Causal Models for Clinical Decision Making
1st AIxIA Summer School on Artificial Intelligence for Healthcare
07-11 July 2025, Trento, Italy
1st AIxIA Summer School on Artificial Intelligence for Healthcare
07-11 July 2025, Trento, Italy
Why do we need causality in clinical decision making? The point of view of a clinical oncologist
S. Provenzano
Medical decisions require an understanding of cause-and-effect relationships to ensure that interventions lead to the desired outcomes. Causality is particularly relevant especially when dealing with complex patients (like cancer patients) because it allows us to choose a specific treatment instead of another according to prognosis. Without causation, decision-making relies on incomplete or misleading information, potentially leading to ineffective or harmful interventions.
Elements of causal models: definition, inference and discovery
A. Zanga
This talk will provide participants with the technical tools needed to design and implement a causal network. We start by introducing the concepts of graphs, probabilistic graphical models and causal graphs. Moreover, participants will learn how to take advantage of experts’ knowledge to derive the cause-effect relationships, shifting from an associational analysis to a causal approach. Then, we will explain the role of causal discovery algorithms to learn new relationships from healthcare data.
Causal networks and their application in healthcare: an overview of a case study on cancer survivors
A. Bernasconi
In this talk we will present the process behind the development of causal networks built using real-world data. To do that we will take as a practical example a causal Bayesian network to study cardiovascular diseases in young breast cancer survivors as a consequence of cancer treatment. This talk will focus on both the process behind the development of such a model and on how it may help clinicians in properly tailor personalized follow-up strategies for these patients.
Practical session - Implementing a causal network: building a causal network from experts’ knowledge and data
Participants will be divided into mixed inter-disciplinary groups and will be asked to answer a practical research question building a causal network model.
SALVATORE PROVENZANO, MD
Department of Medical Oncology, Fondazione IRCCS Isituto Nazionale dei Tumori di Milano
Senior Physician in Medical Oncology with expertise and commitment on soft tissue & bone sarcomas. Currently working on the clinical and translational research in a rare cancer Medical Oncology Unit, with a focus on health networking in cancer care.
ALESSIO ZANGA, PhD
Department of Informatics, Systems and Communication (DISCo), University of Milano-Bicocca
Postdoctoral researcher at the Models and Algorithms for Data and Text Mining Laboratory (MADLab), Department of Informatics, Systems and Communication (DISCo), University of Milano-Bicocca. His main research interest is in the application of (federated) causal discovery in medicine and healthcare.
ALICE BERNASCONI, PhD
Department of Epidemiology and Data Science, Fondazione IRCCS Isituto Nazionale dei Tumori di Milano
Senior researcher at the Evaluative Epidemiology Unit since 2017. Her main research areas are incidence, prognosis and assessment of long-term effects of cancer in adolescents and young adults, with a particular interest in causal models.