This is my story of designing DMCoach, a mobile app that helps people with diabetes to better manage their condition, and empowers health professionals to deliver better care for their patients. Check also my account of the evaluation of DMCoach.
The project took place between January and December 2018. I was part of a team made of developers, researchers and junior designers, alongside Engineering (R&D IT Systems for Health) and Imec. I was responsible for the user research, co-design activities and the formative and summative evaluation, producing the major deliverables on design and evaluation activities by December 2018.
The World Health Organizations reports that there are about 60 million people with diabetes in Europe, and that worldwide, high blood glucose kills about 3.4 million people per year. Of the 60 million people with diabetes, almost 90% have diabetes mellitus type 2 (DMT2). Substantial evidence suggests that DMT2 can be prevented and improved with physical activity
and proper nutrition.
Our challenge was to develop a mobile service to help patients with DMT2 to adopt a healthy lifestyle, and physicians to deliver better care. The project was fast-paced and included the whole design process from user research to the final summative evaluation with patients. It also required efficient project management and communication skills to manage remote teamwork and a leading role in the design and evaluation while simultaneously contributing to other two main research projects. I used AbstractSpoon to manage multitasking and keep track of the tight schedule.
- Benchmarking. First, we did a competitor analysis to identify the best approaches and UX/UI practices for supporting people with DMT2.
- User research. We also spent considerable time researching the academic literature for novel approaches on diabetes management and behaviour change. Then we focused on Italian patients and physicians. We sent out three surveys to doctors and patients, and held a focus group with doctors.
- Design. With the results of the user research in mind, we ran a design sprint and designed the app with a number of design rationales in mind.
- Develop and test. We did several iterations of developement and testing, using both expert evaluation and user testing.
- Evaluate. Then, we ran the summative evaluation in three different cities (two in Italy and one in the Netherlands). And finally, we analysed the data and handed out the deliverables.
We did a benchmark analysis and reviewed 127 mobile apps for diabetes management and 52 food diaries (Android+iOS). We found that the vast majority of them allowed patients to keep track of their own data, like blood sugar, physical activity and nutrition, but few of them provided tailored motivational feedback. We also noticed the lack of health practitioners in these apps: only 5 out of 127 apps for diabetes reported to have involved them in the design of the service, and, although 20 apps allowed patients to share their data by email, no app provided full in-app communication with one’s family doctor.
We extracted and classified the best practices and UX/UI approaches in mobile apps for diabetes management.
We then focused on health practitioners and patients:
- We sent out a short survey to 24 physicians to understand what kind of information they deemed more important when monitoring diabetic patients, their preferred interaction patterns with patients, to what extent they relied on national and international guidelines to recommend lifestlyle advice and, how much they tailored lifestyle advice on individual patients.
- We asked these physicians to have their patients fill in a short questionnaire: the goal was to explore technology familiarity in our specific target users. Unexpectedly, the number of patients we were able to reach with this questionnaire wasn’t enough to get meaningful insights, fast. And we needed to quickly move to the data analysis to inform the design. So:
- I planned and designed a longer, online questionnaire that not only investigated technology familiarity, but also self-management practices, patients’ relationship with caregivers, tracking habits and needs. After a quick internal pilot testing, we sent out the questionnaire both in English and in Italian. However, since I’d learnt in my previous experience that senior users often require a different approach to participant recruitment, we not only posted the questionnaire on online forums about diabetes, but also contacted national organizations for diabetes management, who agreed to directly send the questionnaire to their associates, increasing our chances to get enough respondents. I supervised an undergraduate student on this, and we were able to collect substantial data from 132 patients. I used LimeSurvey to manage this questionnaire because it was a bit more complicated than the previous ones and contained a few logic jumps. This approach speeded up the process and made it more efficient, allowing me to analyse the data in time for the co-design sessions. It also brought an unexpected and fortunate discover: analysing the data I found out that Italian patients had rather peculiar habits regarding data tracking , compared to English-speaking respondents. This was useful to our business partner, since it clearly suggested cultural differences that should be taken into account when addressing non-Italian markets.
- We held a focus group with 2 doctors to explore pain points and desiderata in the most common practices of patient management.
We found that Italian patients…
…are low tech: they don’t use smartphone apps for diabetes management, they don’t use technology to share data with the doctor, but notes on paper and by memory.
…don’t track their data (mostly), but when they do, they tend to use paper.
…don’t share their data with doctors between visits.
We found that Italian patients long for correct information about diabetes and would be happy to share their data with doctors to be monitored.
Based on these data, one would seriously doubt that a mobile app for data tracking would be useful. But we looked deeper and found that our patients:
- Would have been happy to track, and share, their data with doctors to be monitored, have directions on how to behave, have feedback on their exams.
- Longed for correct information: they craved sound information, and specifically, actionable information on nutrition, to avoid sugar peaks, and to know if and when they could “give in to temptation”, to enjoy sometimes pizza or wine.
