Swetha De-Smith
Product Designer
Digital Creator
Ui Designer
Product Designer
Swetha De-Smith
Product Designer
Digital Creator
Ui Designer
Product Designer
Swetha De-Smith
Product Designer
Digital Creator
Ui Designer
Product Designer
Designing an AI prediction tool to reduce missed appointments in NHS children’s services
Designing an AI prediction tool to reduce missed appointments in NHS children’s services
A clinical decision-support tool helping healthcare teams identify patients at risk of missing appointments, prioritise interventions and improve access to care.
TLDR;
TLDR;
Problem: Missed appointments delayed care, disrupted clinic schedules and widened access gaps for vulnerable children and families.
Solution: I designed a web-based AI prediction tool that helped clinicians identify attendance risk, manage caseloads and support earlier intervention.
Impact: The tool achieved 78% prediction accuracy, supported a 25% reduction in DNA rates within the Children’s Hospital Alliance and contributed to a 52% reduction in WNB rates among the most at-risk groups, according to Alder Hey Innovation.
Problem: Missed appointments delayed care, disrupted clinic schedules and widened access gaps for vulnerable children and families.
Solution: I designed a web-based AI prediction tool that helped clinicians identify attendance risk, manage caseloads and support earlier intervention.
Impact: The tool achieved 78% prediction accuracy, supported a 25% reduction in DNA rates within the Children’s Hospital Alliance and contributed to a 52% reduction in WNB rates among the most at-risk groups, according to Alder Hey Innovation.
CONTEXT
CONTEXT
Alder Hey teams were working with patient attendance patterns, but the process of turning that data into timely intervention was slow and fragmented.
The opportunity was not just to show a prediction, but to make the prediction understandable, actionable and useful within a clinical workflow.
Alder Hey teams were working with patient attendance patterns, but the process of turning that data into timely intervention was slow and fragmented.
The opportunity was not just to show a prediction, but to make the prediction understandable, actionable and useful within a clinical workflow.
MY ROLE & CONSTRAINTS
MY ROLE & CONSTRAINTS
MY ROLE:
MY ROLE:
UX Design Lead:
UX Design Lead:
PROJECT CONSTRAINTS:
PROJECT CONSTRAINTS:
Lean budget, technical AI integration, clinical workflows, multi-trust scalability
Lean budget, technical AI integration, clinical workflows, multi-trust scalability
Responsibilities: Discovery, ecosystem mapping, workshop, value proposition canvas, IA, wireframes, UI design, feasibility checks, validation
Responsibilities: Discovery, ecosystem mapping, workshop, value proposition canvas, IA, wireframes, UI design, feasibility checks, validation
Clinical tone creates distance
Button-driven conversation flows can be frustrating. Users consistently tried to type their own queries even when buttons were available, a signal that shaped the entire conversation architecture.
Collaboration: Project manager, internal developers, Alder Hey stakeholders, third-party AI provider
Collaboration: Project manager, internal developers, Alder Hey stakeholders, third-party AI provider
Broken chatbot interactions erode trust fast
In a clinical context, a bot that gets things wrong doesn't just frustrate users, it undermines confidence in the whole service. Accuracy and graceful failure states were non-negotiable.
THE DESIGN CHALLENGE
THE DESIGN CHALLENGE
"How might we help healthcare professionals trust, understand and act on AI-generated attendance predictions without adding more admin burden to an already pressured clinical workflow?"
"How might we help healthcare professionals trust, understand and act on AI-generated attendance predictions without adding more admin burden to an already pressured clinical workflow?"
MAKING SENSE OF THE CLINICAL ECOSYSTEM
MAKING SENSE OF THE CLINICAL ECOSYSTEM
I needed to understand the wider service the WNB tool would sit within. The challenge was not simply to design a dashboard for an AI model. It was to understand how healthcare professionals, admin teams, NHS trusts, patient attendance data and the third-party AI provider would interact in a live clinical workflow.
I needed to understand the wider service the WNB tool would sit within. The challenge was not simply to design a dashboard for an AI model. It was to understand how healthcare professionals, admin teams, NHS trusts, patient attendance data and the third-party AI provider would interact in a live clinical workflow.
