Asian Journal of Probability and Statistics

This study addresses critical limitations in large-scale geospatial health facility datasets arising from taxonomy inconsistency and spatial misclassification. The objective was to develop and evaluate an integrated framework that combines taxonomy harmonisation and administrative validation to improve analytical reliability in Sub-Saharan Africa. A multicountry dataset comprising 98,745 facilities was standardised into harmonised ownership and facility-type categories. Spatial validation was performed via country and first administrative (admin1) boundary containment, alongside multimethod outlier detection. Analytical approaches included descriptive statistics, multilevel logistic regression, and spatial clustering (Moran’s I and Getis-Ord Gi*). It appears that taxonomy harmonisation successfully minimised the fragmentation in the classification of facilities with public ownership, representing 57.2%, while 38.1% were not classified. Spatial consistency was high at the country level (≈98%), and was relatively low at the admin1 level (≈23%). Spatial analysis showed that there was a high concentration (Moran's I > 0.5), and facility density was not uniform across countries. The results were seen to be stable across the different pre-processing conditions and produced strong results. The study showed that data harmonisation and spatial validation in analytical workflows enhance the reproducibility and interpretability. The results in this study underscore the importance of scalable, geospatial health data systems to improve evidence-based planning and address health infrastructure disparities.