"background": "Maternal care facility systems in Kenya are complex, with heterogeneous performance and outcomes. Current evaluation methods often fail to adequately account for this heterogeneity and the hierarchical structure of health system data, limiting the precision of impact assessments and system-level inferences. ", "purpose and objectives": "This intervention study aimed to develop and apply a novel Bayesian hierarchical model to evaluate maternal care facility systems and quantify their association with key clinical outcomes. The primary objective was to provide a robust methodological framework for facility-level performance measurement. ", "methodology": "We conducted an intervention study applying a Bayesian hierarchical model to routine health information system data from a national sample of maternal care facilities. The core model specified the log-odds of a positive clinical outcome (e. g. , uncomplicated delivery) for facility i in county j as (p{ij) = \ + \ Xij + uj + vij, where uj \ N (0, \²) and vij \ N (0, \²) represent county and facility-level random effects, respectively. Model parameters were estimated using Hamiltonian Monte Carlo. ", "findings": "The model successfully quantified substantial variation in system performance attributable to the facility level. A one-standard-deviation increase in the modelled facility system score was associated with a 15% (95% credible interval: 11% to 19%) increase in the odds of a positive outcome. The posterior distributions for random effects revealed distinct clusters of under- and over-performing facilities after adjusting for county-level factors. ", "conclusion": "The Bayesian hierarchical modelling intervention provides a statistically rigorous framework for evaluating maternal care systems, effectively partitioning variance and producing shrunken, more reliable estimates of facility performance compared to conventional methods. ", "recommendations": "Health policymakers and researchers should adopt hierarchical modelling approaches for health system evaluation to enable fairer facility comparisons and more targeted interventions. Future work should integrate this methodology into routine health
Mwangi et al. (Sun,) studied this question.