Blog · · 5 min read

Designing for safety in mental-health apps

What good crisis-safety design looks like in a digital mental-health tool: clear signposting, honest limits, and a person who follows up.

Banner reading: Designing for safety in mental-health apps

A mental-health app will, sooner or later, be opened by someone who is not safe. That is not an edge case to handle at the end of a project. For any tool used alongside counselling, it is one of the main things the design has to get right.

This article sets out what good practice looks like, drawing on published guidance rather than our own opinion, and explains the choices we made in Myndgard Share.

The problem is real and well documented

In 2019, researchers reviewed 69 depression-management and suicide-prevention apps from the Apple and Google stores against six evidence-based suicide-prevention strategies drawn from clinical guidelines. Only 5 of the 69 apps (7%) included all six. Six apps (9%) gave an incorrect crisis helpline number, including two that had each been downloaded more than a million times. Most apps did include emergency contact information (65 of 69), so the failures were not about intent. They were about detail, testing and upkeep.

A wrong phone number is a small bug with a serious consequence. That is the right mindset for safety design: ordinary software quality, applied to the moment it matters most.

Signpost clearly, and to services that are always open

Samaritans' guidelines for online platforms, published in 2020, are written for sites hosting user content, but their advice on supporting people in distress applies to any digital service. They say that all sites should signpost users in distress to:

  • emergency and non-emergency services
  • support that can be reached 24 hours a day, every day
  • trusted helplines that specialise in self-harm and suicide
  • people the user trusts, such as family and friends.

The same guidelines ask that messages to someone in distress be concise, non-judgemental and written in clear, accessible language, because it is hard to take in a lot of information when you are struggling.

In Ireland, that means making sure 112 or 999, Samaritans on 116 123, and the free 24/7 text service 50808 are easy to find from anywhere in the app, not buried in a settings menu. For a university service, the institution's own out-of-hours arrangements belong alongside them.

Be honest that the app is not an emergency service

The most important safety message is often the plainest one: this app is not watched around the clock. If a student believes that typing something worrying into an app will bring an immediate response, the app has made them less safe, not more.

Good practice is to say this early, in onboarding, and again at the moments it matters: next to any question about self-harm, and whenever an answer suggests risk. The message should tell the person what will happen (for example, that their counsellor will be told during working hours) and what to do right now if they need help sooner.

Put a person, not an algorithm, at the end of a risk answer

The World Health Organization's LIVE LIFE approach to suicide prevention, launched in 2021, names four key interventions. One of them is the early identification, assessment, management and follow-up of anyone affected by suicidal thoughts and behaviour. Identification is only the first step. The value is in the follow-up.

NICE's evidence standards framework for digital health technologies makes a related point. It asks developers to define the level of professional oversight a technology needs, proportionate to the risk if it fails, and to provide safeguarding assurances where users may be in vulnerable groups. It also expects technologies to comply with relevant clinical safety standards.

For a tool used in counselling, that translates into something concrete. When a questionnaire answer indicates risk, such as the self-harm item on the PHQ-9, a named clinician should be alerted so they can follow up using their service's own risk procedures. The app should not try to assess or manage risk on its own, and it should not leave the student wondering whether anyone saw their answer.

Follow safe messaging guidelines in every word

Apps contain a lot of words: questionnaire items, notifications, educational content, empty states. Samaritans Ireland's media guidelines for reporting suicide are written for journalists, but their core advice transfers well. Avoid any detail about methods. Avoid dramatic language such as "suicide epidemic". Present suicide as preventable and signpost sources of support.

In practice, that means reviewing every piece of content that touches on self-harm, writing notifications that never reveal sensitive information on a lock screen, and keeping a clinician involved in content review.

Keep the safety details maintained

Helpline numbers change, services merge and opening hours move. Safety content needs an owner and a review date, like any other clinical content. The 2019 app review shows what happens when nobody checks.

How we approach this in Myndgard Share

Myndgard Share is for students who are already in counselling. We have tried to build it around the principles above:

  • Not an emergency service, and we say so. Alerts and messages are not monitored around the clock. The app tells students this and shows emergency numbers, Samaritans and 50808, together with the university's own emergency contacts.
  • A person follows up. Answers that indicate risk, such as thoughts of self-harm, alert the student's assigned clinician so that a person can follow up. The alert supports the clinician's judgement; it does not replace it.
  • Discreet notifications. Notification text is kept generic so that it does not reveal health information.

In practice

If you are a counselling service considering any digital tool, these are fair questions to ask:

  1. Where are crisis contacts shown, and who checks that they are correct?
  2. Does the tool say clearly that it is not monitored 24/7?
  3. What happens, step by step, when a student gives a risk answer, and who is told?
  4. Has the content been reviewed against safe messaging guidelines?
  5. Who is responsible for clinical safety, and how are incidents reported?

None of these needs new technology. They need care, clear words and a named person at the end of the chain.

If you're struggling: Samaritans 116 123, text 50808, or 112/999 in an emergency.

Sources