What the data shows when you look
at the sector, not the trust.
Some patterns only become visible at sector level. These findings come from the TLP NHS panel: public data across 178 to 203 NHS trusts.
Seven of England's ten ambulance services carry chronic workforce distress
the regulator has not formally escalated.
Normalised Fragility identifies trusts carrying persistently high sickness, agency dependency and intention-to-leave across multiple years, without formal NHSE intervention. In plain terms: distress that has lasted long enough that nobody treats it as a crisis anymore.
Eighteen trusts met the criterion. Seven are ambulance services: seven in ten in England. That is a sector pattern, not an outlier.
The Normalised Fragility watch list contains 18 trusts, 6.4% of the 282 English NHS trusts with enough years of data to be scored, including 7 of 10 English ambulance services. These are scored against all English NHS trusts. Ambulance services are over-represented on this measure because sickness absence and intention to leave run higher across that sector as a whole; on a within-sector comparison they are not unusual. Their agency use is in fact well below the national median.
What this means in practice
A trust on this list has carried severe workforce stress for at least two of the past three years without being placed into formal recovery. What we observe is distress that persists across years alongside no formal escalation. Why the two sit together is not something this measure can tell you: whether the chronic state has become the benchmark, or whether escalation is happening in a form the public record does not show, needs local evidence.
When seven in ten services in one sector share the pattern, it is not individual trust failure. It is a structural condition of the sector, with direct consequences for boards, commissioners, and the Department of Health and Social Care.
Validation: trusts we flag sit roughly half a Care Quality Commission rating band worse than trusts we do not, across 184 rated trusts, on the watch-list rule as applied to the scoring panel, which flags 22 to 25 organisations. The separation is stronger than any simpler rule we tested. It is an association between organisations at the same time, not a prediction, and on our own threshold it is strong rather than settled. The full methodology is available on request.
Eighteen trusts, 6.4% of the 282 trusts scored, carry multi-year workforce distress without formal escalation. Seven are ambulance services: seven of England's ten.
When distress is the routine operating state, the escalation reflex switches off. New problems get measured against a baseline that is already broken.
Ask which workforce indicators have been red for so long they no longer get discussed, and whether your escalation thresholds compare you with the sector or with your own chronic state.
A diagnostic report reads the same public measures for your organisation, benchmarked against the right peer cluster, and shows where normalisation has set in. How an engagement works.
When NHS trusts cut agency spend in 2022-23,
sickness absence rose in the six months that followed.
The panel pools three financial-year transitions, FY 2020-21 to FY 2022-23: 519 trust-year observations across 178 trusts. Where agency share fell, permanent-staff sickness absence ran higher across the following six months, April to September. The measure is a change in the mean monthly sickness rate, compared against the twelve months before, so the agency year has already closed before the sickness window opens.
The table expresses that as a pay-bill-equivalent value: what the additional absence is worth as a share of the pay bill for those six months. It is a valuation of lost time, not a measure of cash leaving the organisation, and the analysis does not establish that the reduction caused the rise.
An important caveat
The coefficient is a pooled estimate across all three transitions, not a single year, and it is not evenly spread across them. Trained on the earliest transition alone the relationship is close to zero (β = −0.001, p = 0.96): those years were dominated by pandemic-related agency increases rather than deliberate cost-reduction, and it is the first clean post-pandemic cutting cycle that carries the result.
These numbers are therefore a historical observation of a period that ended in FY 2022-23, not a forward prediction of what will happen in 2025-26. They tell a finance director what comparable NHS trusts recorded alongside a similar decision the last time it was made at scale.
What this is not. The analysis is observational: lower agency share in one financial year and higher permanent-staff sickness absence in the six months after it appear together. It does not establish that one caused the other.
And the pounds are not cash. Staff are generally paid while off sick, so their salary already sits in the pay bill. What the £ columns show is the additional absence valued at pay-bill rates. Real cash depends on how the absence is covered, by bank, agency, overtime or not at all. It is not a net saving either: a pay-bill-equivalent value cannot be subtracted from a cash saving.
