ClarityCarea concept study
AI patient prioritization · built to be trusted

ClarityCare

A risk score is a black box that says “trust me.” I designed the glass box around it.

Care coordinators have to catch the patients most likely to fall through the cracks before the next crisis — but the tooling hands them an undifferentiated wall of EHR data and no idea where to start. ClarityCare is a self-directed concept for turning an explainable AI risk model into a workflow a clinician can actually trust and stay in charge of.

6 core features2 surfaces · desktop + field4 data sources
RM
Robert Maddox
MRN 4471-092 · 64y
HIGH
Why this score

High risk due to 2 missed follow-ups, 1 ER visit this month, and unfilled medications.

missed follow-upsER utilizationmed adherence
Accept suggested action
Override
01
Predict risk
rank the wall so the riskiest surface first
02
Explain it
render the model's reasoning as a sentence
03
Keep the human in charge
a structured override, never a rubber stamp
04
Learn from the override
the correction becomes the next training signal
Explainable-AI (XAI) designSDOH-informed healthcare UXMulti-surface designData-heavy dashboardsHuman-in-the-loop workflowsClarityCare · concept project
The brief

A self-directed concept, scoped like a real brief.

No client handed me this. I wrote the brief myself and held it to the same bar as paid work: a real user with a real deadline, a genuinely predictive — and genuinely opaque — model, and a product that has to earn a busy clinician’s trust in seconds.

A concept project — the ML model, SDOH indices, and named patients are illustrative sample data, assumed to make the argument concrete. Nothing here is staffed, validated, or measured.

01The problem

A wall of data, and no idea where to start.

Care coordinators own the patients most likely to fall through the cracks — missing appointments, cycling through the ER, quietly losing housing or food access. They’re meant to catch them before the next crisis. The tool hands them an undifferentiated EHR and a to-do list that never ends.

“I know I’m missing someone important, but I can’t see who — and I don’t have time to guess.”— Illustrative care-coordinator persona, written for the concept
01

The important patient is invisible in the data.

02

No way to see who it is.

03

No time to reason it out by hand.

02The goal

Enhance judgment. Never replace it.

The north star: the AI proposes, the clinician disposes. If a screen ever made the coordinator feel like a rubber stamp for the algorithm, it was wrong.

The AI · proposes

Predicts risk, surfaces the reasons, and suggests a next step — always with its own uncertainty on the table.

The clinician · disposes

Accepts, overrides with a structured reason, or adds context the model never had. The last word is human.

03Why SDOH sits at the center

The biggest health risks often aren’t medical.

Two kinds of input feed the SDOH layer — what a coordinator asks the patient, and public data about the neighborhood they live in. All four are validated, publicly documented standards; the design work was translating them into something usable in the field.

Asked of the patient

PRAPARE

A national, standardized questionnaire covering housing, food, transportation, employment, and social support.

nachc.org/resource/prapare
Pulled from public data

CDC / ATSDR SVI

Social Vulnerability Index — ranks how vulnerable a neighborhood is using census data: poverty, vehicle access, crowded housing.

atsdr.cdc.gov/…/svi
Asked of the patient

CMS AHC HRSN

Accountable Health Communities screen — a short 10-item form for utility, personal-safety, and interpersonal needs.

cms.gov/…/ahcm
Pulled from public data

ADI

Area Deprivation Index — ranks neighborhood disadvantage from income, education, employment, and housing quality.

neighborhoodatlas.medicine.wisc.edu
The thesisDesign with AI, not around it

The real work wasn’t scores. It was trust.

The model reads four sources — and every feature answers a skeptical clinician’s question:

EHR events — missed appts, ER visits, refillsClaims & utilizationCommunity SDOH indices — SVI, ADIPatient intake surveys — PRAPARE, HRSN

→ a genuinely predictive model, and a completely opaque one.

“Where do I start?”Smart Triage Queue
“Why should I believe this?”Insight Cards
“Am I still in control?”Override flow
“What if I just learned something?”Real-time SDOH
“How did we get here?”Timeline view
“What am I missing right now?”Notifications
smart triage queue — ranked by risk (concept)
ClarityCareTriage QueuePatientsInsightsReports
CH
Filters
Risk level
Risk driver
Care status
Smart Triage Queue
24 patients · sorted by risk · updated 2 min ago
Sort: Risk ▾Assign outreach
PatientRiskTop driversCare statusNext action
RM
Robert Maddox
4471-092 · 64y
HIGH2 missed follow-ups · 1 ER visit · unfilled medsUnassignedCall re: med access
DA
Denise Alvarado
3820-115 · 58y
HIGHHousing loss · transport gap · A1c risingIn outreachConnect housing nav
MB
Marcus Bell
2910-844 · 71y
RISINGFood insecurity · 1 missed follow-upUnassignedScreen food access
YP
Yolanda Pierce
5182-338 · 49y
RISINGUtility assistance flag · refill gapIn outreachReview meds
AW
Aaron Whitfield
1177-286 · 55y
STABLENo active flagsMonitoring

The Smart Triage Queue answering “where do I start?” — the ranked wall, severity and its cause on every row.

