Every dataset refresh, in time order.

Welsh lifestyle figures move forward three years, via Public Health Wales

Welsh smoking, higher-risk drinking, physical-activity and 5-a-day figures now come from Public Health Wales’s Public Health Outcomes Framework (accredited official statistics, derived from the National Survey for Wales), using the 2021/22–2022/23 pooled period for all 22 principal areas — replacing the 2019-20 values that had been frozen since the StatsWales platform migration took the LA-level lifestyle tables offline.

One caveat, disclosed rather than hidden: PHOF publishes these as age-standardised percentages, while the equivalent English and Scottish figures are crude prevalence. The definitional thresholds are identical (>14 units/week, 150 minutes/week, current smoker, 5+ portions); the standardisation typically moves a figure by a point or two. Each Welsh metric’s source note says so explicitly.

• Overweight incl. obesity stays on NSW 2019-20 (BMI ≥ 25, crude) — PHOF publishes no all-adult weight measure. It is now the only remaining 2019-20 Welsh value.

England’s smoking figures now come from the same kind of source as Scotland’s and Wales’s

England’s smoking column now uses the Annual Population Survey estimate (OHID Fingertips indicator 92443, self-reported, adults 18+, latest period 2024) instead of the QOF GP-register prevalence (91547, latest 2022/23). Two years fresher — and, more importantly, all three covered nations’ smoking figures are now self-reported survey estimates: England’s APS, Scotland’s Scottish Surveys Core Questions and Wales’s National Survey for Wales. GP-recorded and survey-reported smoking rates measure genuinely different things; a cross-nation column should not quietly mix them.

• England values shift modestly (Westminster 13.4 → 12.7, Manchester 17.0 → 14.2); the England reference line moves from 14.7 to 10.9.
• Ribble Valley has no APS estimate and now shows “—” rather than a substituted GP-register value.

Scottish data moves to SHeS 2021–2024, and the obesity column finally means the same thing everywhere

All 32 Scottish councils’ higher-risk drinking and physical-activity figures now come from the Scottish Health Survey’s 2021–2024 pooled window (published October 2025), replacing the 2019–2023 pool.

More importantly: Scotland’s overweight-and-obesity column now uses the BMI ≥ 25 “overweight including obesity” measure — the same threshold England and Wales use — instead of the stricter BMI ≥ 30 “obesity only” series. The threshold mismatch we’ve disclosed since April is gone; the column is now directly comparable across all three covered nations. Scottish values jump accordingly (Edinburgh reads 58% rather than 21%) because the measure now includes the overweight band — not because anything changed in Scotland.

• 5-a-day stays on the 2016–2019 pool: re-checked against the new dashboard — no later window exists for this indicator because the survey’s COVID-era switch to telephone interviewing broke comparability.
• Smoking stays on the Scottish Surveys Core Questions 2023 (a much larger pooled sample than SHeS alone).
• The only remaining threshold mismatch on the site is Northern Ireland’s HSC Trust figures (BMI ≥ 30), which are shown as labelled Trust-level context, never as council values.

Six councils were showing "UK average" instead of their own figures

A full audit of the lookup path and data pipeline found — and fixed — two silent failures. First, postcodes in St Helens, King’s Lynn and West Norfolk, Bournemouth Christchurch and Poole, Herefordshire, Armagh City Banbridge and Craigavon, and Newry Mourne and Down (around 1.4 million residents) were falling back to the UK-average panel: postcodes.io returns those names with commas, apostrophes and full stops, and the site’s name-matching didn’t fold punctuation the way the dataset’s keys do. All six now resolve to their real records; verified against live postcodes.io.

Second, Barnsley and Sheffield had been quietly frozen at April data vintages. The Government Statistical Service re-coded both councils after their April-2023 boundary changes; ONS workbooks use the new codes while OHID Fingertips still publishes under the old ones, so every monthly rebuild lost their lifestyle figures and fell back to preserved values. The pipeline now recodes on every fetch, and both areas track upstream releases again.

Also in this audit

• Embed widget now serves all 380 councils (it previously charted only a 33-area sample). Metrics that aren’t published for an area are listed as such — never charted as a zero.
• Labels caught up with definitions: the map and embed now say “Overweight incl. obesity” (BMI ≥ 25) with scales matched to the real 40–80% range, and lifestyle reference lines use England national values labelled “Eng avg”.
• Story figures re-verified against the ONS 2022–2024 healthy-life-expectancy series: the Blackpool–Westminster healthy-years gap reads ~10 years on current data.
• Pipeline hardening: inequality-breakdown rows can no longer be mistaken for area values, and the test suite now runs in CI on every change and before every monthly data refresh.

NI lifestyle data now surfaces via HSC Trust adjacency

The 11 Northern Ireland LGD records had null lifestyle fields since launch. Adult lifestyle prevalence is published only at HSC Trust level (5 trusts), never at LGD2014 council level — per-LGD samples are too small. Confirmed against HSNI, NISRA Continuous Household Survey, Sport NI, HSCIMS, Census 2021 and APS that no UK statistical body publishes at LGD level.

