Smart Hematology Triage System (Mayo Clinic)

SMART HEMATOLOGY TRIAGE SYSTEM

Mayo Clinic

Prototype

ROLE

CX Designer · UX Researcher · Human-AI Interaction

TIMELINE

16 Weeks, Fall 2025

TEAM

5-person team

SCOPE

AI product strategy · research · prototyping · validation

TOOLS

Figma · Miro · Figma Make

At a Glance

In a hematology care setting filled with complex referrals and scattered patient data, Smart Hematology Triage helps nurses see the most important patient information in one place. By highlighting key clinical signals, it reduces the time spent searching through records by 40% and supports faster, safer triage decisions while keeping nurses in control.

Prolem

Healthcare loses $90B+ annually to administrative inefficiencies, while hematology triage nurses can spend an estimated 4–6 hours per shift managing high-volume referrals across fragmented EHR, lab, and imaging systems. In a specialty where delays can postpone cancer diagnosis by weeks, Smart Hematology Triage targets this fragmentation to support faster, safer high-stakes decisions.

Quantified reality

20-30%

Time spent searching, not deciding

~45%

Referrals arrive incomplete

12-18 min

Lost per referral to data hunting

50-60%+

Nurse burnout linked to workload

Solution

Smart Hematology Triage brings scattered patient data into one clear view, using AI summaries, risk flags, and clinician co-design to reduce search time and support faster triage while keeping every decision in clinicians’ control.

• Co-designed for real triage workflows• AI summaries + risk flags• Unified EHR, lab, and imaging view• AI suggests, clinicians decide

Design Process

1

Research

2

Synthesis

3

Ideation

4

Design

5

Final Design

6

Reflection

01.Research

The Triage Burden:

Understanding Referral Workload, Time Pressure, and Clinical Risk

“I’m trained to make clinical decisions, but most days I’m just trying to keep up with referrals.”

Hematology Triage Nurse

We worked with 6+ hematology physicians and triage nurses through workflow observation, interviews, and surveys to understand how referral decisions happen under real clinical pressure. The research showed that the challenge was not lack of expertise, but sustained cognitive load: nurses process back-to-back referrals for hours, balancing speed, accuracy, and safety while trying not to miss critical clinical signals.

Qualitative Research

Observation and interviews showed that nurses often reach clinical reasoning only after navigating scattered records. In hematology triage, one missed signal could delay diagnosis or affect patient outcomes.

"By the time I'm ready to make a decision, I'm already mentally exhausted."

— Hematology Triage Nurse

"I want to think clinically, but most of my time is spent just getting to a point where I can."

— Hematology Triage Nurse

78%

of nurses described triage as mentally exhausting before clinical judgment begins

72%

reported high anxiety about missing critical information during referral review

67%

said decision confidence drops significantly later in long shifts

83%

expressed the need for clear, visual patient summaries before deciding

Quantitative Research

Surveys validated our interview findings, showing triage workflow inefficiencies around time, usability, and decision confidence. Clinicians wanted AI support, but only if it stayed transparent and kept them in control.

60–70%

of triage time is spent reviewing and assembling referral context not deciding

3–5 systems

accessed on average per referral during triage

82%

reported that contextual & visual patient timelines & EHR would significantly reduce triage effort

100%

agreed that AI must support, not approve clinical decisions

02.Synthesis

How might we save clinical time by designing a triage system that is fast, trusted, and easy to use?

What emerged was not clinical complexity, but fragmented systems. Clinicians were losing critical time across notes, labs, and history, showing the need for transparent AI support that brings context together while keeping decisions in their control.

Time Is the Bottleneck

Clinicians lose critical time searching across records before they can make triage decisions. The design needed to reduce navigation and bring key patient context into one place.

A System, Not Just a Tool

Triage depends on labs, notes, timelines, and handoffs. Clinicians needed a connected workflow that organizes information around decisions, not scattered screens.

Accountable AI Support

AI should assist, not decide. Clinicians needed clear summaries, visible supporting data, and easy control over every AI suggestion.

Personas

This highlighted a core design challenge:

How might we reduce triage time by bringing efficiency and required clinical data together in one coherent system?

Hematology Triage Nurse

Handles first-pass triage for incoming referrals and determines urgency under time pressure.

"I spend most of my time finding labs and notes before I can even start thinking through the case."

Attending Hematologist

Reviews complex referrals while balancing clinic responsibilities and limited time.

"A lot of my time goes into piecing together charts instead of making the actual decision."

Triage Coordinator

Tracks referral status, lab completion, and follow-ups across patients and providers.

"The information is in the EHR, but I have to jump across tabs and reports to understand what's going on."

03.Ideation

Using affinity mapping, concept mapping, and clinician discussions, I explored ways to reduce triage time while keeping decisions efficient and clinically sound.

Initial Concepts• AI-assisted triage with priority and risk indicators• AI summaries highlighting key referral findings• Visual EHR data and patient timelines

These concepts focused on improving speed and clarity by organizing information around real hematology triage workflows.

