Design Thinking in healthcare: what changes

Design Thinking is now common product vocabulary. The five phases (Empathise, Define, Ideate, Prototype, Test) are the soundtrack of many workshops. In healthcare it’s no different, but there are nuances that shift the weight and rhythm of each phase.

This post pairs the framework with what I lived applying it in my IBS thesis. The notes hold for any health project, not just conversational AI.

Before the phases: HCD as a principle

Design Thinking works best inside Human-Centred Design, not instead of it.

HCD is the philosophy: start from explicit understanding of people, involve them throughout, iterate based on user-centred evaluation (ISO 9241-210:2019).

Design Thinking is the method: the five phases that operationalise those principles.

In healthcare this combo is particularly relevant because users aren’t just “users”. They’re vulnerable people, with long histories, and little patience for unnecessary iteration. HCD makes sure they’re not forgotten.

Phase 1: Empathise

Empathise in healthcare is slower and deeper than in consumer product.

In consumer product: 6-10 interviews of 30-45 min, directed questions, panel ranked by relevance.

In healthcare: 5-8 long interviews (60-90 min), open narratives, focused on personal history rather than features.

Reasons:

• The story is the data. Symptoms, appointments, decisions, doubts, system failures.
• The person rarely articulates what they need in a Q&A format. The narrative opens space for the real thing to surface.
• Time-to-empathy is time-to-trust. Short sessions get short answers.

Phase output: anonymised transcripts, emerging patterns, list of topics to explore deeper.

More on running this phase with patients in Designing with patients, not for them.

Phase 2: Define

Define in healthcare needs more artefacts than in consumer.

The typical ones:

• Personas based on real patterns (not invented).
• Empathy maps per participant (what they say, think, feel, do).
• Consolidated patient journey.
• Problem statement with at least one phrase indicating the user’s vulnerability.

For IBS, the final problem statement read: “People living with IBS struggle with self-management during acute flare-ups, particularly outside clinical hours, with no structured support and at risk of adopting harmful coping strategies.”

Each part of that sentence came from define: “self-management during acute flare-ups” (not diagnosis, not routine), “outside clinical hours” (1-3am is a critical moment), “harmful coping strategies” (Skrastins & Fletcher, 2016, on maladaptive coping in IBS).

On how to map the journey, see Patient care journeys.

Phase 3: Ideate

This is where healthcare diverges most. In consumer product, ideate is often an internal exercise: the team generates ideas, prioritises, picks.

In healthcare, ideate is most valuable as co-design. A structured session with patients where:

1. You present the problem statement.
2. You show the consolidated patient journey.
3. You ask them to imagine the “perfect product” for that moment of crisis.
4. You work with concrete material (cards, sketches, role-play).

For my thesis, three patients took part in a co-design workshop. The ideas that came out were better than mine in three ways:

• More anchored in reality. I thought in features. They thought in moments.
• More granular. I saw the problem as a block. They split it into micro-moments.
• More empathetic. I thought in utility. They thought in feeling.

Phase output: at least 3-5 viable concepts prioritised together with patients.

Phase 4: Prototype

Prototyping in healthcare follows the consumer pattern (wireframes → mid-fi → hi-fi → interactive) but with extra layers when there’s conversational AI.

For the thesis the sequence was:

1. Prompt flows: mapping possible conversations in a tree (input → expected response → next turn). Before wireframes.
2. Wireframes of the screens wrapping the conversation (state, controls, fallback).
3. High-fidelity designs with colour, type, microcopy.
4. Interactive prototype in Figma with clickable flows.

Note that prompt flow comes before wireframe. In CAI, the conversation is the product; the wireframe is what wraps it. Covered in Prompt engineering as design work.

Phase 5: Test

Test in healthcare has two stark differences:

1. Cost of error is higher. In a productivity app, a false positive causes irritation. In a health assistant, it can drive the wrong decision in a crisis. Testing has to cover adversarial scenarios with care.

2. Patients catch things others don’t. During the test phase of the thesis, five IBS patients flagged issues my team hadn’t seen: temporal ambiguity in suggestions, lack of emotional acknowledgment after pain reports, clinical jargon at stressful moments.

Output: prioritised issues list, iteration decisions, next test criteria.

What changes overall

To sum up, compared to Design Thinking in consumer product:

• Empathise: longer, more narrative, less task-focused.
• Define: more artefacts, problem statement with explicit vulnerability.
• Ideate: co-design >> internal brainstorm.
• Prototype: prompt flows before wireframes (in CAI), increasing fidelity, mid-step validation.
• Test: adversarial scenarios, real patients, care at critical moments.

And one difference that runs across all phases: iteration is slower. It’s not method friction, it’s respect for people. Worth it.

More on the background in the Design for Health guide. On how to recruit patients for these phases, see Designing with patients, not for them. On patient journeys (a critical define output), see Patient care journeys. On Grice’s Maxims as an evaluation rubric in test, see Grice’s Maxims as conversational design heuristics.