The frontend now surfaces the parent HSC Trust’s five-indicator block on every NI LGD result panel and area page, with explicit “These belong to the HSC Trust, not [LGD name]” framing. Same editorial principle as the English district → county adjacency: parent values are never copied into the child record.

Trust-level data sourced from HSNI 2024/25 Trend Tables

• Smoking, obesity, physical activity, 5-a-day: 2023/24 period.
• Higher-risk drinking: 2022/23 period (question not asked in 2023/24).
• Two LGDs straddle Trust boundaries: Mid Ulster (Northern + Southern) and Newry, Mourne and Down (South Eastern + Southern). Both Trust figures are surfaced for these two with a footnote.

Welsh data refreshed, definitions harmonised

Welsh lifestyle data refreshed from Welsh Health Survey 2014–15 (11 years old) to National Survey for Wales 2019–20 (6 years old) for all 22 Welsh principal areas. The two definition mismatches that have been flagging Welsh values for months — pre-2016 daily-units drinking and 5×30 min/wk physical activity — both close in this refresh.

What changed

• Higher-risk drinking — Welsh column now uses >14 u/wk above-guideline (matches Fingertips 92778 and SHeS “Hazardous/Harmful drinker”).
• Physical activity — Welsh column now uses 150 min/wk MVPA (matches Fingertips 93014 and UK CMO).
• Obesity — Welsh column now uses BMI ≥ 25 “overweight or obese” (matches Fingertips 93088); Scotland still uses BMI ≥ 30, disclosed.
• Smoking and 5-a-day — values refreshed; definitions already matched.
• Source: StatsWales hlth5002 via the 14 Sep 2025 Wayback snapshot. Live URL retired in the 2025 stats.gov.wales migration.

District-level data, Scotland refresh, schema cleanup

The single biggest accuracy improvement since launch. Coverage on the four LTLA-publishing lifestyle indicators (smoking, obesity, physical activity, 5-a-day) jumped from 40 % to 97 % by adding a parallel Fingertips fetch at lower-tier-authority level. Scotland was refreshed end-to-end via the Scottish Government’s council-area Shiny dashboard. The dataset shape was tightened: non-council records dropped, country totals moved to a sibling block.

Coverage end state

What changed

• 164 English districts (E07) now carry their own Fingertips lifestyle figures rather than null. Canterbury shows Canterbury, not Kent.
• 32 Scottish councils refreshed across smoking (SSCQ 2023) + obesity / drinking / physical activity (SHeS 2019–2023, scraped via headless Camoufox) + 5-a-day (SHeS 2016–2019, the most recent council-area period).
• 91 non-council records dropped — NHS Integrated Care Boards, combined authorities, regions, and devolved-nation NHS health boards. None were ever returned by postcodes.io and none rendered. Country aggregates moved to _meta.country_totals.
• Definitions harmonised across England / Scotland / Wales for drinking (>14 units/wk) and physical activity (CMO 150 min/wk MVPA). Obesity remains BMI ≥ 25 in England Fingertips vs BMI ≥ 30 elsewhere — disclosed wherever shown.
• Pipeline updated — build_dataset.py now performs dual UTLA + LTLA Fingertips fetches plus an ONS LAD→CTY lookup, so future rebuilds preserve the backfill.
• Frontend — new county-adjacency block on postcode result panel for English districts. Strict editorial principle: parent-county values are NEVER copied into a district row in the data file. Inheritance is UI-only with explicit “belongs to Kent, not Canterbury” framing.
• New pages — /methodology/, /glossary/, /changelog/, plus per-area landing pages at /area/ (380 indexable URLs).

HLE backfill, deprivation lens, sources page

Lifemap goes public at lifemap.org.uk. Major data backfills bring HLE coverage to 49 %; nine post-2019 English UTLAs (Cumberland, Westmorland & Furness, post-2023 North Yorkshire and Somerset, Buckinghamshire, North and West Northamptonshire, Barnsley, Sheffield) are added via Fingertips indicator 90362 (2021–2023). A new deprivation lens on the homepage uses ONS’s 15 April 2026 release; a public /sources/ page audits every external URL.

Highlights

• HLE source switched for English UTLAs from ONS HSLE 2016–2018 (frozen) to OHID Fingertips 90362 (2021–2023). Welsh, Scottish and NI UTLAs continue on the older series until newer LA-level data is published.
• Higher-risk drinking switched from indicator 94185 (Government Office Region only) to 92778 (above-guideline, UTLA-level). Coverage jumped from 8 % to 40 %.
• Choropleth boundaries updated from 2022 to December 2024 ONS UTLA boundaries; map.js performs slug-fallback joins for the 2024 South Yorkshire renumbering.
• Story page, press kit, deprivation chart, brand mark, mobile padding, embed widget, OG images all shipped.

Initial dataset and site

Earlier builds focused on the data pipeline (LE / HLE / Fingertips), the static-page frontend, the postcode lookup, the comparison view, and an early choropleth map. See the full engineering history in CHANGELOG.md for line-level detail.