01
Framing the right questions before designing
Framing the right questions before designing
I started by formulating a set of strategic questions to guide early conversations with the client team and technical stakeholders. These questions helped me clarify what healthcare professionals needed from the prediction tool, what decisions the AI output should support, and where the existing process created delays or uncertainty.
The aim was to avoid designing around assumptions. I wanted to understand:
I started by formulating a set of strategic questions to guide early conversations with the client team and technical stakeholders. These questions helped me clarify what healthcare professionals needed from the prediction tool, what decisions the AI output should support, and where the existing process created delays or uncertainty.
The aim was to avoid designing around assumptions. I wanted to understand:
Who would use the tool and at what point in the appointment workflow?
Who would use the tool and at what point in the appointment workflow?
What admin teams needed to manage access across trusts?
What admin teams needed to manage access across trusts?
What information clinicians needed before deciding whether to intervene?
What information clinicians needed before deciding whether to intervene?
How much explanation was needed for the AI output to feel useful and trustworthy?
How much explanation was needed for the AI output to feel useful and trustworthy?
How prediction progress should be communicated?
How prediction progress should be communicated?
This gave the project a clearer direction and helped keep design decisions tied to real clinical and operational needs.
This gave the project a clearer direction and helped keep design decisions tied to real clinical and operational needs.
02
Mapping the wider service ecosystem
Mapping the wider service ecosystem
I then created an ecosystem map to visualise the relationship between Alder Hey, clinicians, admin users, other NHS trusts, the WNB portal and the third-party AI provider.
This helped me understand the product as part of a wider healthcare service, not just a standalone tool. It made the key value exchanges clearer: patient attendance data needed to move into the system, the AI provider needed to return predictions, clinicians needed to interpret the output, and admin teams needed a way to manage users and trusts at scale.
The ecosystem map helped identify where the experience needed to support:
I then created an ecosystem map to visualise the relationship between Alder Hey, clinicians, admin users, other NHS trusts, the WNB portal and the third-party AI provider.
This helped me understand the product as part of a wider healthcare service, not just a standalone tool. It made the key value exchanges clearer: patient attendance data needed to move into the system, the AI provider needed to return predictions, clinicians needed to interpret the output, and admin teams needed a way to manage users and trusts at scale.
The ecosystem map helped identify where the experience needed to support:
Data flow between Alder Hey, the AI provider and the WNB portal
Data flow between Alder Hey, the AI provider and the WNB portal
Clinician decision-making around patient attendance risk
Clinician decision-making around patient attendance risk
Admin control for onboarding trusts and managing access
Admin control for onboarding trusts and managing access
Cross-trust collaboration and alignment
Cross-trust collaboration and alignment
Clear audit trails for accountability and shared understanding
Clear audit trails for accountability and shared understanding
03
Aligning product value with user needs
Aligning product value with user needs
To bring stakeholders into alignment, I used a value proposition canvas to connect the product idea back to the users it was meant to serve.
This helped shift the conversation from “what should the tool show?” to “what value should the tool create for clinicians and admin teams?”
For clinicians, the value was not just seeing a prediction. It was being able to identify risk earlier, manage their caseload more efficiently and take informed action before an appointment was missed.
For admin teams, the value was operational control: setting up trusts, managing access and enabling multiple services to use the tool consistently.
This exercise helped define the product priorities:
To bring stakeholders into alignment, I used a value proposition canvas to connect the product idea back to the users it was meant to serve.
This helped shift the conversation from “what should the tool show?” to “what value should the tool create for clinicians and admin teams?”
For clinicians, the value was not just seeing a prediction. It was being able to identify risk earlier, manage their caseload more efficiently and take informed action before an appointment was missed.
For admin teams, the value was operational control: setting up trusts, managing access and enabling multiple services to use the tool consistently.
This exercise helped define the product priorities:
Make appointment risk easier to identify and act on
Make appointment risk easier to identify and act on
Reduce manual effort in generating and tracking predictions
Reduce manual effort in generating and tracking predictions
Support individual logins and role-based workflows
Support individual logins and role-based workflows
Enable trust-level scalability from the start
Enable trust-level scalability from the start
Provide guidance and audit trails to support confident decision-making
Provide guidance and audit trails to support confident decision-making
KEY DESIGN PRINCIPLES
KEY DESIGN PRINCIPLES
The discovery work helped define four principles that guided the product experience.
