License: Apache-2.0 TypeScript Tests CI
Open-source HEOR (Health Economics and Outcomes Research) and medical device market-access skills for AI agents. Nine skills covering the full reimbursement journey — from pathway selection to application drafting — backed by a deterministic economic modeling engine and six official reimbursement databases across Switzerland, Germany, France, and the USA.
Works with Claude Code, Codex, Cursor, Gemini CLI, and any Agent-Skills-compatible harness. Skills are harness-neutral: no reliance on Claude-specific tools in core workflows.
Status: pre-release. Not yet published to any marketplace. The engine and data pipeline are production-ready; skill workflows are being refined.
graph LR
A["dossier.yaml<br/>+ documents/"] --> B["regulation-navigator<br/>which pathway?"]
A --> C["tariff-scout<br/>existing code?"]
C --> B
B --> D["prisma-review<br/>literature search"]
D --> E["economic-modeling<br/>CEA, BIA, PSA"]
E --> F["ch-migel-application<br/>or eurhta-report"]
F --> G["hta-quality-check<br/>score + consistency"]
A dossier is a plain directory. Each skill reads from it, writes to it, and
picks up where the previous one left off. Start with just a dossier.yaml
describing your product, and work through the pipeline.
| Skill | What it does |
|---|---|
| regulation-navigator | Maps a product to its reimbursement pathway across CH, DE, FR, US, UK with prerequisites and timelines |
| tariff-scout | Searches official reimbursement lists for existing codes and comparable products |
| prisma-review | Runs a PRISMA 2020 systematic literature review: PICO, search, screening, diagram, synthesis |
| economic-modeling | Builds and runs cost-effectiveness and budget-impact models through a deterministic engine |
| ch-migel-application | Drafts a Swiss MiGeL (device list) application form |
| ch-analysenliste-application | Drafts a Swiss Analysenliste (laboratory test) application form |
| ch-klv-forms | Drafts Swiss KLV forms: Meldung, Antrag, Umstrittenheit |
| eurhta-report | Drafts a EUnetHTA Core Model HTA report chapter by chapter |
| hta-quality-check | Scores a draft against BAG rubrics and checks cross-section consistency |
The @heor/engine package is a pure-TypeScript economic modeling engine with
zero runtime dependencies. Six model types (decision tree, Markov chain,
budget impact, state transition, partitioned survival, discrete event
simulation), Monte Carlo PSA with CEAC and tornado diagrams, batch scenario
comparison, Excel export with live formulas, and structured run comparison.
npx tsx packages/heor-engine/src/cli/run-model.ts examples/markov-chain.json npx tsx packages/heor-engine/src/cli/run-psa.ts examples/psa-markov-chain.json npx tsx packages/heor-engine/src/cli/export-excel.ts models/runs/baseline.json
Six official reimbursement lists are normalized to versioned JSON and refreshed monthly by CI:
| Country | List | Source |
|---|---|---|
| Switzerland | MiGeL (devices and aids) | BAG/FOPH |
| Switzerland | Analysenliste (laboratory tests) | BAG/FOPH |
| Germany | DiGA (digital health apps) | BfArM FHIR API |
| Germany | HMV (assistive devices) | GKV-Spitzenverband API |
| France | LPP (medical products and services) | CNAM / ameli |
| USA | HCPCS Level II (procedure codes) | CMS |
git clone <this-repo> cd heor-skills pnpm install # Run the engine tests pnpm -r test # Explore the demo dossier ls examples/demo-dossier/
Install as a Claude Code plugin (once published):
/plugin marketplace add <owner>/heor-skills
/plugin install heor@heor-skills
plugins/heor/skills/ 9 skills, each with SKILL.md + references
packages/heor-engine/ economic modeling engine (npm-publishable)
data/{ch,de,fr,us}/ normalized reimbursement data (JSON, versioned)
tools/data-pipeline/ fetch + normalize scripts (run by CI cron)
examples/demo-dossier/ fully populated example dossier (SereniCBT)
docs/cookbook/ recipe guides for common workflows
- Country first, language second. Data and templates are organized by jurisdiction. Language is a variant within a jurisdiction, not the other way around.
- The engine does the math. Economic model results come from
@heor/engineCLI runs, never from the LLM. - Drafts, not submissions. Every skill ends with a human-review disclaimer. Output is a starting point for a qualified professional.
- No servers, no databases. Markdown, JSON, and one TypeScript engine. The data pipeline runs in CI and opens a PR when official sources change.
Apache-2.0. Drafts produced with these skills require review by qualified professionals. Nothing here is regulatory, clinical, or reimbursement advice.
- HEOR professionals building market-access dossiers for medical devices and digital health interventions across European and US markets
- AI agent developers who need deterministic, auditable economic modeling rather than LLM-generated arithmetic
- Regulatory affairs teams navigating Swiss (MiGeL, Analysenliste, KLV), German (DiGA, HMV), French (LPP), or US (HCPCS) reimbursement pathways
- Health technology assessment researchers running PRISMA literature reviews and cost-effectiveness analyses
HEOR, health economics, outcomes research, HTA, health technology assessment, market access, medical devices, digital health, reimbursement, cost-effectiveness analysis, CEA, budget impact analysis, BIA, Markov model, decision tree, partitioned survival, discrete event simulation, probabilistic sensitivity analysis, PSA, ICER, QALY, PRISMA, systematic literature review, PICO, EUnetHTA, MiGeL, Analysenliste, KLV, DiGA, HMV, LPP, HCPCS, Claude Code skills, AI agent skills, Switzerland, Germany, France, USA