UK Lead Risk Map — Data sources & methods

This map combines several open datasets to give a constituency-level picture of likely childhood lead-exposure risk in the UK. Each indicator is sourced and caveated below. Where coverage is partial (Scotland and Northern Ireland in particular), the limitation is documented alongside what was used as a fallback.

Geography

Parliamentary constituency boundaries (2024)

Westminster Parliamentary Constituencies July 2024 (UK BGC), from the ONS Open Geography Portal. 650 features, post-Boundary-Review names (PCON24NM, e.g. "Aberafan Maesteg" not pre-2024 "Aberavon"). Geometry simplified to ~2 MB at 0.002° tolerance using Shapely topology-preserving simplification — visually identical at country zoom.

Indicators

Blood lead ≥ 5 µg/dL (BLL)

Source: IHME Global Burden of Disease 2021 — Lead Exposure Estimates. Mean and 95% uncertainty interval for the proportion of the population with blood lead ≥ 5 µg/dL, by age group.

IHME publishes this at county / unitary-authority resolution in the UK, not at parliamentary-constituency level. The constituency map shows the value of the parent county / UA for each constituency (centroid-in-polygon match). Where no local data exists, falls back to nation-level (England, Scotland, Wales, NI) shown with reduced opacity.

Adjusted IHME — modelled children with BLL ≥ 5

The default map layer and the rankings table. A best-data-per-area allocation: each area's IHME GBD 2021 blood-lead total is re-allocated across its constituencies by modelled risk. The baseline area is the finest IHME geography available:

• England — the parent county / unitary authority (keeps local variation; e.g. Blackpool ≈ 3.5%, Surrey ≈ 1.1%).
• Wales, Scotland, Northern Ireland — the nation (IHME publishes no sub-national estimate). Nation baselines: Wales 4.7%, NI 3.4%, Scotland 3.2% vs England 1.6%.

The within-area risk weight is each constituency's predicted IHME blood-lead level, from a regression fitted on the England county/UA units: BLL% ~ pre-1945 housing + IMD deprivation. So the weights are fitted to UK data, not imported odds ratios; both predictors enter positively. The area's IHME child total is then allocated proportionally to predicted-risk × child population, so each area's total is preserved — this re-levels the model onto IHME baselines rather than inflating it. Built by build_adjusted_ihme.py (constituency → county mapping by build_constituency_mapping.py). Illustrative; not a clinical estimate.

Re-leveled to GBD 2023. The 2021 round is the last with UK sub-national lead detail, so it supplies the within-UK pattern. The level is then brought to the latest round: every area is scaled by the UK national ratio GBD 2023 / GBD 2021 = 1.06% / 1.81% = 0.586, lowering the whole map by ~41% while keeping the 2021 pattern. GBD 2023's own UK nation splits are not usable as a spatial signal — they read England 0.03% and Wales 21.0% above 5 µg/dL (Wales CI 5–48%), against 1.6% / 4.7% in 2021 — so only the UK-wide level is taken from 2023, not the per-nation values. Source: IHME GBD 2023 — Lead Exposure Estimates (Proportion above 50 µg/L, ages 0–19, 2023).

Topsoil Pb and historic mines were tested as predictors and left out of the headline figure (regress_ihme_predictors.py): mines show no positive association with IHME blood lead, and topsoil Pb predicts it only outside London — the signal collapses once London (highest soil Pb, but average modelled BLL) is included — so it would extrapolate poorly. Both remain available as standalone map layers.

The raw IHME layer (England county/UA proxy; devolved nations shown flat at the nation value) remains available as a toggle.

Housing built pre-1900 / pre-1945 / pre-1972 / pre-1992

The map exposes four cutoffs because UK lead-exposure pathways span different eras:

Three sources are stitched together by nation:

England & Wales — VOA CTSOP 4.0 (2023)

Valuation Office Agency Council Tax Stock of Properties, table CTSOP4.0 (build period × council-tax band). LAD-level dwelling counts across detailed bands: pre-1900, 1900–1918, 1919–1929, 1930–1939, 1945–1954, 1955–1964, 1965–1972, 1973–1982, 1983–1992, etc. Aggregated to constituency via the ONS ward → constituency → LAD lookup, weighted by ward count.

Approximation: split-LAD constituencies (most of London, big metros) inherit each parent LAD's overall age profile in proportion to ward count. LSOA-level VOA data (CTSOP4.1) would give a sharper split-constituency result.

Scotland — SHCS 2017–2019

Scottish House Condition Survey: Local Authority Analysis 2017-2019, Stock Profile sheet. Sample-based estimates at LA level. Only the pre-1945 vs post-1945 split is published at LA level — the other cutoffs (pre-1900, pre-1972, pre-1992) are not available for Scotland and show as "no data" on those layers.

SHCS uses sample-based 3-year rolling estimates with confidence intervals. The 2017-2019 release is the most recent LA-level publication: 2020 had no survey, 2021 had non-comparable data, and the next LA release (2022-2024) won't appear until 2025.

Northern Ireland — NIHCS 2023 (NI-wide)

NI House Condition Survey 2023 Preliminary Report, Table 4. Bands: pre-1919, 1919-1944, 1945-1964, 1965-1980, 1981-1990, 1991-2002, 2003-2023. Only NI-wide values are published (the 2011 district-level report covers other measures, not age). The same NI-wide values are applied to all 18 NI constituencies.

