Communication, Support & Grief Models
Relevant source files
The following files were used as context for generating this wiki page:
- artifacts/api-server/src/routes/anthropic/systemPrompt.ts
- artifacts/api-server/src/intelligence/hospice/catalog.ts
- artifacts/api-server/src/intelligence/hospice/planner.ts
- artifacts/api-server/src/intelligence/hospice/evaluationFixtures.ts
- artifacts/mobile/data/resources.ts
- artifacts/api-server/src/intelligence/hospice/documentationCoach.ts
The Communication, Support & Grief Models within the Hospice Roadmap project provide a layered intelligence framework designed to support patients, caregivers, and family members through the end-of-life journey. These models go beyond simple information retrieval, offering role-aware guidance, emotional anchoring, and actionable scripts for difficult conversations. The system is designed to provide "wise counsel" that is clinically accurate yet deeply compassionate, mimicking the presence of an experienced hospice nurse.
Sources: artifacts/api-server/src/routes/anthropic/systemPrompt.ts:7-15, attached_assets/Pasted-Build-the-next-major-intelligence-architecture-for-Hosp_1774852812580.txt
Communication Framework and Personas
The system utilizes a structured communication model that adapts based on the user's role (e.g., caregiver, patient, physician, child). A central component of this is the "Ragna" persona, a 58-year-old Norwegian hospice nurse who provides a gentle, steady, and unhurried voice to users.
The Ragna Persona Rules
Ragna's responses follow a strict emotional and practical hierarchy:
- Acknowledge Emotion: Validate the user's feelings before providing data.
- Offer Comfort: Prioritize warmth over cold clinical information.
- Practical Counsel: Provide guidance in short, clear sentences.
- Open Door: End with an invitation to share more.
Sources: artifacts/api-server/src/routes/anthropic/systemPrompt.ts:18-36
Role-Based Communication Playbooks
The system adapts its tone and goals based on the identified role of the user. For instance, communication with children requires concrete, honest language, while communication with medical staff uses professional frameworks like SBAR.
| Role | Tone/Approach | Primary Goal |
|---|---|---|
| Caregiver | Supportive, grounding, practical | Empower with next steps; reduce burnout |
| Child | Honest, simple, concrete | Explain death without confusing euphemisms |
| Physician | Professional, concise (SBAR) | Effective clinical reporting and advocacy |
| Patient | Dignified, autonomous, gentle | Support wishes and comfort-focused goals |
Sources: artifacts/api-server/src/routes/anthropic/systemPrompt.ts:400-450, artifacts/api-server/src/intelligence/hospice/planner.ts:133-136
Support and Advocacy Models
The support architecture includes a "Care Gap Model" designed to recognize when a hospice provider is failing to meet standards of care. The system coaches users on how to document these failures and escalate them to administrators or state agencies.
This diagram represents the flow of identifying care gaps and the subsequent escalation steps suggested to the user.
Sources: artifacts/api-server/src/intelligence/hospice/planner.ts:25-30, artifacts/api-server/src/intelligence/hospice/catalog.ts:310-330
Clinical Reporting (SBAR)
To ensure effective communication between caregivers and hospice staff, the system generates scripts using the SBAR (Situation, Background, Assessment, Request) format.
- Situation: What is happening right now?
- Background: Patient diagnosis and recent history.
- Assessment: Severity of symptoms (e.g., pain scores).
- Request: What specifically is needed (e.g., a nurse visit or medication change).
Sources: artifacts/api-server/src/routes/anthropic/systemPrompt.ts:540-565, artifacts/api-server/src/intelligence/hospice/documentationCoach.ts:33-40
Grief and Bereavement Models
The project incorporates a multi-phase grief model that recognizes the nonlinear nature of loss. It covers anticipatory grief, acute bereavement, and "complicated" or prolonged grief.
Phases of Grief
- Anticipatory Grief: Mourning that begins before death (at diagnosis). The system validates this as a healthy preparation for loss.
- Acute Bereavement: Focused on the immediate 24-72 hours post-death, emphasizing shock, numbness, and the lack of a need to rush logistics.
- Extended Grief: Support for the weeks and months following loss, addressing "grief waves" and the loss of the caregiver identity.
Sources: artifacts/api-server/src/routes/anthropic/systemPrompt.ts:480-505, artifacts/api-server/src/intelligence/hospice/catalog.ts:250-280
Grief Types by Relationship
The system provides specific insights into how grief manifests based on the relationship to the deceased:
| Relationship | Common Grief Characteristics |
|---|---|
| Spouse | Loss of daily companion; disorientation; practical household disruption. |
| Adult Child | Loss of childhood witness; removal of the "generational buffer." |
| Parent | Most devastating loss; intense guilt; often lifelong intensity. |
| Sibling | "Disenfranchised" grief; loss of shared family history. |
Sources: artifacts/mobile/data/resources.ts:930-975
Intelligence Integration and Planning
The planner.ts module acts as the orchestrator for these models. It analyzes user input against the knowledge catalog and patient context to build a ResponsePlan.
This sequence diagram shows how the intelligence planner synthesizes disparate models into a single response plan for the AI.
Sources: artifacts/api-server/src/intelligence/hospice/planner.ts:12-25, artifacts/api-server/src/intelligence/hospice/planner.ts:138-150
Must Say / Must Avoid Directives
The planner enforces safety and emotional alignment through strict directives:
- Must Say: "Hearing is often the last sense to go," "You do not need to rush," "Call 911 NOW" (if red flags detected).
- Must Avoid: Vague reassurance, cold clinical language, defending a hospice when a care gap exists, or optimistic minimizing of symptoms.
Sources: artifacts/api-server/src/intelligence/hospice/planner.ts:80-105
Conclusion
The Communication, Support & Grief Models represent the "emotional intelligence" layer of the Hospice Roadmap. By combining structured clinical knowledge from the catalog.ts with a highly specific persona and a deterministic planning engine, the system ensures that users receive not just information, but the advocacy and emotional support necessary to navigate the complexities of end-of-life care.
Sources: artifacts/api-server/src/intelligence/hospice/planner.ts:150-160, artifacts/api-server/src/routes/anthropic/systemPrompt.ts:600-615