verdict and next-step spine.
• Clinical Intelligence MCP only adds bounded hidden-risk or contradiction findings.
• Hidden-risk findings must be cited and materially relevant.
• No component claims autonomous discharge authority.

Decision check

Ask:

1. Does this improve the answer to "Is this patient safe to discharge today?"
2. Does it sharpen blocker clarity or next-step actionability?
3. Does it belong in the deterministic MCP, the hidden-risk MCP, or the orchestrator?
4. Does it preserve the 3-prompt demo?

Do not

• drift into diagnosis or treatment planning
• widen the system into a generic hospital copilot
• move hidden-risk prompting into the deterministic MCP
• let narrative reasoning replace structured evidence