How Better Design Reduced Hospital Admissions by 29%
INDUSTRY
US Healthcare
CLIENT
US-based healthcare analytics company
TEAM
1 UX Lead, 2 UX/UI Designers, 1 Project Manager
MY ROLE
UX/UI Designer
DEVICE
Desktop & Tablet
Overview
An AI-powered care management platform built for the US health insurance industry. Used internally by insurance companies to help their care management teams proactively manage high-risk members and prevent avoidable hospital admissions.
1.Discovery
1.1 Industry Understanding
The US healthcare insurance system has three legs:
Payers
(Insurance Companies)
Payers collect premiums from members and pay hospitals for treatments. Their goal is to keep members healthy so expensive treatments are avoided.
Patient
(Consumer)
Consumers want good care and financial protection when they fall sick.
Providers
(Hospitals / Doctors)
Providers want steady patient flow and efficient care delivery while maintaining quality.
The financial tension at the centre of this ecosystem is straightforward. Payers pay out every time a member is hospitalised. The more admissions, the less profit. So payers hire Care Managers, typically registered nurses, to proactively check in on high-risk members, close care gaps, and keep them out of hospital.
PRODUCT GOAL
Design a product to make work easier for Care Managers.
Care Managers were juggling 150 to 200 patients a month across disconnected tools, with no way to know who needed attention first or how they were performing as a team.
1.3 Persona Development
Three personas were developed across the discovery calls.
Below Avg
Tech Savvy
Qualification
Registered Nurse
Susan
Care Manager
40 Yrs, Female
Responsibilities / Daily tasks
-
Short list which patients to talk to.
-
Review critical profiles, add notes, set reminders
-
Call patient and make them understand the care gaps
-
Reschedule calls and appointments
-
Mark the requirement as "closed" once the call is done
-
Coordinating with PCP whenever required
-
Flag the patient if their risk is increasing.
Pain Points
-
All information is not at the same place
-
Softwares are normally designed for physicians, not care managers
-
Different softwares are used for different actions
-
Performance metrics aren't clear and tracking is ambiguous
-
Difficult to know whether a patient is being worked on or open for allotment
-
Job insecurity due to competition from younger, more tech-savvy peers
Moderate-High
Tech Savvy
David
CM Admin
42 Yrs, Male
Responsibilities / Daily tasks
-
Onboarding, reshuffling and removal of CM'S to and from the system
-
Working as a CM sometimes due to staff shortage.
-
Assigning patients to CM - it can be one-to-one assignment, bulk or auto-assignment.
-
Reassigning patients in case anyone is on leave or for balancing load of CMs lagging behind.
-
Review the care managers performances. Derive insights and formulate strategies
-
Running reports to extract data at program level, care manager level or team level.
-
Data extraction, downloading
Pain Points
-
Tracking performance of teams and individuals
-
CM availability tracking is difficult
-
Assigning individual patients is tedious
-
Multiple CMs taking up the same patient due to open self-allotment
-
Accessing old patient records and archives
High Tech Savvy
Daniel
CXO
48 Yrs, Male
Responsibilities / Daily tasks
-
Check on the KPI and create new tactics to mitigate the gaps in organisational efficiency
-
Be updated on the industry standards.
-
Cross check if the deadlines are met.
-
Programs are functioning as expected
-
Evalute overall customer satisfaction/ employee satisfaction and attribution rate
Pain Points
-
Difficult to read bulky data
-
Decision making takes a lot of time
-
Compliance tracking and gaps identification
-
All important information not at one place
-
Too busy to concentrate on less important tasks
-
Initiatives and new technologies are difficult to implement
1.4 Competitor Analysis
Six platforms were analysed, a heuristic analysis was run across all six against Nielsen's usability principles.
Platform
Key Highlights
A patient detail focused platform
-
Name pronunciation included as a personal connect element.
-
Health and wellbeing trackers laid out clearly with iconography.
-
Best call times and family members visible upfront. Notes with full edit history.
A claims and analytics
-
Strong population insights dashboard.
-
Filterable by region, payer, line of business and plan type.
-
Clean graphical data representation.
risk stratification platform
-
KPIs surfaced on the member listing page.
-
Card based programme dashboard with average age, membership and risk drivers.
-
Multiple filters available.
care management platform
-
Alerts surfaced immediately.
-
Tab based layout reduced cognitive load.
