Skip to content
Expert Guide Series

What Regulations Must Your Mobile Health App Comply With

A client came to us with a healthcare and health insurance product built for African markets, and the first real decisions on it had nothing to do with features. They were about what we were allowed to hold, how long we had to keep it, and what the product was allowed to say to someone who opened it while frightened. Anyone building in this space has to work out which health app regulations UK bodies apply to them, and that question lands far earlier than most teams expect.

Regulation decides what you're allowed to build, so it belongs in week one, not week forty.

Health products sit on top of several rule sets at once. Medical device law, data protection, clinical safety standards, consumer and accessibility duties, app store policy, and if money moves through the product, financial rules on top of all that. None of them wait politely until you're ready.

What follows walks through each area in turn, using work we've actually done. An anonymous messaging app where we set the retention policy. A baby monitor where we wrote the contract between app and hardware. A dating app where skipping one piece of discovery cost £15,000. A peer to peer currency exchange that Apple read as a money laundering risk. The rules always turn up in the same place, which is the detail of a single screen.

Deciding Whether Your App Is a Medical Device

This is the first fork in the road, and it's settled by intended purpose rather than by technology. If your product is meant to diagnose, prevent, monitor, predict, treat or relieve a condition, the MHRA treats the software itself as a medical device under the UK Medical Devices Regulations 2002. You then need to register, hold technical documentation, run post market surveillance and carry UKCA marking. A step counter sits outside that. A tool that reads the step data and tells someone their heart is at risk does not.

What catches teams out is that the wording on a marketing page can move you across the line on its own.

What the product doesHow it reads to a regulatorWhat follows
Logs symptoms for the user to read backRecord keepingUsually outside device rules
Tells the user what those symptoms meanDiagnosisAlmost certainly a device
Suggests changing a dose or a treatmentTreatment supportA device, and a higher class

Wording is a regulatory decision

On the genetics wellness app we worked on, we rewrote result copy so variation was framed before anyone saw it, with lines like this is to be expected and everyone's different. That started as a way to stop people reading their own results as an error. It did regulatory work too, because it kept the product describing patterns rather than handing out verdicts.

UK Data Protection Law and Special Category Health Data

Health data is special category data under UK GDPR, which means consent or another Article 9 condition on top of your normal lawful basis, a Data Protection Impact Assessment before you build, and a much shorter tolerance from the ICO if you get it wrong. Everything downstream gets heavier. Logging, analytics, crash reporting, third party SDKs, support tooling. Anywhere the data lands is in scope.

We saw how physical that can get on a fitness social product. The original design put everyone's live location on a map so people could meet for shared workouts. Research showed a high proportion of users were going to be women, so precise location sharing turned into a safety problem rather than a feature. We rebuilt it with randomised location offsets, so you see a vague area, talk, read a profile, and only then choose to share exactly where you are.

Run your DPIA against the data your third party SDKs collect, not only the data your own code writes. That's where most of the surprises sit.

Leakage is rarely dramatic. NowSecure's mobile privacy benchmark found that 70% of analysed mobile apps can leak personal data through storage, APIs, logs and SDK traffic. In health, that's not a privacy nuisance, it's a disclosure of someone's medical history.

Start your app project the right way

We deliver the complete blueprint before a line of code is written. User research, psychology-driven design and full technical specifications. You choose who builds it.

See how we work Get started

No commitment

Operating Across Borders: EU, US and Emerging Frameworks

The healthcare and health insurance product we designed for African markets taught us to stop thinking of compliance as one checklist. Every market has its own rules about where health data can sit, who can hold it and what has to stay inside national borders, and those rules shape architecture, not just policy documents. Deciding on your hosting regions late is expensive. Deciding early is free.

In the EU you've got GDPR, the Medical Device Regulation for anything with a clinical purpose, and now the European Health Data Space, which brings duties to make health data available in standard formats.

A product sold in four countries is four regulatory products wearing one set of screens.

Research in npj Digital Medicine looked at 100 health apps from the app store charts and found 21% could qualify as health data holders under the EHDS anonymous data pathway, based on their published privacy policies alone. Plenty of teams are in scope without having worked it out yet.

The US is different again. HIPAA only bites if you're a covered entity or working for one, so a direct to consumer wellness app often sits outside it and lands instead under the FTC's Health Breach Notification Rule and a growing pile of state privacy laws. Assuming HIPAA covers you, or assuming it doesn't, are both ways of getting it wrong.

