Skip to content
Lizzie Teo

Healthdirect Symptom Checker

Overview

Healthdirect is Australia's government funded health information service, reaching millions of Australians across digital and phone channels.

This project was a full redesign of their legacy product, Symptom Checker. It's a triage tool used to help people assess their symptoms and decide on next steps for care.

Role
Product Design Consultant
Impact area
Public health
Focus
  • Content Strategy
  • User Experience
  • User Interface Detailing
  • User Testing

healthdirect logo
The symptom checker landing screen on a phone: Symptom Checker, then 'Who is this symptom check for?' with Me or Someone else, above a Start symptom check button.
The Symptoms step on a phone: 'What are your symptoms?' asking for current symptoms one at a time, with a search field and a link to select from a list instead.
The Assessment step on a phone: 'What is your body temperature?' with four answer options and Back and Next.
The redesigned Symptoms step on desktop: Vomiting and Sore throat added under My symptoms, one at a time.

What people said

User response

“It definitely narrows things down really well. And even though I wouldn’t self diagnose anything usually, it gave perspective on how I might feel if I was properly ill.”

The problem

What success had to look like

Health literacy
People understand medical terms and instructions unaided
Confidence
People feel guided and supported throughout
Completion
60–70%Reach the industry benchmark
Task success
78%+Every severity 3 and 4 issue resolved
Step loss
<20%No single step loses more
Accessibility
AAWCAG 2.1, compliant in full

The approach

AI engine audit

The brief was to rebuild the Symptom Checker around Infermedica, a new AI triage engine, inside Australia's strict clinical standards, legal constraints, and Healthdirect's remit to guide people toward care without diagnosing them. Before any design, we audited what the engine could and couldn't do.

Infermedica triage results on a tablet: a 'Consult a doctor within 24 hours' recommendation above a list of possible conditions
Infermedica, the engine we audited

What held up

  • Strong clinical reasoning
  • Structured symptom assessment

Where trust could break

  • Outputs can read as a diagnosis
  • Sounds overly certain without supervision

The design agenda

  • Reframe recommendations for the Australian clinical and policy context
  • Clearer signals on when to move from AI guidance to self-care

The black note marks the core tension. Advice this confident could read as exactly the diagnosis the service is never allowed to give. The design agenda shows how the redesign answered that.

Landscape review

I ran a competitor analysis on a few symptom checkers, and Ada was one of them.

Ada Symptom Check home screen: 'Hi, I'm Ada. I can help you learn more about your health' above a 'Start symptom assessment' button
Ada's home screen, with the conversational framing our review kept coming back to.

What worked

  • Conversational, human voice
  • One question at a time
  • Visual diagrams demonstrate what a question means
  • App shows possible causes for the user's symptoms
  • Motion design helps with user engagement
  • “Tell me more” on each possible cause removes cognitive load from the report screen
  1. What worked: Conversational, human voice; One question at a time; Visual diagrams demonstrate what a question means; App shows possible causes for the user's symptoms; Motion design helps with user engagement; “Tell me more” on each possible cause removes cognitive load from the report screen.
  2. The friction we noted: Poor readability, with small fonts and low contrast on some small text; Tapping an answer is a challenge for big thumbs; Answer options placed too close together; Scanning multiple options not optimal due to right alignment; Some latency in question transitions.
  3. What we took forward: The patterns that reduced cognitive load, like one question at a time and plain language explanations, became conventions for our own flow; The readability and tap target failures became acceptance criteria for it.

User journey

Aliya

A time-poor parent researching a child's symptoms

“I take online search results with a grain of salt but I like smart tools that save me time and get to the point quickly and have constructive and progressive ideas on treatments and prevention.”

Situation

User actions

  • Child experiencing symptoms

Feelings and thoughts

  • I want to help my family
  • Researching online is me time
  • I want control over things

Needs

  • Learn more — symptoms and possible conditions

Pain points

  • Don't have time
  • Busy with children, parents and work
  1. Situation. User actions: Child experiencing symptoms. Feelings and thoughts: I want to help my family; Researching online is me time; I want control over things. Needs: Learn more — symptoms and possible conditions. Pain points: Don't have time; Busy with children, parents and work.
  2. Consideration. User actions: Visually engaging but not overwhelming; Human touch, feels a connection; A health tool I can trust. Feelings and thoughts: Supports me moving forward as I answer questions; Hits on every symptom I enter; Intelligent, flexible search. Needs: Simple tools, quick reads; Easy to understand questions; Reassurance and guidance. Pain points: If tool is difficult to use; Reading through long text.
  3. Complete questionnaire. User actions: Complete symptom check questions. Feelings and thoughts: This test feels too long. Needs: Plain language to understand questions; To the point. Pain points: Medical terms unfamiliar; Not user friendly; Too many questions; Doesn't feel intelligent.
  4. Care advice. User actions: Read care advice. Feelings and thoughts: Care advice is not giving me info on possible conditions. Needs: Possible conditions to my symptoms; Constructive ideas on treatments, prevention; Feedback, explanations, dialogue. Pain points: Time spent on questions only to get inadequate care advice.
  5. Get help. User actions: Seek help. Feelings and thoughts: Tells me my child should see a doctor and nothing more, it is pointless. Needs: Help me find a doctor near us; Should tell me more about what is happening to my child. Pain points: Not getting tailored advice; Not telling me about possible conditions.

