"background": "District hospitals are critical nodes in healthcare delivery, yet systematic evaluations of their system-level performance and its impact on clinical outcomes in sub-Saharan Africa are methodologically limited. Existing assessments often rely on aggregate data, failing to account for the hierarchical structure of patient care within hospital systems. ", "purpose and objectives": "This study aims to methodologically evaluate the performance of district hospital systems in Kenya by developing and applying a multilevel regression framework to analyse variations in clinical outcomes attributable to hospital-level system factors. ", "methodology": "We conducted a retrospective cohort analysis using anonymised patient records from a nationally representative sample of district hospitals. A three-level hierarchical logistic regression model was specified: (p{ijk) = \0 + \ Xijk + ujk + vk, where pijk is the probability of a positive outcome for patient i in department j in hospital k, X represents patient-level covariates, and ujk and vk are random intercepts for department and hospital, respectively. Model inference was based on 95% confidence intervals derived from robust standard errors. ", "findings": "Hospital-level system factors, notably pharmacy stock-out rates and specialist clinician density, explained 22% (95% CI: 18 to 26) of the variance in risk-adjusted inpatient mortality. A one-standard-deviation increase in specialist density was associated with a 15% reduction in the odds of mortality. ", "conclusion": "The methodological approach demonstrates that a significant proportion of variance in clinical outcomes is attributable to modifiable hospital system characteristics, moving beyond patient-level explanations. ", "recommendations": "Hospital performance evaluations should adopt multilevel modelling to isolate system-level effects. Policy should prioritise investments in human resources and supply chain reliability to improve clinical outcomes. ", "key words": "health systems evaluation, multilevel modelling, clinical outcomes, district hospitals, sub-Saharan Africa", "contribution statement":
Wanjiku Mwangi (Thu,) studied this question.