Clinical Safety, DTAC and Selling Into the NHS

If the NHS is your buyer, the bar moves. Clinical risk management runs under DCB0129 for manufacturers and DCB0160 for the organisations deploying your product, and you need a named clinical safety officer who is a registered clinician. That person signs a hazard log and a clinical safety case report. Neither of those can be written retrospectively in a week, because they describe decisions you made during design.

On top of that sits the Digital Technology Assessment Criteria, which trusts use to assess products before they buy.

  1. Clinical safety, with your DCB0129 evidence
  2. Data protection, including your DPIA and DSPT submission
  3. Technical security, including penetration testing
  4. Interoperability, usually meaning FHIR and NHS login
  5. Usability and accessibility against WCAG 2.2 AA

Hazards hide in copy

A hazard log tends to get written as though risk lives in the algorithm. A lot of it lives in wording. The genetics wellness work showed us how much a single sentence changes what somebody does next, and a clinical safety case has to account for the user who reads a result at eleven at night and panics. Write the risky screens into the hazard log, not just the risky calculations.

Consent, Withdrawal and the Right to Be Forgotten

Consent for health data has to be specific, informed and freely given, which rules out one bundled tick box covering treatment, research, marketing and analytics. Each use needs its own decision. Withdrawal has to be as easy as giving consent was in the first place, so if signing up took two taps, leaving can't take an email to support and a five day wait.

That sounds like a burden. In practice it's one of the few pieces of regulation that improves the product on its own.

Asking permission properly, and saying plainly why you want something, is a framing and tone of voice change rather than a technical one. People end up psychologically more bought in when they feel they hold control over what happens to them. They're more engaged and better retained, even though the end result, the notification permission or the data access, is identical to the version that just demanded it.

Put a permissions screen in the account area that shows what the user has agreed to, in their words, with a toggle beside each one. It satisfies the regulator and it doubles as a trust signal.

The right to erasure follows the same logic. Build deletion as a product feature with a clear confirmation and a stated timescale, rather than a legal process bolted on after launch.

Data Retention When Deletion Conflicts With Legal Duty

We hit this head on while working on an anonymous messaging app. GDPR gives people the right to have their data removed. Separately, there's a legal standing around retaining data in case of criminal activity. Both apply at once, and in a messaging product they pull hard in opposite directions, because the person most motivated to delete their account is sometimes the person who just sent an abusive message.

We settled on a retention policy of around six months. Deleting your account stopped the data being visible or usable in the product, and it didn't wipe the record instantly. If the police opened an investigation, the data was still there to hand over.

Health products have the same shape of problem with different names on it. Clinical records carry statutory retention periods. Safety reporting for a medical device has its own. Insurance claims have theirs. None of them vanish because a user tapped delete.

Write the policy before the delete button

A retention schedule should say, per data type, how long you hold it, why, what triggers the clock and what happens at the end. Then the deletion flow can tell the user the truth, which is that their account is gone, their records are locked away for a stated period, and here's the reason. People accept that explanation. What they don't accept is finding out later.

Identity, Authentication and Safeguarding Obligations

We normally argue for deferring login, because making someone create an account before they've seen any value is one of the fastest ways to lose them. On the anonymous messaging app we did the opposite and put authentication right at the front. The nature of the product meant we couldn't let people in and start sending messages without being able to tie every message back to a particular user on the back end.

That was a legal and safety call rather than a product preference. Messages stayed anonymous to other users while remaining traceable internally, which is what made abuse and harassment cases actionable.

Anonymity to other users and anonymity to the system are two different products with two different legal positions.

Health apps carry their own version. If you're serving under 18s, the ICO's Age Appropriate Design Code sets defaults you have to meet, and the Online Safety Act adds duties where users can contact each other. If a clinician is giving advice through your product, you need to know their registration is real, which means verification at onboarding and a way to revoke it.

Work out who you're legally answerable for before you design the sign up flow, because the answer changes where authentication sits and how much friction is justified.

Payments, Insurance and Financial Crime Rules

We built a peer to peer currency exchange where travellers swapped leftover foreign notes with each other at interbank rates, avoiding commission. The transfer mechanism worked well. What we hadn't factored in was anti money laundering, and Apple took one look at the unlimited transfer capability and flagged the product as a possible laundering vehicle.

We went back and retrofitted several layers. More stringent KYC checks, so we knew who both parties were. Stronger transfer security. Hard limits on how many transfers were allowed between any two parties. All of it was buildable. All of it cost more after the fact than it would have cost in the original design.

