Summary: This paper outlines appropriate statistical methods for conducting cost-effectiveness analyses alongside cluster randomized trials, where ignoring clustering can substantially underestimate uncertainty. Using data from a South African pragmatic trial, the authors demonstrate five valid analytic approaches: joint modeling of costs and effects using Bayesian hierarchical models, two-stage non-parametric bootstrapping that resamples clusters before individuals, and net benefit regression using least squares with robust errors, hierarchical linear models, and Bayesian regression. Across methods, results were broadly consistent, but all showed greater uncertainty once clustering was properly accounted for—illustrated in the wider ICER confidence intervals on page 5. The paper emphasizes the importance of handling skewed costs, correlated outcomes, and intra-cluster correlation when estimating incremental cost-effectiveness. It concludes that several theoretically coherent methods are available in common software such as Stata and WinBUGS, and that analysts should avoid approaches that treat individuals as independent, as doing so yields spuriously precise estimates. 

Summary: This study demonstrates how statistical choices can substantially alter the results of trial-based economic evaluations. Using two empirical datasets (REALISE and HypoAware), the authors perform 14 sequential analyses per study, progressively addressing four common statistical challenges: baseline imbalances, skewed cost data, correlation between costs and effects, and missing data. As shown in Tables 3 and 4, ignoring these issues produced markedly different incremental costs, QALYs, and ICERs compared to analyses that accounted for all challenges. Missing data methods and baseline adjustment had the greatest influence on both point estimates and uncertainty, while accounting for skewness and correlation had smaller but still meaningful effects. The cost-effectiveness acceptability curves on pages 8–9 illustrate how probabilities of cost-effectiveness can shift notably depending on analytic approach. The authors conclude that misalignment between data characteristics and statistical methods risks misleading decisions, and they provide Stata code to support more rigorous, challenge-aligned analyses. 

Summary: This tutorial provides practical, step-by-step guidance for conducting high-quality trial-based economic evaluations in R. It focuses on four major methodological challenges—missing data, skewed costs and effects, correlated cost-effect pairs, and baseline imbalances—and demonstrates how to combine appropriate statistical methods to address them simultaneously. The paper explains the rationale for using multiple imputation, non-parametric bootstrapping, regression-based adjustment, and seemingly unrelated regression (SUR), then illustrates their implementation using a simulated dataset. The included R code shows how to calculate total costs and QALYs, estimate incremental cost-effectiveness, construct uncertainty measures, and generate visual outputs such as cost-effectiveness planes (page 7) and CEACs (page 7). The authors also discuss extensions for clustered designs, linear mixed models, multi-arm trials, and Bayesian alternatives. Overall, the tutorial equips analysts with a ready-to-use workflow for producing statistically robust, transparent, and reproducible economic evaluations alongside randomized trials. 

Summary: This scoping review examines how artificial intelligence can enhance health technology assessment, particularly its economic components. Drawing on 38 studies, it finds that AI strengthens HTA by improving data collection, evidence synthesis, cost-effectiveness analysis, and forecasting. Applications include predictive modeling, simulation tools, analysis of real-world data, and integration of clinical, genetic, and social datasets to support more accurate assessments of costs, outcomes, and resource allocation. AI also contributes to precision medicine, diagnostic innovation, behavioral modeling, and real-time analytics. However, the review highlights significant challenges: data quality and interoperability, privacy and security risks, bias, lack of standardized evaluation frameworks for AI-driven technologies, and regulatory gaps. Robust governance, ethical safeguards, and improved infrastructures are required to support responsible implementation. The authors conclude that AI offers substantial potential to transform HTA, but its benefits depend on deliberate planning, stakeholder collaboration, and the development of consistent methodological and regulatory standards.