Skip to main content

The Causal Modelling Group comprises statisticians and methodologists at King’s College London interested in drawing causal inferences from study data. Many empirical investigations aim to make causal inferences from observational studies or experiments. Causal questions range from “What would be the benefit of a new treatment?” or “What would be the harm of exposure to a risk factor?” to “How does this treatment/exposure change an outcome?” (effect mediation) or “Who would benefit/be harmed?” (effect moderation). The latter is of particular relevance in trials of mental health interventions (therapies) where a person’s ability to benefit might depend on baseline characteristics (predictive markers informing stratified medicine) or on aspects of therapy (process variables). More recently there has been an interest in applying methods developed in the field of causal inference to emulate trials using observational data including routinely collected data from electronic health records. We were also able to utilise such method during the pandemic to evaluate causal effects of policies, for example the impact of the lockdown interventions on mental health service use.

In our group we discuss how to conceptualise the causal question of interest, review principled methods for estimating such causal parameters – that is explicitly state the assumptions under which inferences are valid, and consider new methods that might allow us to draw causal inferences under weaker assumptions.


PhD students

          Project title: -  Emulating cardiovascular trials using KCH’s EHRs and CogStack)

         Project title -  Virtual Trial Emulation using Real World Electronic Health Care Record Data)

         Project title - A unified approach for the statistical analysis of post-randomisation variables in clinical trials)

         Project title - Causal Mediation analysis for 2-level data structures in clinical trials)