About hantavirusmap.io

A free, source-cited, public-good map of every documented hantavirus outbreak from the 1993 Four Corners discovery to today.

What this is

hantavirusmap.io exists because hantavirus surveillance data — even when officially published — is fragmented across the WHO, CDC, ECDC, PAHO, national health ministries, and peer-reviewed literature. We curate, source-cite, and visualize every documented event on one map so journalists, researchers, public health professionals, and the curious public can see the historical and current picture in one place.

Educational use only. Not medical advice. Read the full disclaimer.

Methodology

Every event on the map is an entry in a structured dataset that conforms to a published JSON schema. Each event records:

A primary country and optional multi-country list
Location coordinates with a precision tier (exact / approximate / estimated)
Start date, optional end date
Case counts: confirmed, suspected, deaths, last-updated date
A 50–2,000-character narrative summary
An ordered list of sources with publication tier and access date

Source tiers

Sources are classified into four tiers. The tier determines what the ingestion pipeline is allowed to auto-publish:

Tier 1 — WHO Disease Outbreak News, CDC MMWR/NNDSS, ECDC, PAHO, national ministries of health. Auto-publishes with confidence: verified.
Tier 2 — ProMED-mail, CIDRAP, peer-reviewed literature. Auto-publishes with confidence: preliminary.
Tier 3 — major science journalism (Reuters, AP, BBC, Bloomberg, AFP). Queues for human review.
Tier 4 — local news, blogs, social media. Signal only; never auto-publishes.

Curation pipeline

Updates arrive through three channels:

An automated cron polls RSS feeds from CDC, ECDC, PAHO, and Google News proxies for WHO and ProMED, every two to six hours depending on whether an outbreak is currently active.
A Perplexity web search broadens discovery to outlets that RSS misses, with targeted regional queries for the Americas, Europe, East Asia, and Southeast Asia + MENA.
Public contributions via /contribute are queued for human review before publication.

For each candidate article, Anthropic's Claude (Haiku 4.5 by default, escalating to Sonnet 4.6 for ambiguous cases) extracts a structured outbreak record using a JSON-schema-constrained prompt. Every numeric claim must be supported by a verbatim quote from the source article — the pipeline literally substring-checks that the quote appears in the fetched text before any record is accepted. Records that fail the guard are rejected, not silently corrected.

Records are deduplicated against existing events by a signature of country, region, pathogen, and onset date within a 14-day window — matched candidates update the existing event rather than creating a duplicate.

License

Code is MIT-licensed. Data is published under CC-BY-4.0. You are free to use, share, and build on this dataset for any purpose, including commercial, as long as you credit hantavirusmap.io.

The full raw dataset is available at /data/outbreaks.json and per-event Markdown shadows are at /raw/. An OpenAPI schema is at /openapi.json. AI agents and LLM crawlers — see /llms.txt and /ai.txt.

Team and contact

Built and maintained by Atlas Development Global. Corrections, additions, and reports of missing events at /contribute or by issue at the public GitHub repo.