PRM120 - GRAPHICAL USER-FRIENDLY INTERFACE FOR COMPLEX HEALTH ECONOMIC MODELS BUILT IN R

Abstract

As health economic (HE) models become more complex, the level of expertise required to use R-based models increases, which is a limiting factor for using such models since often users are not experienced in R. Our main objective was to develop a friendly graphical user interface (GUI) in R for a recently developed HE model, which can be used without having experience in R. We developed an interactive dashboard in the R package Shiny for the COPD model presented by Hoogendoorn et al. (Value Health 20(9):A644). This is a complex patient-level model which includes a wide variety of COPD patient (sex, age, BMI-class, smoking, diabetes, depression...) and disease characteristics (reversibility, emphysema, eosinophil-level...). The model supports probabilistic analyses, which adds complexity since it requires a double loop. All these features were translated into a flexible GUI in which all combinations of characteristics can be selected to define subgroups for which deterministic and probabilistic simulations can be run. Binary patient characteristics are selected with radio buttons and continuous characteristics by moving sliders along pre-defined ranges. Mean values for all characteristics are displayed on the screen. The model can perform single-arm and two-arm simulations. Clinical results can be produced separately, which gives the opportunity to run the model afterwards with a change in cost only and save running time. After indicating common input parameters (effects, costs, discount rates...), the user decides whether the simulation will be deterministic or probabilistic. Result tables and figures are presented for intermediate (lung function, symptoms, quality of life, exacerbations...) and final (mortality, cost-effectiveness) outcomes, which can be downloaded in MS-Word format. Despite model complexity, the GUI can be used regardless R user’s experience, enhances model use and, in combination with the raw R-code, facilitates quality control. The results provided are useful for researchers and decision-makers.