This page records where each figure comes from, which geography and year it belongs to, and what it cannot tell you. It is written to be read alongside the map rather than instead of it.
The words used on this map
Official statistics use short names for areas and measures. Where the map uses one, this says what it means. Where a plain word is clearer, the map uses that instead and this gives the official name for it.
| Plain word | Official name | What it is |
|---|---|---|
| Borough | Local authority district | One of the 33 councils in London. |
| Ward | Electoral ward | The area one or more councillors are elected for. |
| MSOA | Middle Layer Super Output Area | A statistical area of about 7,500 people, built from whole LSOAs. London has 1,002. |
| LSOA | Lower Layer Super Output Area | A statistical area of about 1,500 people, and the smallest area this map holds figures for. London has 4,994. They hold roughly equal populations rather than following anything on the ground, so they are small where London is dense and large where it is not. Deprivation deciles are published at this level. |
| Deprivation | Index of Multiple Deprivation, or IMD | The government measure of how deprived an area is. It combines income, employment, health, education, crime, housing and environment. |
| Decile | Decile | A tenth. Decile 1 is the most deprived tenth of England. Decile 10 is the least deprived tenth. |
| Ratio (England = 100) | Indirectly standardised admission or mortality ratio | Admissions or deaths compared with the number expected if England’s rates applied to the area’s own mix of ages and sexes. 100 is as expected, 120 is 20% more and 80 is 20% fewer. It is not a rate per 100 people. Most of the Local Health admissions and mortality figures are these. |
| Charity | VCSE organisation | A charity registered with the Charity Commission. |
Every data source
Everything the map draws comes from one of the sources below. Each row says who publishes it, what it covers, the geography it is published at, the period the figures describe, and how often the publisher releases a new one.
The last column is not written down here. It is read from
data/meta/manifest.json, which the pipeline rewrites on every run,
so it reports when this site last actually pulled that source and whether the
pull worked. A source that has started failing says so here rather than quietly
serving older figures with no sign of it.
| Source | What it provides | Published at | Period covered | Publisher updates | Last fetched here |
|---|---|---|---|---|---|
| Loading the fetch log… | |||||
Unless noted otherwise these are published under the Open Government Licence v3.0 and contain public sector information licensed under it. Postcodes are resolved through postcodes.io, which derives from ONS and Ordnance Survey data. A postcode typed into the map's search box is looked up on postcodes.io at that moment; only the postcode is sent. The basemap is CARTO, built from OpenStreetMap, with Esri's World Light Gray canvas used instead if CARTO's tiles cannot be loaded. Air quality is modelled by Defra and carries its own citation: © Crown copyright Defra via uk-air.defra.gov.uk, licenced under the Open Government Licence (OGL).
Station and bus stop measures use Transport for London open data, licensed under version 2.0 of the Open Government Licence with amendments for Transport for London. Powered by TfL Open Data. Contains OS data © Crown copyright and database rights 2016. Geomni UK Map data © and database rights 2019.
Geographies, and which year each one is from
The four geographies on the map come from different censuses and boundary reviews. They do not nest inside one another, and that has consequences worth understanding before comparing them.
| Level | Vintage | Count in scope | Source |
|---|---|---|---|
| Borough (local authority) | May 2025 | 33 | ONS LAD_MAY_2025_UK_BGC_V2 |
| Ward | May 2025 | 704 | ONS WD_MAY_2025_UK_BGC_V2 |
| LSOA | 2021 census | 4,994 | ONS Lower_layer_Super_Output_Areas_December_2021_..._BGC_V5 |
| MSOA | 2021 census | 1,002 | ONS Middle_layer_Super_Output_Areas_December_2021_..._BGC_V3 |
MSOAs can now be drawn as well as aggregated through. They are a layer of their own in the sidebar rather than a third position on the ward and LSOA switch, because they nest into neither: an MSOA is built from whole LSOAs, but 41% of LSOAs cross a ward line, so an MSOA edge agreeing with a ward edge is a coincidence rather than a rule. The layer is fetched only when you switch it on. An MSOA is in scope when at least one of its LSOAs is, which is the same definition the rest of the pipeline works to.
