How we measure patient information
The method behind the tracker, published before any table is: what we read, how each measure is scored, where the evidence comes from, how results are kept honest, and how a trust disputes one.
What we read
Only what a patient can already see: the patient information a trust publishes on its own public website, HTML pages and PDFs alike, read the way a patient would reach them. No logins, no access to any system, no patient data anywhere in the process.
A trust is measured on its whole estate: every leaflet link its site lists on its own domain is checked for loading. For the measures that need the text, every leaflet is read, or at least 200 of them where the estate is larger, on the same instrument as every other trust. Where a measure cannot be taken (a document library our reader cannot open) the trust is not yet scored on that measure. A single document we cannot extract text from is counted, reported by name and left out of the text measures, never guessed at. A partial score is never shown, because partial measurement ranks the least-measured estates highest.
Every estate is re-read on a rolling cycle, and a subscribed trust's estate is re-measured every month. Each result carries the date it was taken, and every finding links to the page it was taken from, so it can be re-checked by anyone by opening the link.
The seven measures
Each measure is a property of a published page, scored the same way for every trust. The weights are ours, set against the evidence below; they are the same for every trust and change only with a dated note on this page. The weighted total out of 100 is what we call the medrics score. Six of the seven are measured today, and a score is scaled to 100 from those six (90 points); coverage joins the score when it is measured for every estate, and this page will say so on the day it does. No trust is named with a score on this site.
Changed 13 September 2026. Two definitions above were tightened, and the page says so because the weights and definitions only change with a dated note. Currency is now one rule for every trust: a review date that has not passed, or, where a page prints none, a review within the last three years; a page with no date at all counts against, which it did not before. Driving and flying advice now has to be advice, not a passing mention of a car. Figures already published were measured under the previous definitions and are being re-measured trust by trust; each trust's report states the date and the instrument its own figures were taken with.
Changed 15 September 2026. A method review with a second, independent model found that reading a leaflet's whole text for driving and flying topics counted a passing mention of an anaesthetic, sedation or a symptom as the leaflet's subject. A leaflet now counts as being on a DVLA or CAA condition or treatment when its title names one, or when its text is plainly about one; a word in passing no longer does. Two smaller corrections went in at the same time: the word “audiology” no longer counts as an audio format, and where a leaflet's currency rests on a publication, approval or issue date rather than a printed review date, that is now recorded with the measurement. Every earlier result is being re-measured on the corrected reader before any figure is compared or published. From this date a sample of the reader's own judgements is re-read every week by a second model of a different family (an answer from the reader's own family is discarded, not counted), which must quote the page verbatim, and on the point in dispute, to disagree; a disputed judgement stays open, across reader changes, until a person has looked at it or the corrected reader no longer makes it, and while any is open the next release is held.
| Measure | Weight | What counts | Why it matters |
|---|---|---|---|
| Keeping the library up to date | 30 | The share of leaflets whose printed review date has not passed. A leaflet with no printed date counts against. Where a page prints no review date but does print when it was written (a published, issued or approved date: the day the trust last signed the content off, which is what a review date records) or last reviewed, a date within the last three years counts as current, and the label that date came from is recorded with the measurement. Every trust is measured on that one rule. | A printed review date is the trust's own commitment. The PIF TICK criteria, which NHS England's content standard builds on, require a process for reviewing and updating information within set timeframes. |
| Readable | 20 | The reading age of each leaflet's extracted text (for a long PDF, its first twelve and last two pages), estimated with the Flesch-Kincaid formula, and how far the estate's median sits from the NHS aim. Hardest pages are named for a human to read; the formula is an estimate, never a verdict on one leaflet. | The NHS service manual: “We aim for a reading age of 9 to 11 years old.” The Patient Information Forum's position statement says the same. |
| Links load | 10 | The share of the trust's own listed leaflet links that return a page. Every link on the trust's own index is requested; the score uses the most recent read, taken within the last month, and the pages a trust's report names are re-requested when that report is built, so a page fixed since the last read is not held against the trust. Scored steeply: an estate where one listed link in ten fails scores nothing on this measure. | A dead leaflet link is a published service that has stopped. The Public Sector Bodies Accessibility Regulations 2018 and GOV.UK's accessible documents policy treat published documents as services to be maintained. |
| Formats signposted | 10 | The share of leaflets that tell the reader other formats or languages exist: easy read, large print, braille, audio, BSL, interpreting. Site-wide template lines count, and are reported as such in the trust's report. | The Accessible Information Standard: “All organisations that provide publicly funded NHS care or adult social care must have regard to this standard.” |
| Phone-translatable | 10 | The share of leaflets published as web pages rather than PDFs; a web page that is only a link to the leaflet's PDF counts as a PDF. A phone can translate a web page and a screen reader can navigate it; a PDF, unless it is carefully tagged, usually defeats both for most patients. | GDS: “PDFs are not designed to be flexible in their layout.” NHS England: “People who speak little or no English are more likely to be in poor health.” |
