demographic-modeling

Demographic Modeling

The RM detail in the steps below is a working summary. The specs/rm-demographic guide loaded in Step 1 is authoritative — if they disagree, follow the guide (or confirm against type_specification_get).

Conflict Resolution

When guides conflict, apply this priority (highest first):

1. Rules and structural constraints
2. Privacy and separation principles
3. Anti-patterns
4. Principles and examples
5. Convenience

Step 1: Load Guides (MANDATORY)

Before any demographic modeling work, load the authoritative guides:

guide_get("openehr://guides/specs/rm-demographic")
guide_get("openehr://guides/archetypes/principles")

Load additional guides as needed:

• guide_get("openehr://guides/specs/rm-ehr") — for EHR/demographic separation context and cross-referencing patterns

Step 2: Clarify Use Case

Before designing, gather requirements:

• Entity types: Which PARTY subtypes are needed — PERSON, ORGANISATION, GROUP, AGENT?
• Roles: What roles do parties play? What capabilities and time validity apply?
• Relationships: What relationships exist between parties? What is the directionality?
• Deployment context: Is this a standalone demographic service, a PMI wrapper, or embedded within an EHR system?
• Privacy requirements: What level of PARTY_SELF identification is appropriate for the deployment?

Step 3: Research Before Creating

Before designing new demographic archetypes, ALWAYS search CKM first:

ckm_archetype_search("person")
ckm_archetype_search("organisation")
ckm_archetype_search("party identity")

Reuse-first principle: If a suitable demographic archetype exists, use it. Only create new archetypes when no existing archetype covers the concept. If a close match exists, consider specialization instead.

Use ckm_archetype_get to retrieve and review candidate archetypes in full before deciding. For a deeper reuse survey across varied phrasings (demographic concepts are easy to phrase several ways), dispatch the ckm-scout agent — it runs parallel searches and returns a ranked reuse/specialize/new recommendation without filling the main context with raw hits.