Carbon — one blue, sharp corners, 2px grid
Carbon components
Primary button
Label
Text input
taghighstable
Sharp corners · one blue · 2px grid
05Choosing the product’s design system

Built on IBM Carbon — on purpose.

The product shouldn’t look like the tools it replaces. I chose Carbon — IBM’s open-source, publicly available design system — as the foundation for every ClarityCare screen, for four defensible reasons.

Fit

Purpose-built for dense, data-heavy enterprise tools — exactly what clinical triage is.

Access

Audited contrast & keyboard support out of the box — non-negotiable in a clinical setting.

Tone

Sharp, quiet, grid-driven — reads trustworthy and clinical, not consumer-playful.

Continuity

Same IBM Plex type family as this deck — the product threads through the whole story.

The process

One week, blank page to system.

The move that made a week possible: settle the judgment first, produce second. Decide the hard things up front, and production moves without second-guessing.

Day 1 · Frame
The coordinator's problem, the trust thesis, the six skeptic questions.
Day 2 · Structure
Validated SDOH frameworks; the desktop ↔ mobile contract.
Day 3–4 · Draft
Lo-fi sketches → wireframes for queue, insight card, override.
Day 5–7 · Build
High-fidelity Carbon screens, both surfaces — and this deck.

The Smart Triage Queue, three days apart · sketch → wireframe → hi-fi Carbon.

Sketch · Day 3
Wireframe · Day 4
Hi-Fi Carbon · Day 6
06Smart triage queue · “where do I start?”

Lead with severity — and its cause.

Oriented in five seconds. Every row carries the risk level, the drivers behind it, and the next action — so the coordinator reads who and why at a glance. Everything else lives behind filters.

07Insight cards · “why should I believe this?”

SHAP output, rendered as a sentence.

No coefficients. The model’s reasoning is a plain-language sentence, the score breakdown sits beside it, and the confidence cue is honest — a stale data source is stated outright, and it invites the override rather than hiding behind a number.

insight card — model explanation (concept)
← Triage Queue  /  Robert Maddox
CH
RM
Robert Maddox
MRN 4471-092 · 64y · Medicare
HIGH↑ from Rising · 6 days ago
Active SDOH flags
Transportation gap
Lives alone · limited support
Food security · stable
+Insight Card · Model Explanation

This patient is high risk due to 2 missed follow-ups, 1 ER visit this month, and unfilled medications — compounded by a transportation gap.

What drove the score
Missed follow-upshigh
Recent ER visithigh
Medication adherencemed
Transportation (SDOH)med
⚠ Data confidence: Moderate — 3 of 4 sources
Claims data is 40 days stale. If this doesn’t match what you know, override it — your correction trains the model.
Accept suggested actionOverride
override — a structured reason (concept)
Override suggested action
Suggested: Call patient to assess medication access
Reason for override
Add a note (optional)
Reached by phone Tuesday — meds now filled…
Cancel
Submit override
08Override flow · “am I still in control?”

The override is the training signal.

Control and learning, made the same gesture. When the clinician overrides, the structured reason doesn’t just log a disagreement — it becomes labeled feedback the model learns from.

Coordinator overrides

“This is wrong because…” — a required, structured reason.

Captured as labeled feedback

Structured reason + free-text note, attached to the case.

Model improves

The humans correcting it are the humans improving it.

09Timeline view · “how did we get here?”

Context, not just a number.

A single risk number is a snapshot. The timeline shows the story behind it — clinical events, the AI’s re-scores, and staff actions on one thread, color-coded by who acted, ending with the override that fed back into the model.

care timeline — clinical · AI · staff (concept)
← Robert Maddox  /  Care timeline
ClinicalAIStaff
Mar 2 · Clinical
Missed cardiology follow-up
Mar 14 · Clinical
ER visit — chest pain, discharged
Mar 15 · AI
Risk raised Rising → High
Driver: ER visit + missed follow-up
Mar 16 · Staff
Assigned to outreach queue
Mar 16 · AI
Suggested: assess medication access
Mar 18 · Staff
Override — already contacted, meds filled
→ fed back to model as labeled feedback
Notifications · “what am I missing right now?”