User Flow

The user flow reduces search effort by bringing referrals, labs, history, and timelines into one clear triage sequence.

100%+–

04.Design

The design focused on reducing referral review time by organizing key triage signals into one clear view.

It centered on AI priority predictions, transparent decision factors, and visual timelines to help nurses understand context faster.

Branding and Style guide

High-fidelity screens focused on reducing cognitive load by bringing labs, referrals, history, and documentation into one clear triage workflow.

Initial Design

Dashboard

Triage Overview

Decision Flow

Timeline

EHR

Clinical Notes

User Testing

The initial testing consisted of 5 doctors and the ML researcher at Mayo Clinic.

Through observed usage, think-aloud walkthroughs, and post-testing interviews, key gaps in the triage flow surfaced, guiding improvements to clarity and efficiency in first-pass triage.

Triage Overview Density

The overview worked well for familiar users, but first-time users needed a simpler first view to focus faster.

Workflow Order

Clinicians naturally moved between related sections, showing the need to combine steps into one smoother flow.

Familiar Clinical Visuals

Charts already used in practice helped clinicians orient faster and reduced the learning curve.

Integrated Documentation

Adding notes within the triage overview supported more continuous, in-context decision-making.

05.Final Design

Login & Access

A clean, secure entry point enables clinicians to access the triage system quickly with minimal friction.

Triage Dashboard & Referral Queue

Centralized view of incoming referrals with status, priority, and last action at a glance.

Global search and filters support quick sorting and first-pass triage.

Contextual TooltipsUsed to explain key components, support different levels of technical familiarity, and improve accessibility without adding clutter.

Triage Overview & Decision Workspace

Primary workspace for reviewing patient context and making triage decisions.

Allows clinicians to accept or modify AI-assisted triage recommendations.

Keeps key clinical information visible to support quick, informed decisions.

Triage Actions - 4 Simple steps

Clinical Triage Guide - Provides three quick-access views that allow clinicians to make fast triage decisions when detailed review is not required, supporting efficient first-pass triage.

Clinical SummaryConsolidates referral details, lab results, EHR data, and past medical history into a single, high-level overview.

Summary View Options - Allows clinicians to switch between focused views to review each summary area in more detail as needed.

Decision Flow

Aligns with the clinical decision charts nurses already use, showing the AI’s predicted outcome with confidence percentages and the key contributing signals (labs, trends, referral data) to support verification and trust.

Charts in use/AI Algos

Decision FlowPatient Records Across Timeline, Visual EHR, and Detailed EHR

Patient Timeline

Displays the patient’s medical record over a selected time period in a chronological timeline.

Highlights laboratory tests, clinical visits, procedures, symptoms and diagnoses, and referrals in one continuous view.

Shows disease progression status to help clinicians quickly understand how the condition has evolved over time.

Visual EHRCondenses condition-specific EHR data into focused charts and trends, enabling quick assessment of progression and abnormalities without scanning full records.

Detailed EHR View

Provides access to the full EHR in a familiar, structured format.

Allows clinicians to perform detailed review using the same process they are already accustomed to.

Integrated MessagingProvides an email-like interface for internal team communication and patient outreach, keeping conversations organized and accessible.

Quick Patient ContactAllows direct communication from the patient card with messages automatically recorded in the same thread for continuity and reference.

AI AssistantA patient-specific assistant that helps clinicians to get instant answers to key questions for that individual case, supporting faster, informed triage decisions.

My Report

Performance & LearningTracks speed, accuracy, and AI agreement to support continuous efficiency gains.

Operational InsightsHighlights volume, delays, and recurring issues to reduce workflow bottlenecks.

AI TransparencyDisplays match rates, overrides, and confidence signals to guide system improvement.

Accessibility Standards

Designed with a WCAG AA–compliant color system supporting users with color vision deficiencies (~8% of males, ~0.5% of females).Includes adaptive modes for Protanopia (red-blind), Deuteranopia (green-blind), Tritanopia (blue-blind), and Achromatopsia (no color), allowing users to personalize visibility.

Protanopic

Achromatopsia

Normal Vision

Deuteranopia & Tritanopia

Prototype

Result

30-40%

Faster triage process

40%

Reduction in time spent for patient information

3X

Faster identification of high-risk cases requiring immeadiate attention

12 - 18 min

Average time saved per referral review

Future Innovation and Growth Opportunities

This project deepened my understanding of how UX can support healthcare teams working under intense time and cognitive pressure. Observing physicians and triage nurses firsthand showed me that future clinical tools must reduce friction, not add complexity. It reinforced the value of human-in-the-loop AI: systems that save time, support decision-making, and keep clinicians accountable and in control.

Human-AI Collaboration

Continued research on optimal balance between AI assistance and human autonomy in high-stakes clinical decisions

Ethical AI Frameworks

Developing governance models for transparent, accountable AI in healthcare settings

Cross-Specialty Expansion

Adapting the triage framework for cardiology, oncology, and other specialties facing similar challenges

Peoduct Video