Session (CTI)

Separate registration required.

Health economic models developed in R are playing an increasingly prominent role in reimbursement and health technology assessment decisions. While numerous resources exist for building models in R, comparatively little attention has been devoted to reviewing, validating, and adapting existing models. In practice, decision-makers and analysts frequently encounter complex R-based models that they must evaluate, verify, and modify, often with limited familiarity with the original codebase. This course aims to address that gap, drawing on the instructors' direct experience working with and advising HTA bodies including the UK National Institute for Health and Care Excellence (NICE), the Irish National Centre for Pharmacoeconomics (NCPE), and familiarity with the processes of the Dutch Zorginstituut Nederland (Zin) and the Canadian Drug Agency (CDA).

Using a Markov cost-effectiveness model in Atrial Fibrillation as the case study, based on a model developed for UK National Institute for Health and Care Excellence (NICE) guidelines, participants will work through a structured sequence covering model execution, code quality assessment, manual and AI-assisted validation, and practical modification for sensitivity and scenario analyses.

The course begins with an introduction to decision modeling in R, establishing foundational concepts and workflow conventions for reproducible health economic analyses. Participants will then examine the Atrial Fibrillation model's structure, inputs, and outputs before executing it in R to generate base case results.

With the model running, the course turns to coding practice. Participants will learn to recognize well-structured R code, including clear naming conventions, modular organization, documentation standards, and reproducibility safeguards. They will assess whether the case study model adheres to these standards.

Validation progresses through two complementary approaches. Participants will first perform manual extreme-value and unit tests, designing targeted checks that probe model behavior at boundary conditions and verify that individual components produce expected outputs.

The course then introduces an agentic artificial intelligence approach that automates the same structured validation process, demonstrating how AI can replicate and extend what participants learned to do manually. This pairing illustrates the progression from understanding validation principles to scaling them efficiently, while the analyst retains focus on substantive judgment. R's transparent and readable code provides a natural advantage for AI-assisted review compared to spreadsheet-based models where logic is dispersed across cells and tabs.

The course also covers NICE's position statement on the use of AI in evidence generation and reporting and shows it in action: as participants use AI to quality-control the case study model, they will see how to declare and describe AI use, keep the analyst accountable, and maintain transparency and reproducibility.

Finally, participants will modify the R model to implement sensitivity and scenario analyses, adjusting parameters, restructuring assumptions, and generating alternative results. This exercise consolidates skills from the full course, requiring participants to understand the model well enough to make targeted, purposeful modifications.

By the end of the course, participants will be equipped to independently assess, validate, and adapt R-based health economic models encountered in HTA submissions, academic review, and research collaboration. Participants who wish to gain hands-on experience must bring their laptops with R/R Studio installed. An online version of RStudio will be provided prior to the course as a backup. 

PREREQUISITES: Basic R usage, Health economic decision modeling, modeling in health technology assessment, and Markov models.