Non-decent homes (Decent Homes Standard)

England: English Housing Survey: Local Authority Stock Modelling 2023, LA-level model based on EHS sample data combined with administrative records. A home fails the standard if any of the four criteria fail: HHSRS Category 1 hazard (which includes lead), disrepair of key components, lacks modern facilities, or inadequate thermal comfort.

Topsoil lead — BGS UK Compiled Topsoil Survey (UKTS)

Source: BGS UK Compiled Topsoil Survey (UKTS), released August 2024 under the Open Government Licence. The UKTS combines multiple BGS geochemical surveys (G-BASE and related programmes) into a single gridded product: 57,966 shallow topsoil samples (0–15 cm depth, analysed by X-ray fluorescence) interpolated to a 500 m raster in EPSG:27700, with values in mg/kg dry weight for 41 elements including lead (Pb).

The constituency-level values on this map are the mean Pb concentration across all 500 m raster cells whose centroid falls within each constituency boundary, computed with rasterio zonal masking. The 95th-percentile cell value is also stored (visible in the constituency profile panel) to capture local hotspots within each constituency that the mean may mask.

Coverage: The UKTS covers England, Wales, Northern Ireland and part of Scotland. Approximately 30 Scottish constituencies — mainly northern and north-eastern Scotland (Aberdeen, Highlands, Borders, Fife, Edinburgh and surrounding areas) — fall outside the survey footprint and are shown as no-data (grey) on this layer.

Contextual thresholds used in the profile panel:

Note that the UKTS raster represents shallow topsoil (the A horizon, typically 0–15 cm); deeper horizons and in-building dust/paint are separate exposure pathways. Urban "hotspot" cells in central London reflect Victorian industrial contamination, atmospheric deposition from leaded petrol, and historic paint/pipe residues in soils, not necessarily current indoor exposure.

Historic metal mines

The map combines two open inventories, deduplicated by ~500 m proximity:

• Macklin et al. (2023). WAPHA Global Metal Mines Database — supplementary data for the Science paper "Impacts of metal mining on river systems: a global assessment". UK subset: 137 sites.
• EA Inventory of Closed Mining Waste Facilities (2014) — Environment Agency / NRW open data under the Mining Waste Directive (2006/21/EC). 150 sites in E&W, 132 metalliferous, 146 flagged for water pollution. Open Government Licence v3.0.

Combined: 211 unique sites in 51 constituencies, mapped via point-in-polygon onto the 2024 boundaries. Lead-relevant commodities flagged (Pb, Zn, Cu, Ag, Ba/witherite — Pb often co-occurs in polymetallic veins).

Coverage gap: BGS BRITPITS holds >260,000 records including all historic UK workings (the comprehensive inventory) but is licensed by NERC and not openly downloadable. Cornwall is under-represented here as a result — the Camborne–Redruth historic tin/copper fields had thousands of small workings that aren't priority-polluting closures.

Constituency context

MP — name, party, email

Pulled from the UK Parliament Members API (January 2026 snapshot). Constituency names match the post-July-2024 boundaries.

Children aged 0–19

ONS mid-year population estimates aggregated to 2024 parliamentary constituencies.

IMD decile and deprivation score

Both come from the mySociety Composite UK IMD, which produces UK-wide comparable Index of Multiple Deprivation rankings (the nation-specific IMDs use different methodologies, so a composite is needed for cross-UK comparisons).

• IMD decile: 1 = most deprived 10% of UK constituencies, 10 = least deprived.
• Deprivation score: 0–100, higher = more deprived. Population-weighted percentile rank of LSOAs within UK.

Methodology notes & known limitations

• BLL is a county-level proxy. IHME doesn't publish constituency-level BLL; the England baseline is the parent county / UA, and Wales/Scotland/NI use a single nation value.
• The adjusted figure inherits IHME's cross-nation gradient. The headline that Wales ≈ 3× England rests on IHME's modelled nation estimates from sparse UK blood-lead data; treat the levels with caution. Within Wales/Scotland/NI there is no sub-national IHME, so within-nation variation comes only from housing age and IMD — those nations look fairly even on the map, but now at the right level. Illustrative, not a clinical estimate.
• Housing data is uneven across UK nations. England & Wales have rich VOA data with all four cutoffs; Scotland has only pre-1945 from SHCS LA tables; NI has all four cutoffs but applied uniformly NI-wide (no sub-NI data exists).
• Mines layer is not exhaustive. Only Macklin + EA closed waste — actual UK historic-mine inventory (BRITPITS) is much larger but licensed.
• Topsoil Pb is a constituency mean of a coarse raster. The 500 m grid cannot resolve garden-to-garden or street-to-street variation. The 95th-percentile cell value (in the profile panel) gives a sense of within-constituency spread, but neither metric replaces a site-specific soil test. ~30 northern Scottish constituencies have no UKTS coverage.
• Ward-weighting is an approximation. Constituencies that span multiple LADs are aggregated by ward count rather than dwelling count — fine for stable areas, less reliable in complex London/metro splits.
• Constituency → county mapping. Each 2024 constituency is matched to its IHME county/UA via its local authority (two-tier shire districts resolved to their county); a handful of cross-boundary seats are assigned to the county holding their populated core.

Data files in this repository

Author: Lee Crawfurd, Center for Global Development. Source code: github.com/lcrawfurd/lead-map.