Sensitivity. Measured across the full twelve months after the year end instead of six, the coefficient is −0.037 and the same illustration gives about £73,000. That version has p=0.0041, which does not meet the threshold of p < 0.001 we fixed in advance, so we report it here and never as the headline.
Across FY 2020-21 to FY 2022-23, each £100 less agency spend was associated with about £2.70 of pay-bill-equivalent absence across the following six months, on the whole panel. On a £100m annual pay bill and a two point reduction, about £54,000.
A cost improvement can look achieved while a rise sits in a budget line nobody reads next to it. This is a historical observation, not a forecast, and not a demonstrated cause.
Before the next agency-reduction target, ask what will happen to the work, and put sickness absence next to the saving in the same report rather than in a different one.
A diagnostic quantifies what comparable trusts recorded the last time this decision was made at scale, and where your own exposure concentrates. How an engagement works.
Departure Lag:
the gap your headcount does not show.
The Departure Lag is the gap between the share of staff who tell the NHS Staff Survey they intend to leave and the share who actually leave. Across English NHS trusts, around 22% report an intention to leave and around 10% go the following year.
It is a property of the organisation, not of any individual. The two rates come from different measurements on different bases, and this analysis does not link a survey answer to a leaving date. The gap is a trust-level measurement and nothing below that level was observed.
A trust's position relative to its peers largely held between 2021 and 2023 (rank correlation ρ = 0.62 across 203 trusts: in plain terms, the ranking is sticky). Across single years it is lower, 0.49 and 0.37. This says where a trust sits in the order tends to persist. It does not say the size of the gap stayed the same.
The table below shows how NHS Mental Health trusts distributed on this measure in 2023, from the narrowest gap to the widest.
The calibration that validates the measure
Two NHS trusts with publicly contrasting workforce reputations were used to test whether this measure produces doctrine-consistent results (trusts not named, in line with our policy of not naming organisations in public findings):
A trust in active national recovery intervention ran at 15.1pp in 2023. Well above the sector median.
An acute trust widely regarded as one of the best-run in the NHS ran at 4.6pp in 2023, among the most stable in the country. It is an acute trust rather than a mental health one, so it calibrates the measure and not this table; the most stable mental health trust in 2023 sat at the same 4.6pp.
The gap between them is 3.3 times. In other words, the measure places the most-distressed trust 3.3 times further from stability than the well-run trust. The two trusts' public reputations match exactly this difference. That is the kind of face validity that tells you the measure is reading something real.
A third trust, the subject of sustained national scrutiny in recent years, ran at 24.5pp: the widest gap in the 2023 Mental Health panel.
Why this matters for a workforce plan
A workforce plan is normally built from two numbers that are already settled: who is on the payroll, and who has left. The Departure Lag adds a third that is not settled, the share of staff who told the NHS Staff Survey they intend to leave. Where that share sits well above the leaving rate, the payroll and the survey are describing the same workforce differently, and a plan built on the first does not carry the second.
What produces the difference is not established here. The gap is a difference between two published rates measured on different bases. It does not identify anyone, and it does not show that any individual has decided anything.
A trust's position relative to its peers largely held between 2021 and 2023 (rank correlation ρ = 0.62 across 203 trusts: the ranking is sticky, not one-year noise; across single years it is lower, 0.49 and 0.37). The 2023 Mental Health median was 11.3pp.
The payroll and the staff survey describe the same workforce and can point to different places. A plan built on headcount and turnover does not carry the survey signal at all.
Ask how large your organisation's gap is and which way it has moved, then ask what the workforce plan assumes about it.
A diagnostic places your organisation on this distribution against the right peer group, from the most stable decile to the most exposed. How an engagement works.
Note: the Departure Lag is a structural diagnostic, not a forecast. It describes a difference between two published rates in a given year, not what will happen next. It does not predict future agency spend or turnover rates, and it does not establish why the difference exists.
