Towards better medication adherence apps: My PhD research

Now that I’ve finished my PhD (well, I finished it last year, but I’ve never been good at updating this blog) it would be good to summarise things. There’s a nice summary page back at UCL, but I would like something here as well – it’s my PhD after all. If you have too much free time, you can read my thesis (download PDF). Otherwise, you can read the summary below.

Preventing forgetfulness by facilitating the formation of routine-based remembering strategies

The aim of my research was to explore how we could use approaches from behaviour change and habit research to design an app that could help people remember their medications better. I was specifically interested in long-term preventative regimens, such as oral contraception, because there are no symptoms that one could use as external reminders. When women take the Pill, they do it to keep things as they are – not to get rid of runny nose, fever or pregnancy. This lack of symptoms of course makes things harder to remember.

Understanding the context

The first thing I wanted to do was to understand the context: how women currently remember the Pill, what works, what doesn’t, and whether they use any apps. After analysing nearly 1000 responses to my online survey it became apparent that women tend to make the Pill a part of their daily routine and rely on contextual cues to help them remember; hardly anyone bothers with apps – I’ve written about it in more detail earlier.

These remembering strategies are similar to those used by parents who have to remember their children’s antibiotics and older adults who have to remember multiple medications – we’ve checked, just to be sure. Because medication-taking is a routine behaviour that is also highly contextual, it does make sense that people tend to rely on daily routines, objects and locations as memory cues. This is true across populations and regimens, and studies conducted by other researchers with other patient groups and other conditions confirm that (our paper links to relevant studies).

Smartphone apps

So what about the apps? Why do people not use them? I’ve also checked this. Turns out there are a few issues. First, medication reminders – as the name suggests – focus on reminding. However, this does not mean they help people learn how to remember – they don’t; the point of a reminder is that you don’t have to remember yourself. As a result, there are no features that help people fit the medications into their daily routine or link the apps with the contextual cues they rely on. I’ve written a bit more about this study here. We’ve published this study at CHI (you can download a pdf here).

This made me think about habit formation apps – they are specifically designed (or at least are supposed to be) to help people repeat a new behaviour every day until it turns into a habit. I’ve looked at them as well. Turns out, they aren’t that good either. They mostly provide self-tracking and reminders, and do not support contextual cues or routines – even though routines and cues are also super-important when trying to form a habit! I’ve written a bit more about habit apps here. We’ve also published this at CHI (you can download a pdf here).

Habit formation

This made me wonder: do people not know how to support habit formation or is it simply too difficult to do that using an app? According to the literature, habits form when the behaviour is consistently repeated in the presence of stable contextual cues. After a while, these cues will start to prompt the behaviour until it becomes automatic. Could apps somehow support this process?

To answer this question, I conducted two studies that looked at how habits form. The first study focused on mechanisms of habit formation and the impact of different cues (reminders vs. a routine event) on remembering and automaticity of behaviour (see our paper or download PDF; I’ve also written about tech challenges related to running this study). The second study looked at habit formation in the wild, when people are free to select their own cues (I’ve yet to publish the results, but if you’re curious, check out Chapter 5 of my thesis).

The results from these studies were… frustrating. Turns out, while contextual cues are good at facilitating habit formation, they do not prevent forgetting when the behaviour is still new – people need time to learn their cues. The opposite is true for reminders: while they are great at the beginning and help people repeat the new behaviour, they can actually hinder habit formation in the long term, as people have no incentive to try and remember on their own.

So if we want to help people form medication-taking habits, we need something that supports both memory and habit formation from day one, without introducing app dependency (download PDF). Tricky.

Effective remembering strategies

To figure out how to support memory and habit formation I returned to my earlier studies to find the most effective strategies and their components. I also ran another study to find out how people select their cues in the first place, because from the habit formation studies I learned that some of the cues they selected were not every good.

These studies confirmed that the most effective are combinations of contextual cues: daily routines, meaningful locations and meaningful objects. If technology helped people to select these cues, this would make things easier. But before we help people do that, we need to understand how people select cues in the first place. What makes them decide to put things in a specific place or take medications at a specific time? Sometimes, we have specific instructions how to take things, but with the Pill it is up to a person to decide when is the best time. How do they do that?

Selecting remembering strategies

I couldn’t run a study with oral contraception (that would require a lengthy ethics approval process), so we decided to investigate how people decide how to remember vitamin C tablets (this also involved a long ethics approval process, but two months are better than ~six!). I’ve selected vitamin C, because the study was planned for an autumn so people would be more likely to want to participate. Also, missing a vitamin C tablet doesn’t have serious consequences like missing a contraceptive Pill. (Actually, if you have a healthy diet, you don’t need to take vitamin C tablets at all!)

We asked our participants to take one tablet per day for three weeks. We interviewed them before and after the study. We asked them how they were going to remember and later checked whether they followed their plan (spoiler: almost none of them did). The study confirmed that combinations of contextual cues are the most effective and also showed that reliance on a single cue (regardless of whether it’s a routine, an object, a location) is not a good idea. We also discovered that people can’t always anticipate what will work for them and that selecting the right strategy requires lots of trial and error. Hopefully we will publish the results soon, but if you really want to know all the details RIGHT NOW, they are in my thesis.

Implications for design

Based on the research described above, I’ve come up with a list of design requirements. Even though I was interested in smartphone apps, I’ve kept the requirements technology-agnostic, so they can be applied to other types of technologies.

If you’re planning to build a medication reminder that helps people take advantage of contextual cues, or working on a habit formation app, make sure to:

Help users define a good strategy
People tend to rely on routines and other contextual cues, and clearly defined strategies made of multiple cues are the most effective. The app should help users link their medications (or a new behaviour) with an existing routine task they already do and encourage them to consider additional cues.
Provide examples of good strategies
People do not always know what will work for them, so provide a list of example routine events that can be effectively used as triggers (e.g. meals, getting up) and help users define their own routines.
Support trial and error
Finding the right strategy requires trying out different cues, which means that users may end up choosing weak or unreliable cues. Technology should monitor their progress and suggest changes to the strategy if it’s not effective. Users should also be able to modify the strategies recorded in the app (or whatever tech they use) when they identify better cues. The technology should treat each change in the routine as a starting point of a new regimen and reinforce the new remembering strategy to ensure it fits into user’s daily routine.
Remind about whole strategy, not just the task
Reminders can be effective, but to avoid developing dependency on reminders, make sure to remind about the whole strategy and all cues. For example, instead of reminding to users to “Take your Pill”, remind them the whole plan: “Remember to take your Pill after breakfast”.
After a while, disable the reminders
To further decrease likelihood of developing dependency on reminders, make sure they disappear with time. Cue notifications should be available only when starting a new regimen/behaviour. They should stop when users learn their cues and the behaviour becomes a part of their daily routine.
Allow people to check if they’ve already done it
It is easy to forget if you’ve done something you usually do automatically, so the technology should allow users to check that. Cue notifications (see point 4.) could serve as an additional visual cue for situations when the routine changes: as long as the notification is visible and has not been dismissed, users will know they still have not taken their medication.

Theory into practice

I built a prototype of an app that is based on these recommendations. Then I worked with MSc students who built their own version based on the list. You can read more about both apps here. The point of the apps is to demonstrate how my research could be used in practice.

If you end up using any of my findings when developing your product, do let me know!

Related publications

~Falka, 23 December 17