Near-black notes mark where trust broke down — momentum stalling inside the question loop and care advice that didn't explain itself, the doubts behind most of the 55% drop-off.

Information architecture

  • Screen
  • Modal
  • Menu
  • Decision
  • System
Start

Tool introduction

What the check does — and doesn't do

accept terms and start

Basic questions

Who the check is for — age, sex at birth

select / input answers

Indicate symptoms

Free text or body diagram

Type a symptom

Natural-language search

view

Suggested symptoms

Body diagram

Select a point on the body

select

Symptoms for that body part

Selected symptoms

next

Question

“x of total questions” — one at a time

adaptive loop, until done

Why am I being asked this?

Plain language info on risk factors and medical terms

Answer history

View or edit any earlier answer

done

Disposition + reference number

The outcome, in plain language

  1. 1Self-care at home
  2. 2Speak to a nurse
  3. 3See a GP soon
  4. 4Go to an emergency department
  5. 5Call an ambulance
views

Care advice

choose a channel

Service Finder

Search services by need, fee, and distance

Web chat

Get help via texting

System dials the number

Healthdirect nurse, or ambulance

From here the result can be saved to a user account, or shared by email, print, or SMS.

Sketches

Rough concepts kept early debate on flow and framing, not visual polish.

Hand-drawn sketch of the tool introduction and basic questions flow
The tool introduction and basic questions flow, sketched before any UI.
Hand-drawn sketch of a symptom-entry screen option
Sketched options for symptom entry: free text versus guided selection.

Key design decisions

What user testing changed

See a GP outcome

“What to do” was reworded so the services read as a recommendation, not a paragraph to work through.

It now opens with a single sentence telling you what to do, with the important words bolded so people can scan it instead of reading every line.

The services then follow as numbered options. Each one says what the service is, what it costs and how you get seen, with a link straight to it.

Before

Advice, then one mixed list.

The original 'See a doctor within 2 hours' page, captured whole: the pink outcome banner, a 'What to do' heading, two paragraphs of advice, then one bullet list mixing Alternate GPs, Virtual Care Clinics and Urgent Care Clinics with their explanations, closed by a single 'About these services' link. Below the advice the page continues into a 'Find a service' postcode search, a Virtual Care Clinic panel, helpline guidance, a 'Check another symptom' prompt, and the standard healthdirect footer.
  1. The original 'See a doctor within 2 hours' page, captured whole: the pink outcome banner, a 'What to do' heading, two paragraphs of advice, then one bullet list mixing Alternate GPs, Virtual Care Clinics and Urgent Care Clinics with their explanations, closed by a single 'About these services' link. Below the advice the page continues into a 'Find a service' postcode search, a Virtual Care Clinic panel, helpline guidance, a 'Check another symptom' prompt, and the standard healthdirect footer.
  2. The redesigned page, from the top of the screen to the end of the options: the same outcome banner, a 'What to do' summary card stating you need to see a doctor within the next 2 hours, then three numbered options, 1. Virtual Care Clinics, 2. GP and 3. Urgent Care Clinics, each with bullets on what the service is, what it costs and how you get seen, and its own link.

Service guidance

The modal now covers a single service, with headings set as the questions people actually asked:

Do you need an appointment?How long is the wait?What does it cost?

Answering them clearly helps people pick the right service, which supports the goal of reducing avoidable emergency presentations.

Before

Every service, described in turn.

The original panel, titled 'About these services', shown whole: Virtual Care Clinics, Urgent Care Clinics and Emergency Departments each described in a flat paragraph and a bullet list, one after another.
  1. The original panel, titled 'About these services', shown whole: Virtual Care Clinics, Urgent Care Clinics and Emergency Departments each described in a flat paragraph and a bullet list, one after another.
  2. The redesigned panel, titled 'Learn more' and scoped to Virtual Care Clinics, shown whole: a short description, then question headings including 'Do I need an appointment?' answered with 'No' in the first word.
  3. The same 'Learn more' panel scoped to Emergency Departments, shown whole: a description of severe or life-threatening care, then the same question headings, 'Do I need an appointment?' answered 'No. Walk-ins always accepted', wait times, and cost answered 'Free for Medicare card holders'.

Service distance

The list of nearby health services sorts by nearest instead of grouping by service type, and shows how far each clinic is from the set location.

The search radius is adjustable and starts at 50km. The wide default is deliberate: urgent care clinics are sparse, so a tight radius would hide them.

Anyone looking for other services, like a GP, can pull the radius in to 2km.

Today

Sorted by nearest, radius adjustable.

