Home PublicationsData Innovators 5 Q’s for Peter Lucas, CEO of Hedia

5 Q’s for Peter Lucas, CEO of Hedia

by Eline Chivot
Peter Lucas

The Center for Data Innovation spoke with Peter Lucas, chief executive officer and co-founder of Hedia, a medical devices and equipment company based in Copenhagen. Lucas discussed how Hedia’s personal assistant uses AI to provide patients with type 1 diabetes with recommendations and support in their daily lives.

This interview has been edited.

Eline Chivot: What led you to create Hedia?

Peter Lucas: In 2010, at age 27, I was diagnosed with type 1 diabetes. I had been raised by a father who was diagnosed with type 1 himself when he was 27 years old as well. I’ve seen him going through insulin shock a few times, and I’ve seen these hard times diabetes can give to a family. When I was diagnosed, I did the same mistakes my father did, and I was struggling with the condition. So my co-founders and I really created Hedia based on a personal need. I needed something that could help me live a normal life, do the things I love to do, spend time with my family, enjoy a good dinner without thinking about my condition all the time—about how much insulin there is in my body, about how this will affect my blood glucose levels (BGL), about whether these are too high or too low, etc. Hedia takes away some of those issues.

Chivot: How does the life of a patient change when he or she is diagnosed with type 1 diabetes? How is Hedia transforming that?

Lucas: Speaking first from my own perspective, before I was diagnosed, I was a busy person. I was self-employed, I had a company and was organizing events for big corporations in Denmark. I suddenly started to lose weight, I started to lose my sight. I was angry, bitter, in a constant mood swing. And from one day to the next I went from being Peter, to being Peter with type 1 diabetes. When I was diagnosed, one issue I struggled with straight away was that I was—or at least thought I was—no longer able to do the things that I loved to do, including those work-related. So I ended up selling my company. I couldn’t continue on with a chaotic, stressful, fast-paced lifestyle which wouldn’t fit a life with diabetes. It would not have allowed me to be in control of my “new” life which required me to eat the same thing, at the exact same time, every day. I had to transition from chaos, which I loved, to structure, which I don’t love that much. Hedia has given me the ability to “go back” to the life I was leading before being diagnosed. It helps me lead a more chaotic life, do things with my familyall those things I would do before and can now do again, while being in control of my condition.

Our users include people which have been living type 1 diabetes for many years, and people who have been recently diagnosed. The former category tells us that Hedia helps them regulate and “cut” some of the high BGLs and some of the lows, that it gives them more insights into their conditions, and that it empowers them to make the right choices. For people who have been newly diagnosed, Hedia is a tool they can use to transition into the “diabetes lifestyle”—I use the term “lifestyle” because diabetes involves things you can do and cannot do. With Hedia, in the end, there are only a few things that you cannot do, as long as you make sure that you take the right amount of insulin or take into consideration how much insulin you have in your body.

Hedia is in our users’ pockets, helping them track food intake, activity, mood, and nudging them into making better choices. It produces measurements to make sure that the users are on the right track—meaning that their BGL reflects what has been set as a target. Hedia also “remembers” the insulin level that’s in the user’s body, so if it detects a risk for him or her to have a low BGL at a particular moment, we can recommend the precise food intake to avoid hypoglycemia, which is typically the consequence of a low BGL.

Chivot: Which variables and aspects of a user’s condition do you include in the algorithm? How do you build and improve your application?

Lucas: Currently, we are working with different kinds of machine learning algorithms. There is one in particular that we will take into clinical trial—Hedia is a medical device, so we need to make sure that everything we offer is safe for patients to use. Our machine learning algorithm can predict how much insulin people should have injected and should inject based on our users’ historical data, with all the human errors in it. If I go into more detail, I would be revealing our “secret sauce” and unique selling point. But what I can tell you is that this algorithm has to go through a clinical trial before we can put it to use and provide it to users. Our data only comes from the users and includes the BGL, target BGL, average BGL, insulin intake, type of food intake, type of exercise, time in range, and so on. We have a food database for users that includes the ingredients needed, the ingredients consumed, nutritional information, vitamins, etc.

