Haleh Nobar

Politecnico di Milano · Digital health · 4-month student project, 2022

Digesti.

Making everyday health tracking useful for patients and their specialists.

28

Survey respondents

5

Expert interviews

4

Test participants · prototype tested, not launched

UX/UI designer and UX researcher

In a team of five

Role, team and status 6 facts

The patient app combines photo-based food logging, symptom tracking and a chatbot; the specialist web app brings everything recorded into one history.

My roleUX/UI designer and UX researcher
TeamHaleh Nobar, Bianca Bardela, Kira Bernauer, Olivier Bertrand, Brigitte Anna Stegers
ContextInteraction Design Studio, Politecnico di Milano · 4 months, 2022
PlatformsiOS app, chatbot, specialist web app
ToolsFigma, Landbot, LogMeal API
StatusStudent project; prototype tested, not launched

01 — See

The diagnosis needs a diary. The diary needs daily effort.

For people investigating a possible food intolerance, recording meals, symptoms and daily context can become a repetitive task. Specialists need that history to look for patterns, but incomplete or inconsistent entries make it harder to use. We focused on young adults, who often turn to online resources before seeking professional advice.

20%

Up to one in five people may have a food intolerance.

Healthline

10 of 11

respondents with regular digestive problems had lived with them for more than three years.

Our survey

5 of 11

felt their healthcare professional didn’t take their concerns seriously. Only 3 of all 28 had a diagnosis.

Our survey

One diary, two perspectives

Patient

Less effort.

A diary that’s manageable to keep, every day, for weeks.

“Getting help would almost feel weird, but also very much appreciated.”

Patient 14 · food diary study

Specialist

More context.

Everything recorded, in one place.

“A digital food diary in combination with health specialists is really great and timesaving.”

Medical student · expert interview

Where we intervened

The weeks alone with the diary.

The hardest stretch is repeated self-recording, when patients have to stay motivated without immediate feedback. That’s where we focused.

OnsetInformingGetting helpIntakeTestingAnalysingDiagnosis
DoingSlight digestive problems startReflects on symptoms, researches onlineLooks for a specialist, or goes it aloneTalks to the specialistTests, food diary, exclusion dietWaits for results, reflects on dataDiagnosis, or none
Pain pointsDiscomfort, stress, social limitsFalse self-diagnosis, overwhelmSlow, little empathyLittle time, little empathyTedious, frustrating, motivation dropsCan’t interpret data aloneNo diagnosis, or restrictions
OpportunityEarly awarenessReliable guidanceA trusted connectionShared historyEffortless logging + buddyClear overview for bothFollow-up and support

Patient journey, from first symptoms to diagnosis · the two stages Digesti was designed for are marked · swipe sideways for all seven stages

The finding that set the direction

Finding · survey and 5-day diary study

Recording meals and symptoms became tedious.

Implication

Reduce manual work without preventing corrections.

The evidence behind it method · 2 more findings · 5 closest products · 4 requirements · 4 personas

The challenge wasn’t simply making food logging faster. We needed to understand what made patients stop recording, and which details specialists actually needed. We reviewed 9 papers, benchmarked 20+ products and ran a 5-day diary study with the team and friends.

Finding 02

Patients wanted guidance, but were wary of an app’s diagnosis.

→ Be supportive, and connect to a healthcare professional.

Finding 03

Specialists needed history and context, not single records.

→ Bring meals, symptoms and context into one overview.

The closest productsWhat each one does
mySymptomsFood and symptom journal to find IBS triggers
CaraCareGut-health diary, one of the few linked to a specialist
FoodyPersonal diary for food and possible symptoms
AtlasPhoto diary that rates meals for gut bacteria
Mijn EetmeterNutrients and calories, with tailored advice

Five of the 20+ products we benchmarked

Most products were quick but narrow, or broad but slow to fill in; few were linked to a specialist. Asked about the problems they met with food diaries, survey respondents named time above everything else.

How might we keep patients logging without costing them time, and help specialists review a patient’s history?

Effortless

Log a meal with one photo; nothing is mandatory.

Kind

A chatbot buddy that guides, without judging.

