Population health management · South East Wolverhampton

The neighbourhood already has the data.
It has never had the denominator.

A population health management proposition for the ICB: what South East Wolverhampton would measure, why those measures rather than others, and what the ICB would receive — as aggregate figures with no patient data leaving the neighbourhood.

Why this neighbourhood, why connected

Everything below is already funded.
None of it is counted together.

Wolverhampton does not have a shortage of services. It has ten services around the same person, four life stages with different failure modes, and no shared denominator across any of them. That is a measurement problem before it is a care problem — and it is the reason improvement has been hard to demonstrate here.

60.0

Healthy life expectancy, men, against 63.1 in England. Health is lost inside working life, not after it.

7.8

Years of life expectancy separating the most and least deprived Wolverhampton wards, for men

2,431

Alcohol-related admissions per 100,000, against 1,705 nationally

60.2%

Of the city aged 18–64 — the cohort with the least routine contact of any

47%

Of residents said more money would most improve their wellbeing — the top answer by a distance

£26

Returned over ten years for every £1 spent on alcohol treatment

The four failures, by life stage

StageWhat actually failsWhere the signal already sitsWho currently holds it
Infants 0–2Missed universal contacts generate letters instead of people; infant mortality runs at 5.2 per 1,000 against roughly 3.9 in EnglandThe schedule itself — new birth visit, 6–8 week check, immunisation course, milestone reviewsHealth visiting, general practice and child health information systems — rarely reconciled
Children 0–18Uncontrolled long-term conditions read as attendance problemsSchool attendance, non-collection of medicationSchools and pharmacy — neither read by a clinician
Young people 16–25Discharge from a children's service is not admission to an adult oneThe gap between two service recordsNobody. This number does not exist in the city.
Working adults 25–64No routine contact at all until crisis presentationA&E, alcohol admissions, pharmacy, job centres, care home visitsEach service separately, none linked
Frailty 65+Six records, no synthesis, contradictions unnoticedGP, care home, home care, pharmacy, hospital, social careAll six. Never together.

The measurement failure underneath all four

Wellbeing coverage is far higher in care homes than in people's own homes. What is currently monitored therefore describes the people easiest to reach rather than the population — and anyone with no service contact is invisible by definition. Every improvement figure the system currently produces is calculated on a denominator that excludes the people most likely to be failed.

Skeina's first contribution to population health management is not an intervention. It is an honest denominator.

The logic model

From a connected record
to a changed outcome

Set out explicitly so it can be argued with. Each link is a claim, and each claim has a measure attached in the KPI framework below. Where a link is unproven, it is marked as such rather than assumed.

Where this logic model is honest and most are not

The two right-hand columns are weaker than the two on the left, and we say so. Outputs are directly attributable to Skeina. Outcomes are plausibly influenced. Impact is not attributable at this scale within this timeframe, and any proposal claiming otherwise is inviting a disappointment it has scheduled for itself.

We therefore propose to be held to the output and process tiers, with outcome measures reported but not contracted against in the first two years.

What we propose to be measured on

Four tiers, and only two
we would contract against

Tiers 1 and 2 are within the neighbourhood team's control and available monthly. Tier 3 is influenced but confounded. Tier 4 is the reason for doing it and cannot be attributed.

