One master case brief is better than five disconnected summaries

A safer information architecture for turning a complex care journey into consistent, audience-specific views.

Care coordination improves when every professional view is derived from one longitudinal narrative with visible provenance.

The repeated-story problem

Families navigating long-term care often maintain the integration layer themselves. They retell history to doctors, therapists, schools, and support teams; carry reports between appointments; remember what was tried; and explain which changes mattered. Every retelling can omit context or create a slightly different version of the same journey.

Generating a separate AI summary for each audience looks convenient but can reproduce the same fragmentation. Five generated documents can become five new sources of truth.

System design: one narrative, many views

The master case brief is a structured narrative derived from the profile, timeline, observations, interventions, responses, goals, documents, and open questions. It does not replace medical records. It organizes the care journey so the next conversation begins with continuity.

A view is generated from the master brief, not generated independently from scratch each time.

A therapist may need goals, sensory patterns, communication, current interventions, and response. A doctor may need symptoms, investigations, medication, red flags, and timeline. A school may need strengths, triggers, accommodations, and communication preferences. These are views over the same underlying case, not independent summaries. The architecture keeps audience rules, safety language, and source access explicit. A view can be refreshed when the master brief changes, and the user can see what was included or omitted.

System design: provenance is not optional metadata

Important claims preserve provenance, and provenance means four specific things attached to every observation, not a general sense of "where this came from."

• Who reported the observation — parent, therapist, doctor, or the system itself
• When it applied — the date it was true, not the date it was entered
• Source — which document, session, or conversation supports the claim
• Currency — whether it remains current, or has since been superseded

Conflicting information is represented as conflict instead of silently resolved by generation.

Safety comes from scope and provenance

A care-intelligence product should help people organize, understand, prepare questions, and track change. It should not manufacture certainty or impersonate a qualified professional. The system needs clear escalation rules, uncertainty language, access control, and a reliable distinction between reported observation, retrieved education, and generated interpretation.

• Keep reported facts separate from generated synthesis
• Attach time and source to important observations
• Show uncertainty and unresolved conflict
• Design professional escalation into the journey

From production: the view that almost became a new source of truth

Early in Nuvro, the fastest way to ship a new audience view was to let each one — therapist, doctor, school — regenerate independently from the underlying evidence, because each generation call could be shipped and tested on its own. The failure mode showed up immediately in review: two views describing the same intervention with slightly different dates, because each generation was reconstructing history from a slightly different retrieval slice instead of reading it from one place. That is exactly the fragmentation the product exists to remove, just moved one layer down into the architecture instead of into the family's retelling.

The fix was the decision recorded for the platform: audience briefs are projections of one longitudinal record, not independent generations, even though that made the core schema harder to get right up front. The trade-off was explicit and accepted — more upfront schema discipline in exchange for fewer contradictions and less repeated storytelling, which was the entire point of the product.

Mauri applied the same discipline to a single-user journey instead of a multi-professional one. Assessment, education, tracking, reviews, and graduation share one timeline by design, specifically so a person's fourth check-in isn't reasoning from a different version of their story than their first. And because the product treats recovery as having an end state, graduation and relapse plans were designed in from the start rather than optimized away in favor of permanent engagement — meaningful completion, not retention, was the metric that mattered.

What to carry forward

• Maintain one longitudinal narrative
• Derive professional views instead of duplicating records
• Preserve source, time, and uncertainty
• Keep guidance clearly separate from diagnosis
• Treat two views disagreeing on the same fact as a schema bug, not a content bug

Applied in the portfolio

• Nuvro Care Intelligence — A multilingual care-intelligence platform that turns a fragmented autism journey into a coherent, role-aware case narrative.
• Mauri — A phenotype-led guidance and recovery-tracking experience for women navigating PCOS, endometriosis, and hormonal health.