How the register is built
What counts as evidence here, how each figure is labeled, when it expires, and what the register cannot yet show.
Search protocol
- PubMed and PubMed Central are the backbone. Citations are copied from PubMed metadata, and a claim is used only when it was read in the abstract or the full text.
- WHO and regional atlases, official statistical releases, and government gazettes and statutes are used where the document itself can be opened.
- Web search is used for discovery only. A claim seen only in a search snippet is never recorded as a figure or a finding.
- News is used only for a current-events fact with no better source, and it is labeled tier 4.
Inclusion rules
- Every figure records its value, unit, the year the data refers to, its source and, where the source allows, its method.
- Every source records its tier, and a URL, DOI or PMID when one exists.
- A figure that cannot be confirmed keeps the UNDER REVIEW label and a stated reason, or becomes a launch-gate item with a named owner.
- A value supplied by an audit or review is added only after it is confirmed against a primary source. Until then it is logged as held for review and not shown.
- Country notes may name years and instruments but never state a quantity. Quantities live only in figures.
Source tiers
| Tier 1 | Official statistics, law, regulator publications, and an institution's own documents about itself |
|---|---|
| Tier 2 | Peer-reviewed literature |
| Tier 3 | Reputable secondary sources and NGOs, including atlases compiled by professional associations |
| Tier 4 | Media, commentary and anything not otherwise verified |
Evidence labels
| Label | Definition | Use rule |
|---|---|---|
| SOURCE-CHECKED | An observed value or policy text taken from a tier 1 source (official statistics, law, regulator or the institution itself) or a tier 2 source (peer reviewed), with current data. | Safe to quote with its citation. |
| SOURCE-CHECKED · <year> DATA | As above, with data older than its category threshold. | Quote with the data year. |
| MODELED ESTIMATE · <year> | A number produced by a model, by the source or by the register, including GLOBOCAN, GBD and UN IGME estimates. | State the method and the reference year whenever quoted. |
| SECONDARY SOURCE | An observed value whose best source is tier 3 (reputable secondary or NGO) or tier 4 (media or commentary). | Quote only with the source named, and look for a primary source. |
| CONFLICTING SOURCES | Two credible sources give different values. | Show both values and explain the difference. |
| PROGRAM HYPOTHESIS / PROGRAM JUDGMENT | An interpretation or design choice made by the program, not a finding of any source. | Keep the evidence, the interpretation and the decision implication visibly apart. |
| ILLUSTRATIVE ONLY | Constructed for reasoning from user inputs or cited parameters. Contains no external data of its own. | Never presented as a finding. Applies to the patient composite and the cost skeleton. |
| UNDER REVIEW | A value that could not be confirmed, or that an audit challenged. | Do not quote until it leaves review. |
| DATED (badge) | Added to any label when the figure is past its review date. | Re-check the source before use. |
Expiry policy
- Security and travel status: re-check every 90 days.
- Funding status: re-check every 90 days.
- Internet policy and connectivity: re-check every 90 days.
- Drug availability and pricing: re-check within 90 to 180 days; the register uses 180.
- Legal and policy status: re-check within 90 to 180 days; the register uses 180.
- Program cohorts and service directories: re-check every 12 months.
- Population, financing and cancer burden: re-check at the next official release, with a 12 month checkpoint.
- Historical facts: no expiry; a source link is required.
Evidence adequacy
Each country domain shows an adequacy score from 0 to 1, computed from evidence quality and never from the number of citations. A figure scores its best source tier weight (tier 1: 1, tier 2: 0.9, tier 3: 0.6, tier 4: 0.3) times its confidence weight (high 1, moderate 0.7, low 0.4, unresolved 0) times its currency weight (current 1, dated 0.75, stale 0.4, unknown 0.5). A domain scores the mean of its figures; a domain with sources but no figures scores its best tier weight times 0.8.
These weights are the register’s editorial choices, not an external standard. Because a domain score is a mean, one strong figure can outscore several good ones, so read the score together with the number of figures behind it.
Evidence scope
A note can read as a national finding when the study behind it covered one hospital. Each sourced note therefore carries a scope, decided from the wording of the note and its source descriptions alone: national (law, policy, national statistics, nationwide surveys or registries, or a review of the country’s literature), subnational (a named state, province, region, district or city), single-site (one hospital, clinic, program, hospice, insurer or cohort) or mixed (national evidence combined with narrower studies). A note is also marked self-reported when its authors evaluated their own program or organization. Where the sources do not say enough to tell, no scope is shown.
Evidence age and thin dossiers
Each sourced note shows the publication year of its newest source and is marked dated when that year is more than 5 years before the register’s build year. A dossier is marked thin evidence when it rests on fewer than 8 distinct sources across all its sections. Both thresholds are editorial choices: they flag where a reader should look for newer or more evidence, and they do not change any adequacy score.
Known limitations
- For entries written before evidence schema v2, the last-checked date is the date of the commit that last changed them, not a fresh check against the source.
- Several featured figures cite sources (GLOBOCAN, the national census, National Health Accounts, WHO GHED, WHO EMRO) with no URL on record yet.
- Most country evidence comes from journal papers. Laws, formularies and budgets are mapped in research/primary-documents.md but largely not yet read.
- Opioid sales data is a proxy for availability and does not measure palliative access. It is never compared on one axis with reported consumption statistics.
- World Bank income groups and GDP per capita have not yet been retrieved, so no income ratio is shown.
- Several audit-reported values are held for review because their sources could not be fetched; they are listed on the corrections page.
Field definitions are in the data dictionary in the repository (docs/data-dictionary.md). Every change to a claim is on the corrections page, and the data can be downloaded from open data.