Care Deserts — Methodology

Primary metric

Active registered care-home beds per 100 residents aged 75+. This is the same denominator and numerator as OHID Fingertips indicator 92490, so TreatCompare values can be validated directly against published public-health benchmarks. A nursing variant (beds per 100 aged 75+, nursing-designated locations only) matches Fingertips indicator 92493.

Why not “homes per total population”?

• Care homes vary hugely in size. A 6-bed residential home and a 120-bed nursing village both count as “one home” on the legacy measure, which misrepresents real capacity.
• Care-home demand is driven by residents aged roughly 75 and over. Using total population dilutes the denominator with under-65s who are extremely unlikely to enter residential care this decade.
• OHID and DHSC both publish their headline care-home indicators with an age-adjusted denominator. Aligning with that means outputs are comparable to the public-health literature.

Supporting metrics (kept alongside the primary metric)

• Nursing beds per 100 residents aged 75+.
• Beds at dementia-specialist locations per 100 aged 75+ — see caveat below.
• Total beds per 1,000 residents aged 65+ (broader denominator).
• Homes per 10,000 total population — legacy secondary only, retained for continuity with the old TreatCompare page.

Dementia capacity caveat

CQC publishes a location-level dementia-specialism flag but no bed-level dementia flag. Our dementia metric therefore sums the full bed capacity of every CQC location that lists dementia as a specialism. A mixed-use location that houses both dementia and non-dementia residents is counted at its full capacity. The resulting number is an upper bound on specialist dementia provision — true dementia-only capacity is lower. We label this metric “beds at dementia-specialist locations per 100 aged 75+” rather than “dementia beds” so readers can judge it on the right terms.

“Care desert” classification

Bands are derived from the observed national distribution of the primary metric, not a fixed editorial threshold:

• Severe undersupply — bottom decile (≤ p10).
• Undersupply — bottom quartile (≤ p25).
• Average — interquartile range (p25 to p75).
• Well supplied — top quartile (≥ p75).

Each local authority also carries a z-score and a percentile rank, so consumers can rank, filter, or flag as they choose. LAs with fewer than 50% of active locations reporting a bed count are held out of the distribution and labelled insufficient data, to avoid misclassifying areas whose supply we cannot measure reliably.

Data sources

• CQC Care Directory (Open Government Licence v3.0) — authoritative list of registered care-home locations in England. Bed counts come from the “Number of care home beds” (previously “Care homes beds”) column. We exclude dormant and suspended locations.
• ONS mid-year population estimates (2022) for total, 65+ and 75+ populations per local authority.
• NHS England ASC-FR for weekly residential and nursing unit costs (council rates).
• OHID Fingertips 92490 & 92493 as benchmark validation layer, flagging any LA whose derived value differs materially from the published indicator.
• DHSC Capacity Tracker as an optional occupancy layer (response-based, so treated as pressure-on-stock not stock itself).

Geography and deduplication

All rows are standardised to current upper-tier English local authorities, joined on stable GSS codes (E06/E08/E09/E10). Locations with the same CQC location ID are deduplicated by retaining the most recent directory record. Historic abolished boundaries are mapped forward where the directory still lists them.

What “active” means

A location is treated as active if the directory’s dormant flag is blank or “N” and its registration status is “Registered”. Suspended and deregistered locations are excluded from all supply sums. If the directory does not publish any status flag for a row, it defaults to active — the directory is already filtered to registered locations at source.

Validation against OHID Fingertips

For each LA we compare the derived beds-per-100-aged-75+ against the published Fingertips indicator 92490 value and flag where the absolute difference exceeds 10 beds or the relative difference exceeds 15%. Likely causes are documented on a per-flag basis (dormancy thresholds, small-home beds not reported to CQC, boundary changes between publication dates).

Cost analysis

The cost-vs-supply relationship is re-run using the new primary metric. We report the Pearson correlation between beds per 100 aged 75+ and weekly residential cost, plus the bottom-vs-top supply-quartile cost comparison. We label this descriptive, not causal — London and South East land and labour costs confound the relationship, driving both higher per-week fees and lower beds-per-older-resident in parallel.

QA checks

• No negative or zero age denominators.
• No duplicate CQC location IDs.
• No mixing of provider-level and location-level counts.
• Total population is never used as the primary denominator.
• Home count is never used as the primary supply proxy.
• Missing data is flagged explicitly (insufficient-data band) rather than silently dropped.
• Coverage is reported per metric — how many LAs have a cost figure, how many have adequate bed coverage, how many are included in the distribution.

Deliverables

The same pipeline that powers this page writes four CSV deliverables for press and external use:

• clean_master_care_beds_la.csv — master LA dataset.
• care_beds_validation_report.csv — TreatCompare vs Fingertips.
• care_desert_rankings.csv — LAs ranked by the primary metric with bands.
• methodology.md & qa_notes.md — plain-English methodology and validation notes.