Maladie à Virus Ebola — Situation Report Extraction

Democratic Republic of Congo

Author

Billy J Quilty (Charité Berlin, LSHTM & MSF Epicentre)

Published

10 June 2026

Warning

Automated extraction — verification required. Case counts and figures in this report are extracted from INSP DRC PDF situation reports by an AI vision model (Anthropic Claude claude-sonnet-4-6) without systematic manual verification of every value. Extraction errors are possible, particularly where PDF table layouts are complex or inconsistent. All values should be verified against the original INSP DRC situation reports before any operational, clinical, or policy use. This report is intended for research and situational awareness only.

Introduction

This report documents automated extraction of epidemiological data from 24 situation reports (SitReps) published by the Institut National de Santé Publique (INSP DRC) during the 2026 Ebola (MVE) outbreak in the Democratic Republic of Congo, covering the period through 05 June 2026. The focus is on data completeness, extraction fidelity, and cross-source consistency rather than epidemiological interpretation.

Methods

Data sources

Case counts were taken from the INSP DRC situation report PDFs, downloaded automatically from https://insp.cd/ebola/. Where our PDF copies and the INRB-UMIE copies overlap, file checksums confirm they are identical, so any differences between the two datasets come from how the tables were read rather than which documents were used. The INRB-UMIE series, produced by manual data entry, was used as an independent check on our automated extraction.

Automated PDF extraction

Each PDF is read by Claude (Anthropic, claude-sonnet-4-6) using visual understanding of the page layout, allowing it to handle the complex table structures accurately. Four datasets are extracted from each report: daily new case counts, cumulative case and death totals, patient movement through isolation units (available from SitRep 006 onwards), and Point of Entry screening activity. A full text transcript is also saved for audit purposes.

Data harmonisation

Health zone names were standardised to a consistent spelling across all reports. Suspected and probable case counts were summed into a single “unconfirmed” category, since the column labels vary across reports. Cumulative totals were not permitted to decrease over time, and where more than one row existed for the same zone and date the higher value was retained.

Statistical analysis

All analysis and figures were produced in R (version 4.5.1) using the tidyverse [@tidyverse], ggplot2 [@ggplot2], and flextable [@flextable] packages. This report was rendered automatically with Quarto.

Cross-source validation

We checked our automated extraction against the INRB-UMIE Ebola DRC 2026 manually coded dataset, which extends through SitRep 22 (6 June 2026) in the current build. For SitReps 1–11 both datasets use the same PDF files, so differences reflect extraction method rather than source documents. From SitRep 12 onwards INRB-UMIE note that health-zone data is occasionally sent directly from INSP to INRB and may not feature in the published PDFs; this explains divergence at zone level from SitRep 12 onwards. Two zone name variants (Mongbwalu/Mongbalu and Nyankunde/Nyakunde) were matched manually; values marked “ND” were treated as missing.

Results

The following sections compare our automated extraction against the INRB-UMIE manually coded dataset and document known data quality issues in the source PDFs.

National-level comparison — this repo and INRB-UMIE

Cases agree across sources. Cumulative suspected and confirmed case counts match exactly for every date in the overlap period; divergences are confined to the deaths series.
Banner vs sub-table discrepancy (SitRep 010 onwards). INSP’s front-page banner and the per-zone sub-tables frequently disagree on national totals. This repo and INRB-UMIE (national_* files) both read from the banner. The disagreement reflects a structural feature of the PDFs.
SitRep 012 (26 May) — two PDF versions. The original extraction records 238 deaths; the revised version (SitRep_012_v2) records 246 deaths. No per-zone case table was present in either version to cross-check against the banner.
INRB-UMIE national coverage. The current build (build-2026-06-07) includes national banner figures from SitRep 004 (14 May) through SitRep 022 (6 June 2026), and is consistent with this repo across all overlapping dates.

Table 1: National cumulative totals — cross-source comparison (this repo vs INRB-UMIE). Shading indicates values that differ. INRB-UMIE: INRB-UMIE/Ebola DRC 2026. A dash indicates no data available from that source for that date.

