Target Trial Emulation for Impact Evaluation of Group Health Policies

Read full text Summarise with VIRAT
Authors: Y Xu
Cited by: 0

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

This thesis focuses on the use of the target trial emulation framework to study the impacts of group-level health policies using routinely collected data. Target trial emulation is a framework for evaluating causal effects from observational data whileadopting design principles from randomized controlled trials, helping to avoid common pitfalls in observational studies that can cause bias. Although it has grown in popularity for epidemiological research, there are few examples of its application in health service policy evaluation, where interventions are often complex and implemented at the group level. This work uses target trial emulation combined with causal inference analysis methods to investigate the impacts of two health service initiatives implemented in general practices in England– online consultation systems, which are part of the Digital First Primary Care Model (DFPC), and the Accelerate Access Improvement Programme (AAIP)– which were designed to improve primary care service and access. Analyses are performed using data from a range of sources on interventions, outcomes and confounders recorded at the cluster level of general practice. Overviews are provided of the range of data sources providing the key variables and of the causal inference methods available for estimation of both point treatments and longitudinal treatments. The focus is on three main estimation methods: inverse probability weighting estimation of a marginal structural model, G-computation, and targeted maximum likelihood estimation. To mitigate bias arising from model misspecification, machine learning methods and non-parametric variable selection techniques are considered. The thesis also provides practical examples of addressing key methodological challenges: (1) defining causal estimands for group-level health policies; (2) handling time-varying treatments; (3) accounting for clustering in estimation; (4) addressing bias due to missing outcomes with inverse probability of censoring weighting. The methodological and practical challenges in applying target trial emulation to group-level policies are highlighted through the applications, to estimate the impacts of online consultations on emergency department visits and of the AAIP programme on patient satisfaction. This work demonstrates how target trial emulation can strengthen the evaluation of group-level health services policies using observational data, and how comparing multiple causal estimation methods enhances the robustness of the findings. It also highlights the potential of target trial emulation as a rigorous framework for informing health services policy decisions and improving primary care delivery.

Fields of Study & Subjects

Advanced Causal Inference Techniques Evaluation and Performance Assessment Health Policy Implementation Science

Publication details

Journal / source
LSHTM Research Online (London School of Hygiene and Tropical Medicine)
Publication date
1 January 2026
Language
EN
Document type
Thesis