DECISION MODELLING USING R

Background: Economic evaluations often rely on decision models. As decision modelling advances, more complex models are being designed to better represent the underlying clinical conditions. In addition, decision models increasingly rely on new statistical and mathematical techniques (e.g. model calibration, value of information, evidence synthesis). Current commercially available tools provide limited flexibility in embedding such complex statistical approaches within the decision model framework. R is an environment where statistical analysis on primary or literature data can be combined with decision modelling and the results can be presented in publication ready form. Hence, R can embed most components of decision modelling within a single framework. Among other benefits, this facilitates sensitivity analyses while accounting for model uncertainties. In addition, R can accommodate more complex model structures and improve computational times. Finally, the fact that R is freely available improves model transparency and reproducibility.

Format Requirements: The course will focus on how to implement simple decision trees and Markov models using R. Therefore a good understanding of decision modelling is necessary. The participants will be expected to have some experience with designing decision trees and Markov models and have basic understanding of statistics for economic evaluation. Finally, basic knowledge of R is required. Participants will need to bring their own laptops for practical exercises.

Description and Objectives: The course will cover the basics of building decision models using R. First a very brief review of  basic R functions that are commonly used in decision modelling (import/export data, data handling, basic distributions, "if" and "for" loops etc.) will be provided. Next, a simple decision tree will be constructed using R. A base-case analysis, as well as one way and probabilistic analyses will be conducted. Subsequently, a simple Markov model will be designed using R. Base case and multi-way sensitivity analysis will be conducted. Results of both models will be presented in tabular and graphical form. Principles of good modelling practices using R (e.g. consistency, proper documentation etc) will be outlined. Finally, advanced functions of R in decision modelling will be briefly discussed. Examples include building microsimulation models or integrating  network meta-analyses and decision models using R.

By the end of the course the participants will be able to:

• Build a simple decision tree in R
• Build a simple Markov model in R
• Assign distributions to model parameters and conduct probabilistic analysis in R
• Create and export tabular and graphical representations of the results of the decision model.
• Have a broader understanding of the advantages associated with using R in decision modelling.

All R programming templates for decision modelling will be provided to participants after the course for future use.