What Does It Take to Map a Country? Scaling OpenStreetMap Mapping for Accurate Health Accessibility Modelling in Madagascar

Abstract Comprehensive geographic data are essential to accurately model geographic accessibility to healthcare and to guide equitable health system planning and implementation. In low-income countries, however, incomplete road and building data in global databases such as OpenStreetMap (OSM) limit the precision and operational applications of geographic accessibility models. Following a successful pilot in one district of Madagascar, we evaluated the scalability of an exhaustive mapping approach to produce highly granulated household-level accessibility estimates at regional and national levels. Using satellite imagery and the OSM platform, we mapped all buildings, roads, footpaths, and rice fields across seven additional districts in southeastern Madagascar. We estimated travel routes, distance and travel time between each household and the nearest primary health center (PHC) or community health site (CHS) using the OSM Routing Machine, combined with predictions of travel speed from a locally calibrated statistical model. We then assessed population density and mapping completeness for roads and buildings in our study area and across Madagascar using AI-generated reference datasets (Microsoft and Facebook/MapWithAI) and estimated corresponding mapping times. Finally, we estimated the resources required in person-years to scale this approach across Madagascar using two different extrapolation methods. Nearly one and a half million buildings and 197,000 km of footpaths were added to OSM across the eight mapped districts, for a total area of about 30,200 km 2 . Between 24 % and 65 % of the population lived within one hour of a PHC depending on the district, and 87 %–99 % lived within one hour of a CHS. Most Malagasy districts were classified as having low completeness for both buildings and roads. Scaling up the approach to cover the entire country would require between 220 and 350 person-years depending on the extrapolation method and assumptions used. Mapping an entire country with sufficient detail to precisely model healthcare accessibility for every household is feasible but resource-intensive. Combining human mapping, participatory approaches, and AI-assisted datasets can substantially improve OSM completeness and generate actionable, high-resolution travel-time data for health planning. Our findings provide a roadmap for Madagascar and other countries seeking to develop national-scale geospatial infrastructure for sustainable development and universal health coverage.