R for HTA 2025: Shaping the Future of Health Economic Modelling

Last week, the R for HTA Workshop (6–10 June) brought together over 200 health economists, statisticians, analysts, and policy experts to explore the expanding role of R in Health Technology Assessment (HTA). The sessions highlighted a mix of methodological innovation, hands-on application, and strategic dialogue around adoption and standardisation of R-based workflows in HTA.

The in-person session on Friday 6th June, held in Belfast, enjoyed uncharacteristically beautiful weather. Two themes emerged strongly throughout the day: interoperability and open source. Presentations covered a range of technical and practical topics, including structuring inputs in MS Excel, automating modelling reports and network meta-analysis outputs, exporting R models to Excel, and running discrete event simulations in R.

A review of open-source HTA models (Henderson et al., 2025) noted the growing dominance of R, particularly when combined with collaborative tools such as GitHub.

The conversations surrounding the event were just as valuable as the formal sessions. There are early plans for a monthly R for HTA book club to coincide with the launch of a new book, and a potential hackathon is being considered for the autumn.

A recurring theme in the discussions was software design philosophy - specifically, a shift away from large, multifunctional packages towards smaller, more focused tools that do one thing well and are easier to maintain.

In the Monday 9th June session presentations explored advanced topics such as input correlation in probabilistic sensitivity analysis (PSA), extrapolation of survival curves using Bayesian methods, and cost-effectiveness modelling for real-world screening programmes. These talks illustrated how R, sometimes used in combination with Julia or C++, can deliver both analytical flexibility and computational efficiency.

Several software tools were introduced, including interactive shiny apps for treatment comparison visualisation, secure pipelines for working with sensitive data, and platforms for extracting structured data from published tables using language models and ontologies. The diversity of tools and techniques reflected a growing ecosystem of open-source infrastructure tailored to the needs of health economics and outcomes research.

The day concluded with a panel discussion on how R can serve as a bridge between clinical evidence generation and economic modelling. Panellists highlighted the importance of integrating statistical and economic functions within organisations and discussed how R could serve as a shared language to support this integration.

The Tuesday 10th June session shifted focus from tools to strategy - specifically, how R can be embedded in health technology assessment processes globally.

Discussions centred on agency guidance, technical validation, and long-term sustainability of models and codebases. Case studies illustrated how organisations are developing full economic models in R for areas such as oncology and chronic disease, while still facing challenges in markets that default to Excel.

There was also a deep dive into the use of large language models (LLMs) to assist with summarising model documentation - showing both the promise and the pitfalls of automation in regulatory-facing work.

A recurring theme was the mismatch between the capabilities of R and current institutional expectations. While R offers clear advantages - reproducibility, modularity, and rich visualisation - many agencies still lack formal guidance or established review processes for code-based submissions. This gap will gradually narrow as more graduate health economists -trained in R during their MSc programmes - enter the workforce and join the vanguard championing the clarity and efficiency of script-based models, resisting a return to spreadsheet complexity.

Two concrete priorities emerged from the workshop:

1.  Establishing a standardised package ecosystem for health economics in R - a HEverse to rival the tidyverse - has gained significant momentum. Many presentations featured packages such as assertHE for model visualisation (Smith et al., 2024), BCEA for Bayesian cost-effectiveness analysis (Baio & Heath, 2017), and flexsurv for survival analysis (Jackson, 2016). However, maintaining R packages is resource-intensive. Dividing this responsibility into smaller, manageable components could reduce reliance on key individuals and distribute the workload more sustainably. This approach, however, raises the challenge of interoperability: how can these packages effectively communicate and work together? Building a cohesive ecosystem would be a valuable yet daunting task, and no single organisation currently has the incentive to lead it. Since open-source software is non-excludable and non-rivalrous, it naturally suffers from the free-rider problem.

2.  Developing validation frameworks to support regulatory confidence is critical for the adoption of R in formal submissions. Such frameworks must strike a balance between flexibility to foster innovation and robustness to assure reviewers. The efficiency gains from pre-validated code - compared to repeatedly reviewing thousands of lines of code, or millions of cells in a spreadsheet - offer a compelling rationale for investment in these frameworks. But who pays for this? The free-rider problem remains a significant barrier.

All presentations from the workshop are available on the R-HTA YouTube channel.

Slides and supporting materials can be accessed on GitHub.

For further learning, a curated list of relevant courses is available here.

Additionally, a recommended reading list on health economic evaluation can be found here.

Thanks to Felicity Lamrock, Howard Thom, Nathan Green, Gianluca Baio and the R for HTA scientific committee for making the event happen.

References

Baio, G., Berardi, A. and Heath, A., 2017. Bayesian cost-effectiveness analysis with the R package BCEA (pp. 153-66). New York: Springer.

Henderson, R.H., Sampson, C., Pouwels, X.G., Harvard, S., Handels, R., Feenstra, T., Bhandari, R., Sepassi, A. and Arnold, R., 2025. Mapping the Landscape of Open Source Health Economic Models: A Systematic Database Review and Analysis. Value in Health.

Jackson, C. 2016. flexsurv: A Platform for Parametric Survival Modeling in R. Journal of Statistical Software, 70(8), 1-33. doi:10.18637/jss.v070.i08

Smith, R.A., Samyshkin, Y., Mohammed, W., Lamrock, F., Ward, T., Smith, J., Martin, A., Schneider, P., Lee, D., Baio, G. and Thom, H., 2024. assertHE: an R package to improve quality assurance of HTA models