TierMeasureWhy it earns its placeSourceCadence
Tier 1 · coverageWellbeing coverage by setting and strandThe honest denominator. Everything else is conditional on it, and it will start poor.SkeinaMonthly
Tier 1 · coveragePeople with no routine service contact identifiedThe invisible cohort, counted for the first timeSkeinaMonthly
Tier 1 · coverageEquity breakdown by ward and deprivation decileDesigned into the first data model, not retrofittedSkeinaMonthly
Tier 2 · processReferral loop closure rate, by partnerWhether coordination produced anything. Nobody in the city currently holds this number.Skeina and WIN-listed partnersMonthly
Tier 2 · processTime from need identified to need metThe lived experience of coordination, in daysSkeinaMonthly
Tier 2 · processTransitions where an adult service accepted before the children's service closedThe 16–25 cliff, made visible. Expected to be poor.Skeina and partnersQuarterly
Tier 2 · process6–8 week check completion, and days overdue at completionThe first universal contact after birth, covering baby and mother — and a direct infant-mortality-adjacent measureGP record and CHISMonthly
Tier 2 · processPrimary immunisation completion at 12 months, by wardOverdue in the first year means unprotected in the highest-risk months; MMR uptake is below the WHO targetCHISMonthly
Tier 2 · processContradictions surfaced per 1,000 recordsDirect safety yield, and the case for the synthesis itselfSkeina auditMonthly
Tier 2 · processCoordinator send-back and flag rateWhether the engine is being checked at all. A fall to zero is a warning.Skeina auditMonthly
Tier 3 · outcomeMean anticholinergic burden on the frailty registerDirectly modifiable, auditable, defensible to CQCGP recordQuarterly
Tier 2 · processIncome secured for residents (£/yr, partner-confirmed)The one hard-currency line: real pounds into households in the most deprived wards, verified per case — against £24bn unclaimed nationallyAdvice partnersMonthly
Tier 3 · outcomeEntitlement identified and claimedResidents named money as the top wellbeing factor, at 47%Advice partnersQuarterly
Tier 3 · outcomeAlcohol treatment engagement after admissionThe city's worst indicator and the best evidenced returnRecovery Near You and SUSQuarterly
Tier 3 · outcomeSensory review within 12 months, frailty registerMost reversible, least connected issue in frail populationsGP record and optometryQuarterly
Tier 4 · impactEmergency admissions and conveyances per 1,000Reported for context. Not attributable and not contracted against.SUS · WMASAnnual
Tier 4 · impactHealthy life expectancy gap between wardsThe reason for doing any of this. Moves over a decade, not a contract cycle.OHIDAnnual

What we are asking the ICB to agree, specifically

  • Contract against tiers 1 and 2 only for the first two years, with tiers 3 and 4 reported openly alongside
  • No admission-avoidance or savings target in year one. Setting one before a baseline exists invites the number to be managed rather than achieved
  • Agree the indicator set before we start, with partners in the room — the Gesundes Kinzigtal approach, where 126 indicators were agreed through a Delphi process involving patients, providers and payers
  • Enrol highest need first, deliberately and on the record, so cherry-picking is structurally impossible rather than merely denied
Has this been done anywhere?

Every component has precedent somewhere.
The combination is the hypothesis.

This claim is checkable, so here is the check — the closest systems in the world, what each proved, and what each is missing. The honest reading: the pieces are proven separately, the assembly is the innovation, and one famous null result tells us exactly what not to promise.

SystemScaleWhat it provedWhat it lacks
Whole Systems Integrated Care
North West London
2.4 millionLinked primary, community, mental health, acute and social care data with dashboards for proactive case-finding is buildable and governable in the NHSCommissioner-facing at ICS scale. Identifies cohorts; the action on any one person, and whether anything happened next, sits outside it
Connected Bradford800,000Health, education, social care and environment data can be linked across forty years — the education link Skeina's children's strand needs is precedentedA research and planning accelerator, not an operational tool a neighbourhood team works from daily
Compassionate Frome
Somerset
28,510 — one practiceThe only intervention shown to cut emergency admissions across a whole population: down 14% over 44 months while the rest of Somerset rose 28.5%. Broad identification — “anyone giving cause for concern” — plus care planning plus community connection, using a service directory and connectorsPeople and paper, one practice, one heroic team. No instrumentation: coverage, loop closure and equity were never measured, so nobody can say which part worked or replicate it reliably
Camden Coalition
New Jersey
800 patients, RCTThe cautionary tale: intensive multidisciplinary coordination for the highest-cost patients, tested properly — readmissions 62.3% against 61.7% in usual care. No effectTargeted the top of the risk pyramid and measured admissions. Both choices failed, and Skeina deliberately makes neither
Gesundes Kinzigtal
Germany
~60,000All-age, whole-population integration at almost exactly SE Wolverhampton's size, sustained for two decades; enrolled highest-morbidity first; evaluated against matched controls on 126 co-designed indicatorsBuilt on insurer shared-savings contracts with no NHS equivalent, and no VCS or education layer
Nuka · Alaska~65,000Community ownership of the system itself — recipients became owners — identified as the key to two decades of resultsA full care system with its own workforce; the ownership principle transfers, the model does not
Findhelp · USANational — 550,000+ programmesClosed-loop social referral at scale: screen, refer, track, close, report outcomes — embedded in hospital records, used by 600+ systems and whole US statesVendor-owned, US social-care market, no clinical synthesis and no neighbourhood denominator — the instrumentation without the mechanism
Omaolo · FinlandNationalAssessment with routing done properly: a CE-marked class IIa medical device, state manufacturer, formal clinical validation — covering health and social-service assessmentsNames the price of the assessment layer. We adopt the standard, not the layer — deliberately, until Stage 3
Foundry BC · CanadaProvincial — 12–24sThe 16–25 model: health, mental health, substance use, peer and social support with no referral and no wrong door; 35 centres by 2027 plus a virtual armBuildings-first and youth-only; the principle transfers to our Strand B and family hubs, the estate does not
Social prescribing platforms
UK, various
Practice-levelReferral-outcome tracking into the VCS is commercially provenOne referral type, prescription-shaped, clinician-initiated. No synthesis, no rules, no life-course view, no population denominator