All boundary layers are the BGC generalisation: simplified to 20 metres and clipped to the coastline. Keeping them at the same generalisation matters. LSOAs were previously taken at BSC, 200 metres, which is ten times coarser, and because the two had been simplified independently their shared borders no longer coincided: LSOA edges visibly cut across ward lines.
Coordinates in the files the browser downloads are rounded to five decimal places, which moves no point by more than 0.32 metres and roughly halves the file size.
What can go wrong when you read a figure
Six things on this map are easy to read the wrong way. Each one is a property of how the figures are published, not a fault, and each changes what a number means.
1. LSOAs do not sit inside wards
LSOA boundaries are from 2021 and wards were redrawn in 2025, so the two
genuinely cross. 59% of LSOAs sit entirely within one ward. 41% cross a ward
line. Every LSOA is assigned to exactly one ward, using the ONS best-fit lookup
LSOA21_WD25_LAD25_EW_LU_v2, so a ward figure counts each LSOA once
in the ward it best fits, and not in proportion to the area it shares. LSOA
outlines crossing ward outlines at high zoom is the geography being honest, not
a drawing error.
2. Only counts are added up
Everything else is averaged and weighted by population. LSOA figures reach wards that way, and wards reach boroughs the same way, except for genuinely additive measures such as population, crime totals and service counts, which are summed. MSOA figures reach wards through the LSOAs the two share, so a ward straddling two MSOAs gets a blend in proportion to where its people are.
3. Some health figures are a borough figure repeated
The health data comes from OHID Fingertips in two pulls. The Local Health set is published per MSOA and varies ward by ward. The borough set is published per local authority and is written unchanged onto every ward in that borough, so 18 of the health indicators are identical for every ward in a borough. Ranking wards on one of those produces a real-looking order that is an artefact of the geography rather than a finding. The map flags them, and the indicator list shows how many distinct values an indicator actually has.
4. Missing is not zero
A blank ward can mean the figure was withheld, not that it is zero. Fingertips suppresses MSOA figures where small counts could identify individuals. In Query, a ward with no figure for one of your conditions can neither pass nor fail it, so it is left out of the results and the number left out is reported underneath them.
5. Colour follows the indicator, not the size of the number
The highest IMD decile is the least deprived tenth, and the highest life expectancy is the best outcome, so colouring the largest value red would state the opposite of the truth. For the Local Health set the direction is OHID’s own published polarity, carried through from the source. Where an indicator has no better or worse end, it is left uncoloured. The share of residents over 65, or in social renting, or of any ethnic group, is a fact about who lives there rather than a good or bad result.
6. Service counts are attributed by where the service stands
Pharmacies use the pipeline’s own ward count. Everything else is placed by testing its point against the ward polygon. A service used by a neighbouring ward is still counted in the ward it stands in, so these will not always agree with a count made a different way.
7. The ends of a legend are not the highest and lowest areas
The colours run from the 5th to the 95th percentile of the areas drawn, so one extreme area does not use up the whole scale. Every area beyond either end takes the end colour. Where any area does, the legend marks its end value with ≤ or ≥, so “≥ 37.8” means 37.8 or more. The Top and bottom 20 list gives the real values.
8. Each indicator has its own period
The period beside a legend is the one the publisher gives for that indicator’s figures, and it differs a great deal from one to the next. The “Data updated” date at the foot of the map is when the data was last refreshed, not the date the figures describe. A refresh takes the newest figures published, and for some indicators those are several years old.
Known gaps
- City of London. Its 20 wards have no deprivation score and no Local Health indicators. They are tiny and share very few LSOAs, so there is nothing to aggregate. This is a property of the geography, not a fetch failure.
- 39 renumbered MSOAs. Fingertips still publishes the Local Health set against 2011 MSOA codes. Where a code no longer matches a 2021 one, the value is left missing rather than filled from a neighbour.
- DWP benefits are not included. The figures are on DWP Stat-Xplore, which needs a registered API key. NOMIS holds nothing newer than November 2018 for them.
Rebuilding this
The whole map is rebuilt by one script, fetch_all_data.py, with
no manual downloads. Sources, licences and per-source notes are in the
repository: github.com/Harv334/pophealthmap.