Bayesian Modelling of Healthcare Resource Use in Multinational Randomized Clinical Trials

Abstract

Most cost-effectiveness analyses conducted alongside multinational randomized clinical trials (RCT) are carried out applying the unit costs from the country of interest to trial-wide resource use items with the objective of estimating total healthcare costs by treatment group. However, this approach could confound 'price effects' with 'country effects'. An alternative approach is to use multilevel modelling techniques to analyse healthcare resource use (HCRU) from the trial, and obtain country-specific total costs by applying country-specific unit costs to corresponding shrinkage estimates of differential HCRU. To illustrate the feasibility of this approach, we analysed data from twin multinational RCTs, which enrolled approximately 2000 individuals into three treatment arms for the management of patients with chronic respiratory disease. The models were implemented using Bayesian multilevel models, to reflect the hierarchical structure of the data while controlling for co-variates at the patient and country level. This analysis showed that directly modelling the level of HCRU is a promising approach to facilitate cost-effectiveness analyses conducted alongside multinational RCTs, offering several advantages compared with the modelling of direct costs. It is argued that modelling the level of HCRU within the Bayesian framework avoids confounding the price effects with the country effects and facilitates the estimation of costs for several countries represented in the trial.