Methods
Hover any figure on this page to see the exact sentence from the source document it was extracted from.
What this site does
ituri-sitrep aggregates publicly released situation reports for the 2026 Ituri Bundibugyo virus outbreak and displays them in a health-zone-level map with a provenance-first UI. Every rendered figure on this site carries a source_quote_id foreign key linking it to the exact verbatim sentence in the source document.
Sources ingested include: WHO Disease Outbreak News, WHO AFRO situation reports, Africa CDC bulletins, ECDC Rapid Risk Assessments, DRC MoH press releases, ACLED, HDX, and Pathoplexus/Nextstrain.
What it does not do
If you are a field epidemiologist needing an operational tool, use Go.Data, SORMAS, or DHIS2 Tracker.
This is not a forecasting platform, a clinical system, or a substitute for official communications from the DRC Ministry of Health or WHO Disease Outbreak News.
Data sources
All ingested data is publicly released material. No line-list data, patient identifiers, or contact graphs are processed. Each source row carries a licence tier (open / display-only / non-commercial-verified / excluded) that governs export and redistribution. Display-only sources render aggregated overlays but never appear in any CSV export.
Extraction method
Summary generated by Claude Sonnet 4.6. Extracted numbers are linked to their source sentences below.
The extraction pipeline uses Claude Opus 4.7 via the Anthropic API with prompt caching. Each run is keyed on a prompt_version_hash computed from the system prompt, few-shot examples, and tool schema. Re-processing with the same prompt produces exactly one extraction row per sitrep.
Tool schemas are derived from zod via zodToJsonSchema. Character offsets into the source document anchor each extracted figure to its verbatim sentence.
Provenance model
Every renderable number is stored in case_counts with a non-nullable source_quote_id. The UI renders a dotted underline on any figure that has a linked quote. Hover to preview the source sentence; click to open the full evidence drawer with chain-of-custody metadata, citation, and a link to the original document.
ICD-11 classification
The following ICD-11 codes are used in all JSON-LD schemas, extraction schemas, and rendered copy on this site. Verified against the WHO ICD-11 browser (icd.who.int).
| Entity | ICD-11 Code | Notes |
|---|---|---|
| Bundibugyo virus disease | 1D60.00 | Under 1D60.0 Ebola disease. Active outbreak. |
| Ebola virus disease (Zaire) | 1D60.01 | Classic EBOV species. |
| Sudan virus disease | 1D60.02 | |
| Marburg virus disease | 1D60.10 | Under 1D60.1 Marburg disease. |
Limitations
- Reporting lag: figures reflect the sitrep publication date, not event occurrence date.
- Geographic granularity varies by source. Health-zone attribution is imputed where not stated explicitly.
- Access constraints may cause multi-week data gaps for certain zones — absence of recent figures does not imply absence of cases.
- LLM extraction is probabilistic. All figures carry a review status; unreviewed figures are shown with an AI-generated label.
Author
Built by an MD student at the University of Western Australia as a side project. Evaluated accordingly.
Citation guidance
Cite the primary source, not this tool. Each evidence drawer provides a formatted citation for the underlying document in plain text, BibTeX, and APA formats with a one-click copy button.
Code and licence
Source code is available on GitHub under the MIT licence. Data derived from open-licence sources is available as CSV export from the Sources page. Display-only sources render aggregated overlays only and are excluded from any export or redistribution.