# ADHD Private: standards, registration and market dataset

Version 24 September 2026. Compiled 18–21 September 2026 by ADHD Private (adhdprivate.co.uk), a UK directory of private ADHD clinics operated by Sellframe Ltd (SC472357). Published in full, free to download, at adhdprivate.co.uk/data/uk-private-adhd-standards-2026-09/ alongside the reports it supports: adhdprivate.co.uk/research/uk-private-adhd-clinic-regulation-2026 and adhdprivate.co.uk/research/uk-private-adhd-child-assessments-2026.

This is a companion to the clinic-level dataset of 21 September 2026 (adhdprivate.co.uk/data/uk-private-adhd-clinics-2026-09/), not a replacement for it. `clinic_id` is the same key in both, and not one byte of that package has been changed to make room for this one. Where this package corrects something we published earlier, the correction is set out under "Corrections to earlier releases" below and recorded in `corrections.csv`; the earlier files stay exactly as they were published.

Nine CSV files, UTF-8, comma-separated, one header row. `clinic_id` joins all but `market_adhd_companies.csv`, which is about companies rather than about clinics we list.

| File | Rows | One row per | Answers |
|---|---|---|---|
| `clinic_checks.csv` | 342 | clinic listed on ADHD Private | which regulator, if any, holds a current registration for this clinic |
| `right_to_choose.csv` | 342 | clinic | what the clinic's own website says about NHS Right to Choose |
| `site_claims.csv` | 342 | clinic | what the clinic's own website claims about its credentials and its offer |
| `children_pathway.csv` | 342 | clinic | what the clinic's own website says about how it assesses children |
| `appointment_length.csv` | 112 | clinic that states an appointment length | how long the clinic says the assessment appointment is |
| `wales_hiw_register.csv` | 5 | clinic and HIW register entry pair | registration in Wales |
| `northern_ireland_rqia_register.csv` | 2 | clinic and RQIA register entry pair | registration in Northern Ireland |
| `market_adhd_companies.csv` | 470 | UK company whose name has contained "ADHD" | how many ADHD-named companies exist, and since when |
| `corrections.csv` | 48 | correction we made | every value we changed after extraction, with what it was and why |

Every figure we took from someone else carries a link or an identifier a reader can check without an account: a Companies House page, an HIW register page, an RQIA service ID, or the page of the clinic's own website a statement was read on. Our own judgements are marked as ours in the dictionary below.

To spot-check any coded value taken from a clinic's site, open its `<field>_source_url` and search the page for the text in `<field>_quote`. A coded value with no quote beside it was not published: the build blanks a value whose quote it cannot find in the stored page.

Two spreadsheet notes, both about Excel guessing. The files are UTF-8 with no byte-order mark, which is what pandas, R and Numbers expect; Excel on Windows assumes the local codepage on a plain double-click and renders the curly apostrophes as mojibake. And many company numbers in `market_adhd_companies.csv` begin with a zero or with letters, which Excel strips or reformats. Opening the files through Excel's text import, choosing UTF-8 and setting the columns to text, avoids both.

## Licence and citation

**Our contribution** — the selection, the matching, the coding of clinics' words into categories, the derived columns and the compilation — is released under the Creative Commons Attribution 4.0 International Licence (CC BY 4.0). You may publish, adapt and redistribute it, including commercially, with attribution:

> ADHD Private, standards, registration and market dataset, 24 September 2026. adhdprivate.co.uk

Four parts of the package are **not ours to license**, and the CC BY grant does not extend to them.

- **Companies House and Care Quality Commission fields**, which appear in `market_adhd_companies.csv` and in the CQC columns of `clinic_checks.csv`. Contains public sector information licensed under the [Open Government Licence v3.0](https://www.nationalarchives.gov.uk/doc/open-government-licence/version/3/).
- **`wales_hiw_register.csv`.** Healthcare Inspectorate Wales is part of the Welsh Government and its copyright statement, [©Crown copyright 2026](https://www.hiw.org.uk/ccrown-copyright-2026), read on 21 September 2026, says: "You may use and re-use the information featured in this website (not including logos) free of charge in any format or medium, under the terms of the Open Government License." That page's address carries the year, so it moves annually; if the link 404s, search HIW's site for its current copyright page rather than assuming the permission lapsed. The register wording in these rows is reproduced on that basis, and every row carries `register_url` pointing at the authoritative entry.
- **`northern_ireland_rqia_register.csv`.** We found no reuse statement for RQIA's registers when we checked on 21 September 2026, so **this file is excluded from our CC BY grant**. We reproduce these two rows so that each match can be traced, and every row carries the `service_id` and `source_file` needed to find the entry in RQIA's own published spreadsheet. Anyone wanting to republish that wording should ask RQIA.
- **The `*_quote` columns** in `right_to_choose.csv`, `site_claims.csv`, `children_pathway.csv` and `appointment_length.csv`. Each is a short extract from the named clinic's own website, reproduced under the quotation exception in [section 30(1ZA) of the Copyright, Designs and Patents Act 1988](https://www.legislation.gov.uk/ukpga/1988/48/section/30), with the source page as the acknowledgement. Copyright stays with each clinic.

Healthcare Improvement Scotland material is **not** reproduced in this package. `clinic_checks.csv` records only whether we matched a clinic to an HIS entry; the HIS register wording itself is in the clinic-level package, under the restriction described in that README.

## Read this before using the data

1. **We list clinics. We do not rate clinical quality.** Nothing here is our judgement of any clinic's quality, and no column in this package grades a clinic. Every regulator value is a public register entry or our failure to find one. Every value read from a website reduces the clinic's own published wording to a category and ships with the sentence it came from and the page it was on. What is ours is the matching, the coding, and the judgement columns the dictionary marks as ours: `status_category`, `cqc_match_confidence`, `match_confidence`, `devolved_check_scope`, `register_note`, `length_class`, `basis`, `site_fetch_status`, `has_*_content` and `extraction_notes`.

2. **`not_stated` means the clinic's website does not say it. It never means the clinic does not do it.** This is the single most important line in this README, because most columns here are mostly `not_stated`. A clinic that does not mention Right to Choose on its website may well accept Right to Choose referrals. These files measure what is published, not what is practised, and we have not verified practice.

3. **`no_match_found` means we could not match the clinic to a register entry. It does not mean the clinic is unregistered.** 195 of the 342 clinics are in that category, and that number is not a count of unregistered clinics. Three separate things put a clinic there: we searched and found nothing; the clinic lawfully does not need to register; or a match exists and our matching was too conservative to make it. Shared consulting addresses make conservative matching necessary — the clinic-level package sets out how many postcodes carry more than one listed clinic, and how many clinics sit at them. Matched counts are floors.

4. **Not needing to register is the normal case in England, and the devolved nations draw the line somewhere else entirely.** CQC's own guidance says a provider that only diagnoses ADHD or autism, with no treatment, does not need to register: *"If you only provide diagnosis of ADHD and/or autism (without any form of treatment) you will not need to register for this regulated activity"* ([Treatment of disease, disorder or injury](https://www.cqc.org.uk/guidance-regulation/providers/registration/scope-registration/regulated-activities/treatment-disease-disorder-or-injury)). In August 2026 the Parliamentary and Health Service Ombudsman called that a *"regulatory gap"* and recommended the Government consider bringing *"diagnosis of ADHD and autism into scope of CQC registration and inspection"*. That carve-out is English, and it is guidance about a regulated activity rather than a line in the Act. The three devolved regimes define a registrable clinic in their own governing legislation, and each defines it by the profession of the person providing the service:

   | Nation | Provision | An "independent clinic" is |
   |---|---|---|
   | Scotland | [National Health Service (Scotland) Act 1978, s.10F(2)](https://www.legislation.gov.uk/ukpga/1978/29/section/10F), as amended | *"a clinic which is not comprised in a hospital and in or from which services are provided by a medical practitioner, dental practitioner, registered nurse, registered midwife, registered pharmacist, registered pharmacy technician or dental care professional"* |
   | Wales | [Care Standards Act 2000, s.2(4), as it applies in relation to Wales](https://www.legislation.gov.uk/ukpga/2000/14/section/2) | *"an establishment of a prescribed kind (not being a hospital) in which services are provided by medical practitioners"* |
   | Northern Ireland | [HPSS (Quality, Improvement and Regulation) (NI) Order 2003, art.2](https://www.legislation.gov.uk/nisi/2003/431/article/2) | *"an establishment of a prescribed kind (not being a hospital) in which services are provided by medical practitioners"* |

   None of the three turns on whether the service is treatment, so a diagnosis-only ADHD clinic staffed by a doctor is inside their scope and outside CQC's. Scotland's list has widened over time and is the widest of the three: pharmacists and pharmacy technicians were added to it on 19 June 2024. Wales and Northern Ireland add that the establishment must be "of a prescribed kind", which the regulations under each Act set, and that is a separate narrowing — not a diagnosis-or-treatment test. This corrects a sentence in our July report; see "Corrections to earlier releases".

5. **A registration is not a rating, and a rating is not about ADHD.** Of the 342 clinics, 128 are currently registered with CQC and 39 hold a published CQC rating. A CQC inspection covers a whole registered location, which may provide other services besides ADHD assessment, and some reports predate a clinic's ADHD service. Neither HIW nor RQIA publishes a comparable rating for these entries, and HIS gradings are not set against CQC ratings anywhere in this package.

6. **Right to Choose applies in England only and needs a GP referral.** The legal right runs on an elective referral, which [regulation 38 of the NHS Standing Rules 2012](https://www.legislation.gov.uk/uksi/2012/2996/regulation/38) defines as one made by a general medical practitioner, a general dental practitioner or an optometrist; [NHS England's patient choice guidance](https://www.england.nhs.uk/long-read/patient-choice-guidance/) states that *"The legal rights do not extend to self-referrals or any referral other than from or on behalf of a GP, Dentist, or Optometrist."* `right_to_choose.csv` records what clinics say, not who is entitled to what.

7. **The appointment-length classes are ours, and they are not a pass mark.** `length_class` applies a written rule, set out in the Method below, to contact time the clinic states on its own site. The adult ADHD assessment quality assurance standard (AQAS) says *"the total assessment time should be at least 2 hours in most cases"*, and the classes are named around that line because it is the only published number to name them around. AQAS is a professional standard published in a journal. No body enforces it, nobody audits against it, and a clinic below the line is not breaking a rule. Do not read `under_2h` as a failure or `2h_or_more` as a pass.

8. **`market_adhd_companies.csv` is a floor, not a census.** It counts companies whose *name* has contained "ADHD". A clinic can trade without one — as a sole practitioner, or under a name that says nothing about ADHD — and a company with ADHD in its name need not be a clinic, or trading, or in England. Dissolved companies whose names run the letters together are the hardest to find, which hits the earliest years hardest. Every count from this file is a minimum.

9. **This is a directory census, not a register of every UK provider.** It covers the 342 clinics live on ADHD Private on 18 September 2026. 9 of them pay us for a labelled "Featured" placement, flagged in the clinic-level package's `is_featured_paid_placement`. They were checked and extracted on the same terms as every other clinic, and every one of their rows in every file here was re-read by hand before publication. Paid placement is always labelled and never changes a clinic's position in organic listings or matching results. We take no commission on bookings or referrals.

10. **Every source URL in these files was checked, and 21 of them cannot be checked the same way you would.** We opened all **1,424 distinct URLs** cited across these files on 22 September 2026. **1,403 answered 200.** Of the rest:

    * **20 URLs across 6 clinic websites answer 403 to an automated reader while serving normally to a browser** — `adhdcertify.co.uk` (8), `enlightenedminds.co.uk` (5), `insightdiagnostics.co.uk` (3), `everlief.co.uk` (2), `drmusasami.co.uk` (1) and `emotionsclinic.com` (1). Every one was opened by hand in a real browser on 22 September 2026; all 20 loaded, and each carries the wording quoted against it. They are 403-to-us, not dead, and if you follow the citation yourself you will see the page.
    * **One cited page has been taken down since we read it**: `www.diversediagnostics.co.uk/bupa-adhd-assessment`, cited by that clinic's `speed_statement` row in `site_claims.csv`, now returns 404. The quote was found word for word on that page when it was fetched, and `checked_on` on the row is the date that was true of. We have left the value and its quote in place rather than delete them, because the file records what clinic websites said on a stated date, and a page removed afterwards does not make the record of it wrong — but that one citation can no longer be opened, which is why it is named here.
    * **Two Companies House rows — `04629203` and `06399132`, dissolved in 2007 and 2009 — have no `companies_house_url` at all**, because Companies House publishes no profile page for them; they survive only in its dissolved-company search index, which is where we found them, and `found_by` and `date_source` say so. Neither company appears in any chart.

11. **The website evidence was captured on two days, and a few captures are thin.** Pages were fetched on 18 September 2026 and again on 21 September 2026, and a quote is published when it is found word for word in either capture of the page it cites. Two sites — `asd-adhd-assessors.co.uk` and `infocus-adhd.co.uk` — returned almost nothing to the second fetch (105 and 24 characters against 6,290 and 3,414 on the first), and 65 of the 2,545 stored pages came back shorter the second time. Where the capture behind a file is the thin one, a claim the site makes elsewhere will be missing from its row. Every count of what sites state is a minimum of what sites state.

12. **`children_pathway.csv` covers more clinics than our own directory flags as seeing children, and not all of them were read to the same depth.** The first round read the 244 clinics our directory lists with a child or teenage age group, fetching up to four child-focused pages each. Checking the remaining sites found 32 more whose pages describe or price an assessment for children — one of them advertising a child assessment package by name — so those 32 were extracted too, from the general pages already stored for them. They had no child-page fetch, so their rows rest on less text than the first 244, and a sparse row among them reflects what we collected rather than what the clinic does. `pages_read` shows the difference on every row. The finding that our own age-group data missed those clinics is ours to fix, and it is recorded here because it shapes this file.

## Corrections to earlier releases

Both corrections below are also rows in `corrections.csv`, with `release` naming the release corrected.

1. **The clinic-level README of 21 September 2026 said there was no Welsh or Northern Irish register file to compare against.** Its line read: *"There is no HIW or RQIA file to compare against"*. That was true of the 24 July name searches it described, and it stopped being true when we checked both registers clinic by clinic on 21 September. Both produced matches, and both files are in this package: `wales_hiw_register.csv` and `northern_ireland_rqia_register.csv`. In particular, the one clinic whose site claims HIW registration is corroborated by this package. The clinic-level package is unchanged and still says what it said; read this alongside it.

2. **Our July report said the devolved regulators apply the same diagnosis-only distinction as CQC.** They do not, for the reason set out in point 4 above. The sentence is annotated in place at adhdprivate.co.uk/research/uk-private-adhd-market-2026, and the full correction is in that report's September 2026 update.

A third correction, to a single Scottish registration flag in the clinic-level package, is recorded in `corrections.csv` and described in the Method below.

## What we did not include, and why

- **Any measure of whether a clinic is good.** There is no quality column here and there will not be one. Review ratings are excluded as well: platform terms do not let us redistribute them, and a satisfaction score is not a measure of diagnostic accuracy.
- **A market-wide registration rate.** Dividing registered companies by ADHD-named companies would produce a number that looks like compliance and is not: the denominator contains dormant, new, non-clinical and non-English companies, and a provider that only diagnoses need not register in England at all. We publish the two counts and never their ratio.
- **A comparison of Right to Choose providers against CQC status.** Too few rows, and the same denominator problem.
- **Waiting times, and any trend in them.** Our wait field is a category, and its change log begins on 28 July 2026, so no six-month series exists. NHS England Digital's quarterly [ADHD Management Information](https://digital.nhs.uk/data-and-information/publications/statistical/mi-adhd) is the official series; the August 2026 release is the latest.
- **Price trends.** Two usable monthly snapshots are not a trend.
- **Contact details and named people.** No email address, phone number or personal name appears in any column. RQIA's register publishes a responsible person, a manager and a phone number for each service; the build refuses to run if any of those columns reaches the output file. Names survive only inside a `*_quote`, where the clinic wrote them on its own public page and cutting them would edit the evidence.
- **Anything about the people who use our site.** Aggregate monthly volumes are published separately, at adhdprivate.co.uk/data/adhdprivate-site-demand-2026-09.csv.

## Method

**Who is in it.** The 342 clinics listed on ADHD Private on 18 September 2026, the same list, in the same order, as the clinic-level package. Nothing was added or dropped to suit a finding.

**Regulator status (`clinic_checks.csv`).** Each clinic gets exactly one `status_category`, assigned in this order, so a clinic with two registrations is counted once and under the first that matches: CQC registered with a published rating (39); CQC registered, no rating published (89); registered with HIS, HIW or RQIA (13); devolved registration in progress (2); matched only to deregistered CQC locations (4); no match found (195). 16 clinics hold a current devolved registration, and 3 of those also hold a CQC one, which is why the devolved category shows 13. The CQC columns are carried forward from the clinic-level package's 18 September register pull and re-derive its published funnel exactly: 342 listed, 304 searched, 132 matched, 128 currently registered, 39 with a published rating. The build refuses to write the file if that funnel does not reproduce.

**Devolved registers.** Every clinic with a location outside England was checked by hand against the register of each nation it has a presence in: 46 clinics, covering 27 with a Scottish presence, 13 Welsh and 9 Northern Irish. A further 33 England-only clinics were checked because a name or a postcode drew a candidate; 263 drew no candidate at all. A shared postcode or a similar name is a candidate, never a match. Scotland was re-confirmed against the HIS service pages fetched on 21 September, with "registration in progress" counted as no match. Wales was checked against HIW's service directory, and registration dates were read from each service page. Northern Ireland was checked against RQIA's `all-registered-services.xlsx`, register of 8 June 2026. Rejected candidates are itemised in our working files and are not published, because publishing "we considered and rejected this clinic" invites exactly the misreading this package is trying to avoid. 11 rows carry a `register_note` explaining a judgement.

That work also found one error in the clinic-level package: one clinic's Aberdeen location is on the HIS register and the 18 September match did not find it. `corrections.csv` records it. The clinic-level file is unchanged, because a published file that quietly changes is worse than a published file with a documented error.

**What clinics say (`right_to_choose.csv`, `site_claims.csv`, `children_pathway.csv`).** Three separate extractions over the pages we stored from each clinic's own website. Every value is quote-or-blank: a model proposes a value and the sentence it came from, a script then checks that sentence appears verbatim in the stored page, and a value whose quote cannot be found is blanked rather than published. A second model pass and a hand pass follow, and both are allowed only to weaken a value — to `not_stated`, or down the one ordered ladder — never to strengthen it. The build refuses to run a correction that strengthens a value. Every change either pass made is a row in `corrections.csv`.

For Right to Choose, 153 clinics had at least one page read, 87 of those pages contained Right to Choose text at all, and 71 rows carry a status quote. 337 sites could be read; the 5 that could not are marked `not_collected` in `rtc_status` and are not counted as anything else.

**The Right to Choose recall pass.** Because a clinic that accepts Right to Choose referrals often says so on one page and nowhere else, the September collection did not stop at the pages stored on 18 September. Sitemaps were checked for Right to Choose pages and those pages fetched, at one request per host every second. A clinic still shows `not_stated` if nothing was found: 266 do.

**Appointment length (`appointment_length.csv`).** The rule was written before it was applied, and it codes **contact time only**, because AQAS is about contact time: *"two hours or more of direct patient contact time"*, where a consultation may be *"face-to-face, or remote using video-conferencing"*. Time spent on questionnaires, scoring or report writing is not contact time and does not count towards the class.

| `length_class` | Applied when |
|---|---|
| `under_2h` | the whole stated range is below 120 minutes |
| `2h_or_more` | the whole stated range is 120 minutes or above |
| `straddles_2h` | a stated range crosses 120 minutes |
| `per_session_only` | only one session of several is timed, so no total exists |
| `unclear` | "up to N", a self-contradiction, or a figure that includes non-contact work |
| `child_pathway` | the only figure on the site is for the child assessment |

Sessions are totalled where the site states each one. 112 clinics state something codeable; 99 of those rows are adult-pathway, and 86 of the adult rows carry a number rather than only a class. `basis` records which reading produced the class, and the row carries the quote it rests on. This is our rule applied to their words; a different rule would give different counts, which is why both the rule and the quotes are published.

**Companies (`market_adhd_companies.csv`).** Four sources, all of them Companies House, fetched on 21 September 2026 with each URL, fetch time and SHA-256 recorded: the advanced company search for names containing "ADHD" (all statuses); the free monthly bulk file of live companies dated 1 September 2026, matched on "ADHD" anywhere in a name so that run-together names are not missed; the dissolved-company alphabetical index; and every matched company's own profile page. The whole-word search alone missed names like "MyADHD", which is why the bulk file is in the list.

Each company counts **from the date its ADHD name took effect**, not from incorporation, because a company renamed into an ADHD name in 2025 is not evidence of a 2014 market. `previous_company_names` on each profile gives those dates: 33 of the 470 companies have held a different name, and `adhd_name_from`, `adhd_name_until` and `adhd_name_periods` record when the ADHD name applied. 451 dates come from the company's own profile, 17 from the bulk file and 2 from the dissolved index.

A company is "on the register" on a date if its ADHD name had taken effect by then and it had not ceased. On that basis there were 314 on 21 March 2026 and 346 on 21 September 2026. 3 charitable incorporated organisations carry no usable date and are excluded from every dated count while remaining in the file; they are the reason a count here can be one or two lower than a naive count of rows.

**CQC locations with ADHD in the name** were counted from CQC's active and deactivated location files of 1 September 2026 together, matching on location name: 49 in total. They are counted separately from the company figures and never added to them.

**Hand checking.** Beyond the second and hand passes on every extraction, a stratified sample was re-read against the full stored page, chosen so that every non-blank value of every charted field appears at least once, plus every row belonging to the 9 clinics that pay us. `corrections.csv` records how many rows were re-read for each file and how many of them changed.

Almost every change the hand pass made removed a value or weakened it, and a correction may only weaken one **unless it brings its own quote**. A change that makes a value stronger is accepted only when the person making it supplies the sentence it rests on, that sentence is found word for word on the stored page it cites, and the build refuses to run otherwise. Two changes in this package meet that bar, and both are in `corrections.csv` with their quotes: one row published a single rating scale from a quote that named two, and one row named a profession the site never claims, where the site gives a job title instead. Three further rows carry a new quote without changing their value at all: the recorded sentence named an individual clinician where the same sentence states the role on its own, so the shorter wording is published instead.

**People's names.** This file is about clinics, not clinicians. Where the sentence a value rests on names an individual, the quote is replaced by `[withheld: the line names a person]` and the value still stands; 248 values are published that way. A note gives a clinician's role and never their name, and five notes that did name one were rewritten to the role — each is in `corrections.csv` as a `redact`, with the reason but without the name. Two things are deliberately not redacted. Some clinics trade under a person's name, which is the business's name and is published as such. And a `_source_url` is the clinic's own public page, whose address occasionally contains the clinician's name; four do. Those addresses are kept, because without them the row could not be checked by anyone, which is the point of publishing the evidence at all.

**Sources checked at publication.** Every external URL in this package was requested on the day of publication and its status recorded. Two clinic domains answer an automated request with HTTP 403 and open normally in a browser; those pages were opened in a browser and the values kept.

## Data dictionary

Values are lower-case with underscores unless noted. A blank cell means the field does not apply to that row, and is not the same as `not_stated`, which means we looked and the site did not say.

### `clinic_checks.csv` — 342 rows, one per listed clinic

| Column | Meaning |
|---|---|
| `clinic_id` | joins to every other file and to the clinic-level package |
| `clinic_name` | the clinic's listed name |
| `nations` | the UK nations the clinic has a location in, `;`-separated |
| `status_category` | **ours.** One of `cqc_registered_rated` (39), `cqc_registered_unrated` (89), `devolved_registered` (13), `devolved_registration_in_progress` (2), `cqc_deregistered_only` (4), `no_match_found` (195) |
| `cqc_searched` | `true` (304) / `false`: whether the clinic was searched on the CQC register |
| `cqc_matched` | `true` (132) / `false`: whether any CQC location was matched to it |
| `cqc_currently_registered` | `true` (128) / `false` |
| `cqc_current_location_rated` | `true` (39) / `false`: a rating is published for a currently registered location |
| `cqc_match_confidence` | **ours.** `high` (117) or `medium` (15); blank where nothing was matched |
| `his_register` | `registered` (11), `registration_in_progress` (2) or `no_match` (329) |
| `hiw_register` | `registered` (4) or `no_match` (338) |
| `rqia_register` | `registered` (2) or `no_match` (340) |
| `devolved_check_scope` | **ours.** Why the clinic was checked: `presence_outside_england` (46), `england_only_candidates` (33), `no_candidate_drawn` (263) |
| `register_note` | **ours.** A sentence explaining a judgement, on 11 rows |
| `checked_at` | the date the registers were checked |

### `right_to_choose.csv` — 342 rows, one per listed clinic

| Column | Meaning |
|---|---|
| `website_url` | the site the pages were read from |
| `site_fetch_status` | **ours.** `ok` (337), or why not: `site_under_maintenance` (1), `fetch_failed` (1), `no_own_site` (1), `site_unreadable` (1), `no_website` (1) |
| `pages_read` | how many pages of that site were read for this file; 153 rows are above zero |
| `has_rtc_content` | **ours.** `true` (87) where the stored pages mention Right to Choose at all |
| `rtc_status` | `offers` (39), `not_offered` (30), `paused_or_closed` (2), `not_stated` (266), `not_collected` (5) |
| `rtc_adults` | `yes` (29) where the site says the Right to Choose service covers adults |
| `rtc_children` | `yes` (19) where it says the service covers children or young people |
| `rtc_adhd_named` | `yes` (41) where ADHD is named in the Right to Choose wording |
| `rtc_area_restrictions` | `yes` (28) where the site limits Right to Choose to named areas or ICBs |
| `<field>_quote` | the sentence the value was read from; 71 rows carry a status quote |
| `<field>_source_url` | the page it was read on |
| `extraction_notes` | **ours.** Commentary on 75 rows: what a value rests on, or what the site says that the categories do not capture |
| `checked_on` | the date the pages were read |

### `appointment_length.csv` — 112 rows, one per clinic that states a length

| Column | Meaning |
|---|---|
| `length_class` | **ours.** `2h_or_more` (44), `under_2h` (29), `straddles_2h` (13), `child_pathway` (13), `unclear` (11), `per_session_only` (2) |
| `minutes_low`, `minutes_high` | the stated range in minutes; 86 of the 99 adult-pathway rows carry one |
| `basis` | **ours.** Which reading produced the class: `stated_figure` (74), `child_figure_only` (13), `up_to_120_or_more` (8), `sessions_totalled` (5), `standard_package` (3), `up_to_below_120` (3), `one_session_timed` (2), `versions_differ` (2), `booked_slot` (1), `includes_non_contact_below_120` (1) |
| `quote_1`…`quote_3`, `source_url_1`…`source_url_3` | the sentences the class rests on, and the pages they were read on |
| `note` | **ours.** Why a row was classed as it was, where the quote alone does not show it |
| `v1_stated_duration_mins` | the figure our July survey held, for comparison |
| `decided_on` | the date the class was applied |

### `wales_hiw_register.csv` — 5 rows, one per clinic and HIW entry pair

4 clinics; one clinic holds two entries. `entry_name`, `establishment_type`, `services`, `sector`, `postcode`, `registration_status` and `registered_on` are HIW's own fields. `match_basis` and `match_confidence` are **ours**: how we linked the entry to the clinic, and how sure we are. `register_url` opens the entry.

### `northern_ireland_rqia_register.csv` — 2 rows, one per clinic and RQIA entry pair

`service_id`, `service_name`, `provider`, `category`, `categories_of_care`, `categories_definition`, `postcode` and `registration_status` are RQIA's own fields, from `all-registered-services.xlsx`, register of 8 June 2026 (`register_of`). Both entries are registered as an Independent Medical Agency. `match_basis` and `match_confidence` are **ours**. RQIA's personal columns — responsible person, manager, phone — are not in this file and the build refuses to write them.

### `site_claims.csv` — 342 rows, one per listed clinic

What each clinic's own website says about who does the work and how it is sold. Every value is
read from the site and carries the sentence it was read from. **Counting a claim is not endorsing
it**: we did not test whether a clinic can see you in the time it states, and a claim recorded
here is the clinic's, not ours.

A speed or NHS-wait claim is recorded only where the site makes it about its ADHD assessment. Four
of these clinics sell other private services alongside, and make such a claim about one of those —
a gender clinic, private scans, a private GP service — and not about the ADHD assessment; those
values are cleared, with the reason in `extraction_notes` and a row in `corrections.csv`. Payment
terms are not treated this way: how a business takes payment is not tied to one service.

| Column | Meaning |
|---|---|
| `website_url` | the site the pages were read from |
| `site_fetch_status` | **ours.** `ok` (337), or why not: `site_under_maintenance` (1), `fetch_failed` (1), `no_own_site` (1), `site_unreadable` (1), `no_website` (1) |
| `pages_read` | how many pages of that site were read for this file; 337 rows are above zero |
| `has_claims_content` | **ours.** `true` (324) where the stored pages carry any of the material below |
| `clinicians_named` | `yes` (213) where at least one assessing clinician is named on the site |
| `professional_registration_shown` | which register the site names for its clinicians: `gmc` (77), `hcpc` (105), `nmc` (27), `gphc` (6), or `other:<name>` for any other regulator the site names (1 row). A site may name more than one; the cell then holds them separated by `; ` |
| `memberships_shown` | professional bodies the site says it or its clinicians belong to: `rcpsych` (58), `rcpch` (8), `bps` (71), `ukaan` (7), `acamh` (1), `bacp` (16), `ukcp` (11), or `other:<name>` for any other body the site names (49 rows naming 63 such bodies between them, 43 distinct, the most frequent named 9 times). Membership of a body is not accreditation of a service |
| `accreditation_claims` | `unnamed` (16) where the site says it is accredited without naming a body. Where it names one, the cell holds `body:<name>` as the site writes it — 17 rows naming a body 19 times between them, 15 distinct bodies, the most frequent named 4 times. **No accreditation scheme is specific to ADHD assessment** — see the FAQ in the study. Several of the named bodies run no accreditation scheme at all, and some are insurers or commissioners rather than accrediting bodies; where we could check a claim against a published member list we did, and `extraction_notes` says what we found |
| `other_certifications_shown` | non-clinical certifications the site displays: `iso_9001` (2), `iso_27001` (4), `cyber_essentials` (4), `cyber_essentials_plus` (1), `dspt` (4), or `other:<name>` (0) |
| `speed_statement` | how the site describes how soon it can see you, in its own terms: `period_days` (48), `period_weeks` (88), `period_months` (8), `no_waiting_list` (16), `speed_word_only` (32). The quote holds the wording |
| `nhs_wait_comparison` | `yes` (74) where the site sets its own timing against NHS waiting times |
| `paid_fast_track` | `yes` (18) where the site sells an option to be seen sooner for a higher fee |
| `finance_offered` | `instalments` (42) where the clinic itself splits the fee, `third_party_finance` (42) where a credit provider is named |
| `discount_offered` | `promotion` (18) a time-limited or coded offer, `concession` (10) a reduced fee for a named group, `bundle_saving` (20) a saving for buying assessment and follow-up together |
| `<field>_quote` | the sentence the value was read from; 323 rows carry at least one quote. **One quote per field, not per value:** where a field holds several values, the quote is the line the field was read from and may name only some of them; `extraction_notes` says where the rest were found |
| `<field>_source_url` | the page it was read on |
| `extraction_notes` | **ours.** Commentary on 321 rows: what a value rests on, or what the site says that the categories do not capture |
| `checked_on` | the date the pages were read |

5 of the 342 rows are `not_collected` throughout, because the site could not be read at all.

### `children_pathway.csv` — 342 rows, one per listed clinic

What each clinic's own website says about how it assesses a child or young person. The column
names follow the elements set out in CAAQAS (the consensus standard for child and adolescent ADHD
assessment), so the two can be read side by side. **Describing an element is not meeting a
standard**, and a blank is not a "no": most clinics simply do not publish this level of detail.
Clinics that assess adults only sit in this file with every field `not_stated`.

| Column | Meaning |
|---|---|
| `website_url` | the site the pages were read from |
| `site_fetch_status` | **ours.** `ok` (337), or why not: `site_under_maintenance` (1), `fetch_failed` (1), `no_own_site` (1), `site_unreadable` (1), `no_website` (1) |
| `pages_read` | how many pages of that site were read for this file; 272 rows are above zero |
| `has_children_content` | **ours.** `true` (248) where the stored pages describe assessing children at all |
| `min_age` | the youngest age the site says it assesses, as a whole number of years; 124 rows state one |
| `assessing_roles` | who the site says carries out the assessment: `consultant_psychiatrist` (48), `psychiatrist` (29), `paediatrician` (31), `clinical_psychologist` (54), `psychologist` (33), `specialist_nurse` (13), `nurse_prescriber` (3), `other_clinician` (32). More than one may be named |
| `child_seen` | how the child themselves is seen: `in_person` (67), `video` (24), `in_person_or_video` (64), `not_seen` (0) where the site says the child need not attend |
| `parent_carer_input` | `yes` (170) where a parent or carer is interviewed or completes a measure |
| `school_information` | `yes` (162) where information is sought from the child's school or nursery |
| `school_observation` | someone observes the child in the school setting: `routine` (10) as part of the assessment, `optional_or_extra` (32) as an add-on or on request |
| `developmental_history` | `yes` (124) where a developmental or birth history is taken |
| `co_occurring_conditions` | `yes` (120) where the site says other conditions are screened for or considered |
| `semi_structured_interview` | the diagnostic interview named: `ace` (14), `young_diva` (9), `k_sads` (0), `mini_kid` (0), `dawba` (0), `unnamed` (37) where an interview of this kind is described but not named, or `other:<name>` — 10 rows, all of them a DIVA version written two ways (`other:DIVA-5` 6, `other:DIVA` 4) |
| `rating_scales` | the questionnaires named: `conners` (43), `snap_iv` (16), `sdq` (7), `vanderbilt` (2), `brown` (1), `unnamed` (75), or `other:<name>` for a scale the site names that is not in this list (11 rows naming 14 such scales between them, 13 distinct spellings, one scale appearing under two of them, and no spelling used more than twice) |
| `objective_test` | a computer-based attention test: `qbtest_routine` (11), `qbtest_optional` (9), `qbcheck_routine` (43), `qbcheck_optional` (14), `explicitly_not_used` (0) where the site says it does not use one. Another named test is written `other:<name>` (7 rows, each a continuous performance test under a different name). CAAQAS does not require one |
| `physical_assessment` | when height, weight, pulse or blood pressure are taken: `at_assessment` (11), `before_medication` (23) |
| `written_report` | `yes` (164) where a written report is part of what is bought, `no` (1) where the site says one costs extra or is not provided |
| `post_diagnosis_feedback` | `yes` (139) where a feedback or explanation session follows the assessment |
| `stated_length` | the assessment length in minutes, as the site states it — a single figure or a `low-high` range; 89 rows state one. CAAQAS sets no duration requirement |
| `<field>_quote` | the sentence the value was read from; 229 rows carry at least one quote. **One quote per field, not per value:** where a field holds several values, the quote is the line the field was read from and may name only some of them; `extraction_notes` says where the rest were found |
| `<field>_source_url` | the page it was read on |
| `extraction_notes` | **ours.** Commentary on 254 rows: what a value rests on, or what the site says that the categories do not capture |
| `checked_on` | the date the pages were read |

5 of the 342 rows are `not_collected` throughout, because the site could not be read at all.

### `market_adhd_companies.csv` — 470 rows, one per company

Not joined to `clinic_id`: these are companies, most of which we do not list and some of which are not clinics.

| Column | Meaning |
|---|---|
| `company_number`, `company_name` | as Companies House holds them |
| `adhd_name_from` | the date the company's ADHD name took effect. Every dated count in our reports runs from this, not from incorporation |
| `adhd_name_until` | the date it stopped, where it has |
| `adhd_name_periods` | every period an ADHD name applied, for the companies that have held more than one |
| `company_status` | `active` (357), `dissolved` (110), `liquidation` (2), `converted-closed` (1) |
| `company_type` | as filed, including 17 charitable incorporated organisations, 3 of them Scottish |
| `date_of_creation`, `date_of_cessation` | incorporation and cessation; 111 rows carry a cessation date |
| `previous_adhd_names` | earlier ADHD names, on 33 rows |
| `sic_codes`, `sic_86_human_health` | SIC 2007 codes as filed, and whether any is in the 86xxx human health group (254 rows) |
| `jurisdiction` | `england-wales` (430), `scotland` (24), `northern-ireland` (11), `wales` (3); blank on 2 rows Companies House holds no jurisdiction for |
| `found_by` | **ours.** Which search found it: `advanced_search;bulk_file` (301), `advanced_search` (111), `bulk_file` (56), `dissolved_index` (2) |
| `date_source` | **ours.** Where `adhd_name_from` came from: `profile` (451), `bulk_file` (17), `dissolved_index` (2) |
| `companies_house_url` | the company's public page. **Blank for the 2 rows whose `date_source` is `dissolved_index`** (dissolved 2007 and 2009): Companies House serves no company profile page for them — its web service and its API both answer 404 — so they survive only in its dissolved-company search index, which is where we found them. Neither is in any chart: both left the register long before the first year we count. |
| `checked_at` | the date it was fetched |

### `corrections.csv`

One row per change made after extraction, plus the release-level corrections described above. `release` names the release corrected; `action` is `correct` or `append`; `pass` is which pass made it; `previous_value` and `corrected_value` say what changed; `reason` says why. The build refuses to write a correction that strengthens a value.