Propensity Score estimation and matching for multiple treatment comparisons in observational studies

Gastrostomy feeding is commonly used to support patients with amyotrophic lateral sclerosis who develop severe dysphagia. Data is available from a large, longitudinal, prospective cohort study (ProGas), which enrolled patients with a diagnosis of definite, probable, laboratory supported, or possible amyotrophic lateral sclerosis who had agreed with their treating clinicians to undergo gastrostomy at 24 motor neuron disease care centres or clinics in the UK. The aim of the ProGAs study was to compare three gastrostomy insertion approaches (percutaneous endoscopic gastrostomy (PET); radiologically inserted gastrostomy (RIG) and per-oral image-guided gastrostomy (PIG) in terms of safety and clinical outcomes. The primary outcome of the study was 30-day mortality after gastrostomy. Since ProGas was not a randomised controlled trial, we addressed the issue of treatment indication bias by undertaking a post-hoc propensity score analysis.

The use of propensity scores (PS) to control for pre-treatment imbalances on observed variables in non-randomized or observational studies examining the causal effects of treatments or interventions has become widespread over the past decade. Until recently, the PS method has been applied exclusively for two treatment comparison settings (e.g. treatment vs. control) despite that it is frequently of interest to compare more than two treatments or interventions in health research. PS covariate adjustment, inverse probability weighting (IPW) estimator, and PS matching are the three PS approaches commonly seen in two treatment comparisons.

The proposed research plan would be to undertake a review of the statistical literature on methods for estimating propensity scores for multiple treatment comparisons in observational studies; and the methods for matching these PS with multiple treatment comparisons; followed by an audit of recently published multiple treatment comparison observational studies which used propensity scores to determine what PS estimation and matching methods are commonly used. The PS estimation methods and PS matching and their effect on statistical analysis and conclusions will be compared using data from several observational studies with multiple treatment comparisons. The project would then involve some computer simulation and analysis to compare the different methods of PS estimation and matching with the view to developing guidance on how to estimate Propensity Scores for multiple treatment comparisons and how to use or matching on the propensity score in the analysis of the outcomes in multiple treatment comparisons in observational studies.

Eligibility:
Candidates should possess either one of the following:
A First class honours degree in mathematics or statistics, or
An Upper Second class honours degree in mathematics or statistics and a Masters qualification in statistics.

How to apply:
Follow the link in this advert to our online application form.
Select a start date of 1 October 2019 and ensure that you complete the scholarships section of the application form by the deadline.
Please ensure that the project title and supervisors’ names are included in the application.