From July 2025, NHS England’s demographic WLMDS publication made it possible, for the first time, to see at provider level whether waiting list pressures are distributed equally across age, sex, ethnicity and deprivation groups. This project brings together the most recent releases of data to support commissioners, trust boards and regional teams in identifying performance gaps and targeted support needs.
Over 7 million people are waiting for elective treatment on the NHS. The 18-week referral-to-treatment (RTT) standard has not been met nationally since 2016, and the pressures that produced the current backlog, a decade of demand growth, workforce constraints, and the profound disruption of COVID-19, are well understood.
The WLMDS data highlights distinct patterns of waiting time variation by specialty, geography and trust, and demographic characteristics:
Each dimension allows waiting time variation to be examined separately through the lens of specialty, geography and demographic characteristics.
| Finding | What the data shows | |
|---|---|---|
| 01 | Across all months, patients from more socioeconomically deprived areas are over-represented in the longest waiting time bands. | Patients in IMD Decile 1 are over-represented in the longest waiting time bands in every month. Over-representation increases steadily with deprivation, from the least to the most deprived deciles. Although the association is small (Cramér’s V = 0.005-0.009), it is consistently non‑random across all eight monthly tests. |
| 02 | Patients from Indian, Pakistani and Bangladeshi ethnic groups are consistently over‑represented in the 18-52 week waiting time band, with the pattern concentrated in specific specialties. | This pattern is observed in all eight months, with a consistent direction of effect. The association is small (Cramér’s V = 0.009-0.013) but persistent. Over‑representation is concentrated in a subset of specialties, including Dermatology, Neurosurgical and General Internal Medicine. |
| 03 | East of England shows the largest and most persistent regional gap | In all eight months, the East of England records the highest breach and severe breach rates. The overall breach rate runs approximately 4-5 percentage points above the national average, equating to around 32,000-40,000 additional patients in breach at any given snapshot. |
| 04 | Mid and South Essex illustrates the upper end of trust‑level variation, with consistently high breach rates across all months. | The trust has approximately 175,000 patients, with 50-52% in breach, around 10-13 percentage points above the national average. It appears among the ten highest breach rates in every one of the eight months, indicating a sustained trust‑level gap. |
| 05 | Specialty shows the largest variation in waiting times. | There is an approximately 30 percentage point gap between the highest and lowest specialty breach rates. The ordering of specialties is stable across all eight months, and the magnitude of variation by specialty is greater than that observed across regions or demographic groups. |
| 06 | National waiting list performance is broadly stable at the 18-week threshold, while severe breaches (patients waiting 52+ weeks) have declined over time. | At each snapshot, around 2.8 million patients are in breach, with the overall breach rate remaining within a 1 percentage point range across the period. In contrast, the severe breach rate declines from 2.79% to 1.92%, indicating improvement in the longest-waiting cohort while overall breach levels remain stable. |
Across all eight monthly snapshots (July 2025 to February 2026), patients from more deprived areas are consistently over-represented in longer waiting time bands relative to their share of the overall waiting list. This pattern is monotonic across deprivation deciles and is observed in the same direction in every month without exception.
Patients in IMD Decile 1 (most deprived) consistently account for a higher proportion of long waits than their representation in the total waiting list. This over-representation declines stepwise across deciles through to IMD Decile 10 (least deprived), forming a stable gradient across the full period.
The individual-level association is small (Cramér’s V = 0.005-0.009), based on chi-square tests of independence between waiting band and IMD decile. P-values are adjusted using the Benjamini-Hochberg procedure across 32 tests, with all eight IMD-adjusted p-values < 0.0001. Absolute differences between the most and least deprived groups range from 0.3 to 1.2 percentage points
Taken together, the findings indicate a highly consistent but low-magnitude association. The significance lies in its stability and direction across all months rather than the size of any single effect.