The live service results for Urgent Care in Bowden: care-type filter chips with Urgent Care selected, a Filter button and a 'Sort by: Nearest' control, then a 'Search results' heading reading '50 results within 1166km', followed by result cards led by Total Urgent Care Norwood at 3.0km, each card carrying accessibility, bulk billing, opening hours and its distance.

  1. The live service results for Urgent Care in Bowden: care-type filter chips with Urgent Care selected, a Filter button and a 'Sort by: Nearest' control, then a 'Search results' heading reading '50 results within 1166km', followed by result cards led by Total Urgent Care Norwood at 3.0km, each card carrying accessibility, bulk billing, opening hours and its distance.
  2. The Filters panel over the results: an 'Open now' toggle, a 'Search radius' group offering 2km, 5km, 10km, 20km and 'more than 50km' with the widest option selected, then Billing and Appointment type groups, above a 'Show results' button.

Accessible by design

Every question is one row and one decision, with three targets big enough for a thumb. No dropdowns, no typing, nothing that needs a steady hand.

'Don't know' sits beside Yes and No, so nobody has to guess their way forward and the engine gets an honest gap instead of an invented answer.

The four stages stay named at the top of every screen. Tested with people with disability and with users wary of technology, and built to meet WCAG for screen readers and keyboard use.

The Background step on mobile with a four-stage progress stepper and health background questions, each answerable in one tap
  1. The Background step on mobile with a four-stage progress stepper and health background questions, each answerable in one tap

Guided help

Where a question could trip people up, a 'More info and how to check it' link opened a plain-language explainer.

It spelled out what the question meant and how to check it, so people could answer with confidence without leaving the flow.

The Assessment step asking whether you often have problems organising regular tasks or activities, with a 'More info and how to check it' link and Yes, No, and Don't know answers
  1. The Assessment step asking whether you often have problems organising regular tasks or activities, with a 'More info and how to check it' link and Yes, No, and Don't know answers
  2. The 'More info' modal explaining what the question means and how to check it in plain language

Actionable advice

There was a bigger goal behind the results page: take pressure off emergency departments by steering people toward virtual care and urgent care clinics when those could safely treat them.

Results surfaced the information people actually weigh when choosing care: wait times, cost, and appointment requirements.

Clinics were numbered in order of recommendation, with essentials like 'No fee' and distance highlighted on each service card.

'Seek immediate medical care' outcome page listing care options in order of recommendation — Virtual Care Clinic (no fee), Urgent Care Clinics, then Emergency Departments — above a Find a service search

This prototype is tappable. Info icons open an explainer for each care option, and the Find links open that service's search.

  1. 'Seek immediate medical care' outcome page listing care options in order of recommendation — Virtual Care Clinic (no fee), Urgent Care Clinics, then Emergency Departments — above a Find a service search
  2. 'Learn more' explainer for Virtual Care Clinics covering how care is delivered, whether an appointment is needed, wait times, and cost (free for Medicare card holders)
  3. Service Finder results for Virtual Care in Bowden, with care-type filter chips and a single virtual care clinic result
  4. 'Learn more' explainer for Urgent Care Clinics covering in-person care that cannot wait for a GP appointment, walk-ins, wait times, and cost (many bulk-billed)
  5. Service Finder results for Urgent Care in Bowden, ordered by nearest, each clinic card showing accessibility, bulk billing, opening hours, and distance
  6. 'Learn more' explainer for Emergency Departments covering severe or life-threatening injuries or illness, walk-ins always accepted, wait times, and cost (free for Medicare card holders)
  7. Service Finder results for Emergency department services in Bowden, ordered by nearest, each hospital card showing opening hours and distance

Keeping momentum

Usability testing showed the encouragement screen added a human voice and propelled users forward.

Steppers and clear instructions eased anxiety through a long clinical questionnaire.

Background step on desktop with a four-stage progress stepper across Background, Symptoms, Assessment, and Results, above the health background questions
  1. Background step on desktop with a four-stage progress stepper across Background, Symptoms, Assessment, and Results, above the health background questions
  2. 'First step done' encouragement screen thanking the user before the symptoms questions

The outcome

People finished the check, and left knowing what to do next.

Completion
84%Up from 49%, past the 60–70% target
Reach
2MChecks a year, across jurisdictions
Adoption
2×More used than the nurse helpline
Right care, right place
Only offers the pathways available to that person's location, time, and jurisdiction
Connected to care
Links straight to real-time appointment availability and booking through Service Finder
A front door
Became the foundation Healthdirect now calls Gen 2, its first step toward a national virtual front door
Read Healthdirect's announcement of Gen 2

Designing symptom checker in Healthdirect’s native app

The Symptom Checker was prototyped for Healthdirect's mobile native app as a future build, featuring an additional “Follow-up Care Plan”, a guided experience that helps users manage their symptoms at home.

As part of this work, I developed motion design prototypes to enhance user engagement and bring the interface to life. The concept was pitched to the business as a case for hiring a developer specialising in interactive animation.