Soigneur FHIR Implementation Guide
0.1.0 - ci-build
Soigneur FHIR Implementation Guide - Local Development build (v0.1.0) built by the FHIR (HL7® FHIR® Standard) Build Tools. See the Directory of published versions
This page walks the FHIR R4 data model resource-by-resource. It graduates from
docs/plan-maf-azure/04-data-model.md (how resources reference each other) and
05-terminology.md (the code systems). The Artifacts index has the
generated StructureDefinitions; this page explains how they fit together.
Everything is FHIR R4 on Fire Arrow (ADR 0001). The athlete owns their record; the coach
never writes Patient (04 §15.1). All athlete-facing metric labels are trademark-safe
(Load / Weighted Power / Intensity / Fitness / Fatigue / Form); the cc-* codes are
internal machine keys.
gender and
birthDate carry the sex and age demographic axes; a timezone extension drives
athlete-local scheduling. At least one provider identifier (the cc-provider system)
ties the record to imported provider data.73985-4), not a Procedure (ADR 0002). Per-metric values hang off it via
hasMember, the raw file via derivedFrom → Binary, and adherence via basedOn →
the CarePlan / RequestGroup it was prescribed under.core.metrics), so a summary can never disagree with its stream.method is required (cc-inv-ftp-method) so an accepted/set value is
distinguishable from an estimate.Composition split into an
immutable Core and an editable
Live layer (ADR 0003).intent = plan) with strategy / scheduling-mode / coach-managed
extensions. A phase must not exceed 18 weeks (cc-inv-plan-max-18wk; warn > 12).DiagnosticReport (ADR 0009).Basic (ADR 0021). These are eval-realism tunables, not numeric coaching
thresholds (which live in the corpus / frame, 13 §2.5).The simulator (below) exercises two sex-asymmetric mechanisms the coach must reason about:
24373-3) — the marker for the hidden iron-deficiency sub-phenotype: a
plateau/fatigue that mimics overreach but needs iron screening, not rest.Imported provider data carries a provider Provenance (download/translate, 04 13/13/Device. Skills that reason about data provenance and
sufficiency (cold-start vs veteran) can therefore be tested honestly.
The IG's example Instances are lifted from SoigneurSynth, the physiological athlete simulator (ADR 0021), so every profile is exercised against real generated data. Two phenotype exemplars are machine-checked here:
synth-iron-01: FTP stalls while ferritin falls
and overreach debt stays low.synth-overreach-01: observable Fitness keeps rising
while true FTP has already peaked and turned down — the trap the coach must catch.These exemplars double as eval fixtures (08 §H); the full generated cohort is published as a separate versioned artifact.