Each country is coloured by how its most recent ILI/ARI rate compares to its own history for the same week of the year — not to the raw numbers other countries report. The five bands (very low → very high) share the ERVISS band names but are locally computed percentiles, not MEM thresholds.
The ERVISS feed mixes two denominators: some countries report per 100 000 population, others per 100 000 consultations. Same disease, same week, no flag in the file to tell them apart — Austria's 2817 and Iceland's 5.5 are not on the same scale. Banding each country against itself sidesteps this.
ILI is heavily seasonal, so a plain all-history percentile would just paint the map by month. The bands here compare against the same ISO week in each of the previous 5 seasons — answering "is this week high for this time of year?".
The 5-year cap is deliberate: it excludes the acute COVID period whose disruption to consultation patterns would flatten the bands. As older ERVISS data accumulates the cap will start to bite; today most countries have 2–4 seasons available.
The band says the current value is unusually high for that country at that week. It cannot distinguish a genuine surge from a change in how that country counts (denominator switch, sentinel-network change, reporting-cadence shift). Treat a lone very-high country as a signal to check the source, not as a comparative claim against its neighbours.
As a partial safeguard, values that sit well beyond the country's own historical range for the same week (more than ~2.5× its 98th-percentile) are treated as likely source-side artefacts and greyed rather than banded very-high.
Source: EU-ECDC ERVISS weekly data. Bands are computed by this site's extractor; see the JSON's band_config field for the exact thresholds.
The US layer shows CDC's ILINet: ~4,000 sentinel outpatient providers report weekly visits meeting the ILI syndrome definition — fever ≥100°F (37.8°C) plus cough and/or sore throat — as a percentage of all outpatient visits (weighted ILI, or wILI).
ILINet is a syndromic count: cases are recorded by clinician observation, not by influenza testing. Since October 2021 the definition also no longer excludes "known cause other than influenza", so ILINet captures anything presenting with those symptoms — influenza, COVID-19, RSV, common cold, other respiratory viruses. A high wILI reading means "more people are turning up with fever + cough/sore throat", not "more people have flu".
Japan reports clinician-diagnosed influenza specifically (usually rapid-test-supported); the US count is syndromic ILI from any cause. A US "low" and a Japan "low" are not the same signal.
Unlike the UK's population-based rate, this is a care-seeking denominator — so raw values aren't comparable across layers.
Each state is coloured by its wILI against the same ISO week in the previous 5 seasons (percentile cuts p50 / p75 / p90 / p98). Grey = reporting but no meaningful seasonal reference. States reporting more than 4 weeks behind are dropped (currently ~8, through the summer trough).
Source: CDC ILINet via the Delphi Epidata API (Carnegie Mellon University); ILINet is a programme of CDC's US Influenza Surveillance System. Bands are computed by this site's extractor; see the JSON's band_config field for the exact thresholds.
The Japan layer shows JIHS (formerly NIID) IDWR sentinel surveillance: ~500 designated sentinel clinics report weekly clinician-diagnosed influenza as cases per sentinel (定点当たり報告数).
The case definition is influenza-specific: (a) sudden onset, (b) fever >38.0°C, (c) upper respiratory symptoms (cough, sputum, breathlessness), and (d) general fatigue — OR strongly suspected via laboratory diagnosis such as rapid antigen testing. Rapid antigen tests are used very heavily in Japanese clinics, so the reported influenza cases are usually RIDT-supported. They are not PCR-confirmed, but they are flu-specific — this is a narrower, more influenza-specific signal than the syndromic ILI counts the UK and US layers report. A Japan "low" and a US "low" are not the same signal.
Each prefecture is coloured against its own same-week history (percentile cuts p50 / p75 / p90 / p98). JIHS only exposes weekly CSVs from 2023, so history is currently 2–3 seasons — during the summer trough most prefectures will show grey (flat baseline, no meaningful band); bands come alive in autumn/winter when the same-week spread widens. Bands here are locally computed anomaly signals, not Japan's own fixed epidemic-onset (1.0/sentinel) or alert (30/sentinel) thresholds.
Neither a population rate (like the UK) nor a percent of visits (like the US) — so raw values aren't comparable across layers.
Source: JIHS Infectious Diseases Weekly Report (IDWR). Bands are computed by this site's extractor; see the JSON's band_config field for the exact thresholds.
In Development
Please make contact if you have any suggestions/comments.