GTPCNZ

CI Quarto Pages

GTPCNZ is a public policy-research and translation package on primary care funding architecture in Aotearoa New Zealand and Australia.

Repository: https://github.com/edithatogo/gtpcnz

Website: https://edithatogo.github.io/gtpcnz/

Substack series: https://rareinsights.substack.com/

What This Is

  • A public-data anchored benchmark for exploring primary care funding architecture.
  • A Quarto website and reproducible report.
  • A Plotly Dash interactive model lab for Hugging Face Spaces, with Streamlit retained only as a legacy compatibility surface during migration.
  • A public audit trail of assumptions, caveats, and launch materials.

What This Is Not

  • It is not an endorsed policy position.
  • It is a public-data anchored benchmark, not a linked-data calibrated or patient-level forecast.
  • In repo-health terms, the aggregate validation lane is public_aggregate_validated / empirically_supported_if_gated, and remains bounded by the documented claim boundaries.
  • It should not be used to claim precise fiscal savings, ED reductions, hospital-demand reductions, workforce effects, implementation impacts, or causal effects.
  • Full caveat: This is a public-data anchored benchmark and educational explainer. It is not linked-data calibrated and not a patient-level forecast. It should not be used to claim precise fiscal savings, ED reductions, hospital-demand reductions, workforce effects, implementation impacts, or causal effects.

Design Option Explored Here

The design option explored here retains capitation for continuity, enrolment, baseline viability, and population responsibility. It also tests an uncapped, scheduled, rules-based fee-for-service stream for eligible primary medical activity, controlled through item rules, provider scope, clinical governance, documentation, audit, co-payment protections, and place-based accountability.

Short version:

Uncapped does not mean uncontrolled.

Public Surfaces

Run Locally

Install the repo-local Prefix.dev Pixi runtime:

python scripts/bootstrap_prefix_pixi.py

Run tests:

python scripts/run_pixi.py run -e dev pytest -q

Check the auditable repo-health score:

python scripts/run_pixi.py run -e dev python scripts/check_repo_health.py

Render the Quarto website:

quarto render --to html

Run the Dash model lab:

python scripts/run_pixi.py run dash

Run the Dash migration and runtime gates:

python scripts/run_pixi.py run -e dev test-dash
python scripts/run_pixi.py run -e dev test-runtime
python scripts/run_pixi.py run -e dev test-public-gates

Run the legacy Streamlit compatibility surface only when checking migration parity:

python scripts/run_pixi.py run -e dev streamlit run streamlit_app.py

Deploy

GitHub Pages is deployed from .github/workflows/pages.yml and remains the public front door.

The Hugging Face Dash Space is deployed from .github/workflows/huggingface-space.yml to:

  • Space: edithatogo/gtpcnz-dashboard
  • URL: https://edithatogo-gtpcnz-dashboard.hf.space/
  • Runtime: Prefix.dev Pixi via Docker

Streamlit Community Cloud is no longer the canonical interactive target. The legacy compatibility deployment remains:

  • Repository: edithatogo/gtpcnz
  • Branch: main
  • Entrypoint: streamlit_app.py

If GitHub Actions jobs fail before a runner starts, use docs/operations/github-actions-recovery-v1.8.3.md before changing workflow code.

License And Citation

See LICENSE for the mixed code/content licensing terms and CITATION.cff for citation metadata.

Public calibrated-model programme

The current public runtime remains a public-data anchored benchmark and educational explainer. The public aggregate lane now reports public_aggregate_validated / empirically_supported_if_gated for registered aggregate gates only. The programme separates public registries, source snapshots, calibration targets, structural uncertainty, VOI, cockpit, and release gates from future/private templates. Unsupported precision and causal claims remain out of bounds: precise fiscal savings, ED reductions, hospital-demand reductions, workforce effects, implementation impacts, and causal effects.