The Normalised Fragility watch list
Eighteen NHS trusts, 6.4% of the 282 trusts scored, show persistent multi-year workforce distress without corresponding regulatory escalation: flagged in two or more of three consecutive survey years. When distress becomes the routine operating state, the system stops escalating it. These are the trusts where it has.
Sector mix of the eighteen
- 7 of the 10 English ambulance services
- 5 acute trusts
- 5 mental health trusts
- 1 community trust
The ambulance signal is the headline: seven of ten English ambulance services on a single eighteen-trust list, 70% of a sector against 6.4% overall prevalence. Independently verifiable against published NHSE and CQC data.
The Normalised Fragility watch list contains 18 trusts, 6.4% of the 282 English NHS trusts with enough years of data to be scored, including 7 of 10 English ambulance services. These are scored against all English NHS trusts. Ambulance services are over-represented on this measure because sickness absence and intention to leave run higher across that sector as a whole; on a within-sector comparison they are not unusual. Their agency use is in fact well below the national median.
The watch list is a descriptive, structural diagnostic of current conditions, not a forecast. It is built from the NHS Staff Survey and workforce data across a four-year panel, 2021 to 2024, covering 178 to 203 NHS trusts depending on the measure.
Eighteen trusts flagged in two or more of three consecutive survey years. One sector, ambulance, supplies 70% of its services to the list against a 6.4% overall prevalence.
This is the system-level view: where chronic distress has stopped generating escalation, board assurance and oversight are reading a gauge that has been stuck for years.
Boards and commissioners should ask whether any organisation they oversee fits this pattern, and what evidence would surface it if it did.
The watch-list method is reproducible from public data. A briefing walks a board or ICB through the criteria, the flags and what they mean for oversight. Book a discovery call.
The evidence standard.
Every validated claim on this site comes from a public-data panel of 178 to 203 NHS trusts, 2021 to 2024 depending on the measure: the NHS Staff Survey, NHS workforce statistics and published provider accounts. Every measure was tested against pre-specified pass or fail gates, meaning the pass mark was written down before the tests were run. Results are published whether they passed or failed.
What passed
Pressure Displacement. β = -0.054, p = 0.0003, N = 519 trust-year observations. In plain terms: a small but statistically solid relationship between agency reduction and pay-bill-equivalent permanent-staff sickness absence in the observed period. Not cash, not a net saving, not a cause. Validated as a structural diagnostic of that period, not a forecast.
Departure Lag. Sector median gap of 11.5 percentage points across 203 NHS trusts, with a rank correlation of ρ = 0.62 across 203 trusts, 2021 to 2023. In plain terms: an organisation's position in the order tends to persist rather than move at random. Measured across single years the correlation is lower, 0.49 and 0.37.
Normalised Fragility. Eighteen trusts, 6.4% of the 282 trusts scored, flagged in two or more of three consecutive survey years. Externally corroborated: across 184 rated trusts, on the watch-list rule as applied to the scoring panel, which flags 22 to 25 organisations, those it flags sit roughly half a Care Quality Commission rating band worse than those it does not, and the same pattern appears against NHS England oversight. The comparison is between organisations at the same time, not over time.
What failed, published rather than buried
Capability Collapse by Cluster. The measurable version, a capability concentration risk score, failed its pre-specified gate: the clean effect size came in at Cohen's d = -0.22 against a pre-specified threshold of -0.30. In plain terms, the effect pointed the right way but was too small to pass the bar we set in advance. The construct stays in the doctrine library as a named mechanism; it does not ship as a validated model.
Forecasting. A forecasting version of the indicators was tested and did not beat a simple benchmark, last year's agency spend. That is why nothing on this site makes a forecasting claim: the validated diagnostics describe present structure, and are sold as exactly that.
These are the two headline failures. Claims that did not survive re-testing have been retracted and are documented in the methodology record. The full methodology, including the dated pre-specification records and the retraction log, is available on request, and the test-by-test register is available under NDA.