We build and improve our app primarily based on user feedback, so that includes surveys, interviews, and focus group tests, which take place at our offices. Our users can also freely provide us with feedback, just feeding us information about what they would like to see improve or a feature they would like for us to add. They can do so through the app, but also directly by email, phone, social media, and our website’s chat. When we have enough feedback, and it starts clustering up, we follow up and integrate some of their requests into our software. Let’s say for instance that some users want a food database with X amount of food items. When enough users have expressed this request, we start building it. We do have our own to-do list, which we try to pair up with our users’ needs and “wish-list.” If their needs and our priorities are not pairable, we will prefer building upon the things our users need. Good feedback is great for us as a team, but it isn’t what improves our product. Our product improves when people open up and share their true issues, concerns and pains.

Our main source of innovation isn’t always this 1-to-1 communication between us and our users, but we listen to them because their feedback can be pretty critical in terms of protecting their health. For example, Hedia is only available in English for now. A Danish user asked us to have it developed in Danish, and we waited until enough users came up with the same request, suggesting it was a significant issue and a relevant change to make. We are currently converting the app and translating it to offer a version in Danish. It was important to use our complaint handling to assess whether there was a higher risk for Danish users to make mistakes using the app in English rather than in their native language. Chances were we needed to lower that risk, and we will be able to do that by translating the app, which is the “medical” way to improve Hedia. The other way is “technical,” based on user feedback. We try to combine both.

Chivot: How does Hedia differentiate itself from other digital coaching tools?

Lucas: I have never thought of Hedia as a digital coach actually, that is the first time I hear we may be part of this space. Hedia is a medical device. The app is what enables our users to use the software, but the algorithm in Hedia requires for all our software to be developed as medical software, meaning we have to follow an EU Directive when build our app. This involves a number of boundaries. Hedia is approved as a medical device by CE (Europe’s certification mark), so we’re allowed to market our company in Europe only—not in the United States or China. If we want to take Hedia to the United States, we would need to have it approved by a U.S. federal agency, the Food and Drug Administration, according to its own standards. That can be a costly affair. So our entry barrier is higher than if we were a normal digital coach or a welfare tech company. Our competitors would be companies that develop diabetes apps as medical devices. Hedia is a full-blown medical device, our space only includes Europe, and so far we have identified only one competitor, which was acquired two years ago by a bigger company. As far as we know, we are the only active CE-approved diabetes app in Europe.

Chivot: How else do you anticipate using AI to help patients?

Lucas: We currently work with our “bolus” algorithm, the insulin algorithm behind Hedia. This is the one we’re going to clinical trial with, which is based on historical data and can predict how much insulin users should have injected based on the input Hedia gets from them. We’re working with different types of machine learning in our food database where we can pair up our users with other users who may look alike but who actually are in better control. We then try to nudge users to take some of those steps other users are taking to lead a better life. For example, we can pair up a man who weighs a 100 kilograms and has high BGLs with another man who also weighs 100 kilograms but has a lower BGL. Users can check what it is that these other users do in their lives which makes them in better control of their condition, compare that with their own activities and food habits, and then consider making better choices.

Development could be applied to Hedia’s exercise feature: we currently have five different exercises in our app which users can choose from and schedule. We are looking into how exercise affects a person’s BGL. Based on that knowledge, we want to start automating the app’s insulin recommendations, so that users know how they need to adjust their intake. Users would be able to check their BGL before exercising, for instance before they start a 5-kilometer run, so that they won’t put themselves at risk of hypoglycemia when running—which can happen in case of a low BGL.

We are working with machine learning on various areas, both supervised and neural network. Some of our work must go through clinical trials, and we can implement some other parts that do not require those trials in a more “techy way” to be able to move forward faster—but always remembering who we are and what we are: a medical device. So if that was to have a direct effect on our insulin recommendations, we always need to make sure there is still clinical validation behind what we do.

You may also like

Show Buttons
Hide Buttons