Congo's health ministry counted 3,262 confirmed Ebola cases and 1,437 deaths in an overnight update carried July 28, a successor record that arrives as responders report abandoning or failing to reach communities after attacks by residents and rebels. [1]
The figure succeeds the July 28 account of surveillance teams facing workloads beyond the case counts. That account held that field workload is not incidence — the strain on the response did not establish a number. The ministry's overnight update supplies the number, and the World Health Organization's Africa regional office supplies the official successor table in its Situation Report 11, with data as of July 26. [2]
The reconciliation caveat
WHO's weekly situation report — the eleventh, with data as of July 26 — flags that the July 22 case spike was a data-reconciliation exercise, not new incidence. [2] That caveat is the one the fastest-on-record frame tends to skip. AP leads with pace — the fastest Ebola outbreak on record, set to pass the 2018–20 outbreak's 3,317 cases within days. [1] The pace framing is true as a cumulative statement. It is not true as a statement about onset, because a cumulative ministry count and a reconciliation spike are not acceleration.
Reconciliation is not a euphemism; it is what a mature surveillance system does when it catches up with its own backlog. Suspect cases are reclassified, duplicates removed, community deaths retroactively confirmed. Each of those operations moves the cumulative number without moving the epidemic. WHO's decision to flag the spike publicly is the organization marking its own homework — a credit to the table and a warning to everyone reading it as a speedometer. [2]
The death toll deserves the same care as the case count. One thousand four hundred thirty-seven deaths against 3,262 confirmed cases is a cumulative ratio shaped by the same reconciliation operations — deaths confirmed late arrive beside cases confirmed late, and the ratio between them says as much about the ledger's catching-up as about the virus. [1] What it does not say is anything about the communities responders can no longer reach, where both numbers are still being written.
The access loss cuts in both directions. If responders cannot reach communities, the true count may lag reality — cases go uncounted. It may also distort downward: a shrinking response footprint can make the ledger look calmer than the epidemic is. [1] The July 28 locked table — 3,200 confirmed, 1,405 deaths — is not revised; 3,262 and 1,437 is a successor record, not a correction.
That distinction is this paper's standing discipline on the outbreak file, and it matters more as the numbers approach the 2018–20 benchmark. Surpassing 3,317 cumulative confirmed cases will be reported as a record; whether it is a record of incidence or of bookkeeping depends on a split nobody has published. [1]
The unknown chains
WHO's framing, carried by UN News on July 14 with data as of July 11, is blunter than the record-pace headline: "this is a fire," and the majority of new cases are coming from unknown transmission chains. [3] Unknown chains are the part of the ledger that reconciliation cannot fix. A reconciled count tells you what was already found. An unknown chain tells you what is still spreading unobserved.
The arithmetic of the advance deserves the same care. The move from 3,200 to 3,262 is 62 cases; how much of that is reconciliation versus new onset is unpublished, and WHO has not committed to a split. [2] The question matters because the two readings point at opposite responses: reconciliation argues the surveillance net is working; new onset argues the epidemic is outrunning it.
Access loss as an instrument error
The access reporting is the second half of the story and the half most likely to corrupt the first. AP reports responders abandoning or unable to reach communities after attacks by residents and rebels. [1] Contact follow-up in those communities does not pause gracefully; it breaks. Cases there enter the count late, or never, and deaths there may enter only when a burial team is reached. The ministry's overnight cadence has so far persisted — but a count produced by a shrinking footprint is an instrument reading its own degradation, and July 29's table carries no error bar for it.
X is unobserved this session — the search provider rate-limited the query — so the frame here is absence, not silence. The Congolese health-journalism community that would carry the local record has not been captured. What remains is the ministry's overnight cadence and the shrinking footprint, and the open question of whether access loss will degrade the table itself.
-- KENJI NAKAMURA, Tokyo