PRM62 The Impact of Centre Selection on the Generalisability of Economic Evaluation Results from Multi-Centre Randomised Controlled Trials

Abstract

Economic evaluation (EE) estimates for individual centres in multi-centre randomized controlled trials (RCTs) can differ significantly from the trial-wide result. The existing methods addressing the generalisability of EE results from RCTs (e.g. bivariate hierarchical modelling) assume that the recruiting centres are representative for their jurisdictions, but this assumption has not been generally verified. No explicit method of selecting centres and their recommended sample sizes has been described, despite having been suggested in the literature. The working hypothesis is that transparent centre selection is a crucial step in assessing the generalisability of EE results from RCTs. Two questions arise: 1) What criteria underpin the current practice of selecting centres for RCT-based EEs? and 2) Can a valid quantitative algorithm be formulated to assist the centre selection process at the trial design stage? First, the use of modelling-based methods addressing generalisability has to be supported by evidence that centres are representative for the jurisdiction under scrutiny. There is, thus, a need to assess the current practice of selecting centres for RCT-based EEs. Second, a quantitative methodology for purposively selecting centres for RCTs coupled with EEs has to be devised in order to underpin an objective centre selection process. The proposed operational measure is a generalisability index (GIx) which aggregates relevant generic and intervention-specific covariates and can be formulated at both jurisdiction and centre-level. The GIx can be validated against centre-level cost-effectiveness estimates. A successfully validated GIx will provide evidence towards the legitimate use of existing generalisability techniques. The GIx will allow an objective generalisability assessment for centres that did not participate in the RCT. Describing the rationale for centre selection must become a standalone item in reporting checklists for RCTs and EEs. Furthermore, such a methodology will bridge policy and research by correlating jurisdictional interests with RCT design.