Integrating real-world data and discrete event simulation cost-effectiveness analysis: an introductory tutorial.
TL;DR
Combining real-world data and DES can enable cost-utility studies to be conducted, thus streamlining the process of economic evaluation, and additional tools are required, such as probabilistic sensitivity analysis.
Abstract
INTRODUCTION This study provides an introductory tutorial on conducting a simplified cost-utility analysis of a new treatment for a generic cariovascular disease, combining discrete event simulation (DES) and real-world data (RWD). To help readers understand the proposal, we present an application of this approach to a RWD database, along with files containing its implementation in Excel and R and the explanation in Spanish in the Supplementary material.
Method
This dataset was used to calculate quality-adjusted life years (QALY) and costs for each individual in the observed alternative (standard of care) by combining longitudinal and survival data. Simulated alternative (counterfactual) results were obtained by adding hazard ratios (HR) from published randomised clinical trials to the parametric survival analysis within a discrete event simulation (DES) model. Bootstrapping allowed to assess uncertainty. Life expectancy was estimated using parametric survival analysis in order to extend the analysis to the patient's death.
Results
The exercise files in Excel and R are available in the OSF repository (https://osf.io/gyuq9/overview?view_only=748bc600f59646f1b144e21458c50a42). These outline the steps for calculating the QALYs and costs for the standard and new alternatives based on the dataset. The incremental cost-effectiveness ratio is then calculated and the uncertainty of the result analysed using the cost-effectiveness plot and acceptability curve obtained via bootstrapping.
Conclusions
Combining real-world data and DES can enable cost-utility studies to be conducted, thus streamlining the process of economic evaluation. In order to apply this proposal to research studies, additional tools are required, such as probabilistic sensitivity analysis.