What Frome and Camden prove together

Camden aimed intensive coordination at the costliest patients and measured admissions: null. Frome worked the whole population with broad identification and community connection, and is the one documented admissions success. The lesson is architectural: go population-wide, connect people to community capacity, and do not stake the case on admissions. Skeina is, in one sentence, Compassionate Frome with instrumentation — the mechanism that worked, made measurable, repeatable and ownable by the neighbourhood rather than dependent on one exceptional team.

So the USP, in four checkable claims

ClaimWho else does itWhy it matters to a commissioner
1 · The unit of account is a closed loop, not a risk score — every referral, including into the VCS by WIN identifier, chased to a recorded outcome, published by partnerSocial prescribing tools do it for one referral type. Nobody does it cross-sector against a population denominatorIt is the difference between “we identified 400 people” and “173 needs met” — the number a commissioner can actually buy
2 · Deterministic, published rules — no model of any kind — 52 rules, each with a named owner, readable by a DPO in a sittingEveryone else risk-stratifies with proprietary or statistical modelsDeployment speed and cost: no medical-device pathway, a light DPIA, and an IG conversation measured in weeks not years
3 · Life-course strands on one register — infants’ universal contacts, school attendance, the 16–25 cliff, working-age outreach and frailty in one architectureEverything comparable is single-cohort (frailty, SEND, super-utilisers) or all-age but data-only (Bradford)One coordinator function and one governance serve four cohorts; the marginal strand costs rules, not a new platform
4 · Neighbourhood-owned, asset-locked — a CIC held by the providers and charities it servesNuka, in Alaska. No UK coordination platformPartners share data with infrastructure they co-own in a way they never do with a vendor — ownership is the data-sharing strategy
The realistic path

Deliberately boring technology,
in five funded steps

Nothing below requires a platform procurement, a data centre or a national integration programme. The first year runs on searches the practices can already do, records the care homes already keep, and a public read-only feed of WIN.

StepWhatData neededNew IG requiredFunded how
0 · now–12 wksRetrospective Phase 0: one care home, 30 residents, the frailty rules run against last yearGP record + consented care home recordsNone new — direct care basis existsIn-kind, H&B
1 · months 3–9Prospective frailty + infants strands; the dashboard's first real monthly return; alcohol rule live via discharge listsPractice searches, CHIS reports practices already receive, WIN public feedCare home DSA; WIN feed agreement with the councilPCN capacity monies + ICB place innovation
2 · months 6–12Family hub front door for 0–18 and 16–25; Project SEEN becomes Strand A's delivery armHub programme data; consented per familyHub operating procedure + consent modelFamily hub / Start for Life programme
3 · months 9–18The education link: attendance for consented children with LTCs — the long pole, started earlySchool attendance, per consented childArticle 26 arrangement with the LA as education authority — Bradford proves it is doableCouncil public health / DPH priorities
4 · months 12–24Independent evaluation against the co-designed indicator set; decision to scale, change or stopThe audit log itselfEvaluation agreementNIHR (ARC / HDRC route) or charitable

What it costs, honestly

Year one is a team, not a technology: one full-time care coordinator, a fractional clinical lead, a fractional analyst, and the rule engine — which is a website and a set of practice searches. A low six-figure annual cost, in the range of a single band 8a post plus sessions, against which the first deliverable is a number the system has never had: how many people the neighbourhood found, and how many needs it met.