We also found that Italian physicians…
…Needed to administer personalized treatment, not only regarding medicine intake and frequency of blood sugar measurements, but also regarding nutrition and physical activity.
…Wanted patients to track physical activity, sigarettes, alcohol.
…Wanted patients keep a detailed food diary only for a limited time at the beginning of treatment, to help set correct habits.
…Wanted to monitor patients using concise reports on a few indicators, receiving prompt alerts if a patient deviates from an optimal range, and give positive feedback for good habits.
…Had many patients and little time: general practitioners in Italy can have up to 1500 patients, and the doctors we interviewed confirmed that they usually could allocate only 10 minutes per patient.
The design sprint
We prepared the sprint analysing the data from user research to derive meaningful takeaways. We used these data to design personas to guide design decisions and to foster empathy amongst our users and the team. To comply with time constraints of the team members, we did two iterations, the first a two-day long design sprint with the whole team to converge on the main approach and lay the foundations of the mobile app, the second a one-day follow-up to focus on the open question “How might we show user’s progress toward lifestyle goals?”. Here’s how it worked:
- The morning of the first iteration we met up with two UX and HCI researchers, two developers experienced in diabetes management and two junior designers. We introduced the key takeaways from benchmarking, user research, psychological theories of behavior change. We outlined our personas, to their needs and context:
- Mario, a 78-years-old overweighted widower with bad eating habits and who has been recently diagnosed with diabetes;
- Roswitha, a younger lady with a positive attitude towards a healthy lifestyle, who has been living with diabetes for several years and has already put in practice some healthy behaviour change;
- Giancarlo, a diabetologist who in a typical day can dedicate only 10 minutes to each patient and has no time to individually explain positive lifestyle change.
- We attached all this material and resources on the wall to act as our shared distributed external memory for the team, and to help our poor short-term memories. It kept everything that mattered smultaneously visible and easily accessible.
- With all this resources in plain sight, we elicited users’ paint points: on diabetes management for patients and on patient management for doctors. To guide priorities and design, we attached all these pain points on the wall, next to our personas.
Based on our personas and pain points, we identified two strategies to foster behaviour change
(lower stages of TTM)
Short educational pills to increase the cognitive dissonance between attitudes and bad habits and trigger behaviour change, plus rewarding feedback for baby steps towards healthy eating and physical activity.
(higher stages of TTM)
Tailored feedback and motivational messages to keep intrinsic motivation high, and a direct access to a dedicated area in the app where she can keep track of her progress towards lifestyle goals.
- In these strategies, all motivational feedback is tailored on the user’s tracked behaviour, stage of change, or sent by the doctor.
- We started to define user journeys to visually map the main stages of user interaction with the system. Keeping in mind that our user research pointed to low technology familiarity, we envisioned a great deal of interaction to be initiated not by the user, but by the system through notifications, adopting a conversational metaphor in which the system (or the doctor) sends messages to the user providing educational content, tailored feedback and asking for data.
- Following the user journeys, we outlined a few low fidelity sketches to converge on a shared approach.
- Conversational metaphor.
- Tailored feedback and motivation based on individual data and stage of change.
- Interaction initiated by the system.
- Simple interaction, few levels, clear labels.
- One-way communication from doctor to patient: patients receive direct feedback from doctors, doctors decide when to initiate communication.
The design rationales for the mobile app (patients)
Be there when users need it.
Respect users’ time, stay out of the way if not needed.
Avoid complex interaction.
Use a conversational metaphor
Adapt the content based on the user’s journey to healthy lifestyle.
Motivate with tailored messages.
Give useful educational content.
Give actionable and personalized feedack.
Reward and motivate.
Deliver value, convey the benefit.
ADD A HUMAN TOUCH
Connect patients with doctors.
Send doctor’s advice.
The design rationales for the web app (physicians)
BE CONSIDERATE OF DOCTOR’S TIME
Allow one-way communication.
Provide a quick configuration of patients’ profile.
Provide pre-written messages that are easy to modify.
ACT AS A PERSONAL ASSISTANT
Assist doctor in the management hundreds of patients.
Allow definition of patients’ personalized lifestyle plans and goals.
Convey essential, clinically relevant, pre-analyzed information.
We used Google Material design guidelines. We developed and shared a style guide to agree on a color palette and a set of standards regarding the formatting, the design and the interaction of the app. We then designed tha app with our design rationales and our personas in mind. We used Google Drive and InVision to share this work in progress.
The development and user testing
As the design was proceeding, our colleagues at Engineering started to work on the back-end, developing the core features of the app, integrating the front-end as it progressed. As we had a minimum viable product, we started right away with the user testing, integrating an expert evaluation when the first usability problems were solved.
Once we solved all the major problems, we ran the final evaluation with the patients in two Italian cities, plus a pilot evaluation of an advanced wristband for the unobtrusive monitoring of physical activity and physiological data in Eindhoven.
DMCoach was funded by EIT Digital. In collaboration with: