Methodology and evidence standards

Source-backed observations

A price observation should point to a public clinic page, published PDF, official listing, provider correction or reviewed user-submitted evidence. Where no usable source exists, the price is treated as missing rather than estimated.

Checked dates

Public pages show snapshot or source-check dates where available. Older records can remain visible with caveats, be excluded from medians, or be rechecked before being used in statistics.

Provider verified vs source backed

Provider or practice verified data means the clinic or representative has submitted or confirmed an update. Source-backed data means a public source supports the observation. Both still need review before affecting public comparisons.

Missing data and outliers

Missing prices are labelled or omitted from calculations. Outliers are checked for source mistakes, treatment scope, units, finance wording, ranges and comparability before being included in medians or ranges.

Data sources

Source types are stored explicitly: practice websites, public PDFs, NHS.uk or devolved NHS sources, GOV.UK, CMA publications, GDC/CQC references where relevant, user submissions and practice corrections.

• Practice websites and PDFs can support published private prices.
• NHS and official sources can support charge, policy and access-context pages.
• GDC signals are supplemental identity evidence, not a canonical practice directory.
• User submissions create review tasks only.

Price extraction

Prices are stored with source document ID, observed date, confidence, review state, price kind, billing unit and comparability label. Median calculations use exact, reviewed, stats-eligible observations.

• Median is preferred over mean because it is less sensitive to outliers.
• From, range, finance and per-unit prices are excluded from like-for-like medians.
• Anomalous prices are flagged for review before use in statistics.

NHS acceptance rules

Acceptance is never a single binary. The model stores adults, children, urgent slots, referrals and waiting-list status separately with source type, observed date, confidence and freshness window.

• User reports trigger review; they do not become published facts automatically.
• Phone calls are reserved for stale, contradictory or high-value records.
• Accepting status should usually refresh inside 7 to 14 days.

Freshness and expiry

NHS acceptance status has a 90-day maximum display window without re-verification. Private prices have a 180-day maximum display window. Pages with no verified data inside 180 days are demoted to noindex.

• Freshness badges appear beside data points.
• Expired NHS status is labelled not recently verified.
• Stale or disputed claims are suppressed from sitemaps.

Page-quality gates

Sitemaps and robots directives use the same gates shown on each page. Weak pages can exist for operations but are suppressed from indexing.

• City pages need at least 10 mapped practices and 10 reviewed price observations.
• Cheapest pages need at least 8 exact stats-eligible prices.
• Constituency packs stay noindex until local evidence is strong.

Commercial firewall

Corrections are free, acceptance placement is never sold, and organic ranking logic must not use affiliate value or paid listing status.

Scores are shown only with component counts, confidence bands and excluded-data notes. User reports are triggers, not facts.

Dataset snapshots

Public reports and stat pages should cite a frozen dataset snapshot, methodology version, source count, publication date and limitations. CSV or JSON distributions are exposed only where the sample is defensible.

Version history

access-index-v0.1 published 30 Apr 2026: Pilot methodology for city-level evidence pages. It is deliberately conservative and suppresses thin local pages.

Any future change to score weights, price eligibility, freshness windows or source hierarchy should be logged here before public reports use it.

Corrections are reviewed through the corrections page.