Healthcare apps and behavioural science: maybe it’s time?

You are inherently irrational.

Before you leave the page, consider two decisions:

People often avoid the gamble in the first scenario and choose it in the second, even when the values are equivalent. Prospect theory explains why: people respond differently to perceived gains and losses. Our decisions are not perfectly rational, but the ways in which they depart from rationality can be surprisingly consistent.

Behavioural scientists have documented many of these patterns. Finance, technology, social media and gaming businesses have become particularly good at using them to shape how people interact with products.

Healthcare has an engagement problem

Healthcare has a long-standing problem with patient engagement. The World Health Organization's report on adherence to long-term therapies estimated that adherence among people with chronic conditions in developed countries averaged about 50%.

Behavioural science may be able to help. Even in 2020, tens of thousands of health apps were trying to deliver more personalised care. Some were already using behavioural techniques to improve engagement and retention, although the quality of implementation varied widely.

Three ideas stood out to me at the time:

1. Understand what the patient already knows.
2. Segment by behaviour and motivation, not only diagnosis.
3. Use incentives that respond to the individual.

1. Understand prior knowledge

People do not always perceive their own health risk accurately. A person may describe their health as good while living with a condition that creates substantial future risk. An intervention built for somebody already looking for help will not necessarily work for somebody who does not yet see a reason to change.

That means a health product needs to understand what the user knows, what they believe and what gap the intervention is trying to address. Education should start from that point rather than assuming every user arrives with the same level of awareness.

2. Segment by behaviour

Healthcare often categorises people by diagnosis. On a ward, somebody becomes the patient with a nosebleed in one bed or the patient with heart failure in another. That shorthand is useful for clinical work, but it can easily become a one-size-fits-all model for engagement.

Digital products can go further. Demographics and diagnosis are only part of the picture. Motivation, confidence, routine, previous attempts and personal circumstances all affect whether somebody can follow a plan. Behavioural segmentation gives the product a better basis for deciding what kind of support may be useful.

3. Match incentives to the person

The Behaviour Change Technique Taxonomy provides a common language for describing techniques used to support behaviour change. These can include giving feedback, setting a specific goal, planning an action or recognising progress.

In a health app, that might mean:

1. giving clear feedback about a person's health or progress;
2. agreeing goals that depend on a specific behaviour; and
3. recognising meaningful achievements rather than rewarding empty activity.

The point is not to add badges to a clinical product. It is to choose support that fits the person's knowledge, motivation and circumstances.

The ethical limit

Behavioural design can become manipulation when it hides information, makes alternatives difficult or serves the product at the user's expense. That risk matters in healthcare, where consent, autonomy and trust are central.

A defensible approach should meet several basic tests:

1. Do not build an intervention on an untruth.
2. Do not make the alternative deliberately difficult to choose.
3. Look for unintended consequences as well as the intended effect.
4. Make consent visible and explain what the intervention is doing.
5. Use methods you would be comfortable defending in public.

What I believed then

Behavioural science had clear potential to improve engagement and support preventative care. The difficult part was not identifying another bias or adding another nudge. It was understanding the individual well enough to choose an appropriate intervention, then using it without compromising autonomy.

At the time, I was building EczemaDoc as a self-management app for eczema. We co-designed the product with patients and used these principles to think about how people recorded symptoms, understood triggers and stayed engaged long enough to learn from the data. EczemaDoc later became Proton Health, but the core question remained useful: how can a health product help somebody act without trying to control them?