Bundibugyo virus, Democratic Republic of the Congo and Uganda, 2026

How to read.The three lines are the dated public-reporting anchors carried in this snapshot's source trail [1][2][8][9][3][10]. Official sources are preferred for headline values; the x-axis uses each source's reported data date when available, while later snapshot/capture timing is shown separately in the data-latency table. Where a row mixes publisher cadences, the timeline refs above identify the source used for each metric. Suspected counts use the clinical case definition, deaths are attributed deaths, and confirmed are laboratory-PCR positives. The orange band sits directly above the confirmed-case endpoint and shows the inferred underlying laboratory-confirmable count after correcting only for reporting completeness. The band applies to the confirmed series alone; the method does not model an inferred range on deaths or on suspected counts. The vertical dashed arrow is the ascertainment gap, the implied count of confirmable cases not yet captured in lab data, not a forecast of future spread.

How to read the two bands. The orange band comes from LOVS Module C: each publicly-confirmed point divided by the reporting-completeness 50% uncertainty interval. It estimates only the laboratory-confirmable count, because the completeness posterior was fit against confirmed cases. The blue-slate band is a different quantity: total cases (clinically compatible, most never lab-confirmed). It applies Imperial Method 2, C = D · (1 + r/β)^α / CFR, with r = ln(2) / 14d, the Rosello 2015 onset-to- death gamma (α = 4.42, β = 0.388/day), and the BDBV CFR 95% CI from US CDC outbreak history (CFR_LOW = 26%, CFR_HIGH = 40%, 55 deaths / 169 cases across the 2007-08 Uganda and 2012 DRC outbreaks); upper-band uses CFR low, lower-band uses CFR high. The (1 + r/β)^α term is the gamma moment-generating function evaluated at the growth rate; it back-corrects the right-censoring bias from infections that will die but have not yet died at the observation date. Naive D / CFR omits this correction and systematically undershoots in a growing epidemic, which is why the band below is wider and higher than a naive ratio would imply. At the as-of date this yields 612–942 total cases from 144 deaths. The dashed horizontal lines at 400 and 900 mark the joint WHO + Imperial College MRC GIDA estimate from May 20, 2026, which uses the same Method 2 formula plus a population-movement model (Uganda export counts vs. observed importations). At the as-of date, the blue-slate band (612 to 942) brackets and extends above the upper end of the Imperial range (400 to 900). The two methods do not have to agree: Imperial integrates an export-flow signal that pulls the central estimate down, while the band here only uses deaths and CFR at the central doubling time. Both methods agree the true total is several multiples of the 53 publicly confirmed cases. The Imperial "values over 1,000 not excluded" tail tracks with the upper edge of the blue-slate band. Honest caveats: (1) the band carries CFR uncertainty only; the doubling-time uncertainty (τ₂ ∈ {7, 14, 21} days per the Imperial sensitivity set) is presented separately in the sibling sensitivity grid so the two assumptions are not blurred into one width; (2) no BDBV-specific published doubling time exists; the 14-day central scenario is anchored to the 2014 West Africa NEJM range (Guinea 15.7 d, Liberia 23.6 d, Sierra Leone 30.2 d) on the fast side, consistent with Imperial's central choice; (3) the as-of completeness posterior is applied across the whole window for the orange band, assuming approximate stability; (4) both bands are latent-quantity uncertainty, not forecasts of growth.

How to read the grid. Each cell shows the projected cumulative-case estimate on the same denominator scale as the assumed CFR; it is not a laboratory-confirmed count. The calculation uses 144 deaths at that (CFR, doubling-time) pair and the formula total_cases = deaths × (1 + r/β)^α / CFR with r = ln(2) / doubling_time and the Rosello 2015 onset-to-death gamma (α = 4.42, β = 0.388 per day). Faster doubling (left columns) implies more cases per observed death because each endpoint death reflects a case that occurred ~11.4days ago, when the outbreak was smaller; the column corrects for right-censoring in a growing epidemic. Lower CFR (top rows) implies more total cases per observed death because each death indexes proportionally more survivors. The orange tint is a readability aid: darker cells carry larger inferred totals. The grid's spread is the honest range to hold in mind, not a rank ordering of cells by likelihood.

How to read. Country outlines and lakes are Natural Earth 1:10m public-domain geodata, Douglas-Peucker simplified for inline display (border tolerance 0.025°, lake tolerance 0.01°). Zone markers are placed at verified WGS-84 decimal-degree coordinates from data/zones.json in the source repository; visible labels show names only to keep the dense border cluster readable. The map foregrounds 18 corridors whose adjusted 50% upper bound reaches at least 10%; 0 lower-signal corridors remain in the audit list rather than the map. The 4 pre-committed calibration corridors are highlighted with their ascertainment-adjusted 50% upper bounds. The corridor lines are watch points, not predictions of where the outbreak will spread, and not recommendations to restrict movement between these zones.

How to read. Each bar is the method's 50% uncertainty range for the binary outcome "at least one new laboratory-confirmed case appears in the target zone within 30 days, given continued reporting in the source zone." The tight clustering of upper bounds across the review set is the headline signal: the method has enough structure to identify a surveillance review set, but not enough to rank corridors as deployment priorities. Corridors below 10% are retained for auditability rather than foregrounded.

How to read. Each line is one review-set corridor from the watchlist bar chart, drawn from source-zone centroid (orange dot) to target-zone centroid (blue dot) at WGS-84 decimal-degree coordinates from data/zones.json. Wider lines indicate higher watchlist-row upper bounds. The 4 mapped pre-committed calibration corridors are dashed orange with their registry upper bound in a chip; the other mapped corridors are descriptive blue. Background corridors below 10% stay in the interval chart audit list. All review-set corridors in the bar chart are georeferenced and shown here. The cluster of mapped upper bounds within the 50%–53% band shows what the bar chart says numerically: the method does not yet discriminate corridors above chance.

On Goma. Goma is a major commercial hub in North Kivu, roughly 700 km south of the outbreak core in Bunia. It already shows one laboratory-confirmed BDBV case, a traveler infected in Ituri, so it is rendered as a spillover-case site on the Outbreak geography map above. The most parsimonious epidemiological reading of that case is one of two: the virus traveled directly from the outbreak core via Goma's commercial air or lake links, or community transmission has gone undetected along the Bunia, Beni, Bundibugyo, Kasese corridor. Either reading raises surveillance priority for those intermediate zones. Goma is not in this snapshot's pre-committed corridor calibration set, because each snapshot freezes its calibration zones at build time and the 20 May snapshot was pinned without a Goma corridor. Goma should only join a later snapshot as a candidate calibration zone if that snapshot pins its own scoring contract before outcome assessment.