The heat map is the public face of the Frailty Emergence Probability (FEP) engine: a daily, district-level estimate of where emerging frailty is concentrating across Kent and Medway, built entirely on open data with no patient identifiers.
The heat map renders a single composite FEP score per district on a colour scale, alongside the economic cost of inaction, discharge-delay intelligence and prescribing signals. It is a population-level prioritisation instrument: it answers the commissioner's question, "of my thirteen districts, where should I look first?"
It is explicitly not a clinical tool. It does not score individuals, it does not diagnose, and it does not replace the Electronic Frailty Index (eFI) that every GP system already runs. The eFI tells you where an identified patient sits today; the heat map tells you which populations are drifting toward crisis, and how fast.
Frailty is the strongest predictor of avoidable admission, long length of stay and delayed discharge in older adults, yet it is identified almost entirely reactively, at the front door of the hospital, after the event the system most wanted to prevent. Every Primary Care Network in England is contractually required to identify and manage frailty under the Directed Enhanced Service, but no commissioned tool covers the space between appointments and ahead of presentation.
That gap is the rationale. A daily, open-data population signal lets a system act on geography before it has to act on a person in an ambulance.
Each contributing signal is normalised to a 0–100 scale within signal, against England averages as the comparator, so that a district's score expresses relative standing rather than a raw count. Normalised signals are then combined as a weighted sum and the result is itself rescaled across the thirteen districts for display. Because both the inputs and the output are relative, the map is a ranking instrument first and an absolute-risk instrument second, a distinction we return to under weaknesses.
The live engine ingests a broad signal set (the index page describes 21 signals across NHS Fingertips outcomes, NHSBSA prescribing data and ONS synthetic measures). The interactive configurator exposes a deliberately smaller, commissioner-adjustable six-signal model so that a non-technical user can see how the ranking moves as priorities change. Both are legitimate; they trade completeness against transparency.
Ecological inference. District-level signals describe places, not people. A district can score high because of a concentrated pocket of need that the average obscures, which is precisely why the Rural Access Vulnerability Index exists to drill below it.
Proxy drift. Several inputs are proxies (prescribing as a marker of clinical frailty, deprivation as an amplifier). Proxies can move for reasons unrelated to frailty, a formulary change, a coding shift, and the composite cannot tell the difference on its own.
Weight subjectivity. The default weights encode a defensible but contestable clinical prior. We expose them precisely so they can be argued with rather than trusted blindly.
Lag asymmetry. Some signals refresh daily, others monthly or annually, so the composite blends data of different vintages. The map is current to its slowest meaningful input, not its fastest.
These are the defaults exposed by the interactive configurator. They are a starting clinical prior, not a fixed truth, every weight is adjustable, and the ranking recomputes live so a commissioner can test their own assumptions.
| Signal / proxy | Default weight | Why it earns its place |
|---|---|---|
| Over-75s living alone | 25% | Strongest single demographic frailty proxy; isolation compounds every other risk. ONS Census 2021. |
| Unplanned admissions rate | 20% | Outcome signal, a district already on a crisis trajectory. NHS Fingertips. |
| Polypharmacy prevalence (5+ meds) | 20% | Well-validated clinical frailty marker available at population scale. NHSBSA dispensing. |
| Deprivation (IMD) | 15% | Access and vulnerability amplifier; modifies how quickly risk converts to crisis. MHCLG. |
| DWP Attendance Allowance | 10% | Functional-limitation indicator independent of health-service contact. DWP geographic data. |
| Care-home capacity gap | 10% | High need against low provision marks where the system has least slack. CQC register + population ratio. |
The choice to make over-75s-living-alone the heaviest single weight reflects the consistent finding that social isolation is both an independent risk and a multiplier of clinical risk; the choice to cap any single signal well below 50% reflects a deliberate refusal to let one proxy dominate the composite.