The discovery work helped define four principles that guided the product experience.
Make risk actionable
Help clinical teams identify which patients may need support before an appointment is missed.
Support timely intervention
Design the experience around what teams could do next, from reminders to flexible scheduling and practical support.
Build trust through clarity
Make the AI process easier to understand through clear feedback, guidance and progress visibility.
Design for multi-trust scale
The product needed to work for Alder Hey first, while supporting wider NHS trust adoption.
Make risk actionable
Help clinical teams identify which patients may need support before an appointment is missed.
Build trust through clarity
Make the AI process easier to understand through clear feedback, guidance and progress visibility.
Support timely intervention
Design the experience around what teams could do next, from reminders to flexible scheduling and practical support.
Design for multi-trust scale
The product needed to work for Alder Hey first, while supporting wider NHS trust adoption.
Make risk actionable
Help clinical teams identify which patients may need support before an appointment is missed.
Build trust through clarity
Make the AI process easier to understand through clear feedback, guidance and progress visibility.
Support timely intervention
Design the experience around what teams could do next, from reminders to flexible scheduling and practical support.
Design for multi-trust scale
The product needed to work for Alder Hey first, while supporting wider NHS trust adoption.
INFORMATION ARCHITECTURE
INFORMATION ARCHITECTURE
The ecosystem mapping showed that clinicians and admin users had different goals, so I separated the product into two focused areas.
The ecosystem mapping showed that clinicians and admin users had different goals, so I separated the product into two focused areas.
CLINICIAN PORTAL
CLINICIAN PORTAL
ADMIN PORTAL
ADMIN PORTAL
WIREFRAMES
WIREFRAMES
Early wireframes helped define the core workflow: Upload data, generate predictions, review outputs and maintain a record of activity.
Early wireframes helped define the core workflow: Upload data, generate predictions, review outputs and maintain a record of activity.
FINAL INTERFACE DESIGN
FINAL INTERFACE DESIGN
Once the core workflow and information architecture were defined, I moved into the final interface design. The goal was to create a clear and structured experience that helped healthcare teams upload data, review prediction outputs, access previous records and maintain visibility through audit trails and resources.
Once the core workflow and information architecture were defined, I moved into the final interface design. The goal was to create a clear and structured experience that helped healthcare teams upload data, review prediction outputs, access previous records and maintain visibility through audit trails and resources.
Landing/Dashboard screen
A clear starting point for users to access uploads, prediction activity, audit trails and resources.
Landing/Dashboard screen
A clear starting point for users to access uploads, prediction activity, audit trails and resources.
Upload screen
Users can upload historic data, upload future prediction data and download predicted outputs from one focused area.
Upload screen
Users can upload historic data, upload future prediction data and download predicted outputs from one focused area.
Prediction insights screen
Charts, filters and analytics help users interpret prediction trends.
Prediction insights screen
Charts, filters and analytics help users interpret prediction trends.
Previous predictions screen
Users can revisit previous prediction runs, sort table columns and download earlier outputs.
Previous predictions screen
Users can revisit previous prediction runs, sort table columns and download earlier outputs.
Audit trail screen
The audit trail gives teams visibility of key system activity, supporting traceability and confidence in the workflow.
Audit trail screen
The audit trail gives teams visibility of key system activity, supporting traceability and confidence in the workflow.
IMPACT
IMPACT
9 NHS children’s services
9 NHS children’s services
The tool was deployed across Alder Hey and other NHS children’s services.
The tool was deployed across Alder Hey and other NHS children’s services.
52% reduction in WNB rate
52% reduction in WNB rate
Trusts reported a reduction in WNB rates among underserved patient populations.
Trusts reported a reduction in WNB rates among underserved patient populations.
Faster decision-making
Faster decision-making
Real-time tracking helped healthcare professionals manage prediction activity more efficiently.
Real-time tracking helped healthcare professionals manage prediction activity more efficiently.
Better collaboration
Better collaboration
The tool supported clearer alignment between staff and NHS trusts.
The tool supported clearer alignment between staff and NHS trusts.