Nothing critical waits for a login.

Every alert is actionable in place — a risk jump, an overdue outreach, an override flagged for review — each with the one action it needs, not a feed the coordinator has to re-triage.

notifications — actionable in place (concept)
← Triage Queue  /  Notifications
7 new
Show
All · 7
Score changes · 3
Overdue actions · 2
Flagged overrides · 2
Robert Maddox · risk raised Rising High
Driver: new ER visit · 8 min ago
Open patient
Denise Alvarado · outreach action overdue
Housing nav connection · due yesterday
Reschedule
Override flagged for review · data-issue reason
Marcus Bell · 3 coordinators reported the same driver
Review
Yolanda Pierce · risk raised Stable Rising
Driver: refill gap · 1 hr ago
Open patient
10Designing for the field

Thirty seconds of attention, in someone’s living room.

Mobile wasn’t a shrunk-down desktop — it was re-architected for the home visit: a coordinator with thirty seconds of attention, needing to re-score risk on the spot after entering a new SDOH fact. Optimized for speed, clarity, and a single next action.

Condensed queue

Stacked filters, high-priority flags, one line of reasoning per patient.

Real-time SDOH · “what if I just learned something?”

Flip a toggle → a banner fires → the model re-scores on the spot.

field surface — two mobile postures (concept)
9:41
Triage Queue
High · 6RisingSDOH
Robert MaddoxHIGH
2 missed follow-ups · unfilled meds
Call re: med access
Denise AlvaradoHIGH
Housing loss · transport gap
Connect housing nav
Marcus BellRISING
Food insecurity · 1 missed follow-up
Screen food access
QueueVisitsAlertsMe
9:47
← Denise Alvarado
Update social needs
Stable housing
Changed — lost housing this week
Reliable transportation
Food security
✓ Risk re-scored: Rising → High
Housing loss added to model input just now.
Save & re-score
11The desktop ↔ mobile contract

Same product. Two postures.

Documented as an explicit contract — the analyst at a desk, the coordinator on a doorstep — so neither surface drifts from the shared logic.

FunctionDesktopMobile
Patient queueFull list + filters + panelsCondensed list, stacked filters
ExplainabilitySHAP breakdown in side panelPlain text + collapsible sections
OverrideModal or inlineSlide-up sheet, big tap buttons
SDOH entryForm-basedToggle-friendly, field-safe
AlertsDashboard feedPush notifications
TimelineSide panelInline swipeable strip
12How the AI shaped the UX

Each model capability created an interface obligation.

Risk scores→ demanded

Visual severity indicators

SHAP explainability→ demanded

Plain-language insight cards

Data-confidence levels→ demanded

Transparency cues that invite override

Multi-model support→ demanded

Risk-type filters & segments

Model training→ demanded

Structured override capture

Real-time inputs→ demanded

Live re-prioritization

The model’s honesty about its own limits is the UX.
The hard calls

Three tradeoffs I’d defend in a critique.

Density vs. overload

Lead with severity + its cause; filter the rest.

Oriented in five seconds beats complete. A spreadsheet is complete and useless.

Hide doubt vs. show it

Show confidence honestly; invite the override.

A tool that hides its uncertainty earns less trust, not more.

Chart vs. plain language

The sentence a colleague would say.

No clinician should decode coefficients mid-shift to act on a score.

Every fork came back to one question: would a coordinator trust this at 4pm on a hard day?
Responsible AI

A point of view on AI in care.

01

Never make someone obey a model they disagree with.

02

Never waste the disagreement — it’s the training signal.

03

Be honest about uncertainty; encourage overriding a shaky call.

04

The clinician is accountable — so the clinician stays in control.

Not compliance checkboxes — the reason it’s a second opinion, not an autopilot.
Reflection · where it landed

With AI in the loop, the interface’s job is to make the model accountable to the human using it.

Judgment first

Settle the hard calls, produce second — sequencing I’ll keep.

Explainability isn’t a feature

It’s the whole relationship between the clinician and the AI.

Trust is the deliverable

A black box turned into a glass box the clinician stays in charge of.

The deliverable is the argument and the high-fidelity system that embodies it: a triage queue, an explainable insight card, an override flow that keeps the clinician in charge, real-time SDOH entry, a timeline, and a field surface — a black box turned into a glass box.

A concept project with no clinician validation or live model — so there are no metrics, and I don’t invent any. Named patients are illustrative sample data.