Trusts and regions serving more deprived populations also tend to have higher breach rates, raising the question of whether the observed deprivation gradient is driven primarily by geography. The WLMDS publication does not allow this relationship to be formally disentangled, because deprivation and geography are correlated and the data are published as aggregated cross-tabulations rather than linked patient-level records, preventing multivariable adjustment.
However, the gradient is observed across organisations with differing overall performance levels, suggesting that deprivation is not solely a proxy for geography within this dataset. It should therefore be considered alongside geographic variation as a related but distinct dimension, while recognising that the two are correlated.
In every monthly snapshot from July 2025 to February 2026, patients recorded as Indian, Pakistani or Bangladeshi make up a larger share of the 18-52 week wait band than their share of the total waiting list. The same groups appear in the same direction in all eight months. The aggregate Cramér's V for ethnicity is 0.009-0.013 - negligible at the individual level. Most patients in every ethnic group wait similar lengths of time. The finding is noted for its consistency across time and its concentration in specific groups, not for the size of the individual-level effect.
The pattern is not explained by concentration in the highest-breach specialties. Representation is not elevated in Ear Nose and Throat, Oral Surgery or Plastic Surgery, all of which show representation index values at or below parity throughout the period. Instead, over-representation is observed in Dermatology, Neurosurgical and General Internal Medicine, which sit at or below average breach rates overall. This suggests the pattern is not simply driven by entry into the most pressured pathways.
Across all eight months, the persistence of the same directional pattern across the same groups is not plausibly explained by random variation alone. However, the data does not indicate any issue of individual or organisational conduct. It identifies where consistent patterns are observed, not why they arise. Possible explanations such as referral pathways, appointment uptake, or treatment sequencing, cannot be determined from waiting list counts alone.
Regional breach rates show a consistent ordering across all eight months. No region switches position materially. East of England leads every regional breach rate chart and every severe breach rate chart throughout. Its waiting list of 757,000-859,000 is the fourth largest of seven regions, so list size alone does not account for the gap. Its breach rate runs approximately 4-5 percentage points above the national average in every month. As a descriptive illustration of scale: if East of England's breach rate were equal to the national average, approximately 32,000-40,000 fewer patients would be in breach at any given snapshot. This figure is offered as a translation of the gap into patient terms, not as a projection or target.
North East and Yorkshire carries a list of 870,000-994,000, comparable in size to East of England, and sits below the national average on breach rate in every month. The observation that two regions of similar size sit at opposite ends of the observed distribution, consistently across eight months, is recorded as a descriptive fact. The data does not establish what accounts for this difference.
London carries the largest regional list (consistently 1.2-1.3M) and sits near the national average on breach rate throughout, with below-average severe breach rates in most months from December onwards. North West and South East are persistently above the national average on severe breach rate. South West is below average on both metrics in every month with the smallest regional list.
Mid and South Essex NHS Foundation Trust carries a breach rate of 50-52% across the window, running 10-13 percentage points above the national average at every snapshot. Its severe breach rate (52 or more weeks) exceeds 7.5% in every month. At approximately 175,000 patients and a severe breach rate of around 8%, approximately 14,000 patients have been waiting more than a year. It appears in the top ten for breach rate in all eight months and holds the highest position in six of them. The WLMDS does not explain what produces this gap.