-
Calendar and reminders widget built in.
2. Define
2.1 Synthesising the Research
Coming out of discovery,we had
three personas
six competitor analysis
existing product sketches
a feature list across three user roles
THE SINGLE MOST IMPORTANT INSIGHT FROM SYNTHESIS
Who do I call first today?
The existing product gave no answer to that question. It gave a flat list of 150+ patients and left the nurse to figure it out herself.
2.2 Design Direction
Fragmented patient information
Everything a Care Manager needs in one place. No switching between tools to complete a single task.
Care Manager pain point: Different softwares are used for different actions
Show less, reveal more
Show a summary first. Let users drill into detail on demand. Never show everything at once.
CXO pain point: difficult to read bulky data.
Assistance over reporting
Tell users what to do next, not just what the data says.
Group by task, not by data type
Group data by the user's task and context, not by data type.
Data always paired with visualisation
Every significant metric needs a visual anchor alongside it.
Consistent semantic colour
Colour must mean the same thing everywhere.
Smart filters, not just search
Better UX for column selection. Reminders-choose date range, more filters like upcoming 7 days.
Gentle learning curve
Interface must be learnable without training for a nurse with below-average tech savvy.
2.3 Information Architecture
2.3 Visual Direction
Before any screen design, a visual direction was established.
The moodboard was anchored on four words:
Modern
Subtle
Formal
Serious
TYPOGRPAHY
COLORS
Aa
Fira Sans
Type
H1
H2
H3
H4
H5
H6
B1
B2
Caption
Font(px)
32/Bold
28/SemiBold
24/SemiBold
20/SemiBold
18/SemiBold
16/SemiBold
14/Regular
12/Regular
12/Regular
Primary
Success
Warning
Error
Brand
Neutrals
3. Develop
3.1 Key Wireframe Decisions
Wireframes were produced for each module and reviewed with the client before high-fidelity work began. This stage was where the most structural debates happened.
Member Listing
Work out a member assigning system for unassigned members.
Why: Without a assigning mechanism, two CMs can simultaneously open the same unassigned patient, creating duplicated effort and care gaps elsewhere.
A user-level column selector letting CMs choose which data fields to display
Why: Different care programmes surface different data as relevant. A CM working a diabetes cohort cares about different columns than one working a cardiac cohort.
The member list default to priority order, instead alphabetical or date order.
Why: the Care Manager's first question every morning is who needs attention most urgently
Member Detail
The wireframe introduced a tabbed structure: My Dashboard, Overview, Actionable Insights, Risk Score, Contact Details, Social Determinants, HRA Questionnaire
Customisable “My Dashboard” Tab as the default landing tab on member detail
Why: Different CMs need different information first depending on their patient's condition, programme, and current care status.
The tab structure
Why: A single scrolling page puts equal visual weight on all information. Tabs creates a hierarchy.
Notes and note history
Why: CMs call patients repeatedly over months. They can review previous interactions and enter the call with context
4. Deliver
4.1 Final Design System
Before screens were finalised, the design system was locked. Every component was documented in the style guide.
BUTTON/PRIMARY
BUTTON/SECONDARY
BUTTON/TERTIARY
SEARCH INPUT
RADIO BUTTON
Search
Search
Anthony Daniel
Label
Label
Label
Eight tabs, each with a defined purpose
4.2 Final Screens
Member Listing
Smart filtering letting CMs save their preferred table configuration.
KPI summary bar at the top
Member Detail
My Dashboard
Overview
Actionable Insights
Risk Score
Contact Details
Social Determinants
HRA Questionnaire
Encounters data
Medical History
Care Manager Dashboard
Care Manager Dashboard
Power BI Dashboard
It is the CXO's primary screen. It shows programme performance across the entire member population, not individual patients.
Challenge:
Power BI has significant styling constraints.The decision to design the Power BI layer to match the platform's visual language. Executive Dashboard felt like part of one product, not a separated integrated reporting tool. It must follow:
-
same type scale
-
same colour system
-
same card patterns
Handoff
Final designs delivered via Figma
A working prototype
Component Library and Style guide
29% Reduction in Hospital Admissions
Improved early identification of high-risk patients helped reduce avoidable hospital visits.
74% Better Risk Identification
More high-risk patients were identified compared to baseline methods.
Outcomes
These outcomes reflect the performance of the complete platform:
AI/ML model
Care management team
Product Design