This post is for the product owner, UX lead or marketing manager at a hospital whose "Find a doctor" page has a long row of filters and no evidence about which ones patients use. Filters are cheap to add and expensive to keep. Each one adds code, test cases, data to maintain, URL complexity and visual noise. Hospital teams tend to add them in response to requests: a director wants a "Gender" filter, the insurance desk wants "Insurance", a campaign needs "Available today". Nobody removes them.
It answers two questions. Which filters tend to earn their place on a doctor search, and how do you find out for your own site? The first is a reasoned shortlist from how patients decide; the second is an analytics method you can run in a few weeks. We do not report usage numbers from client sites here. The right answer depends on your city, specialties and patient mix, and the method is what transfers.
For the search side of directory design, including which filter combinations should be indexable, see our companion material on medical SEO keywords patients search; this article covers the interface and the measurement.
How patients actually choose a doctor
Before choosing filters, consider the decision. A patient looking for a specialist usually has some of these constraints:
- A need: a symptom, a condition or a procedure, not a specialty name. Patients often do not know which specialty to pick.
- A place: a city, a locality or "near me", driven by travel time.
- A time: "this week" or "tomorrow", especially for new problems.
- A language: comfort speaking about health in a particular language. In multilingual cities this is a hard constraint. The Chennai piece on winning local search when patients search in two languages describes the same pattern from the search side.
- Cost or coverage: whether the hospital works with a particular insurer or scheme, and what a consultation costs.
- Trust: qualifications, years of practice, and sometimes gender, particularly for gynaecology, urology and some therapy services.
A filter earns a place when it matches one of these constraints and is answerable from data you can keep accurate. That second condition kills many filters. "Available today" is worthless if the data updates weekly. "Accepts insurance" is dangerous if the list is stale, because a patient who relies on it may face a bill. Never publish a coverage or availability filter you cannot maintain, and have the billing desk confirm the data owner.
A first-pass shortlist
Here is a reasoned starting point, not a measurement.
| Filter | Why it may earn use | Condition for keeping it |
|---|---|---|
| Specialty | Primary way to enter | Needs patient-friendly labels and synonyms (for example "heart" mapped to cardiology) |
| Location or branch | Travel is a hard constraint | Needs accurate placements per branch |
| Availability | Urgency drives choice | Only if updated at least daily; otherwise show the weekly timetable |
| Language | Hard constraint in many cities | Needs verified data, reviewed by the doctor |
| Gender | Hard constraint for some specialties | Show where relevant, hide elsewhere |
| Insurance or scheme | Cost is a major concern | Only with a named owner and review date |
| Consultation mode | Video or in person | Only if both are actually offered |
| Experience, rating, price | Often requested internally | Test before keeping; may add noise |
Two cautions. Rating filters on doctors need care: public star ratings for individual doctors raise reputation and compliance questions, and the provider's compliance owner should review the approach. For doctors in India, the NMC advertising rules limit what can be claimed about doctors; our NMC advertising guidelines explainer covers the current position. Sorting doctors by "best" or "most experienced" invites exactly the kind of comparison that needs review.
Second, filters that look like personalisation (such as "recommended for you") should not be built until you have data and a clear explanation for the patient.
Measure before you prune
Instrument the filter panel so that every filter interaction becomes an event. With Google Analytics 4, a minimal event design:
| Event | Fires when | Parameters |
|---|---|---|
doctor_search_view | Search page loads | entry_source (nav, campaign, organic landing) |
doctor_filter_apply | A filter value is selected | filter_name, filter_value_group (not free text), filters_active_count |
doctor_filter_clear | A filter is removed or reset | filter_name |
doctor_results_view | Results render | result_count_bucket (0, 1 to 5, 6 to 20, more than 20) |
doctor_card_click | A doctor card is opened | result_position_bucket, filters_active_count |
doctor_book_start | Booking is started from a result or profile | source (card or profile) |
doctor_book_complete | Booking is submitted | source |
Parameter names are yours to choose; the structure is the point. Event parameters must be registered as custom dimensions before they appear in reports, as Google describes in its help page on custom dimensions and metrics: the dimension is created in the Admin area under custom definitions with the scope set to Event and the exact event parameter name, and it can take 24 to 48 hours to appear after data is sent. Register the parameters on day one, not after the test period, because data sent before registration is not available under that dimension.
Rules for the data:
- No personal data in events. Send filter names and grouped values. Never send typed search text that could contain names, phone numbers or conditions. If you log free-text queries for search improvement, do that in a separate system with its own review.
- Bucket the values. Counts and positions in buckets keep cardinality low and avoid hitting dimension limits.
- Respect consent. Tracking should follow the consent mode you have configured, and the provider's compliance owner should approve the event design.
- Use server logs as a cross-check. If filters call an API with parameters, API logs show usage without relying on the browser tag.
Our analytics tracking build covers how we set up this layer.




