Lessons from designing digital health for patients, with patients
What designing a disease management app for patients living with chronic conditions taught me about the friction between people and technology.
Disclaimer: these learnings are based on qualitative observations and secondary research. I have omitted and obfuscated confidential information, and the learnings are my own and do not necessarily reflect the views of a client or a partner.
Designing products and services for chronic disease management is challenging, for two main reasons: First, the healthcare industry is a highly regulated, complex ecosystem, with many moving parts that rarely communicate effectively. Designers in healthcare must not only understand patient’s needs and frustrations but also consider the interests and motivations of different players in the ecosystem, for example, physicians, insurance companies and governments. This article by Chris Kiess is a deep dive into the complexity of designing within the healthcare ecosystem. Second, motivating a new routine is not an easy task. Designers should learn about digital behavioral design and psychological models, The 6 stages of behavioral change, for example.
According to the CDC, chronic conditions such as diabetes, hypotension or mental illness cause significant health and economic costs. Ensuring that patients proactively manage their illnesses is key to reducing these costs and improving people’s lives. Disease management apps have shown the potential to engage more people in managing their health, but they have yet to live on to their full potential.
Last year, I designed a disease management app for people living with chronic conditions. Through immersive patient interviews, secondary research and co-creation sessions, I learned about people’s emotional and functional barriers that may restrain them from proactively managing their health, and their relationships with technology. Here are my top 5 learnings.
#1 The patient-provider dynamic is changing
The old paradigm of the paternalistic model of medicine is transforming into an equal partnership between patients and doctors, where patients are empowered to take control of their health. The role of technology, in this case, is to facilitate a conversation between patients and their doctor, discussing a medication plan, for example. For this to work, patients must comply with a treatment plan and commit to self-management.
To establish trust, technology should augment rather than replace patient-provider relationships.
A core component of the doctor-patient relationship is the patient’s faith in the advice and treatment recommended by the doctor. In a recent survey, adults with chronic conditions were more likely to put high trust in their doctor, compared to those who consider themselves healthy.
Trust is individual and context-dependent. However, for most patients who participated in our interviews, there seemed to be a direct connection between trust and compliance, meaning that patients who trust their doctors will be more likely to use an app, as long as it has been ‘prescribed’ to them.
While data privacy was a general concern for many, patients were willing and expecting to share their data with their healthcare providers in exchange for a more personal treatment plan and improved health.
In any app that collects medical data, it’s important to design a simple, easy-to-discover feature for patients to share data with their doctor.
Apple’s Health allows users to download and share their entire health data. Fitbit recently announced a partnership with Google to improve its ability to share data with medical professionals, although it is currently only available from the web app. In many data collecting apps, the data sharing feature is “buried” under Settings (Flo app, for example for women’s health) or Profile (Apple health). eMoods, a mood tracking app to manage bipolar symptoms, is a better example. It allows data sharing directly from its Graphs tab, a more intuitive, in-context placement.
#2 Healthcare is personal
Every patient journey is unique to them and patients’ needs may vary based on their unique conditions, including permanent or temporary disability, lifestyle, and personality. Looking into a patient's entire journey may reveal the reasons why they fail to follow up on a treatment plan. For example, a missed medical appointments, for many patients, is the result of challenges in coordinating the logistics of special transportation needed to accommodate their chronic condition, as described in this HBR article.
A good principle to keep in mind is the importance of design for extreme users. Inclusive Design considers from the very beginning how something might be used by everyone. Consider those who are least likely to engage because of their age, low-digital literacy, a permanent or temporary disability, or any other contextual or environmental cause.
To make treatment plans more personal, doctors collect and analyze patients' medical data. A truly personal view of patients will not only include “medical” data, but also information about sleep, exercise, and nutrition. For example, OneDrop, a diabetes management app, collects nutrition and exercise data in addition to glucose level and medication.
An even more personalized approach also collects information about lifestyle indicators such as travel, life transitions (e.g divorce), and emotional well-being (e.g stress level). Stress level tracking is important given that many chronic conditions, including depression, heart conditions, and diabetes, are worsened by stress. Breeze, a mood tracker, collects data about pretty much any lifestyle activity related to the user’s self-reported mood. Flo, a women’s health tracker, tracks mood as well as lifestyle data such as travel and alcohol consumption.
#3. Patients don’t believe in “digital magic”
Some of the reasons for poor adoption of wearables are related to poor usability and perceived accuracy, as found in one AARP study from 2015. Since then, wearables’ functionality and usability have improved significantly, although perceive accuracy is still a barrier for adoption.
Patients interviewed had two main concerns when it comes to accuracy: the first is sensor accuracy, especially in cases where data is being collected passively. The other is the accuracy of the analysis and insight. This was especially true when we presented concepts of predictive analytics and diagnostic features.
Every sensor used in tracking devices comes with an accuracy level, but accuracy level information does not have a primary location on UI. On the Garmin app, for example, there is a written section about sensor accuracy found down the More Menu. Similarly, the Apple Watch does not communicate the sensor accuracy, but it does provide information on how to make the reading more accurate.
As with in-person interactions, digital health interactions needs to be designed with transparency, using clear communications to promote trust. Like humans, machines aren’t perfect decision-makers, and medical predictions are based on information collected by both sensors and humans. Showing the data behind the prediction is one way to increase trust.
One Drop, a diabetes management app recently expanded its AI-powered Predictive Insights to include 8-hour blood glucose forecasts. The app provides an option to see more information about the prediction, which is directly linked to information previously provided by the user.