| Month | Waiting List | Breach Rate | vs National (pp) | Severe Breach Rate |
|---|---|---|---|---|
| Jul 2025 | 171,965 | 50.39% | +10.51 | 7.75% |
| Aug 2025 | 174,720 | 49.95% | +9.99 | 8.15% |
| Sep 2025 | 176,715 | 50.02% | +10.01 | 8.53% |
| Oct 2025 | 179,405 | 50.29% | +11.00 | 8.75% |
| Nov 2025 | 178,750 | 51.18% | +12.05 | 8.27% |
| Dec 2025 | 177,565 | 52.49% | +12.66 | 8.19% |
| Jan 2026 | 175,290 | 52.56% | +12.67 | 8.57% |
| Feb 2026 | 175,275 | 51.75% | +12.64 | 7.98% |
Mid and South Essex is the most visible example, but the wider pattern is one of stable trust-level variation rather than a single outlier. United Lincolnshire Teaching Hospitals and University Hospitals Sussex appear in the top ten in seven of eight months each. James Paget University Hospitals appears in all eight months with a list of approximately 32,000-35,000 patients; a notably different scale to Mid and South Essex, which reinforces that high breach rates are not confined to the largest trusts. Liverpool Women's records a severe breach rate of 12.02% in December at approximately 8,000 patients, illustrating that the headline breach rate does not always surface the most acute concentrations of very long waiters.
Ear, Nose and Throat has the highest breach rate of any specialty in every month, at approximately 47-50%, consistently 7-10 percentage points above the national average. Oral Surgery is consistently second or joint-first, followed by Plastic Surgery at approximately 42-48%. At the lower end, Elderly Medicine remains around 17% throughout the period, with Other Mental Health and Other services also consistently below the national average.
The ordering of specialties is stable across the eight-month window, with no material movement between high and low-breach categories. The scale of variation indicates that differences in waiting times are greatest when examined by specialty, although the underlying causes are not identifiable from WLMDS data alone.
The national 18-week breach rate opens at 39.87% and closes at 39.11%, moving within a band consistent with normal variation rather than any directional trend. Approximately 2.8 million patients are in breach at every snapshot. This is consistent with published RTT statistics over this period and is noted here as the context within which the other findings sit.
The severe breach rate, the proportion of patients waiting 52 or more weeks, shows a clearer change, falling from 2.79% in July to 1.92% in February. At a list size of approximately 7 million, this corresponds to approximately 61,000 fewer patients waiting more than a year by February compared to July. The WLMDS records this change but does not indicate what actions or factors produced it.
This page documents the data sources and processing pipeline behind the analysis, and sets out the analytical constraints that shape how the results should be interpreted.
The Waiting List Minimum Dataset (WLMDS) is a weekly data collection submitted by NHS providers to NHS England, in scope since April 2021. Every trust submits data weekly covering open pathways (patients currently waiting), clock starts (new referrals), and clock stops (treatment commencing or patient leaving the list).
NHS England began publishing aggregated WLMDS data monthly in April 2024. The demographic breakdowns, by age, sex, ethnicity and deprivation, were added in July 2025. The demographic publication takes the form of three CSV files released monthly.
| File | Content | Geography | Time Coverage |
|---|---|---|---|
| Geography | Waiting list counts by age, sex, ethnicity and IMD decile, split by waiting band. One row per organisation, demographic category, and waiting band. | England, 7 regions, 42 ICBs, 134 NHS acute trusts (183 organisations total) | Monthly snapshot - most recent week only |
| Specialty | The same four demographic dimensions cut by the 24 RTT treatment functions. Enables specialty-mix analysis of demographic inequality. | England level only | Monthly snapshot - most recent week only |
| Timeseries | Historical weekly England-level data by age and sex from September 2021. Ethnicity and IMD are not available in the timeseries - a significant limitation for longitudinal equity analysis. | England level only | Weekly from September 2021 to present |
We have assembled the complete published history: all eight monthly releases from July 2025 to February 2026. NHS England does not maintain a public archive on the publication page. The historical files were retrieved by reconstructing the URL pattern used for each monthly release on the NHS England server. This gives us 24 CSV files in total: 8 Geography, 8 Specialty, and 8 Timeseries snapshots. By stacking the Geography and Specialty snapshots we build the most complete longitudinal picture currently possible of how the demographic composition of the waiting list has evolved.
