Abstract
{ "background": "District hospitals are critical nodes in national health systems, yet their operational reliability in low-resource settings is poorly quantified. Existing evaluations often lack methodological rigour in handling the hierarchical nature of hospital system data, where facility-level outcomes are nested within regional administrative structures.", "purpose and objectives": "This study provides a methodological evaluation of multilevel regression modelling for measuring and predicting system reliability in district hospitals. It aims to assess the suitability of this analytical approach for capturing variance at different organisational levels and identifying key predictors of reliability.", "methodology": "We conducted a longitudinal analysis of administrative and performance data from a nationally representative panel of district hospitals. System reliability was operationalised as a composite score of service continuity metrics. A three-level random intercepts model was specified: $y{ijk} = \\beta0 + \\beta X{ijk} + uk + v{jk} + e{ijk}$, where $i$ denotes time, $j$ denotes hospital, and $k$ denotes state. Estimation used restricted maximum likelihood with robust standard errors.", "findings": "The methodological evaluation demonstrated that 68% of the variance in reliability scores was attributable to differences between states, while 24% was between hospitals within states. A one-unit increase in predictive maintenance expenditure was associated with a 0.15 increase in the reliability score (95% CI: 0.11 to 0.19), holding other factors constant.", "conclusion": "Multilevel regression is a robust methodological framework for analysing hospital system reliability, effectively partitioning variance across hierarchical levels and providing unbiased estimates of predictor effects.", "recommendations": "Health systems researchers should adopt multilevel modelling for facility-based assessments to inform targeted interventions. Policymakers should prioritise state-level institutional factors, which our model identifies as the dominant source of reliability variation.", "key words": "health systems research, multilevel modelling, health facility management, operational research, health metrics", "contribution statement": "This paper provides a novel methodological framework for