	Suspected cases		Confirmed cases		Deaths
Date	Suspect. (ours)	Suspect. (INRB-UMIE)	Confirmed (ours)	Confirmed (INRB-UMIE)	Deaths (ours)	Deaths susp. (INRB-UMIE)	Deaths conf. (INRB-UMIE)
14 May	—	246	—	8	—	—	4
17 May	—	246	—	13	—	—	4
18 May	516	516	33	33	131	131	4
19 May	575	575	51	51	148	148	4
20 May	671	671	61	64	160	160	6
21 May	746	746	83	83	176	176	9
22 May	867	867	91	91	204	—	10
23 May	904	904	101	101	119	220	10
24 May	906	906	105	105	223	223	10
25 May	998	998	106	106	238	238	12
26 May	1,077	1,077	121	121	246	246	17
27 May	906	906	125	125	223	223	17
28 May	349	349	210	210	17	236	17
29 May	—	349	263	263	42	236	42
30 May	—	321	282	238	42	236	42
31 May	—	220	321	321	48	—	48
01 Jun	—	289	344	344	60	242	60
02 Jun	—	206	363	363	62	—	62
03 Jun	—	—	381	381	64	—	64
04 Jun	—	153	452	452	82	—	82
05 Jun	—	119	488	488	86	—	86
06 Jun	—	117	515	515	91	—	91
07 Jun	—	94	550	550	101	—	101
08 Jun	—	—	598	—	115	—	—

Data quality notes — case counts

Mongbwalu zone-level data starts late. SitRep 001 records 246 suspected cases for all of Ituri province combined; no Mongbwalu-specific count appears until SitRep 004 (18 May, 302 suspected). The Mongbwalu line therefore starts later than in INRB-UMIE — this is a reporting gap, not a case-onset gap.
Mongbwalu confirmed-case revision (22 May). Both datasets show cumulative confirmed cases falling from 14 (21 May) to 10 (22 May). Our series holds the value at 14 (counts cannot decrease); INRB-UMIE shows the raw figure. This is likely a downward revision in the source PDF.
SitRep 014 (28 May) — no zone breakdown. Only a national aggregate was published during an official case-count revision. The dotted vertical line in zone panels marks this gap.
SitRep 016 (30 May) — province-level only. Zone-level breakdowns return in SitRep 015 (29 May) but are absent again in SitRep 016, which reports only Ituri / Nord-Kivu / Sud-Kivu provincial aggregates.
SitRep 009 (23 May) — deaths dip. Cumulative suspected deaths fell from 160 (20 May) to 119 before recovering to 223 in SitRep 010. This is reproduced faithfully from the PDF and likely reflects an interim partial recount.

Suspect and probable case reporting discontinued. From SitRep 015 (29 May) onwards, INSP stopped publishing cumulative suspect/probable breakdowns in the zone tables. From SitRep 016 (30 May) onwards they are absent from both cumulative and daily new-cases tables at all geographic levels. SitRep 014 stated explicitly that suspected deaths were temporarily excluded pending investigation results (“Les décès suspects ont été temporairement exclus du comptage dans l’attente des résultats des investigations en cours”). The orange-shaded region in the national “Suspected deaths” panel marks this ongoing series break.

INRB-UMIE suspected cases after SitRep 016 are not cumulative. From SitRep 016 onwards, INSP’s banner split suspect reporting into ‘Cas suspects en cours d’investigation’ (under investigation) and ‘Cas suspects en isolement’ (in isolation). INRB-UMIE sum these two active categories as their national suspected case figure. This means the INRB-UMIE suspected case series from SitRep 016 onwards reflects active suspects at a point in time, not cumulative ever-suspected — explaining the declining trend.

Daily new cases (national)

Figure 1: Daily new cases reported nationally by case classification (from ‘Nouveaux’ rows in each SitRep). Suspect/probable cases absent from SitRep 016 onwards.

Cumulative counts by zone and source

Figure 2: Cross-source comparison of cumulative confirmed cases, suspect/probable cases, and outbreak deaths by health zone. Solid lines: automated extraction (raw); dashed lines: INRB-UMIE (manual extraction). For SitReps 1–11 both sources use the same PDFs; from SitRep 12 onwards INRB-UMIE zone data may include figures sent directly from INSP to INRB that do not appear in the published PDFs. From SitRep 16 onwards INRB-UMIE national suspected cases reflect active suspects (under investigation + in isolation), not cumulative totals.

Table 2 summarises numerical differences at the most recent date available in each source per zone.

Table 2: Cumulative case counts at the latest available date per zone: automated extraction vs INRB-UMIE. Δ = ours minus INRB-UMIE. Dashes indicate no data in that source. Note: INRB-UMIE zone data from SitRep 12 onwards diverges from the published PDFs (exact supplementary source not documented in their public repository).

