babamaiyaki200016-prog/Nigeria-State-TB-Incidence

# Nigeria State-Level TB Incidence and Case-Detection Gap, 2023 The first **state-level (sub-national) tuberculosis incidence** estimates for Nigeria, with an explicit estimate of the **undetected fraction** in each state — a layer that neither WHO nor the NTBLCP publishes. Written entirely in **R**; every figure and number is reproduced from public sources by the scripts here. ## Headline Against WHO's 2023 national estimate of **499,000 incident TB cases** (95% UI 326,000–709,000), **371,019 were notified (74%)** and roughly **a quarter went undetected**. The undetected burden concentrates in the deprived, under-served **North-West and North-East** (Jigawa, Kano, Katsina, Bauchi, Borno, Kebbi). The same high notifications mean opposite things in different states — a barely-caught epidemic in Sokoto versus a well-run programme in Lagos — so the estimates are framed as a **policy typology** (invest in case-finding / sustain / audit the data). **These are model-based estimates, not measurements.** Nigeria has no sub-national TB prevalence survey, so the incidence/detection split cannot be identified from data alone. Uncertainty is propagated honestly (per-state 95% intervals dominated by WHO's national uncertainty); the ranking is robust, individual absolute values are ranges. ## What the model does Notifications = incidence × detection (one equation, two unknowns). We triangulate a **structural-risk** estimate (incidence tracks deprivation/undernutrition) with a **detection-adjusted** estimate (case-finding tracks GeneXpert access), benchmark to the WHO national total, and propagate WHO's national uncertainty plus parameter uncertainty through a **3,000-draw Monte Carlo**. ## Reproduce ```r # install once install.packages(c("sf","ggplot2","rnaturalearth","rnaturalearthdata", "viridis","patchwork","ggspatial","exactextractr","terra","pdftools")) # run in order (each writes to data/) for (f in sprintf("analysis/%02d_*.R", 1:12)) source(Sys.glob(f)) ``` | Scr …