ihmeuw/facility_readiness

Public repo for "Health facility readiness for maternal and newborn health care and its link with service utilisation: a geospatial analysis of four African countries, 2008–2023" from The Lancet Regional Health - Africa # Facility Delivery-Readiness Analysis Code This repository contains the analysis code for the paper's facility delivery-readiness index, small-area-estimation (SAE) geospatial models, readiness-adjusted coverage estimates, and exhibits (figures/tables), posted in accordance with GATHER (Guidelines for Accurate and Transparent Health Estimates Reporting). Four countries are covered: Nigeria, Burkina Faso, Kenya, and Ethiopia. ## Repository structure and pipeline order Scripts are organized into four stages, run in this order: 1. **`1_index_prediction/`** — Combines facility survey data across countries, constructs the facility delivery-readiness index (several weighting approaches: simple mean, grouped weights, PCA, SEM), and predicts the index for facility-surveys/years missing full indicator data using multiple imputation and prediction models. - `combine_data.R` — merges per-country facility datasets, builds the delivery-readiness index and its component weights. - `mi_comparison.R` — compares imputation approaches for missing index components. - `pred_index_rev_functions.R` — shared functions for predicting the readiness index from partial indicator data. - `launch_pred_index.R` — launches the index-prediction jobs. 2. **`2_geospatial analysis/`** — Fits small-area-estimation (INLA) models to the facility-level readiness index to produce smoothed, small-area estimates of delivery readiness over time, pooling all five countries in shared models. - `00_collapse_for_sae.R` — collapses facility-level data to the area/year level for SAE modeling. - `1_prepare_covariates_for_sae.R` — prepares geospatial covariates. - `2_Model_Selection_Pooled.R` → `3_Best_Model_Pooled.R` → `4_Fit_Best_Model_Pooled.R` — candidate model fitting, best-model selection, and final model fit, in that order. - `prep_rake.R` — post-estimation raking of small-area estimates. - `UsefulFunctions/` — shared helper functions sourced by the scripts above (INLA model-fitting helpers, VIF/stepwi …