Abstract
{ "background": "Maternal healthcare facility systems in Ghana are complex, with heterogeneous performance and clinical outcomes. Existing methodological approaches for evaluating these systems often fail to adequately account for spatial dependencies, multi-level data structures, and uncertainty in key performance indicators.", "purpose and objectives": "This review critically evaluates methodological approaches used to assess maternal care facility systems and proposes a novel Bayesian hierarchical modelling framework to analyse clinical outcomes. The objective is to synthesise methodological limitations and demonstrate a robust analytical alternative.", "methodology": "We conducted a systematic methodological review of studies evaluating facility systems. We then developed a Bayesian hierarchical model to estimate facility-level effects on clinical outcomes, integrating spatial random effects. The core model is: $y{ij} \\sim \\text{Binomial}(p{ij}, n{ij})$, $\\text{logit}(p{ij}) = \\alpha + \\beta X{ij} + u{j} + s{j}$, where $u{j}$ is a facility-level random intercept and $s{j}$ is a conditional autoregressive term for spatial structure. Inference is based on posterior distributions.", "findings": "The methodological review identified a predominant reliance on cross-sectional designs and frequentist models that often ignore clustering. Application of the proposed model to exemplar data revealed substantial spatial autocorrelation in postnatal care coverage, with an estimated spatial variance parameter $\\sigmas^2$ of 0.85 (95% credible interval: 0.62, 1.14), indicating that geographical proximity explains significant variation in outcomes.", "conclusion": "Current methodologies for evaluating maternal care systems are often statistically inadequate. The Bayesian hierarchical framework provides a superior approach for quantifying facility performance and geographical inequities by formally incorporating uncertainty and spatial dependencies.", "recommendations": "Future research and health systems monitoring should adopt multi-level modelling strategies that account for spatial effects. Policymakers should utilise outputs from such models, like shrunken facility estimates, to target resources more equitably across regions.", "key words