Methodology

Hantavirus Live aggregates official publications from public-health authorities and presents them with source attribution and timestamps. This page describes how that process works, so you can judge the data for yourself.

Source tiers

Not all sources carry the same weight. Hantavirus Live organises them into tiers:

Tier 1 — Supranational authorities

European Centre for Disease Prevention and Control (ECDC), World Health Organization (WHO).

Tier 2 — National authorities

U.S. Centers for Disease Control and Prevention (CDC), Robert Koch Institute (RKI), Friedrich-Loeffler-Institut (FLI), and equivalent agencies in affected countries.

Tier 3 — Media signals

News reports tracked via GDELT and similar aggregators. These are displayed separately and never mixed with official figures. They serve as early indicators that a new official publication may be forthcoming.

When sources conflict, the higher-tier source takes precedence. If two sources at the same tier disagree, both values are shown with their respective timestamps until the discrepancy is resolved by a subsequent official publication.

Deterministic parsing

Official publications are fetched by an automated worker at regular intervals. The worker uses deterministic, rule-based parsing — regular expressions and structured HTML extraction — to read numbers from known page layouts. There is no AI, machine learning, or large-language-model processing anywhere in the live data pipeline.

Content hashing

Each time a source page is fetched, the worker computes a hash of its content. If the hash matches the previous fetch, no update is recorded. This avoids duplicate entries and ensures that the Telegram channel only posts when something has actually changed in the official publication.

Raw snapshot retention

Every fetched page is stored as a raw snapshot alongside the parsed result. If a parsing error is discovered later, the original content is available for re-processing without needing to re-fetch a page that may have changed since.

The needs-review state

When the parser encounters an unexpected page layout, a value that falls outside plausible bounds, or a conflict between sources, the affected data point enters a "needs review" state. In this state the previously confirmed value remains displayed, and the new value is held back until the site operator verifies it manually. This prevents bad parses from surfacing incorrect numbers.

Correction policy

If a displayed value turns out to be wrong — whether due to a parsing error or a correction issued by the originating authority — it is updated as soon as the corrected figure is confirmed. The previous value and the correction are both retained in the database with timestamps, so the change history remains transparent.

What this site does not do

Hantavirus Live does not interpret, forecast, model, or editorially comment on the data it aggregates. It does not produce its own case counts, estimates, or risk assessments. All analytical judgment remains with the originating authorities.