Biobuild-africa/cocci-alert

USSD/WhatsApp poultry-health triage and informal-sector data system for Nigerian smallholders. # CocciAlert **A privacy-protected USSD/WhatsApp poultry-health triage and informal-sector data system for Nigerian smallholders.** CocciAlert converts smallholder poultry health events into anonymized, consented, LGA-aggregated food-system resilience data, addressing the lack of formal disease records in Nigeria's informal poultry sector. This repository contains the **TRL 3-4 evidence pack** prepared in support of the FCI4Africa Open Call 1 proposal (Horizon Europe HORIZON-CL6-2024-FARM2FORK-01-11), submitted under **Option B · Challenge 5: Limited transparency and data obstructing the integration of the informal sector**. --- ## Why this exists Smallholder poultry farms contribute roughly 65 to 77% of total chicken production in Nigeria (Bamidele et al., 2022, *Vet Med Int* 2022:7746144), and operate almost entirely outside formal disease-recording systems. Coccidiosis is a major and largely preventable cause of flock mortality, yet: - No structured event-level data exists at LGA scale - Farmers default to inappropriate over-the-counter antibiotic use, contributing to AMR - Buyers cannot distinguish risk-managed from unmanaged sources - The FCI4Africa Food Systems Dashboard has no informal-sector poultry-health layer CocciAlert is a **triage tool, not a diagnostic tool**. It helps a farmer decide whether a flock needs no action, immediate management changes, or a veterinarian, and in doing so, generates one of the first structured, anonymized, LGA-level poultry-health event datasets for informal smallholder poultry systems in Nigeria. --- ## What's in this repository The complete **TRL 3-4 evidence pack**: | Folder | Contents | |---|---| | `docs/` | Clinical decision logic, TRL validation memo, expert review pack | | `engine/` | Deterministic Python rule engine encoding the clinical decision tree | | `validation/` | 31-scenario validation harness + run logs | | `integration/` | End-to-end controlled prototype: input → engine → consent → anonymized log …