The platform is built end-to-end on Microsoft Fabric. A monthly automated pipeline downloads each new release from NHS England, validates it against expected schema and volumes, and loads it into the Fabric lakehouse. The modelling layer runs in dbt, transforming the raw CSVs through staging and intermediate layers before materialising the mart as tables in the Fabric Warehouse.
| Layer | Materialisation | Schema | Role |
|---|---|---|---|
| Staging | Views | dbt_staging | Raw extraction, type casting, text standardisation, schema validation |
| Intermediate | Views | dbt_intermediate | Parsing of demographic dimensions, separation of pre-computed total rows, suppression flagging, additivity enforcement |
| Mart | Tables | dbt_mart | Star schema dimensions and fact table, ODS hierarchy enrichment, SCD Type-2 geography tracking |
The dimensional model is structured as a single fact table holding waiting list counts by demographic group, with dimension tables for provider, specialty, date, waiting band, and demographic category. The provider dimension is enriched with ODS hierarchy data to link each trust to its ICB and region, enabling geographic roll-up and drill-down. SCD Type-2 tracking on the geography dimension ensures that organisational restructures do not silently rewrite historical relationships.
The Power BI semantic model connects to the mart via an Import mode connection to the Fabric Warehouse SQL analytics endpoint. All percentage calculations are computed in DAX with enforced denominator consistency.
| Constraint | Detail | Impact |
|---|---|---|
| Management information status | WLMDS is subject to less central validation than the monthly RTT official statistics. Trust-level figures should be treated with appropriate caution and validated against RTT totals where possible. | Confidence in trust-level figures is lower than for national figures. |
| No specialty breakdown at trust level | The Specialty file provides demographic-by-specialty at England level only. Trust-level specialty-demographic intersections are not available in the published data. | Specialty-mix analysis of trust-level inequality is not directly possible from the published files. |
| Timeseries ethnicity and IMD absence | The Timeseries file contains age and sex only. Ethnicity and IMD longitudinal analysis is limited to the eight Geography snapshots. | Trend analysis for ethnicity and deprivation covers eight months only, not the full WLMDS history from September 2021. |
| Suppression threshold | Count values below the suppression threshold are shown as * rather than a figure. Small demographic groups at small trusts may be systematically suppressed. | Equity analysis for minority ethnic groups at smaller trusts is limited. |
| Count rounding | All count values are rounded to the nearest 5. Percentage calculations using small counts are affected disproportionately. | Low absolute impact at trust level for major demographic categories; significant for small groups. |
This project is designed to support equity-aware performance analysis in a public-sector context, where data is incomplete, reporting conventions change over time, and many of the factors behind observed outcomes sit outside the dataset itself.
The emphasis throughout is on identifying consistent patterns rather than maximising statistical significance, distinguishing scale effects (specialty, geography) from distributional effects (demographic representation), and being explicit about what management information can and cannot support.
The analysis is not intended to optimise operational performance metrics or to rank organisations in isolation. Its purpose is to surface where differences in waiting time experience are persistent, directional, and potentially relevant to equity monitoring, and to do so in a way that is robust to the known limitations of NHS administrative data.
The equity findings in this analysis should be read as signals rather than precise measurements. Several features of the data and publication format affect how demographic patterns can be interpreted.
For these reasons, demographic over or under-representation in long wait bands should be interpreted as an indication of where further investigation may be warranted, not as a definitive measure of inequity or performance.
Where ethnicity is not recorded in the WLMDS submission, the Secondary Uses Service (SUS) ethnicity record is used instead. This is intended to reduce systematic under-representation of patients with missing ethnicity data, which would otherwise bias equity analysis toward trusts with better recording practices rather than better outcomes.
Using SUS improves overall coverage but introduces heterogeneity, as the degree of augmentation varies by trust and is not published. For that reason, ethnicity findings are interpreted cautiously, unrecorded ethnicity rates are surfaced alongside results, and comparisons are framed in terms of consistency and direction rather than precise magnitude.
Leaving ethnicity unaugmented would produce cleaner provenance but materially weaker analytical signal for equity monitoring at national scale.