Weekly global outbreak briefing — 2026-W38

What materially changed across the ENCLAI surveillance record in 14–20 September 2026, measured against 7–13 September 2026.

Every figure re-derived from the production database on 22 September 2026 at 13:23 UTC · authoritative sources only · download the figures (CSV)

What changed in the week of 14–20 September 2026. Measured against the preceding week (7–13 September 2026), the largest movement was Norovirus gastroenteritis in Germany, which rose from 257 to 390 reported cases (+133); 11 disease–country pairs began reporting with nothing on record in the previous 90 days (Acute poliomyelitis in Kenya, Anthrax in Zambia, Chikungunya fever in Hong Kong). 1 of the 36 disease–country pairs that reported in both weeks moved far enough to clear this briefing's materiality threshold; the rest moved within the range that ordinary reporting variation produces.

Why the headline totals are not the story. Summed across everything that reported, period case counts were 6,844 this week against 7,519 the week before — but those two figures do not describe the same set of countries and diseases. 45 sources published in this window and 20 in the previous one; the largest pair reporting in the earlier week and not in this one was COVID-19 in Mexico, at 2,073 cases — by itself about 307% of the difference between those two totals. The comparable figure is the matched one: on the 36 disease–country pairs present in both weeks, reported cases went from 4,830 to 5,105. Every movement discussed on this page is a matched-pair movement.

How much of the week this is. Surveillance records arrive long after the events they describe: across the last 90 days the median authoritative record reached ENCLAI 25 days after its own event date, and one in ten took more than 73 days, because several national systems publish weeks in arrears. This briefing therefore uses a fixed cut-off — only records that had arrived by 22 September 2026, 2 days after the week closed, on both sides of every comparison (the comparator week is cut off at 15 September 2026, the same age). Without that, the earlier week would look larger for no reason other than having had an extra week to fill: ENCLAI now holds 805 records for this window and 1,193 for the previous one, against the 805 and 567 used here. Read this as an early view of the fast-reporting part of the week, not as the week's final record.

What is absent is not zero. 33 disease–country pairs reported a case count in one of the two weeks but not the other. Those are reporting gaps — a monthly bulletin, a public holiday, a feed that has not landed yet — and this briefing does not treat them as a rise from or a fall to zero. They are listed below with their one observed value and no comparison, which is all the record supports.

Cumulative totals are shown as levels. 136 outbreak totals were restated during the week. A cumulative figure is a running total to a date, so it cannot be added to the weekly counts above and cannot be differenced across restatements (agencies revise the base). They appear in their own table, as of their own dates.

Reporting coverage, both weeks

Prior week: 567 records, 56 diseases, 29 countries, 20 sources. These count publications, not cases — a busier week may only mean more agencies published. A further 1,957 records this week carried a lower credibility grade (unverified signal, estimated or corroborated news); none of them contribute a single figure on this page.

Movements, matched disease–country pairs

Only pairs that reported a period case count in both weeks appear here. A movement is marked material when it changed by at least 100 cases and by a factor of at least 1.5. Those thresholds are ours, stated so you can disagree with them: the underlying figures are in the CSV, unfiltered.

Showing the 12 largest absolute movements of 36 matched pairs. All of them are in the CSV.

Reported in only one of the two weeks

Newly reporting disease–country pairs

Pairs with an authoritative record this week and none in the previous 90 days. This says the record started, not that the disease arrived: a first report can equally mean a new outbreak, a source coming online, or an agency changing what it publishes.

Countries with no authoritative record in the previous 90 days that reported this week: Sierra Leone, Zambia.

Cumulative totals restated this week

136 cumulative restatements in total; all are in the CSV.

Standing 30-day elevation signals

Context on a different clock: these compare the last rolling 30 days with the 30 before, so they are not a statement about this ISO week. Pairs where either side is zero are excluded — a zero there means no reporting in that window, which is not the same as no cases.

95 flagged pairs in total; all are in the CSV.

Method, and what this cannot tell you

• Window. The last complete ISO week (14 Sep to 20 Sep 2026), compared with the seven days before it. Fixed boundaries, so this URL means the same thing whenever it is re-run.
• Vintage. Both windows are cut off 2 days after their own week ended (22 Sep and 15 Sep 2026), so neither side has had longer to fill than the other. Records that arrived later exist and are not lost — they are simply outside this edition, and a rebuild at a later vintage would show more of them.
• Scope. Only records graded AUTHORITATIVE — published by an official public-health body — after de-duplication. Unverified signals are counted but never quantified here.
• Two kinds of number. Period counts (new cases in a stated period) and cumulative totals (a running total to a date) are separated throughout and never summed together.
• Reporting is not incidence. Everything above describes what agencies published. A rise can be a real outbreak, a new source, or a change in publishing cadence, and this briefing cannot tell you which.
• No denominators. Sources rarely publish populations at risk, so nothing here supports a cross-country rate comparison.
• Case definitions differ by country and change over time; syndromic entries (marked syndromic) count presentations matching a case definition, not confirmed diagnoses, and must not be read as a disease count.
• Revisions happen. Agencies restate figures after publication. This page is the record as it stood at the generation time stamped above; a later week's briefing may reflect a revised past.