OBJECTIVES: Complex Excel-based cost-effectiveness models in health technology assessment (HTA) suffer from prohibitive computational run-time to perform probabilistic sensitivity analysis (PSA). Previous work aimed to shift Excel users to the R programming language through tutorials, example models, and R tools that automate PSA. Our objective is to circumvent this process with the REEEVR tool that can convert Excel models into R automatically.

METHODS: Using concepts from programming language interpreters we parse, tokenise, generate abstract syntax trees, and translate each cell of complex Excel-based models into R. This output is then ordered, streamlined, and combined with libraries. The process is wrapped into a graphical user interface.

We tested our tool on an Excel-based 4-state Markov model comparing four implants for hip replacement surgery, using 1000 PSA samples. This followed models from academia, industry, and the UK National Institute for Health and Care Excellence (NICE) Guidelines Development Group (GDG). We compared total costs, quality adjusted life years (QALYs), and net monetary benefits (NMB) at £20,000/QALY to verify the conversion. A hand-coded R model was also used for comparison.

RESULTS: The output of the converted R model exactly matched those of the Excel-based model for all implants; in both models, the NMB for the best implant was £209,790 (95% credible interval £201,143 – 205,310). The run-time was reduced from approximately 20.5 seconds in Excel to 0.26 seconds in converted R, an 80 times reduction. This was also substantially lower than approximate 1.37 second run-time of the hand-coded R model.

CONCLUSIONS: Our REEEVR tool can exactly convert a probabilistic Markov model from Excel into R with a greatly reduced run-time. The converted R model can also be more efficient than hand-coded R models. The example was informed by academia, industry and the NICE GDG, suggesting the REEEVR tool could have wide applicability in HTA.