Abstract: Missing data is a common issue in cost-effectiveness analysis (CEA) of randomised trials. Methods such as multiple imputation are now commonly used to account for the missing values, assuming the data to be ‘missing at random’ (MAR). In many settings, it seems however plausible that the data may be ‘missing not at random’ (MNAR, or ‘informative’). For example, patients whose health status is relatively poor may be less likely to return quality-of-life questionnaires. In these circumstances, guidelines recommend assessing whether conclusions are robust to different missing data assumptions. But this is rarely done in practice, perhaps due to a lack of clear guidance on how to conduct such sensitivity analysis.
In this presentation, I will start by a review of the current practice for addressing missing data in trial-based CEA. I will then outline several possible approaches for conducting sensitivity analysis, and focus on one particularly accessible approach based on multiple imputation. Its’ implementation will be illustrated with a trial of a brief intervention for weight loss in primary care. I will finish by discussing an alternative method for longitudinal data, where the missing data are imputed assuming a distribution borrowed from a reference group.
Biography: Baptiste Leurent is an assistant professor in medical statistics and epidemiology at the London School of Hygiene and Tropical Medicine (LSHTM). He is currently conducting a NIHR-funded PhD on missing data in cost-effectiveness analysis, in the Medical Statistics Department. Previously he worked as a statistician in different research institutions in the UK and Thailand (PHPT, UCL, MRC), on clinical trials and epidemiological studies in HIV, mental health and primary care. Before starting his PhD, he worked at LSHTM with the ACT Consortium, looking at improving the use of artemisinin-based combination therapy for malaria. He is particularly interested in applied methods to improve the design and analysis of randomised trials.