Smartphones assist us in many aspects of our lives – from keeping in touch with friends and family to scheduling work and helping us to lead a healthy lifestyle. Many of us also use health apps to track our run, join a Zoom yoga class or log our meals.
But are these health apps as effective as they can be? And given the popularity and wide reach of smartphones, can they be leveraged to deliver affordable and effective health care at a large scale?
One-size-fits-all doesn’t work
My research with Macquarie uni has indicated that the current one-size-fits-all approach in the health system is not enough to help people start and maintain healthy habits.
In other words, simply advising someone to walk 10,000 steps a day, without giving them specific advice on how to incorporate this in their personal life, is unlikely to change their behaviours.
Research from the US shows that each of us has a different lifestyle, needs and preferences that influence our decisions and health behaviours.
My PhD addresses this problem by focusing on “personalisation”, which aims to deliver the right health support to each person in the right moment, in a way they would be personally most receptive to.
Taking AI into fitness
To better understand how this can be helpful, let’s take a hypothetical example of the use of health apps. A student, say ‘Lila’ is 19 years old and recently moved out of home to attend university. To support herself financially, Lila also takes on a part-time job as a medical receptionist, in addition to her full-time study.
With such a busy schedule, and the fact that most of her work and studying involves working in front of a computer, Lila finds it difficult to stay active.
There are millions of students like Lila across Australia who undergo many life and academic changes as they transition from high school to university and work. European research has found that university students find it difficult and overwhelming to stay healthy while taking on these new responsibilities.
My own research into health apps serves to help people by Lila by combining smartphones with AI and similar tools to get to know the users – who they are, what they do and what their lifestyle patterns are.
Leveraging data to motivate
Smartphones store a wide range of information, including our activity, calendar and availability, and Internet search habits. All this information can be fed into an algorithm, which can learn about the users’ life pattern and identify the most appropriate moment to provide health advice.
While many existing health apps like Fitbit use novel algorithms to deliver motivating comments or suggest interesting activities, few actually consider the user’s thoughts and preferences.
This lack of user involvement means that some health suggestions might not fit into the users’ lives, making it difficult for people to incorporate healthy behaviours in their routine.
To overcome this problem, our multidisciplinary team at Macquarie University are developing and evaluating a personalised mobile app which generates activity suggestions based on user preferences and needs.
1. Gathers information from the users’ smartphones to understand their activity patterns and barriers to physical activity;
2. Suggests three choices to help the users be more active, and;
3. Allows the user to pick a suggestion that will be most suitable for them.
This approach promotes healthy behaviours by respecting the user’s autonomy and letting them choose the most suitable course of action for their life.
To ensure that we develop an evidence-based, effective mobile app that can be integrated in the larger health systems, our team of experts come from several disciplines, including medical doctors, machine learning experts, software engineers and user experience (UX) designers.
Our team members come from institutions across Sydney, including Macquarie University, University of Sydney, University of New South Wales, and University of Technology Sydney.
Here’s how it works
Here is an application of how this would work for Lila. By 3pm on Friday, based on Lila’s phone sensors, our mobile health app knows that Lila has been mostly sitting down during the week as she is studying for an upcoming exam.
The app will then send a notification, prompting Lila to check out the top three suggestions to be more active, such as “Why not go for a short 15 minute walk during your study break?”, or “Doing a set of 10 push-ups quickly can help freshen your mind”.
Subsequently, Lila chooses to go for a walk and become more active as a result. Lila gradually builds up healthy habits, incorporating more exercise in her busy life.
Our team has pilot-tested this novel personalised app amongst 23 students, and found an overall increase of more than 1300 steps in their daily step count. This initial result shows the promising potential of our approach, as past research has linked an increase in step count to reduce mortality and morbidity risks.
Given the promising results of our first trial, I hope to extend this research by incorporating advanced AI and machine learning techniques. For example, it is possible for our personalised mobile app to get more information about Lila by connecting to her calendar app.
Our be.well app can also connect to external sources to get information about the weather, or air quality.
With more information, the activity suggestion can become more specific and actionable, such as saying “Hey Lila, the sun is shining so why not go for a short walk and pick up a coffee from your favourite shop?”
While our approach is promising, some of our users had expressed concern over the privacy of their data. So, our future work will also investigate how to deliver personalised support without invading user privacy, in an ethical, safe and effective way.
Ly Tong is a PhD candidate and a Research Officer at Centre for Health Informatics at Macquarie University. Her project looks at the development and evaluation of personalised digital interventions for behaviour change, under the supervision of Dr Liliana Laranjo, Professor Enrico Coiera and Dr Juan Quiroz. Her research is supported by the International Macquarie University Research Excellence scholarship.