Complete

Meals, symptoms, stress and sleep in one history.

Connected

Linked to a specialist; the app doesn’t diagnose.

Constraints: needs a connection and a third-party food-recognition API; health data must stay private; specialists can’t be flooded with raw data.

The hardest stretch is repeated self-recording, when patients have to stay motivated without immediate feedback. Four personas, four barriers to keeping a diary: Bill, 22, the effort-avoider; Sheldon, 25, the self-diagnoser; Karen, 27, the worried researcher; Deborah, 18, the forgetter.

02 — Understand

Three decisions behind the product.

What is easiest for the patient isn’t always most useful for the specialist; each decision balances the two.

Decision 1 · Food logging

Scan, then correct.

We dropped the prototype’s “Am I correct?” step, saving a tap on every meal: the photo proposes the dish, and patients fix ingredients on the next screen. Everything else stays optional, because our nutritionist warned that strict logging can encourage obsessive behaviour.

The trade-off

Accuracy against effort: recognition can be wrong, but confirming every meal costs time. Corrections stay possible, only where they are needed.

Prototype

  1. Scan
  2. “Am I correct?”
  3. Edit ingredients

Final design · one blocking step fewer on every meal

  1. Scan
  2. Edit ingredients, only if needed
  1. The scanner: the camera recognises a pizza and labels it.

    1 · Scan

    The photo proposes a dish.

  2. Edit food: the recognised ingredients as removable chips, with a Finish editing button.

    2 · Correct

    Ingredients stay editable before saving. Fix only what’s wrong.

Decision 2 · Chatbot

A guided conversation.

A chatbot buddy sets patients up without forms or judgement.

The trade-off

Guidance against authority: patients wanted help but distrusted an app’s diagnosis, so the chatbot guides and collects, and the judgement stays with the specialist.

Connects

Links the patient to their specialist with a personal code.

Learns

Asks about symptoms with predefined answers.

Prototyped

Built in Landbot for testing.

In the first test

Trust in the chatbot 5.25 of 7; perceived empathy 3.33 of 5, a moderate score.

A concept sketch of a phone showing a chat between the patient and Digesti.
Concept sketch · a conversation instead of a form

Decision 3 · Specialist experience

From individual entries to a patient history.

The specialist dashboard brings symptom trends, dietary context and patient history together in one interface for review.

The trade-off

Patient convenience against specialist needs: quick entries only become useful when they add up to trends, concerns and a timeline a specialist can scan.

The specialist web app on a laptop, with the patient app on a phone in front.
The patient app on the phone, the specialist web app on the laptop

1

Symptoms over time

A chart compares reported symptoms across dates, so the history reads as trends rather than isolated records.

2

Points of concern and current diet

Summary cards flag relevant observations, with the patient’s current diet alongside.

3

Patient timeline

A chronological feed of symptoms, meals and diet events, for reviewing individual records in context.

Concept and flows 3 scenario sketches · 11 flow diagrams · the system architecture

Sketches and scenario videos led to one system: photo logging, a guiding chatbot and a shared history. The overview answers “how am I doing?” first; logging stays one level deep, and the chatbot is reachable from every screen.

  1. Concept sketch: a phone photographing a pizza.

    Scenario 1

    Snap a meal, done.

  2. Concept sketch: a chat with Digesti.

    Scenario 2

    A conversation instead of a form.

  3. Concept sketch: a specialist reading a patient overview on a monitor.

    Scenario 3

    The specialist reads the diary.

Quick mode takes returning users straight to the food scanner. New users register and set up their profile through the chatbot, then reach the main menu. When the specialist changes the diet or sends feedback, a notification opens the chatbot; it explains the change and answers questions.

