{ "background": "The evaluation of health system innovations in low-resource settings requires robust methodologies to measure real-world adoption. Randomised field trials (RFTs) are considered a gold standard for impact evaluation, yet their specific application and methodological rigour in assessing district hospital system adoption in sub-Saharan Africa remain under-scrutinised.", "purpose and objectives": "This systematic review aims to critically appraise the methodological approaches, strengths, and limitations of RFTs used to evaluate the adoption of integrated systems within district hospitals. It seeks to identify common design flaws and best practices to inform future trial design in similar contexts.", "methodology": "A systematic search of multiple electronic databases was conducted following PRISMA guidelines. Peer-reviewed studies employing an RFT design to evaluate a system-level intervention within the district hospital tier were included. Data were extracted on trial design, randomisation unit, adoption metrics, and analytical methods. Quality was assessed using a modified Cochrane Risk of Bias tool for cluster-randomised trials.", "findings": "Of the screened records, a small proportion met the inclusion criteria, indicating a scarcity of high-quality RFTs in this domain. A dominant theme was the frequent use of hospital or catchment area as the unit of randomisation. The primary analytical model employed was a generalised linear mixed model accounting for clustering: $logit(P(Adoption{ij})) = \\beta0 + \\beta1 Treatment{ij} + uj + e{ij}$, where $u_j$ is the random intercept for cluster $j$. Inference was typically based on 95% confidence intervals, with several studies failing to account for intra-cluster correlation, potentially inflating Type I error rates.", "conclusion": "While RFTs offer a powerful design for causal inference, their application in this specific context is limited and often methodologically compromised. The evidence base would benefit from more rigorously designed trials with appropriate sample sizes and analytical techniques that address the hierarchical nature of health system data.", "recommendations": "Future