[enrichment] ncd-risk-factor-surveillance-global source inventory - #61

Issue linked

No GitHub issue linked. This completes the in-repo scoped task source-inventory for public-health/ncd-risk-factor-surveillance-global.

Files changed

• problem-packs/public-health/ncd-risk-factor-surveillance-global/evidence.json
• problem-packs/public-health/ncd-risk-factor-surveillance-global/datasets.md
• problem-packs/public-health/ncd-risk-factor-surveillance-global/claims.json
• problem-packs/public-health/ncd-risk-factor-surveillance-global/evidence.md
• problem-packs/public-health/ncd-risk-factor-surveillance-global/problem.json
• problem-packs/public-health/ncd-risk-factor-surveillance-global/problem.md
• problem-packs/public-health/ncd-risk-factor-surveillance-global/tasks.json
• problem-packs/public-health/ncd-risk-factor-surveillance-global/AGENTS.md
• Generated indexes/wiki: tasks-available.json, agent-radar.json, docs/wiki/Agent-Radar.md, docs/wiki/Index.md

Evidence added or changed

Replaced the prior broad/inferred evidence set with seven dated WHO-facing source records:

• WHO STEPS systems-tools page
• WHO NCD surveillance data and reporting page
• WHO NCD microdata repository
• WHO Global Health Observatory NCD theme
• WHO-linked NCD data portal
• WHO Global Adult Tobacco Survey
• WHO Global School-based Student Health Survey

The dataset inventory now classifies eight source families as usable, limited, or rejected. The central claim is narrowed to source fragmentation and provenance readiness, not prevalence, burden, or country ranking.

Validation method

• pnpm build passed
• pnpm validate passed
• pnpm reproducibility:check passed
• git diff --check passed
• pre-push hook reran validate and build, passed
• pnpm verify:sources failed repo-wide on three pre-existing unrelated URLs: merra2-climate-reanalysis-wbgt-source, bgs-dphe-arsenic-survey-2001, ehp-arsenic-exposure-review-2024. All seven changed NCD evidence URLs returned HEAD 200 in that run.

Known limitations

• This is a source-surface inventory, not a country-by-country extraction of every low-income-country STEPS wave.
• It does not prove sub-national data absence for any country.
• It does not rank NCD surveillance gaps or imply funding priority.
• It treats dashboards as routing/context sources unless indicators are traced back to source-level records.

Kill condition

A reviewer identifies a current public WHO or WHO-endorsed table that, for low-income countries, lists adult NCD risk-factor survey waves with survey family, year, public microdata access status, measured versus modeled indicator provenance, population age range, and sub-national identifier availability in one normalized source.

Reviewer type needed

Domain reviewer.

Protocol Notes

verify:sources currently fails repo-wide on unrelated existing URLs even when the changed evidence URLs pass. The protocol would be easier to review if source verification could optionally run in changed-files-only mode while retaining full scheduled checks.

Daily run checklist

• Read AGENTS.md, SKILL.md, SAFETY.md, all relevant schemas
• Chose mode: enrichment
• Picked scoped task from agent-radar: public-health/ncd-risk-factor-surveillance-global/source-inventory
• Produced expected artifact: datasets.md update plus evidence.json records
• Every claim has a kill condition
• Every evidence record has dated source, stable URL, method, limitations
• pnpm validate passes
• pnpm reproducibility:check passes
• pnpm verify:sources run; changed URLs pass, repo-wide pre-existing unrelated failures remain
• pnpm build passes, regenerated files committed
• PR opened with required body fields
• Protocol Notes included