gpaasi/scd-screening-site-selection-uganda

Reproducible workflow to construct HbS×PfPR co-risk, rank newborn SCD screening sites with deterministic greedy selection, and generate coverage, equity, and collocation diagnostics for Uganda # Prioritizing Sickle Cell Newborn Screening Sites in Uganda using HbS and PfPR with Greedy Selection ## Summary This repository implements a geospatial siting workflow for newborn screening and clinic placement in Uganda. The pipeline assembles HbS allele frequency and PfPR2–10 rasters, aligns them on a 5 km analysis grid, builds a co‑risk surface as HbS × PfPR, and selects a ranked list of proposed sites using a deterministic greedy algorithm with a 25 km service radius. It also produces coverage curves, district equity summaries, radius sensitivity, and collocation of proposed sites to health facilities using two‑step floating catchment area (2SFCA) polygons. Key outputs are GeoTIFF rasters, a GeoPackage of proposed sites and buffers, district‑level equity tables, and figures ready for a manuscript. ## Folder layout ``` . ├─ config/ # central parameters ├─ data/ │ ├─ raw/ # place inputs here │ └─ processed/ # all outputs land here ├─ R/ │ ├─ greedy.R # greedy selection utilities │ ├─ utils_spatial.R # CRS, alignment, masks, stats │ └─ plot_helpers.R # map and chart helpers ├─ scripts/ │ ├─ install_pkgs.R │ ├─ 00_setup_paths.R │ ├─ 03_pfpr_stack_and_mean.R │ ├─ 04_build_corisk.R │ ├─ 05_mask_unserved_25km.R # optional │ ├─ 06_greedy_select_25km.R │ ├─ 07_collocation_2sfca.R │ ├─ 08_coverage_curves_and_milestones.R │ ├─ 09_equity_lorenz_gini.R │ ├─ 10_radius_sensitivity.R │ └─ 12_export_maps.R ├─ figures/ ├─ docs/ │ ├─ METHODS_methods_draft.md │ ├─ RESULTS_outline.md │ └─ SUPPLEMENT_methods.md ├─ tests/ │ └─ test_greedy.R ├─ CITATION.cff ├─ LICENSE └─ README.md ``` ## Required inputs Place under `data/raw/` unless you edit paths in `config/parameters.yaml` or `scripts/00_setup_paths.R`. Essential - `EBK_HbS3_UGA.tif` HbS allele frequency prediction, about 1 km, values 0 to 1 - `PfPR_mean_2015_2024.tif` Ten‑year mean PfPR2–10 on t …