Case studies:

Pump & CGM

Pump & CGM

 

Overview

Creating data visualization across multiple devices using first of its kind data integration between insulin pumps, continues glucose monitors, blood glucose devices, food and exercise data.

Responsibilities
I was the Lead Designer on this project and responsible for wires, user testing, prototyping, and all design. I worked alongside one Product Manager to make sure all goals were reached.

 

Three Goals

 

Web Research

We started by going over Medtronic, Dexcom, Insulet, Tandem, CapturAGP and many other software platforms graph by graph, chart by chart. We leveraged our Medical Advisory Board (a group of widely accredited diabetes doctors we consult with) to get feedback on what they liked, what they didn’t like and how the software they use now can be better. After a week of interviews we had a directions for the web platform that focused on 3 key features.

 

Mobile Research

With the patient app we already had a good starting point for what we thought patients wanted to see. CGMs and Insulin Pumps would be specific to one set of users, those with Type 1 diabetes. Based on past user testing we outlined two new sections we thought would satisfy most T1 patients.

 

Personas

I identified several key personas that I would target in our initial launch. For doctors I focused on those still working with PDF reports or that use some sort of graphing software. For patients I decided to focus on first adopters, that means the data nerds and the highly motivated parents with a T1 children. This platform was our first crack at the CGM and Insulin Pump market I wanted to make sure the early adopters would become advocates.

 

Wires & Mocks

I started sketches solutions while we were still discussing with doctors about which direction we should go and doing our research. Between the Medical Advisory Board, the PM and myself we had enough information for me to put a full color mock together in a day or two. It wasn’t too pretty but it was something we could start getting feedback on instead of a concept we had to describe.

 

Weekly User Testing & Development

Each week we met with patients and doctors and each week our mock got better and better. While working on finalizing the mocks we were getting our developers rolling on proving out the concepts we had in mind. So when it came time to bring our concepts to life the devs were already building it and just needed the finishing touches. We had some time to work with since the backend engineers were still hard at work on the data integration piece. 

 

Get Out of the Building

The Product Manager and I traveled to Boston and Kansas City to sit down with patients and doctors to get first hand feedback. Because of the number of patients and doctors (30) and short amount of time (4 days) the interview process needed to be very efficient.

Testing Sites

Joslin Diabetes Center
Joslin Diabetes Center is the world’s largest diabetes research center, diabetes clinic, and provider of diabetes education. It is located in Boston, Massachusetts, USA.
Children's Mercy
Children's Mercy Hospital is a comprehensive pediatric medical center in Kansas City, Missouri that integrates clinical care, research and medical education for patients ages birth to 21.

Testing Method

  1. Lifestyle and behavior
    Learning about a person’s current habits is one of the most important steps to understanding a products success or failure in a person's life.
  2. Pain point ranking
    We singled out 6 common pain-points to people with diabetes and asked which was most important. This would help guide feature prioritization.
  3. Upload personal data
    Having our MVP run with personal data made patients more engaged with our testing. There was a personal stake in making sure we got things right with their data.
  4. Tell us what you see
    We let users explore the system and describe what they see without interfering to explain. It was very beneficial to let the user dive deeper into the product for answers.
  5. Tasks
    Asking users to complete tasks using the system to gauge ease of use. (ex. How much did you bolus last night? Can you tell us the last time you went high? etc)
  6. General feedback
    An opportunity for users to voice things they liked, didn’t like and things we might be missing.
  7. If we could make one thing for you what would it be?
    Boiling down all of their feedback into one big ask. This consistent question helped us group patients by interest for future product testing.

Key Takeaways

Patterns
People saw patterns in their diabetes data but wanted to know what to do next.
Many solutions
Every doctor had a different preference for displaying patient data and one view we didn't include was highly requested.
Context
The ability to easily see the food they ate or miles they ran is crucial for making sense of glucose events.
Pain points
Adults need help figuring our bolus amounts for food while kids struggle finding the correct basal rates.

 

RESULTS

Web Platform

24 hour Ability to scroll left and right between days and view blood glucose, medication, carbohydrate, bolus, basal and exercise data.

24 hour
Ability to scroll left and right between days and view blood glucose, medication, carbohydrate, bolus, basal and exercise data.

Calendar Months worth of blood glucose and bolus data. Reduced amount of information to essential to spot patterns or odd days to dive into.

Calendar
Months worth of blood glucose and bolus data. Reduced amount of information to essential to spot patterns or odd days to dive into.

