Search

Multilevel models provide a flexible modelling framework for cost-effectiveness analyses that use cluster randomised trial data. However, there is a lack of guidance on how to choose the most appropriate multilevel models. This paper illustrates an approach for deciding what level of model complexity is warranted; in particular how best to accommodate complex variance-covariance structures, right-skewed costs and missing data. Our proposed models differ according to whether or not they allow individual-level variances and correlations to differ across treatment arms or clusters and by the assumed cost distribution (Normal, Gamma, Inverse Gaussian). The models are fitted by Markov chain Monte Carlo methods. Our approach to model choice is based on four main criteria: the characteristics of the data, model pre-specification informed by the previous literature, diagnostic plots and assessment of model appropriateness. This is illustrated by re-analysing a previous cost-effectiveness analysis that uses data from a cluster randomised trial. We find that the most useful criterion for model choice was the deviance information criterion, which distinguishes amongst models with alternative variance-covariance structures, as well as between those with different cost distributions. This strategy for model choice can help cost-effectiveness analyses provide reliable inferences for policy-making when using cluster trials, including those with missing data. [ABSTRACT FROM AUTHOR]

Background: The interpretation of trial results can be helped by understanding how generalisable they are to the target population for which inferences are intended. INTERVAL, a large pragmatic randomised trial of blood donors in England, is assessing the effectiveness and safety of reducing inter-donation intervals. The trial recruited mainly from the blood service's static centres, which collect only about 10 % of whole-blood donations. Hence, the extent to which the trial's participants are representative of the general blood donor population is uncertain. We compare these groups in detail.Methods: We present the CONSORT flowchart from participant invitation to randomisation in INTERVAL. We compare the characteristics of those eligible and consenting to participate in INTERVAL with the general donor population, using the national blood supply 'PULSE' database for the period of recruitment. We compare the characteristics of specific groups of trial participants recruited from different sources, as well as those who were randomised versus those not randomised.Results: From a total of 540,459 invitations, 48,725 donors were eligible and consented to participate in INTERVAL. The proportion of such donors varied from 1-22 % depending on the source of recruitment. The characteristics of those consenting were similar to those of the general population of 1.3 million donors in terms of ethnicity, blood group distribution and recent deferral rates from blood donation due to low haemoglobin. However, INTERVAL participants included more men (50 % versus 44 %), were slightly older (mean age 43.1 versus 42.3 years), included fewer new donors (3 % versus 22 %) and had given more donations over the previous 2 years (mean 3.3 versus 2.2) than the general donor population. Of the consenting participants, 45,263 (93 %) donors were randomised. Compared to those not randomised, the randomised donors showed qualitatively similar differences to those described above.Conclusions: There was broad similarity of participants in INTERVAL with the general blood donor population of England, notwithstanding some differences in age, sex and donation history. Any heterogeneity of the trial's results according to these characteristics will need to be studied to ensure its generalisability to the general donor population.Trial Registration: Current Controlled Trials ISRCTN24760606 . Registered on 25 January 2012. [ABSTRACT FROM AUTHOR]