Health products touch money more often than founders realise. Subscription billing is straightforward enough, and arranging or selling insurance is an FCA regulated activity. Taking payments on behalf of clinicians makes you a payment intermediary. Handling health savings or reimbursement means AML duties, sanctions screening and strong customer authentication.

If money moves between two users rather than from user to you, assume anti money laundering rules apply and design the limits in from the start. Retrofitting them means rewriting your core flow.

Connected Devices, Sensors and the Hardware Contract

We worked on a baby monitor where the client was building the cameras and sensors alongside the companion app. The original plan was off the shelf Bluetooth protocols. The integration turned out to be far more involved than anyone anticipated, so we formally defined the contract between the software and the hardware. It set out exactly how the app turned things on and off, pulled sensor data and got camera feeds, all in a format the app could actually consume.

We also built a prototype that ran a small web server from the device itself. That let us connect to it, see the internals, check that settings applied in the app were really landing on the hardware, and control the device at the same time. It meant software testing didn't sit waiting on finished hardware.

Hardware brings its own rule book

A connected health device pulls in UKCA marking, radio equipment rules, and the Product Security and Telecommunications Infrastructure Act, which bans default passwords and requires a published support period for security updates. The EU adds the Cyber Resilience Act on top. The support period matters more than people think, because it's a public promise about how long you'll keep patching, and it has to survive your next funding round.

App Store Review as a Regulator in Its Own Right

Apple rejected our currency exchange product over money laundering risk before any financial regulator had said a word. That's worth sitting up for. App review is the gate everything passes through, and the reviewers apply their own reading of health and financial risk on their own timetable.

Apple's guidelines restrict dosage calculators unless they come from a recognised institution, require health research apps to carry ethics approval, and ask for documented credentials where clinical advice is given. Google Play runs a health apps declaration with similar ground. Both can pull a live product. Neither offers a hearing.

The quality picture backs up why they're cautious. In an assessment published through PMC, only 33% of 46 health apps met the security and privacy transparency step of the evaluation framework, after those apps had already cleared the first screening step.

Treat review notes as a design input

Write your app store answers while you're designing, not while you're waiting. The questions about data collection, medical claims and user safety are the same questions a DPIA asks. If your privacy label and your privacy policy disagree, that's a rejection and a data protection problem at once, discovered by whichever one finds you first.

Accessibility and Equality Act Duties

The Equality Act 2010 requires reasonable adjustments for disabled users, and a health app is exactly the kind of service where failing to do that excludes the people who need it most. The practical standard is WCAG 2.2 AA. Selling to the NHS or any public body makes it contractual. Selling into the EU brings the European Accessibility Act, which has applied to consumer digital services since June 2025.

On that healthcare and health insurance product for African markets, we stripped medical jargon out of the language and simplified it hard, because people often arrive emotionally heightened and reading is the first thing that goes.

Plain language is accessibility work, and in health it's also clinical safety work.

The same project shaped the visual side. We went for a dark UI with green undertones, green because it carries associations of calm, safety and medical support, while deliberately staying away from a clinical look. Contrast ratios and text scaling were part of that decision rather than a check we ran afterwards.

Test with the screen reader on and the text size at maximum before you ship, not at the end of a QA cycle. Both will find layout problems that no design review catches.

Algorithmic Transparency and AI-Driven Recommendations

Our first question on anything with a model behind it is whether we're being transparent with the user. Are we saying what we're doing, saying that this is AI, and giving people the data behind why we've shown that particular result? Recommend a fitness routine and the product should show the reasoning. This routine, because of these three things you told us.

The part teams underestimate is what that costs on the back end. Showing your working later means recording it at the time, so you need to store which model version ran, what inputs it saw and what it returned, against a timestamp. Without that log you can't reproduce a recommendation somebody complains about six months on, and you can't answer a regulator who asks how a decision was reached.

UK GDPR gives people rights around decisions made solely by automated means. The EU AI Act layers duties on health related systems, and the MHRA treats adaptive clinical software as a device with change control obligations. Health apps regulation is moving fastest here, so build the audit trail before anyone requires it.

A reasoning line beats a disclaimer

One sentence explaining why a suggestion appeared does more for trust, and for your evidence file, than a paragraph of small print nobody reads.

Designing Compliance So It Feels Like Care

The rules keep pointing at the same behaviours. Say what you're doing. Ask rather than take. Explain the reasoning. Hold less. Those read as legal duties on paper and as decency inside a product, which is why the compliant version is usually the version people like more.

We check every feature against one question, which is whether it helps the user or works against their stated intentions. Push notification frequency and in app messaging sit under the same test, so everything that fires has to be timely and relevant to that person. Pressure tactics fail that test, and they also fail consumer protection rules on unfair practices.