Another way is to be transparent about the confidence level of a prediction. In AI, a confidence level is a numerical expression of certainty in percentages. Showing users, the confidence level of insight or a prediction will increase trust. This UI pattern has been used in the past by kayak on their price prediction, however today they only include a description of the data the prediction is based on.
#4 Tone of voice can make a big different
While the overall approach must be practical, empathetic design can go a long way when it comes to healthcare UX. By keeping a positive tone when educating and communicating, patients could subconsciously feel more hopeful about the outcome of their treatment.
Another way to enhance positivity in the patient experience is positive reinforcement. Adding prompts similar to those found in fitness apps (e.g “you’re doing great, keep going!”) can improve adherence, as long as they are designed thoughtfully and in the right context: You may want to avoid sounding too enthusiastic when the data means conditions are worsening, for example.
From those interviewed, some patients appreciated a more casual language at times, but most of them still wanted the overall design and tone to feel “medical” or “clinical”.
eMoods and Breeze are mood trackers that take the opposite approach when it comes to tone. eMoods’ design language and tone are clear, minimal and utilitarian. On the other hand, you’ll find rainbows and clouds on Breeze’s interface, and a friendly, casual tone. After you complete your first entry, the app encourages you to keep going with a smiling cloud.
Some patients might find an app like Breeze too casual to be considered medical, while some may find the eMoods’ interface dry and discouraging.
Moodpath, another mood tracking app, is a better example. It prompts users with subtle notifications, in a neutral, respectful and optimistic tone.
#5 Patient satisfaction will be measured by time saved — not spent
Measuring success for digital health varies. With patient adherence being one of the biggest challenges in healthcare, a good design will be measured by patients’ engagement and participation. Ideally, it will lead to behavioral change that leads to improved health, cost, and efficiency.
Chronic disease management is time-consuming, especially in those cases when providers prescribe multipart protocols. While patients’ needs broadly spread across many spectrums, we found that when it comes to compliance, patients have a common goal: lessen the time they spend on “dealing” with their illness.
Disease management is not something patients choose or want. It’s about what they must do daily. That is why designers should design with effectiveness and efficiency in mind, keeping a clean feel and clear communications.
There is no ‘delight’ or ‘surprise’ in disease management, and patients are looking for tools that will end up saving them time, not add on to their plate.
As a table stake, digital tools should work without friction, and be supported with attentive service. A nice to have will be an added personal touch or a sleek and modern design, for example. Beware of anything that feels too ‘gimmicky’ or any language that sounds more like “marketing talk”. Patients perceived those as a waste of time.
In addition to barriers of adoption resulted from a permanent or temporary disability, people are busy, distracted or simply uninterested. No matter what the barrier is, tracking tools must be designed to collect data with a minimum amount of effort from the user, within medical and technological constraints.
Passive tracking might require some setup or specific tools, but once it is working the actual tracking should be largely automatic and require almost no additional action from the tracker. All you do is turn it on and you are collecting data. Wearable trackers offer this form of passive data collection. Garmin’s sleep tracking only requires users to set a goal, before passively tracking and analyzing sleep.
A good example of passive adherence tracking is the propeller inhaler, that collects data during usage. And then there is tracking from within the body, or “digestible”, for example, the FDA approved Abilify MyCite. These pills include sensors that record when the medication is taken.
But passive tracking is not always possible. Many data points require manual entry, and designers are challenged to make a data entry and simple as possible. We can reduce effort by sending reminders and notifications to enter data, with the simplest question format possible. Moodpath, for example, is using simple yes or no questions to collect self-reported data.
Notifications are also used for critical times, like warnings (for example, heart rate threshold alerts) and timely reminders (”it’s time to take your pill”). When properly executed, push notifications can provide users with the most relevant information at the right time and encourage them to take action.
For predictive notifications, it is important to include an advice on how to act on the insight, so users can feel more in control.
For example, One Drop’s high glucose warnings include advising on how to reduce the risk of a rising glucose level.
Push notifications can be disruptive if not managed properly and thoughtfully. Notifications should be flexible and easy to set and manage by the patient with or without their provider. The content and frequency must be attainable, realistic and time-sensitive. Patients might feel frustrated, annoyed or guilty if they cannot act on an actionable health insight at a given time.
Human-Centered Design is already playing a key role in the future of healthcare. Solutions must be developed out of a deep understanding of patients’ needs, pain points, motivations, and beliefs, and not out of the recent buzz of a technological breakthrough. An innovative, novel solution by itself may attract some enthusiastic early adopters at first but is not what will drive much-needed sustainable adoption or behavior changes.
Patients don’t think technology is here to replace their doctors but to assist with mutual decision making. They will most likely try a tool that is approved by their doctor, as long as they trust their doctor’s advice.
Patients are more than happy to share their medical data with their doctors, in exchange for a more personalized treatment plan. But many barriers prevent people from following up on their plan. These barriers include physical barriers, such as debilitating pain of disability, emotional barriers, such as low motivation, and pessimism, and mental barriers such as lack of trust or poor time management.
There are many ways app design can address these barriers. First, making it easy for doctors to know more about their patients, through holistic data collection that is easy to share. Second, adding more transparency patterns such as confidence level will increase trust in technologies and processes that are relatively new. Third, using an appropriate tone will make patients feel like they are being acknowledged. Finally, we must focus on optimizing interactions to be quick, frictionless, and non-disruptive, as patients wish to spend less time on their illness and more time doing what they love.
