°®ÍþÄÌapp

Due to a change in circumstances, the speaker will now be joining us virtually. The lecture remains hybrid, and attendees are welcome to participate in person or online as originally planned. 

This is the 34th Joint LSHTM-RSS annual lecture on perspectives on statistics in medicine (formerly known as the Bradford Hill Memorial Lecture) which has been running since 1992. The speaker this year is Professor Vanessa Didelez, the Professor of Statistics and Causal Inference at the Leibniz-Institute of Prevention Research and Epidemiology – BIPS, Bremen, Germany, in a joint appointment with the Department of Mathematics and Computer Science of the University of Bremen.

When conducting causal analysis with observational data, potential unmeasured confounding often takes centre stage. In this presentation, Professor Vanessa Didelez will discuss other threats to meaningful causal inference commonly encountered in time-dependent settings, specifically "wrong question bias" and, related to that, problems induced by design choices. A lack of clarity in formulating the research question can lead to study designs or analyses that may not properly address the intended question. In causal inference, the problem often arises from ignoring or not fully appreciating the decision-making context; specifically, focusing on questions that are irrelevant to actionable decisions or policy. These issues can arise in many different settings, especially with longitudinal or time-to-event data, such as in the context of (longitudinal or dynamic) causal mediation analysis, when competing risks are present, in Mendelian randomisation with time-dependent exposures, or when aiming to use prediction to inform treatment decisions. As for design-induced problems, prominent examples include 'prevalent-user' or 'immortal-time' bias. However, more subtle issues can also arise from unfavourable design choices, such as lack of alignment at time zero. In many collaborations, we found that following the principles of target trial emulation raised awareness and helped us avoid or mitigate many problems, both with formulating the research question as well as choosing a suitable design. Professor Vanessa Didelez will review some of the problems illustrating them through a variety of practical examples, such as the evaluation of colonoscopy screening and the analysis of modifiable behaviours in children's health.

Speaker

Professor of Statistics and Causal Inference at the Leibniz-Institute of Prevention Research and Epidemiology – BIPS, Bremen, Germany, in a joint appointment with the Department of Mathematics and Computer Science of the University of Bremen.

Event notices

• Please note that you can join this event in person or you can join the session remotely.
• Please note that the recording link will be listed on the page when available.