Why a computable version is necessary

If the PDF works for clinicians, why build a computable schedule at all? Because the hard questions are not about reading the schedule - they are about evaluating a patient against it, and that is something only a machine can do at scale, correctly, and repeatably.

The schedule is not one document - it is a series of versions

The NHS childhood schedule has changed substantially and repeatedly:

A child born in 1998 has a genuinely different "complete" schedule than one born in 2010 or 2020. Any system that applies today's schedule to all patients will over- or under-flag.

A failure mode that exists in the wild

This is not hypothetical. SystmOne Online's patient-facing childhood vaccination view has been observed presenting a naive grid of the current national routine schedule projected onto a patient's historical record, with due dates filled across antigens and age bands that were not all part of the schedule for that birth cohort. It is an understandable implementation shortcut - a static current-schedule table is much easier to build than a valid-time evaluator - but it is exactly the failure mode greenbook is designed to avoid.

The hard part is not drawing a table. The hard part is answering, for a real person and a real evaluation date: which schedule version applied, which products existed then, which doses were due by then, and whether the recorded history conforms to that historical schedule.

This is a valid-time problem

The schedule is not a single document but a series of versioned snapshots, each with a valid_from date. Evaluating a patient born in 2003 requires the schedule as it stood in 2003, not today's schedule.

greenbook models this directly: one file per schedule version, dated by its valid_from, plus evaluate-auto, which builds a patient-specific effective schedule by selecting each expected dose from the version in force when that dose first became due. The first curated historical slices are now in rules/, with more Green Book revisions being added incrementally.

The vision: invert the pipeline

Today the data flows one way - from a Word document, to a PDF, to hand-written code. What if the computable schedule were the source of truth, and everything else were generated from it?

Experts author the computable Schedule
        │
        ├──→  the published PDF / age-centric table   (rendered)
        ├──→  GOV.UK and other websites               (rendered)
        └──→  clinical decision support in every system (one trusted source)

Vaccination and public-health experts would author schedule changes directly in a computable format. Every downstream publication - PDFs, websites, clinical tools - would be rendered from the same trusted data, instead of each vendor reverse-engineering a PDF and quietly diverging.

That is the long-term goal. The rest of this site shows the domain model that makes it possible, and the engine that proves it can be evaluated.