The Problem: Care managers coordinating perinatal collaborative care don't have a registry dashboard built for how AI could actually help them — existing clinical standards define solid baseline registry functions, but none account for things like disengagement prediction or automated caseload summarization.
What I'm Doing: Running a research process that uses established frameworks (AIMS Center, AHRQ) as the floor, caregiver interviews to localize requirements to Baby2Home, and direct testing of candidate AI features with care managers before any feature gets locked into a design.
Status: Wireframe due December 2026. Currently in literature review and framework analysis.
01. The Scene // "The gap wasn't where I expected it"
Collaborative Care Model registries have been studied for a while — the AIMS Center and AHRQ have both published solid frameworks for what a patient registry dashboard needs to do. So going in, I expected the gap in Baby2Home's dashboard to be a gap in coverage — missing fields, missing views, the usual.
It wasn't. When I mapped the existing literature across collaborative care registries, patient registry frameworks, clinical dashboard/HCI design, and the newer AI-supported learning health system research, the core registry functions were well established and largely consistent across sources. What was consistently absent — across all four literature categories — was any real treatment of AI-driven prioritization, disengagement prediction, automated summarization, intelligent recommendations, workload balancing, or integrated maternal-infant-family tracking.
"Not what does a CCM registry need — that's already well answered. Where could AI actually extend it, and would care managers trust that?" — the reframed research question
02. What I'm Actually Doing // "Five phases, one hard deadline"
Phase 1 — Foundation & Framework Analysis (Jul–Aug 2026, current phase):
- Literature review across four categories: CCM registries, patient registry frameworks (AHRQ), clinical dashboard/HCI design, and AI-supported care management research.
- Framework analysis mapping AIMS Center's workflow requirements and AHRQ's structural requirements directly onto Baby2Home, to confirm exactly which AI-related gaps neither framework addresses.
- Interview protocol design — a semi-structured guide including a dedicated module that walks care managers through six candidate AI features to gauge trust, desirability, and workflow fit for each.
The roadmap ahead:
Phase 2 — Discovery (Aug–Oct 2026)
Recruiting care managers, clinicians, and therapists across partner sites; one-on-one interviews plus workflow shadowing where possible, testing reactions to the candidate AI features alongside standard workflow questions.
Phase 3 — Synthesis (Oct–Nov 2026)
Consolidating interview findings with the framework analysis into core dashboard requirements, then scoring six candidate AI features on desirability, trust, and workflow fit to select 2–3 to actually design.
Phase 4 — Wireframe Design (Nov–Dec 2026)
Low-fidelity sketches, then iterative wireframes tested specifically for whether the AI features read as clear and trustworthy — not just usable. Final wireframe delivered to Baby2Home stakeholders in December.
Phase 5 — Validation & Dissemination (Jan–Apr 2027)
Usability/heuristic review of the locked wireframe, and manuscript writing framing the AI-integration gap as the paper's contribution.
03. The Tension Already in Play // "No pivot yet — but here's the real design bet"
Even at this early stage, there's a real design decision baked into the plan worth naming: I'm not allowed to just pick the AI features that sound most impressive.
04. The Plan // "Not a solution yet — this is what Phase 4 builds toward"
The target deliverable is a single wireframe that does two things at once: meets the baseline CCM registry requirements defined by AIMS Center and AHRQ, and layers in 2–3 AI features that interview data shows care managers would actually trust and use.
Leading Candidate: Disengagement Prediction
Pending validation. Flagging patients at risk of dropping out of collaborative care before it happens.
Leading Candidate: Automated Caseload Summarization
Pending validation. Reducing the manual overhead of reviewing large caseloads.
Leading Candidate: Risk-Based Prioritization
Pending validation. Helping care managers triage who needs attention first.
These three are likely candidates, not decided ones — final selection depends on interview results from Phase 2 and 3. The design principle underneath all of it: AI features earn their place in the dashboard the same way any core registry function would — by being grounded in what the standards require and validated by the people who'll actually rely on it every day.
05. Where Things Stand // "An honest status check, not a scoreboard"
This is deliberately not a results section, because there aren't results yet — and I'd rather say that plainly than dress up a work-in-progress as a finding. As of now, the literature review and framework analysis are underway, and the interview protocol is being developed. Care manager recruitment starts next. Usability validation and manuscript writing are intentionally scheduled after the December wireframe deadline, so they can't compete with it for time.
06. Behind the Scenes // "Director's Commentary (so far)"
07. Souvenirs // "Coming as each phase wraps"
This case study will grow with the project. Artifacts will be linked here as they're delivered:
Structuring five months of research to protect one hard deadline is its own design problem — the plan explicitly pushes validation and writing to after December so they can't quietly eat into design time. Building AI-feature reactions into the interview protocol from day one, rather than testing a finished AI concept later, was a deliberate choice — it means trust and desirability get to shape which features get designed, not just how they look once decided. Something I'm watching for as this continues: whether the "one feedback cycle" constraint ends up being enough, or whether the plan needs to flex there once real wireframes are in front of care managers.