Tutorial: Causal Inference in Biostatistics, Epidemiology and Social Sciences – A Hands-on Introduction to DAGs, g-Methods, Target Trial Emulation, and Cloning

IDTag | ZeitpunktSpracheReferent*innenInstitution
ID183So 08.09.24 | 13.30-18 UhrENL. R. Hallsson,
U. Siebert,
F. Kühne
Department of Public Health, Health Services Research and Health Technology Assessment, UMIT TIROL – University for Health Sciences and Technology, Hall in Tirol

Abstract

The format includes theoretical lectures, hands-on practical training exercises and polls, case examples drawn from the published literature, as well as an interactive breakout session, discussions, and Q&A parts. Participants are invited to discuss their own research problems.

Background and Motivation
Good medical and public health decisions need robust empirical evidence on causal effects of exposures, interventions, or policies. Therefore, biostatisticians, epidemiologists, data scientists and decision analysts need skills and practical tools to apply appropriate methods to derive causal effects instead of merely statistical associations from data and assess when published effect estimates have a causal interpretation and when not.
This tutorial is directed to all stakeholders in medical and public health decision making, researchers from all substance matter fields including statisticians, epidemiologists, outcome researchers, health economists, modelers and health policy decision makers who are interested either in methods or interpretation of empirical studies with causal research questions.

Content
We will introduce causal inference frameworks and methods that are needed for the design and analysis of big real-world observational data and pragmatic trials. We will cover the principles of causation, the use of causal diagrams (directed acyclic graphs, DAGs), and focus on causal inference methods for time-independent confounding (multivariate regression, propensity scores) and time-dependent confounding (g-formula, marginal structural models with inverse probability weighting, and structural nested models with g-estimation). The “target trial” concept and a counterfactual approach with “cloning – censoring – weighting” will be used as tools to avoid self-inflicted biases (e.g., immortal time bias). We will demonstrate how causal methods are applied to big real-world datasets with case examples from oncology, cardiovascular disease, infectious diseases, nutrition, and obstetrics.

Objectives
The objectives of this hands-on tutorial are to draw and interpret causal diagrams, develop a target trial protocol, understand causal study designs, decide which biostatistical/epidemiological methods must be used in different situations, and to be able to apply these methods in specific situations, such as single-arm trials with external control arms (ECAs) or large real-world data linking medical records with claims data.


After attending this tutorial, participants should be able to:

  • Define causal interventions and actions, draw, and interpret causal diagrams, and apply the rules of causal diagrams to distinguish causal from non-causal statistical associations.
  • Develop a target trial protocol and be able to use “clones” in real-world data analysis.
  • Decide which biostatistical/epidemiological methods must be used in different situations to derive causal effect parameters.
  • Design observational studies for valid causal effect estimation and target them to specific settings such as single-arm trials or claims data.
  • Describe how these methods are applied in big real-world data.
  • Understand the potentials, challenges, assumptions, and limitations of causal frameworks.
  • Know how health technology assessment agencies currently handle innovative causal inference methods and target trial emulation.

Tutorial material includes all session handouts, exercises with solutions, a comprehensive background reading library, and software recommendations.

Kontakt

lara.hallson [at] umit-tirol.at