Health zone	Confirmed (ours)	Confirmed (INRB-UMIE)	Δ confirmed	Deaths (ours)	Deaths (INRB-UMIE)	Δ deaths
Bambu	5	5	0	2	2	0
Beni	9	9	0	7	7	0
Bunia	163	152	11	55	15	40
Butembo	6	4	2	4	2	2
Katwa	12	11	1	8	8	0
Mongbwalu	114	98	16	88	29	59
Nyankunde	32	26	6	15	1	14
Rwampara	122	111	11	75	20	55

Response indicators

Data quality notes — response indicators

Rwampara 21 May (SitRep 007). Data are split across three unlabelled sub-columns; both datasets handle these differently. Values for that date should be treated with caution.
20–21 May (SitReps 005–006). An earlier table layout made in-bed count and new admissions columns ambiguous; minor differences between datasets on those dates result from this. Agreement is close from 22 May onwards when the format was standardised.
Coverage. INRB-UMIE (build-2026-06-07) extends through 6 June 2026; our extraction covers through SitRep 023 (7 June 2026). Validation panels overlap through 6 June 2026.

Figure 3: Cross-source comparison of zone-level response indicators. Solid lines: automated extraction; dashed lines: INRB-UMIE. Both series treat the ambiguous 21 May Rwampara sub-columns differently; values for that date should be interpreted cautiously. The two datasets differ slightly on 20–21 May due to an ambiguous table layout in SitReps 005–006, and are in close agreement from 22 May onwards.

Figure 4: Cross-source comparison of daily Point of Entry (POE) screening totals. Automated extraction reports a single aggregate per sitrep date; INRB-UMIE data are summed across individual checkpoints to produce a daily total.

Figure 5: Cross-source comparison of cumulative contacts traced by health zone. Solid lines: automated extraction (raw); dashed lines: INRB-UMIE (manual extraction). ND values excluded.

Discussion

Outbreak trajectory

488 confirmed cases have been reported nationally as of 05 June 2026.
The epidemic is concentrated in Ituri province (Bunia, Rwampara, Mongbwalu), with a smaller number of cases in Nord-Kivu and Sud-Kivu.
INSP discontinued cumulative suspect/probable reporting from SitRep 015 onwards, limiting assessment of total burden and complicating comparisons with earlier reports.

Limitations

Zone-level breakdowns are absent or incomplete in SitReps 001–013 and in SitReps 014 and 016; the zone-level series is therefore not fully continuous.
Response indicator table layouts varied across early reports, introducing minor inconsistencies in trend data.
Automated extraction and INRB-UMIE manual coding provide complementary checks: agreement increases confidence in a figure; disagreement typically reflects genuine ambiguity in the source document.

Appendix — Internal consistency check

Zone sum vs national aggregate

Cumulative zone-row sums are compared with the national banner figure for each date. Highlighted rows (non-zero Δ) arise from three causes:

Early reports (SitReps 001–003) provide only a province-level Ituri total, not individual zone rows.
Some reports omit certain zones from the breakdown table.
The banner vs sub-table discrepancy present from SitRep 010 onwards (described above).

Table 3: Zone sum vs national aggregate. Values show the cumulative total across all extracted health zones versus the national figure from the front-page banner. Δ = zone sum minus national banner. Highlighted rows have a non-zero difference in at least one metric.

	Suspected			Confirmed			Deaths
Date	Zones (susp.)	Nat. (susp.)	Δ susp.	Zones (conf.)	Nat. (conf.)	Δ conf.	Zones (deaths)	Nat. (deaths)	Δ deaths
18 May	515	516	-1	32	33	-1	131	131	0
19 May	574	575	-1	50	51	-1	148	148	0
20 May	671	671	0	56	61	-5	160	160	0
21 May	741	746	-5	75	83	-8	175	176	-1
22 May	852	867	-15	79	91	-12	201	204	-3
23 May	887	904	-17	89	101	-12	215	119	96
24 May	887	906	-19	91	105	-14	218	223	-5
25 May	966	998	-32	91	106	-15	235	238	-3
27 May	874	906	-32	109	125	-16	218	223	-5
29 May				224	263	-39	39	42	-3
01 Jun				227	344	-117	44	60	-16
02 Jun				242	363	-121	46	62	-16
03 Jun				258	381	-123	48	64	-16
04 Jun				320	452	-132	65	82	-17
05 Jun				354	488	-134	69	86	-17
06 Jun				381	515	-134	74	91	-17
07 Jun				416	550	-134	84	101	-17
08 Jun				463	598	-135	97	115	-18