Reward design is where this gets concrete. Think about somebody using a fitness app to lose weight while feeling anxious about it. Gamification built around a stone lost, then two stone lost, ties good feeling to an outcome they can't control day to day. We celebrate the first workout instead, and coming back three times in a week, so the reward sits on the behaviour.

Pre-framing works the same way. On the genetics wellness app, writing variation into the copy in advance stopped people reading a normal result as a fault in the product. Fewer alarmed users, fewer support tickets, fewer complaints that turn into incidents.

Building Compliance Into the Project Plan

On a dating app built around verified profiles and keeping bots out, the client wanted to skip discovery for the messaging component and put everything into onboarding. What got built was a generic messaging feature that allowed automated and fake messages, which contradicted the entire premise of the product. The messaging section had to be rewritten, at roughly £15,000 of extra budget and two months of extra work.

Nothing about that was a technical failure. The gap was that one part of the product was designed against a rule and another part wasn't. Health projects fail the same way, with a consented onboarding flow feeding an analytics pipeline nobody assessed.

Where health app regulations UK projects slip

  1. Decide medical device status before you write the first user story
  2. Run the DPIA while the data model is still changeable
  3. Appoint a clinical safety officer if the NHS is in your plans
  4. Write the retention schedule before you build the delete button
  5. Draft your app store answers during design, not at submission

Budget for it honestly. GoodFirms puts the uplift at 30 to 50% for regulated sectors like healthcare, and that lines up with what we see. Planned in, it's a cost. Discovered late, it's a rewrite.

Conclusion

Every rule in this article comes back to a handful of decisions made early. What the product claims. What data it holds and for how long. Who it's answerable for. Whether it explains itself. Make those decisions in week one and the regulation mostly describes the product you were already building. Leave them and you get the dating app outcome, where a sound piece of work gets torn out because it was designed against the wrong assumption.

The projects we've described here all landed somewhere a founder didn't expect. Apple reading a currency product as a laundering risk. An anonymous messaging app needing login at the very front. A baby monitor needing a written contract between software and hardware before either could be finished. None of those were visible from a feature list, and all of them were visible from a conversation at the start.

Working with the process matters more than any single rule. Teams that resist discovery, iteration and incremental release in favour of launching everything at once are the ones most at risk of never launching at all.

If you're building something in health and want a clear read on what applies to you before the architecture sets, let's talk about your health app compliance plan.

Frequently Asked Questions

How do I know whether my health app counts as a medical device?

It comes down to intended purpose rather than the technology you use. If your product is meant to diagnose, prevent, monitor, predict, treat or relieve a condition, the MHRA treats the software as a medical device under the UK Medical Devices Regulations 2002. A simple step counter sits outside that, but a tool that reads step data and tells someone their heart is at risk does not.

What does being classed as a medical device actually require of me?

You would need to register with the MHRA, hold technical documentation, run post market surveillance and carry UKCA marking. These are ongoing obligations rather than one off tasks, so they affect your team's capacity long after launch.

Can my marketing copy really change my regulatory status?

Yes. The wording on a marketing page or a results screen can move you across the line on its own, because regulators read what you claim the product does. Copy that describes patterns keeps you in record keeping territory, while copy that hands out verdicts reads as diagnosis.

Why is health data treated differently under UK GDPR?

Health data is special category data, so you need an Article 9 condition on top of your normal lawful basis, plus a Data Protection Impact Assessment carried out before you build. The ICO also has much less tolerance for mistakes in this area, so the stakes of getting it wrong are higher.

Which parts of my product are in scope for data protection?

Anywhere the data lands is in scope, not just your main database. That includes logging, analytics, crash reporting, third party SDKs and your customer support tooling, all of which need to be assessed rather than assumed to be safe.

At what point in a project should I be thinking about regulation?

Week one, not week forty. Regulation decides what you are allowed to build, so leaving it until the end means discovering that core features have to be reworked or dropped. The first real decisions on a health product are usually about what you can hold, how long you keep it and what the product is allowed to say.

Are medical device rules and data protection the only things I need to worry about?

No. Health products sit on top of several rule sets at once, including clinical safety standards, consumer and accessibility duties and app store policy. If money moves through the product, financial rules apply on top of all of that.

Can app stores block my product even if I meet the legal requirements?

Yes, app store policy is its own layer and reviewers apply their own reading of what your product does. A peer to peer currency exchange we worked on was read by Apple as a money laundering risk, which shows that store review can stall a launch independently of regulators.