Trends Aggregate Seeing 1 week to 90 days worth of blood glucose data all on one 24 hour graph. Also individual day menu above the main graph gives the user the ability to select which days to include or exclude from the main graph.

Trends Aggregate
Seeing 1 week to 90 days worth of blood glucose data all on one 24 hour graph. Also individual day menu above the main graph gives the user the ability to select which days to include or exclude from the main graph.

Trends Individual The Trends graphs all represent the same data in different ways. We found in user testing that some doctors preferred this way of looking at data where each line represents one day. They are color coded by day of the week.

Trends Individual
The Trends graphs all represent the same data in different ways. We found in user testing that some doctors preferred this way of looking at data where each line represents one day. They are color coded by day of the week.

Trends Individual - hover state Hovering over a day will highlight that day of the week while still being able to see all other days of the week in the background.

Trends Individual - hover state
Hovering over a day will highlight that day of the week while still being able to see all other days of the week in the background.

Pump Insights Joining blood glucose and insulin pump data gave us the unique opportunity to create this view. Insulin Pump injection sites should be changed every 3 days, going beyond this will have an adverse effect on your bodies ability to process insulin leading to high blood glucose. While prior to this doctors would simply have to elude to this fact but in this view the graph above shows blood glucose variability going higher beyond 3 days of changing an insulin pump site and below it every change event has it's own blood glucose graph outlining data hours before and after the change.

Pump Insights
Joining blood glucose and insulin pump data gave us the unique opportunity to create this view. Insulin Pump injection sites should be changed every 3 days, going beyond this will have an adverse effect on your bodies ability to process insulin leading to high blood glucose. While prior to this doctors would simply have to elude to this fact but in this view the graph above shows blood glucose variability going higher beyond 3 days of changing an insulin pump site and below it every change event has it's own blood glucose graph outlining data hours before and after the change.

Temp Basal Insights People on occasion will stop, reduce or increase the flow of insulin from the insulin pump. This is done for many reasons, a common one is exercise the effects of which cause the body to process insulin better. Doctors wanted insights into what happened during these events and since this was done exclusively by the patient how successful they were in keep their blood glucose at a reasonable level. Each temp basal event has a different duration and we solved this by splitting our glucose graph into 15 minute increments. Whether the event was 1 hour or 6 hours the blood glucose during the event would all be accessible with a glance. 

Temp Basal Insights
People on occasion will stop, reduce or increase the flow of insulin from the insulin pump. This is done for many reasons, a common one is exercise the effects of which cause the body to process insulin better. Doctors wanted insights into what happened during these events and since this was done exclusively by the patient how successful they were in keep their blood glucose at a reasonable level. Each temp basal event has a different duration and we solved this by splitting our glucose graph into 15 minute increments. Whether the event was 1 hour or 6 hours the blood glucose during the event would all be accessible with a glance. 

Mobile Platform

Week View Because of the sheer amount of data coming from all connected devices we had to focus user attention on what was most important to help them make a decision. We chose 3 metrics that give users enough information to spot an odd week and dive deeper to see the days in that week.

Week View
Because of the sheer amount of data coming from all connected devices we had to focus user attention on what was most important to help them make a decision. We chose 3 metrics that give users enough information to spot an odd week and dive deeper to see the days in that week.

Day View Once the user has spotted a day they want to review we give them a minimal graph, stats and a list of each individual event for that day. The user can either drag on the graph to see the exact number for blood glucose, carbohydrates and insulin bolus or scroll on the event list below for more context around the event like what type of food they ate.

Day View
Once the user has spotted a day they want to review we give them a minimal graph, stats and a list of each individual event for that day. The user can either drag on the graph to see the exact number for blood glucose, carbohydrates and insulin bolus or scroll on the event list below for more context around the event like what type of food they ate.

Day View - Landscape For users that really want to explore the graph we included a landscape view that includes a y-axis. The user can swipe between days.

Day View - Landscape
For users that really want to explore the graph we included a landscape view that includes a y-axis. The user can swipe between days.

No Data and Quick Info When we don't have data we show the screen on the left. When the user first syncs their devices and goes to the day view screen we show the screen on the right.

No Data and Quick Info
When we don't have data we show the screen on the left. When the user first syncs their devices and goes to the day view screen we show the screen on the right.

Key Learning

Personal

Try often to improve often
Creating a quick first version leads to quicker critique and faster improvement. Subsequent iterations should follow the same rapid feedback cycle to ensure the product is on the right path.

Company

Show and tell
Visualizations of data aren’t enough, both patients and healthcare professionals are looking for easier identification of problem areas and recommendations for treatment.

 
HypoMap

HypoMap