Abstract
{ "background": "Maternal mortality remains a critical public health challenge in many low-resource settings. Current evaluations of maternal care facility systems often rely on aggregate statistics, which mask facility-level heterogeneity and fail to quantify uncertainty in performance estimates, limiting targeted policy responses.", "purpose and objectives": "This protocol details a novel methodological framework for evaluating maternal care facility systems by linking system readiness to clinical outcomes. The primary objective is to develop and validate a Bayesian hierarchical model to estimate facility-specific performance on maternal mortality while formally accounting for structural and resource covariates.", "methodology": "We will utilise a retrospective cohort design, linking facility audit data with clinical outcome records from multiple health facilities. The core statistical model is a Bayesian hierarchical logistic regression: $\\text{logit}(p{ij}) = \\alphaj + \\beta X{ij}, \\quad \\alphaj \\sim \\text{Normal}(\\mu\\alpha, \\sigma^2\\alpha)$, where $p{ij}$ is the probability of mortality for individual $i$ in facility $j$, $\\alphaj$ is the facility-specific intercept, and $X_{ij}$ represents covariates. Inference will be based on posterior distributions with 95% credible intervals.", "findings": "As a research protocol, this paper does not present empirical results. The anticipated findings from applying this protocol include facility-specific posterior estimates of maternal mortality ratios and the identification of key systemic predictors, such as the proportion of facilities where staffing levels are a dominant predictor of outcomes.", "conclusion": "This protocol provides a rigorous, transparent, and reproducible methodological pathway for moving beyond descriptive facility assessments to probabilistic, evidence-driven evaluation of maternal care systems.", "recommendations": "We recommend the adoption of this Bayesian hierarchical modelling framework by health ministries and researchers for routine health system evaluation, as it enables direct probabilistic ranking of facilities and identifies modifiable factors for targeted intervention.", "key words": "Bayesian hierarchical model, maternal health, health systems evaluation, facility readiness, Uganda, clinical outcomes