Causal inference in public health from large observational databases

Charles ASSAAD

The CIPHOD team focuses on the theoretical exploration and development of innovative approaches to the discovery of causal graphs, the identification of total and direct effects, and the search for the root causes of anomalies, with an emphasis on their practical utility for epidemiologists and exploiting large temporal databases and high-level knowledge.

Team presentation

The general research objectives of the CIPHOD team are to propose new theoretical discoveries and develop innovative methodologies in the field of causal inference, focusing on their usefulness for epidemiologists and working in close collaboration with other IPLESP teams to apply them in epidemiological contexts. These objectives are articulated around three axes, with particular emphasis on large databases, temporal data and high-level knowledge (abstractions). The three axes are:

  • the discovery of causal graphs
  • identification and estimation of total and direct effects, and
  • finding the root causes of anomalies.

For the first axis, our aim is to design methods for constructing a causal graph from observational data, with particular attention paid to data from various environments such as different hospitals. We aim to develop approaches that preserve patient confidentiality while being targeted, for example by identifying a part of the graph that is sufficient to answer the questions posed in the second and third axes.

For the second axis, we aim to find the conditions for identifying total and direct effects from observational data, as well as determining the optimal method for estimating them when it is necessary to identify the effects of a particular intervention.