| Driving and flying advice | 10 | Of the leaflets on conditions and treatments the DVLA or CAA rules cover (sedation and anaesthesia included, under the GMC duty below), the share that give the patient driving or flying advice. A leaflet is on such a condition when its title names one or its text is plainly about one; a word in passing does not make it so. Wording that only mentions a car in passing, such as asking someone to drive the patient home, does not count. Presence of advice only: we do not judge whether the advice is right. The rules themselves are quoted verbatim from gov.uk and the CAA in our own tool and re-checked several times a week. | The DVLA's guide for medical professionals, and the GMC's guidance that doctors should alert patients to conditions and treatments that may affect their driving. |
| Coverage | 10 | The share of a checklist of conditions, treatments and procedures that patients commonly need information about that the trust's own leaflet titles cover: across the whole estate, not emergency care alone. The checklist is ours and is published with any table that uses it; today it starts from common emergency discharges and is being widened to the whole patient-information estate before this measure joins any score. | Information has to exist before it can be read. NHS England's content standard and the Accessible Information Standard assume information is available for the patient's condition, whatever it is. |
| The medrics score | 100 | The weighted sum, shown only when every wired measure was taken on the same instrument for that trust. Today that is every measure except coverage: 90 points, scaled to 100. | |
What the evidence says
Each measure rests on the NHS's own published aims and duties, linked above, and on peer-reviewed evidence we keep in a register and re-verify. The studies we lean on most:
- Low health literacy and outcomes. Berkman and colleagues' systematic review, Annals of Internal Medicine 2011: low health literacy is associated with poorer health outcomes and poorer use of care. doi:10.7326/0003-4819-155-2-201107190-00005
- UK leaflets read above the aim. Oliffe and colleagues, BMJ Open 2019, assessed UK medicine information sheets against reading level and patient comprehension. doi:10.1136/bmjopen-2018-024582. Noble and colleagues, Epilepsia 2022, found epilepsy patient information pitched well above the recommended level. doi:10.1111/epi.17299
- Leaving the emergency department without understanding. Engel and colleagues, Annals of Emergency Medicine 2009: most discharged patients had gaps in understanding their care or instructions, and most were unaware of it. doi:10.1016/j.annemergmed.2008.05.016
- Language access changes outcomes. Karliner and colleagues, Medical Care 2017: convenient access to professional interpreters in hospital was associated with lower readmission rates. doi:10.1097/MLR.0000000000000643
- Easy read improves knowledge. Hurtado and colleagues, Journal of Intellectual Disability Research 2013: easy-read health information improved knowledge for people with intellectual disabilities. doi:10.1111/jir.12097
The gaps, stated plainly. We found no study that ties out-of-date leaflets to harm; that measure rests on the trust's own printed promise and the PIF and NHS England process standards. We found no UK audit of driving advice in patient leaflets; that measure rests on the DVLA and GMC duties. Readability is strongly linked to comprehension and only weakly to hard outcomes. The score therefore measures an estate against the NHS's own stated aims and duties, not against patient outcomes, and no measure contains a clinical judgement.
How we keep it honest
One instrument for everyone
Every trust is read by the same code with the same thresholds. When the measuring code changes, every earlier result is re-measured before any comparison is drawn, so two rulers are never blended.
Re-checked at the moment it matters
Pages a report names are fetched again when the report is built. A page the trust has already fixed is not held against it.
Quotes are verified bytes, not memory
Every rule or standard we cite is captured from the source page and re-verified on a schedule. A reworded source page surfaces as a stale claim on our side, not as a fact on yours.
Whole, or not yet scored
A score is shown only when every wired measure was taken on the same instrument, today six of the seven. A blind spot is reported as a blind spot, never filled in.
Who sees what, and when
Results are published in stages, and consent is the only door to a trust's name.
- Now: the method, and anonymous regional figures. This page, and one median per NHS England region where three trusts are measured in full. No trust is named on this site with a result.
- Each subscribed trust: its own private benchmark. Its estate, measure by measure, against the other trusts measured, inside its monthly report.
- Trusts that choose to: a public entry. A trust appears by name on this site only when it consents, after seeing its evidence pack, with two weeks' notice before it first appears. It can withdraw, and leaves the public table at the next refresh.
- Later: the table. A public table of consenting trusts, refreshed monthly with its refresh date shown, this method beside it, and a ‘disputed’ state for any result under review.
These pages carry no advertising. The leaflet tool clinicians use to hand a patient the right page stays free to every clinician and patient.
How we read a site, small numbers, and who is behind this
Reading a site. Public pages only: no logins, no forms, nothing submitted. The measurement reader names itself with a contact address in every request, waits at least 0.6 seconds between requests to the same site, reads one estate at a time, and reads a full estate at most once a month. A trust that would rather its site were not read can write to us and is removed from measurement.
Small numbers. A region shows a median only where at least three trusts are measured in full, and the number of trusts behind every median is printed beside it. No trust is named without its written consent, and no figure that would identify a trust by elimination is published.
Who is behind it. MEDRICS LTD, a small company founded and run by Greta Gapsyte, self-funded, with no advertising, sponsorship, pharmaceutical or NHS funding. The measures rest on the NHS's own published standards and the evidence above; patients have not yet shaped them directly, and we would welcome a patient panel that did. Write to us.
Dispute a result
Any trust may ask for the evidence behind any figure about its estate, and we send it link by link. If we are wrong, we correct it and say so; while we check, the result is marked disputed rather than hidden. Write to info@medrics.co.uk with the page in question.
See the score by region See the trust service
Contains public sector information licensed under the Open Government Licence v3.0.