01
Executive thesis
Patient reviews are treated as a quality signal or dismissed as noise, and both readings are wrong. Reviews are a form of operational evidence about the parts of the experience a patient is competent to assess: how long they waited, whether anyone explained what was happening, whether the reception staff were civil, whether the building was clean, whether the appointment they booked was the appointment they got.
What patients cannot assess is clinical quality. A patient has no way to judge whether a diagnosis was correct or a technique appropriate. A surgeon with excellent outcomes and a curt manner will accumulate worse reviews than a warm clinician with mediocre outcomes. Treating average rating as a proxy for care quality inverts the incentive.
The decision this article supports is what to do with review text. Classified by theme and routed to the department that owns each theme, reviews become an early warning system for access and process failures. Read as a rating, they generate defensiveness and no action.
02
Numbers that frame the issue
| Figure | What it tells us | Type | Source, date and geography |
|---|---|---|---|
| 79% of surveyed adults read reviews before choosing a provider | Reviews function as a selection input, whatever they measure | External evidence, international | Tebra 2025 survey of 3,964 US adults, 2025, United States |
| 25% of respondents read negative reviews to evaluate how organizations handle feedback, with 24 to 48 hours presented as an ideal response window | A share of readers are assessing the response, not the complaint | External evidence, international | Press Ganey, 2025, international and US-weighted |
| The Ministry of Health states its regional patient-experience partner conducts more than 2.5 million surveys annually | Structured patient-experience measurement exists at scale in the region, separately from public reviews | Programme context | Saudi Ministry of Health patient experience programme, current page, Saudi Arabia and region |
The first two are international commercial surveys with US or US-weighted respondents and are used for direction only. The Ministry of Health figure is programme context describing survey volume; it is not a public benchmark and should not be used in marketing claims.
03
What reviews can and cannot prove
Three limits are worth stating plainly, because they determine how the evidence should be used.
Reviews are not a sample. People write reviews after unusually good or unusually bad experiences. The satisfied majority in the middle mostly do not. This produces a bimodal distribution that does not represent the patient population, which is why review averages and structured survey scores frequently disagree. Both can be accurate about different things.
Reviews do not measure clinical outcomes. A patient assesses what they can perceive. Clinical judgement, technique selection and diagnostic accuracy are not among them. A review may describe an outcome the patient experienced, which is useful, but it does not evidence whether the care was appropriate.
Reviews are shaped by expectation, not only by service. The same forty-minute wait produces different reviews depending on whether the patient was told to expect it. A large share of negative reviews describe an expectation failure rather than a service failure, and the fix is communication rather than capacity.
What reviews do prove is that something specific happened to a specific person on a specific day, in a domain they were positioned to observe. That is genuinely valuable, and it arrives faster than any internal reporting cycle.
04
The Patient Review Theme Taxonomy
Reviews resolve into seven themes. Each has a different owner and a different fix, which is the operational point of classifying them.
| Theme | What it covers | Owner | What a spike indicates |
|---|---|---|---|
| Access and waiting | Appointment availability, waiting time in clinic, delays, cancellations | Operations and scheduling | Capacity or scheduling design problem |
| Communication | Explanation of condition, treatment and next steps, listening | Clinical leadership | Consultation time or communication training |
| Administration | Booking, registration, records, referrals, follow-up | Administration | Process or system failure |
| Billing and insurance | Cost clarity, claim handling, unexpected charges | Finance | Information failure before the visit, usually |
| Staff conduct | Reception, nursing, non-clinical interaction | Department management | Local culture or staffing pressure |
| Environment | Cleanliness, parking, facilities, signage | Facilities | Physical or wayfinding issue |
| Clinical experience | Outcome as experienced, perceived competence, trust | Clinical leadership | Genuine clinical review may be warranted |
Only the last theme touches clinical quality, and even there the evidence is perceptual. The first six are all operational, and all are within the organization’s control without any change to clinical practice.
Classify every review by primary theme, with a secondary theme where present. Report theme prevalence by site, by department and over time. That report is more useful than any average rating, and it is actionable by named owners.
05
Access, communication and the themes that dominate
Two patterns are worth anticipating when you run this classification, stated here as expectations to be tested rather than as findings.
The first is that access and waiting themes are likely to dominate negative feedback, because waiting is the most perceptible part of the experience and the most emotionally loaded. A patient in discomfort experiences twenty minutes differently from a patient in a queue at a bank.
The second is that communication themes are likely to dominate the difference between good and excellent ratings. Once access is adequate, what separates a four from a five is usually whether the patient felt heard and understood what was happening.
Both are hypotheses. They should be tested against your own classified review set rather than assumed, and they may differ by specialty and by service model.
06
Arabic and English reviews are not the same dataset
Analysing only the English reviews produces a partial and usually flattering picture, and it is the most common methodological error in Gulf healthcare review analysis.
