CardioCausalAI

Exploring Causal AI to Improve Risk Management and Service Delivery in Cardiothoracic Surgery

 

Description:

Cardiovascular diseases, such as heart failure, remain one of the leading causes of mortality worldwide. Cardiac surgery is one of the therapeutic options available for these patients, but it carries a high risk of serious complications, which may occur not only during hospitalisation but also in the postoperative period.

Effective risk and prognosis assessment is crucial; however, traditional risk models rely on linear methods that reflect the “average patient” and fail to capture the complexity of individual pathophysiology. On the other hand, machine learning (ML) has emerged as a promising alternative, capable of analysing large volumes of electronic health records to predict risks and outcomes in cardiac surgery. Despite its potential, the clinical application of ML models in decision support systems (DSS) is still limited due to their non-actionable and non-interpretable nature. Highly predictive variables in ML models are not necessarily causal, which limits their usefulness in guiding clinical interventions.

To bridge this gap, the CardioCausalAI project aims to develop causal artificial intelligence (AI) models that integrate causal inference with existing DSS to uncover complex cause-and-effect relationships between predictive variables and outcomes. By improving the interpretability of predictions, the project seeks to identify both direct and indirect causal factors of critical outcomes in patients with heart failure, generate evidence-based hypotheses about causal relationships, and determine the relative importance of various factors to guide targeted interventions.

In addition, the project proposes a pilot study to validate a digital platform based on a chatbot designed to collect patient-reported measures from individuals with atrial fibrillation before and after undergoing cardiac ablation. This approach aims to improve clinical follow-up, promote more personalised symptom management, and enable a timelier response to patient needs. The solution could lead to increased efficiency, better health outcomes, and serve as a scalable model for other chronic conditions. CardioCausalAI brings together a multidisciplinary team of researchers from AICOS and Value for Health CoLAB, in collaboration with Hospital de Santa Marta.

 

Consult the project's Spec Sheet here:

Portuguese