Use this URL to cite or link to this record in EThOS:
Title: Full Bayesian methods to handle missing data in health economic evaluation
Author: Gabrio, Andrea
ISNI:       0000 0004 7964 9928
Awarding Body: UCL (University College London)
Current Institution: University College London (University of London)
Date of Award: 2019
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
Trial-based economic evaluations are performed on individual-level data, which almost invariably contain missing values. Missingness represents a threat for the analysis because any statistical method makes assumptions about the unobserved values that cannot be verified from the data at hand; when these assumptions are not realistic, they could lead to biased inferences and mislead the cost-effectiveness assessment. We start by investigating the current missing data handling in economic evaluations and provide recommendations about how information about missingness and related methods should be reported in the analysis. We illustrate the pitfalls and issues that affect the methods used in routine analyses, which typically do not account for the intrinsic complexities of the data and rarely include sensitivity analysis to the missingness assumptions. We propose to overcome these problems using a full Bayesian approach. We use two case studies to demonstrate the benefits of our approach, which allows for a flexible specification of the model to jointly handle the complexities of the data and the uncertainty around the missing values. Finally, we present a longitudinal bivariate model to handle nonignorable missingness. The model extends the standard approach by accounting for all observed data, for which a flexible parametric model is specified. Missing data are handled through a combination of identifying restrictions and sensitivity parameters. First, a benchmark scenario is specified and then plausible nonignorable departures are assessed using alternative prior distributions on the sensitivity parameters. The model is applied to and motivated by one of the two case studies considered.
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available