Three differences to plan for. Arabic and English reviewers may be different populations with different expectations, so theme prevalence can genuinely differ rather than merely appearing to. Arabic reviews frequently mix dialect with Modern Standard Arabic, which defeats keyword-based classification and requires either manual reading or a classifier built for the dialect. And automated sentiment tools trained predominantly on English tend to perform poorly on Arabic, misreading politeness conventions and negation in ways that skew results in the direction of neutrality.
The practical consequence is that classification of Arabic reviews should be done by Arabic speakers, at least until a classifier has been validated against a manually coded sample. Reporting a single blended theme distribution across both languages, without checking whether they agree, hides the group that is usually the larger share of the patient population.
07
Positive and negative review language
Positive and negative reviews are not mirror images and should not be analysed together.
Negative reviews are specific. They name a time, a person, a failure and a consequence. This makes them operationally rich and the reason they are worth reading closely despite being unpleasant.
Positive reviews are usually general. “Excellent doctor, highly recommend” carries almost no information. This creates a problem for prospective patients, who need specificity to recognise their own situation, and it is why a high rating built on general praise can still leave decision uncertainty. The gap is discussed in Reputation Debt: The Gap Between Care Quality and Public Evidence.
The implication for review solicitation: asking for a rating produces general praise. Asking a patient what they came in for and how it went produces specific narrative, which serves both the operational analysis and the prospective patient. The request must remain genuine and non-incentivised, and must not be filtered by expected sentiment.
08
Differences by specialty and service model
Theme distribution should differ predictably, and comparing sites or specialties without adjusting for this produces false conclusions.
Episodic and procedural services such as dentistry and dermatology generate reviews weighted toward cost clarity, comfort and administration, because the clinical event is short and the surrounding process is most of the experience.
Chronic and long-course services such as rehabilitation and endocrinology generate reviews weighted toward communication, continuity and staff conduct, because the relationship extends across many contacts.
Urgent and unscheduled services generate reviews weighted heavily toward waiting and triage, largely irrespective of the care delivered.
High-consideration surgical services generate fewer reviews with far greater length and detail, because the decision mattered more to the person writing.
Comparing an emergency department’s rating with a dermatology clinic’s rating is not a comparison. Benchmark within service model, and be sceptical of any cross-specialty league table.
09
How teams should use review evidence
A working method, in five steps.
- Classify every review by primary and secondary theme, in Arabic and English, using the taxonomy.
- Route each theme to the department that owns it, with the review text attached rather than an aggregate score.
- Track prevalence by theme, site and department over time, and treat a rising theme as a signal before it appears in structured surveys.
- Separate expectation failures from service failures within each theme, since the two need different responses.
- Respond publicly within a defined window, without disclosing any patient information, following the pattern in The Privacy-Safe Healthcare Review Response Playbook.
Two cautions. Do not set targets on review volume or rating for individual clinicians, since it creates pressure to solicit selectively and produces exactly the distortion that makes reviews unreliable. And do not route clinical experience themes through marketing; where a review suggests a genuine clinical concern, it belongs in the clinical governance process, not in a reputation workflow.
Theme prevalence belongs on the operational dashboard rather than the marketing one, a distinction developed in The Healthcare Growth Dashboard: 12 Metrics That Change Decisions.
10
Methodology, bias and limitations
What this article is. A taxonomy for classifying patient review content, with a method for routing it operationally, and a statement of what review evidence can and cannot support.
What it is not. This article contains no review analysis findings. DEMA has scoped a Saudi healthcare review language study covering theme prevalence in Arabic and English. It is not complete. No theme distribution, no percentage and no benchmark from that study appears here, and the two patterns described above are labelled as hypotheses precisely because they have not been measured.
Known biases in review data. Selection bias toward extreme experiences. Recency bias, since recent reviews dominate perception regardless of volume. Language bias, since Arabic and English reviewers may raise different themes and analysing only one language produces a partial picture. Platform bias, since different platforms attract different reviewer populations.
Evidence limitations. The Tebra and Press Ganey findings are international commercial surveys with US or US-weighted respondents. The Ministry of Health reference describes programme scale, not results, and is not a benchmark.
Where review is required. Any handling of review text that could identify a patient requires review under the Saudi Personal Data Protection Law. Reviews suggesting clinical concerns must be routed through clinical governance rather than treated as reputation management. This article is not legal advice.
Sources.
11
Reading your reviews as a rating?
Classified by theme and routed to the right owner, the same reviews become the fastest operational feedback you have. A Healthcare Growth Diagnosis classifies your review set in Arabic and English, shows which themes dominate by site and department, and separates what is a service problem from what is an expectation problem.
Request a Healthcare Growth Diagnosis