And what we will not promise

An admissions reduction. The one rigorous trial of intensive coordination found none, and the one population that achieved one did it over four years with mechanisms we are only now instrumenting. Admissions are reported in tier 4, openly, uncontracted. If they move, the independent evaluation — not us — gets to say so.

What the ICB receives, monthly

The return is a working screen,
not a description of one

Aggregate only — counts, proportions and medians at neighbourhood and ward level. Cells with a denominator below 10 or a numerator below 5 are suppressed; ward rows are rounded to the nearest 5; no line-level data exists in the return and no route back to an individual is possible. Every figure is produced from the rule engine's own audit log and is reconstructable against the register version stamped on the page.

Open the dashboard

Eight headline indicators with month-on-month movement and trajectory plans; coverage by cohort; loop closure by partner with breach counts; the infants' universal-contact schedule with the number of missed contacts recovered by a person rather than a letter; ward-level equity with the gap-to-best column; the engine-assurance panel; and the found-people register — the 420 people in six months who appeared in no denominator anywhere. All figures on the demonstration are synthetic and labelled as such.

Why aggregate is enough, and more would not be lawful

Performance-managing a neighbourhood team needs denominators, trends and comparisons — all aggregate. Line-level flow would require a separate legal basis and would add nothing the ICB needs for this purpose. Settling that now, in the data-sharing agreement, is the single most common failure point this project can remove in advance.

The spine already exists

WIN is the directory.
Family hubs are the front door.

Two decisions that stop Skeina duplicating what the city already runs.

The Wolverhampton Information Network

WIN is the council's own directory, developed with service users, their families and carers, covering social care and support services, children's services, the SEND Local Offer, social and leisure activities, and advice on employment, benefits, keeping healthy and getting about. It already holds what a neighbourhood coordination system would otherwise try to rebuild.

WIN does thisSkeina does this
Holds the listYes — maintained by the council, kept current by providersNever. A second directory is out of date within a year.
Says what existsYes, for the whole cityNo
Knows a person needs itNo — the person has to lookYes, from the record, by published rule
Sends a minimised referralNoYes — need, setting, access, consent. No diagnosis, no narrative.
Knows whether anything happenedNoYes. This is the part nobody currently does.

The integration ask is small

Skeina needs WIN's service records as structured references so a referral can name a real listed service rather than a free-text guess, and so loop closure can be reported back against the same identifier. That is a read-only feed of a public directory — the lightest possible integration, with no personal data involved in either direction.

Family hubs

Family hubs are the only universal front door that working-age adults, children and young people all pass through without being ill. That makes them the natural host for the two strands where clinical contact is rarest — and it is why Project SEEN already sits there rather than in general practice.

StrandFamily hub roleWhat Skeina adds
Children 0–18Early help, parenting, SEND Local Offer, the existing Project SEEN relationshipsAttendance and LTC signals the hub cannot see, routed with consent
Young people 16–25Young parents, care leavers, the point where children's services endThe open-transition register and its named owner
Working adults 25–64The parent at the school gate who has no other contact with the systemOpportunistic Health Check, entitlement and alcohol offers at a non-clinical contact

What family hubs would need in return

Hub staff are not clinicians and must never be placed in a position of clinical judgement. The decision-rights model already handles this — coordinators route by category, clinicians sort by acuity — but it has to be written into the hub's operating procedure rather than assumed, and the escalation route from a hub must reach a duty clinician the same session, not a task queue.

What we are asking the ICB for

Agreement on the measures,
before we measure anything

Declaration of interest

Health & Beyond is both the proposed developer of Skeina and the proposed pilot site. Skeina is being constituted as a community interest company held jointly by Health & Beyond Ltd, local charities and other neighbourhood providers, so that an asset lock — not our good intentions — prevents the neighbourhood's coordination layer becoming any single provider's property. Any evaluation must be independently conducted, and we would expect that to be a condition of support rather than an offer from us.