Transmission dynamics
SEIR, agent-based, and metapopulation models for outbreak response and scenario planning.
Most tools tell you where coverage is falling and stop there. Medilytics goes further: it pulls your data together, models what happens next, and then hands you a ranked, costed shortlist of the moves worth making — the interventions that protect the most people for the least money, worked out across every programme. Less dashboard to stare at, more decision already made.
It's a working prototype. The figures in it are illustrative, built from published NHS and UKHSA data, so you can see how the platform behaves without any real patient records involved. Connecting it to live NHS data and validating the models is the work still ahead of us.
Developed in collaboration with top epidemiologists
Plenty of tools will show you a falling line. The harder question, and the one that costs money to get wrong, is what to do about it. That's what Medilytics is built to answer.
Run an intensive school and GP catch-up across the lowest-coverage London boroughs, lifting second-dose coverage to the 95% threshold.
Every other flagged programme is ranked the same way, by the risk it carries and the return it offers, so the top of the list is always the next move worth making.
Every relevant source, reconciled into one picture of where each programme stands.
Projects how coverage, risk and cost play out if nothing changes, and if you act.
Turns all of that into a shortlist of costed actions, ordered by the ones worth doing first.
A model is only worth as much as the decision it helps you make. Here's what ours are built to do.
The numbers you need are usually scattered across half a dozen systems: ImmForm, HES, ONS, local records. Medilytics brings them into one place, reconciled and ready to use, so nobody is stitching spreadsheets together before the real work starts.
Forecast vaccine demand by cohort and region before the season starts, so the order you place is one you can stand behind rather than a round number.
Project transmission weeks ahead, with the uncertainty made explicit, so you can plan for the curve before it bends.
Model immunity and waning over time, so timing and boosters are set by evidence rather than the calendar.
Analysis prepared to stand up to scrutiny — reproducible, documented, and ready for peer or regulatory review.
See where coverage is slipping, then see what to do about it. The platform ranks the interventions worth making and shows the likely cost and impact of each, across every programme.
Put a pound figure on a coverage gap: the admissions and cases you would avoid, and what you get back for what you spend. Every assumption is yours to change and question.
We start with vaccines, but the same engine applies just as well across public health, pharma and payer decisions.
SEIR, agent-based, and metapopulation models for outbreak response and scenario planning.
Forecast demand and coverage across age cohorts, regions, and rollout schedules.
Model how protection rises and decays over time from trial and real-world data.
Survival, efficacy, and subgroup analysis prepared to stand up in front of a regulator.
Turn routine and registry health data into population-level signal you can act on.
Compare interventions side by side, with costs and uncertainty, before committing budget.
The strongest forecasts rarely come from one model. They come from running several and comparing them. We bring the best together and apply them to the decision you're actually facing.
We run a family of models — our own and the best from research and public health — and show where they agree, and where they don't.
Models run inside your secure environment, on your own data. Only aggregate results leave — aligned to UK GDPR and information governance.
Developed in collaboration with top epidemiologists, bringing rigorous, defensible methods to the decisions regional teams actually face.
We start from the decision the model has to feed — not the data we happen to have.
Ingest, clean, and document every source, so the inputs are traceable later.
Fit, calibrate, and stress-test against held-out data and alternative structures.
Quantify uncertainty and show where the model breaks — not only where it works.
Reproducible code, documented assumptions, and a readout your board can actually read.
Most models are black boxes. Ours are built to be picked apart, because a number nobody can question is a number nobody should rely on.
Every assumption is documented and handed over with the model — nothing hidden.
Every projection ships with its confidence band. You see the range, not just a point.
Code and data lineage travel with the result, so anyone can re-run it.
Tested against held-out data and prior seasons before anyone relies on it.
We model only on anonymised data supplied by health bodies — identifiers never reach us. Aligned to UK GDPR and information governance.
The clearest view of what happens next.
We want to be the modelling partner people turn to when the answer has to be right, and has to survive someone checking it.
We're building Medilytics with a small group of founding partners — health teams with a decision to make, and modellers who want their work put to use. Tell us which you are, and what you're trying to answer.