97,000 Patient Reviews Suggest Digital Health Needs More Than an App Score

A digital health platform can appear to have one customer journey: one app, one login, one brand and, perhaps, one experience score. Patients do not encounter it that way.

They book appointments, check records, request prescriptions, seek advice and try to understand what happens next. Each task comes with a different expectation of the service, and a different consequence when it fails.

That is the useful finding in a study published this month in *PLOS One*, which analysed 97,034 Google Play reviews of Sehhaty, Saudi Arabia’s national digital health platform, gathered between 2019 and 2026. The setting is specific, and its findings should not simply be transplanted into the NHS or another health system. Yet its research design offers a practical challenge to providers, health-tech companies and patient-experience teams everywhere: stop treating app feedback as a single verdict on the product.

The study identified nine dimensions of patient experience, spanning reliability, access, navigation, service efficiency, support, account access, continuity, information accuracy and patient control. More importantly, it found that those issues clustered differently according to the service patients were trying to use.

The service changes the meaning of a poor experience

For appointment booking, access was the overwhelming concern. It appeared in 97.5% of reviews that explicitly referred to that service. That should not surprise anyone who has watched a patient attempt to find a suitable slot: the relevant question is rarely whether the interface is attractive. It is whether the service can get them to care at a moment when they need it.

Records and administrative documents produced another pattern. Feedback on medical reports, sick leave and personal health records focused much more heavily on information accuracy and transparency. Here, the patient is judging whether the information is correct, complete and available in time to act on it. A smooth interface cannot compensate for a missing result, unclear record or document that does not reflect reality.

Prescription and broader health services, meanwhile, were more likely to expose concerns about integration and continuity. Patients were judging whether the platform connected the relevant parts of care: the prescription, the record, the provider and the next action. Instant consultations had a more balanced profile, combining access, communication and the speed of service delivery.

This distinction matters commercially as well as clinically. A supplier can show high adoption or a healthy average satisfaction score while concealing a weak point in a service that carries much greater patient effort and reputational risk. A booking journey can be technically stable but still fail because appointments are unavailable. A record portal can have excellent uptime but still erode confidence if people cannot understand or trust what it shows.

The right unit of analysis is therefore not simply the app. It is the service task within the care pathway.

Technical fixes do not end the experience problem

The longitudinal element of the research is equally instructive. As Sehhaty matured, technical stability and account-access complaints declined. That is progress. But the centre of gravity moved towards access, integration and continuity of care.

This is a familiar pattern in digital services. The early stage is dominated by visible defects: crashes, failed logins and broken processes. Once these are reduced, patients notice the more consequential limitation: whether the digital channel can genuinely help them obtain care, understand their information or move between services without becoming the courier of their own medical history.

The OECD’s recent work on people-centred digital health systems reaches a related conclusion. Patients who have to repeat information that should already be available report markedly poorer trust, experienced quality and person-centred care. Digital infrastructure may exist behind the scenes, but the experience is still fragmented when the person receiving care has to join the dots.

For teams commissioning or designing technology, this changes the research question. “Are users satisfied with the platform?” is too blunt to guide investment. Better questions include:

• Can a patient secure the right appointment without switching channel or starting again?

• Can they see information that is accurate enough to use confidently?

• Does a digital consultation result in a clear next step, rather than another unresolved interaction?

• When care crosses organisations or teams, does the patient experience a handover or a dead end?

Those questions connect product performance to service design. They also make it easier to identify which operational teams need to be involved in solving the problem.

Patient feedback is an early-warning system, not a complete answer

App-store reviews are valuable because they are immediate, voluminous and often tied to a precise moment of difficulty. They can reveal recurring language, newly introduced faults and unmet expectations faster than an annual survey. Used well, they can help a health organisation decide where to investigate.

They are not, however, a representative sample of all patients. People with strong positive or negative experiences may be more likely to post. Reviews rarely explain the full clinical context, and they cannot establish whether a reported problem reflects the interface, capacity constraints, a local workflow or an individual misunderstanding.

The answer is not to dismiss this evidence. It is to connect it to other forms of insight.

A mature listening approach would combine review analysis with contact-centre reasons, drop-off data, appointment availability, accessibility testing, patient interviews and observation of real tasks. If reviews complain that a record is unclear, for example, teams should test what patients believe the information means, whether staff can explain it consistently and what action patients take afterwards. If booking complaints rise, the investigation should distinguish an interface problem from a shortage of suitable appointments.

That combination is particularly important as artificial intelligence enters more patient-facing journeys. New Pew Research Centre findings, published this week in the United States, underline that people make distinctions between tasks. Respondents were relatively open to AI-assisted translation, but far less comfortable with AI making diagnoses or insurance decisions without human review. The transferable lesson is not that every market will respond identically. It is that patient acceptance is shaped by the purpose of the service and the stakes attached to it.

Health-tech teams should apply the same discipline to their evidence. Do not ask whether patients trust “AI”, the portal or the app in the abstract. Ask what they are being asked to do, what they need to know, what can go wrong and whether a person can obtain meaningful help.

From feedback volume to better decisions

The most promising feature of the new study is not that it found a way to mine a large body of reviews. It is that it turned those reviews into service-specific priorities.

That is the standard health organisations should adopt. Patient feedback should be organised around moments of need: booking, understanding, coordinating, deciding and following up. It should then be read alongside operational evidence, not used as a decorative scorecard for a digital programme.

A platform-wide rating may still have a place as a broad signal. But it cannot tell a service leader whether their most urgent problem is appointment capacity, an unclear record, a lost handover or a consultation that leaves the patient unsure what happens next.

Patients are not reviewing software in a vacuum. They are reporting whether a route into care worked. The research function is to preserve that context and turn it into a better decision.