Data Sources

A directory is only as good as the register underneath it. Zavis treats a healthcare facility as real when a government health regulator has licensed it, and treats everything else — ratings, photographs, opening hours, neighbourhood boundaries — as enrichment layered on top of that spine. The distinction matters because the two kinds of data fail differently. A regulator register goes stale. A commercial data feed goes wrong. We would rather publish a facility whose telephone number is six months old than one whose existence we cannot prove.

This page lists every input, what we take from each one, what we leave behind, and the licence we rely on. It is written to be checkable. Every source in the table below is public, and each link goes to the source itself rather than to a summary of it.

UAE facility records are anchored to three regulators, split by emirate. The Dubai Health Authority licenses facilities in Dubai and publishes them through Sheryan. The Department of Health licenses facilities in Abu Dhabi, Al Ain and Al Dhafra. The Ministry of Health and Prevention licenses facilities in the five Northern Emirates: Sharjah, Ajman, Umm Al Quwain, Ras Al Khaimah and Fujairah. Between them, those three registers define the 12,381 UAE facilities in the index.

From each register we take the fields that identify and locate a facility: legal name, licence number, licence type, facility category, emirate, address and published telephone number. We take the Arabic facility name where the authority publishes one. Abu Dhabi publishes Arabic facility names in its open-data portal. The Dubai professional records do not carry Arabic names, which is why Arabic practitioner names are absent from our doctor profiles. We do not transliterate them ourselves.

We leave behind everything a register publishes for internal administration: inspection notes, internal reference codes, and contact details for named individuals that the facility has not published as its public contact. A facility enters the directory only when it appears in one of these three registers. We do not buy facility lists from data brokers, and we do not create a facility record because a search engine or a map has one.

The professionals index covers 31,899 licensed practitioners: physicians, dentists, nurses, pharmacists and allied health staff. Each profile carries the regulator’s unique identifier and the licence type, so a visitor can open the source register and check our entry against the original in under a minute. That is the standard we hold the index to. A field that cannot be checked against the register does not belong on the profile.

Practitioners are matched to a primary facility on a conservative name match. Where the match is uncertain we publish the practitioner without a facility link, because a wrong employer is worse than a missing one. We do not publish a practitioner photograph without explicit consent; profiles without consent use a generated initials avatar. Insurance acceptance, languages spoken and availability appear only when the source dataset carries them.

The index now covers six markets: the UAE with 12,381 facilities, Türkiye with 10,579, Saudi Arabia with 2,360, Bahrain with 820, Kuwait with 725 and Qatar with 657. That is 27,522 facilities in total. Source quality outside the UAE is uneven, and we would rather state that here than imply a uniform standard we do not have.

Our order of preference is the same in every market: a national regulator register first, a national open-data portal second, open geographic datasets last. The National Health Regulatory Authority in Bahrain publishes its licensed hospital and pharmacy lists as downloadable documents and spreadsheets, with no queryable register behind them. Qatari facilities are compiled from the Qatar Open Data Portal alongside the published Ministry of Public Health facility list. Saudi Arabia and Kuwait have no public licensed-facility download, so those markets are built from OpenStreetMap and equivalent open datasets, then de-duplicated by name and coordinate proximity and cleaned by hand. Türkiye is the newest market in the index and its source documentation is still being written; it will be published on this page in the same detail as the rest.

Where a market’s source does not publish a licence number, the facility page does not show one. A missing licence number on a Zavis page means we do not have it, never that the facility is unlicensed.

Area pages need real boundaries. Dubai’s come from Dubai Pulse, the emirate’s open-data platform, which publishes community polygons with both English and Arabic names as GeoJSON. Abu Dhabi districts and sectors come from the Abu Dhabi Open Data platform. For the Northern Emirates, where no equivalent open dataset exists, we fall back to OpenStreetMap administrative boundaries.

Holding geography separately from the facility records is what lets area pages exist without a commercial mapping dependency. Dubai Pulse and OpenStreetMap both require attribution on reuse. The OpenStreetMap contribution is licensed under the Open Database Licence, and the attribution appears on the pages built from it.

Google Places supplies enrichment and nothing structural. Where a licensed facility has a Google Business Profile, we read the rating average, the review count, opening hours, photographs and the Maps URI, and attach them to the record the regulator already established. A rating shown on a Zavis page is a Google rating and is labelled as one. We do not present it as a Zavis score, and we do not fold it into a rating of our own.

We do not republish Google review text. Aggregate ratings are emitted in structured data only once a facility has at least three reviews, because a five-star average drawn from a single review produces a misleading search result. No facility exists on Zavis because Google has it. A Google Business Profile with no matching regulator licence does not become a listing.

Facility records are refreshed against the source registers on a recurring schedule, and every row carries a google_fetched_at timestamp recording when its Google enrichment was last read. A facility that disappears from its source register is flagged for review and drops out of the indexable sitemap on the next generation pass, which runs hourly. Directory pages are statically rendered and revalidated every six hours, so a change written to the database reaches readers well before a search engine re-crawls the page.

The Intelligence editorial layer runs on its own cadence. The pipeline publishes twice daily, and each of the 332 articles carries a dateline. An article’s modified date moves only when a person has re-read the piece.

Not every licensed facility earns an indexable page, and the rule is mechanical rather than editorial. A facility page enters the sitemap and stays indexable only when it has at least one form of unique content — a description longer than eighty characters, or at least one photograph — and at least two of five completeness signals: a Google rating above zero, a telephone number, a website, that same description length, and a non-empty opening-hours record.

Facilities that fail the gate remain in the directory and stay findable through search and browse. They are marked noindex and held out of the sitemap until the record improves, which usually happens when the operator claims the listing and completes it. A single shared function decides both the sitemap entry and the page-level robots directive, so the two signals cannot drift apart and send Google contradictory instructions.

The same discipline governs filter pages. The facet policy engine defines which combinations of city, specialty, insurer, area and language may be crawled and indexed. Anything outside that allowlist is either never generated or generated as a browsing-only page carrying noindex,follow. The reasoning behind the allowlist is set out on our methodology page.

Editorial coverage draws on primary material: regulator circulars and dashboards, federal cabinet decisions, exchange filings, peer-reviewed literature and named on-record interviews. Clinical articles cite their sources in a numbered list at the foot of the page, with DOIs resolved through Crossref where the source has one. Authors and external reviewers who hold an ORCID iD have it shown on their profile, so a reader can check a reviewer’s publication record independently of anything we say about them. The full standard is published at /editorial-policy.

• Buy facility lists from third-party data brokers or lead vendors.
• Create a facility record from a map or search listing that has no regulator licence behind it.
• Generate, synthesise or paraphrase patient reviews. No Zavis-originated review is live on the site today; the design and launch criteria are published at /verified-reviews.
• Present a Google rating as a Zavis rating.
• Scrape commercial property portals, classified sites or competing aggregators for healthcare listings.
• Publish a practitioner photograph without that practitioner’s consent.
• Sell placement in the directory, or let a commercial relationship change where a facility appears on a hub page.