Cost-effectiveness analysis with informative missing data: tools and strategies

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 conditional on the observed data, assuming the data to be ‘missing at random’ (MAR). In many settings, for example when responses to self-reported health-related quality-of-life (QOL) questionnaires are incomplete, it seems however plausible to assume that the data are ‘missing not at random’ (MNAR, or ‘informative’). For example, patients whose true health status is relatively poor may be less likely to return QOL questionnaires, even after conditioning on the observed data. In these circumstances, methodological guidance recommends assessing whether cost-effectiveness inferences are sensitive to different missing data assumptions, but most published CEA studies fail to do this. My PhD aims to develop a practical and accessible framework to analyse, report, and interpret cost-effectiveness analysis of randomised trials with informative missing data.

In the first half of this seminar, we will introduce the issue of missing data and informative missingness in cost-effectiveness analysis, and describe some of the methods that have been proposed to perform analysis under MNAR. We will illustrate these methods in a CEA of the 10 Top Tips trial, which evaluates a primary care intervention for weight loss.

The presentation will be followed by a group discussion, for those interested. The aim will be to discuss the direction of the PhD research and the development of methods for CEA in presence of informative missing data, using either individual patient data or decision models. In particular, we would like to have your views on: what are the barriers to wider use of MNAR sensitivity analyses? Which analytical approaches appear to be the most relevant to CEA? How to present the results in a way that is readily understood by analysts and decision makers?

Note that the seminar will be split into two parts of 45 minutes; you are welcome to attend either or both parts.