Data supporting the research of Risk Mechanisms and Prediction of New-onset Atrial Fibrillation: An Interpretable Model Based on Causal Information Decomposition

This dataset was derived from the MIMIC-IV and MIMIC-III databases and aims to investigate the risk mechanisms and prediction of new-onset atrial fibrillation in critically ill patients. The research is an observational clinical study using routinely collected electronic health record data, including demographic information, vital signs, laboratory measurements, and clinical interventions during ICU admissions. Data were extracted and processed following the MIMIC usage agreements and analyzed using machine learning and causal information decomposition methods to develop an interpretable predictive model supported by SHAP explanations.