  1. Onboarding and navigation, diagram.

    Onboarding and navigation

  2. Registration · flow, diagram.

    Registration · flow

  3. Registration · screens, diagram.

    Registration · screens

  4. Food scan · flow, diagram.

    Food scan · flow

  5. Food scan · screens, diagram.

    Food scan · screens

  6. Register a symptom, diagram.

    Register a symptom

  7. Diary · flow, diagram.

    Diary · flow

  8. Diary · screens, diagram.

    Diary · screens

  9. Specialist feedback · flow, diagram.

    Specialist feedback · flow

  10. Specialist feedback · screens, diagram.

    Specialist feedback · screens

  11. Chatbot, diagram.

    Chatbot

  12. System architecture, diagram.

    System architecture

The team’s flow documentation and the system architecture · select one to open it full size

Iterations greyscale to final · 4 screens · what changed and why

Food capture got quieter, a tab bar replaced the menu, and the specialist page grew into concerns, diet and history. The earlier specialist prototype combined an overview chart with five metric tiles. We reorganized the dashboard around flagged concerns, dietary context and a timeline.

  1. Overview, greyscale prototype.
    Greyscale
    Overview, final design.
    Final

    Overview

  2. Scanner, greyscale prototype.
    Greyscale
    Scanner, final design.
    Final

    Scanner

  3. Edit food, greyscale prototype.
    Greyscale
    Edit food, final design.
    Final

    Edit food

  4. Specialist page, greyscale prototype.
    Greyscale
    Specialist page, final design.
    Final

    Specialist page

On the specialist page we removed five metric tiles and a placeholder patient card, and added flagged points of concern, the current diet and a timeline of meals and symptoms.

Overview

To keep the symptom trend as the focus while making food capture quieter, with the scanner one tap away in the tab bar.

Scanner

To keep logging moving: the scanner suggests the dish, and any correction happens on the next screen rather than in a blocking yes/no step.

Edit food

To keep entries accurate with less typing: a wrong ingredient is removed as a chip, not retyped.

03 — Make clear

Quick to log, structured to review.

Manageable daily logging for patients, and a history specialists can review over time. Patients start a meal entry with a photo, correct the recognised ingredients when needed, and use a chatbot to record additional context.

01 · Patient records

A photo of each meal, symptoms through the chatbot.

02 · Digesti structures

Recognised ingredients and symptoms on one shared timeline.

03 · Specialist reviews

Trends, flagged concerns and diet in one patient history.

What the first test told us

Four participants played Bill, our effort-avoider, in 20–30 minute sessions with the Figma and Landbot prototypes. Four people and one short session each: these are first signals, not proof.

5.25 / 7

Trust in the chatbot

3.33 / 5

Perceived empathy, a moderate score

3 of 6

UEQ scales rated excellent

4

Test participants

Worked

Clear, efficient and attractive: rated excellent on three UEQ scales.

Needs work

Less engaging day to day: stimulation and dependability rated lowest against the benchmark, and the chatbot scored low on being available to talk.

Not yet tested

The specialist app, with specialists.

User Experience Questionnaire · approximate mean per scale, read from the chart in the 2022 report (p. 111)

  1. Attractiveness2.0excellent
  2. Perspicuity2.1excellent
  3. Efficiency2.0excellent
  4. Dependability1.2above average
  5. Stimulation1.2above average
  6. Novelty1.1above average

−30 · neutral+3

Full UEQ scale, −3 to +3 · n = 4, sessions of 20–30 minutes · ratings are the UEQ benchmark categories · first impressions, not generalisable findings

What we established

Established

Logging felt clear and efficient in a short test.

Still unproven

That patients keep logging for weeks, and that specialists review a history faster.

What I learned · Design the input as carefully as the output

Reduce effort without hiding errors

Photo recognition helps only when patients can still correct the result.

Support, don’t diagnose

Patients wanted guidance but distrusted an app’s diagnosis, so Digesti hands the judgement to their specialist.

Design both sides of shared information

What feels like a quick diary entry to the patient becomes part of a longer history for the specialist.

Before launch 3 things still to do

Still open

Test the specialist side

Sessions with doctors and nutritionists, since the specialist dashboard was never tested with the people it was for.

Still open

Design for week three

Stimulation rated among the lowest against the benchmark. Add gentle motivation for daily logging, then run a multi-week patient pilot instead of a 30-minute session.

Still open

Decide the measures first

Diary completeness, continued use to the end of an exclusion diet, and the time a specialist needs to review a history.

Next project · 04

A Highway to Greenness.

How car brands make their models look green.

↑ Top