Orders Service — AI Integration
Status: populated Owner: TBD Last updated: 2026-04-18 Companion: Service Template
1. Current AI Use
Tier: None (S0/S1)
No AI or ML components are active in the current implementation. CDS checks are rule-based, deterministic computations driven by terminology-service rules.
2. Planned AI Capabilities (S3)
2.1 Intelligent Order Suggestions (Advisory — Tier B)
| Aspect | Detail |
|---|---|
| Purpose | Suggest relevant order sets based on encounter context, diagnosis codes, and patient history |
| Input | Active encounter diagnosis codes, patient demographics, previous order patterns |
| Output | Ranked list of suggested order set names (advisory only) |
| Human gate | Clinician explicitly selects and instantiates suggested order set; AI never creates orders directly |
| Fallback | If model unavailable, display no suggestions; order entry functions normally |
| Model host | ai-gateway-service |
2.2 Duplicate Order Detection Enhancement (Advisory — Tier B)
| Aspect | Detail |
|---|---|
| Purpose | Semantic duplicate detection across similar medications (e.g., different brand names, same active ingredient) |
| Input | Current order code, active medication list |
| Output | Advisory CDS warning if semantic duplicate detected |
| Human gate | Same as existing CDS warning acknowledgement flow |
| Fallback | Fall back to exact-code duplicate check |
3. Hard Constraints
- AI must never create, activate, cancel, or modify an order.
- AI must never override a CDS hard-stop determination.
- AI suggestions must be clearly labelled as advisory in all UI surfaces.
- All AI calls route through ai-gateway-service; orders-service does not call external LLM APIs directly.
- AI model inputs and outputs must be logged for auditability.
- CDS hard-stops are rule-based only — AI may not change a hard-stop to a warning.