Statistical analysis plan: Development and validation of prediction models for new-onset atrial fibrillation in patients admitted to an intensive care unit

Introduction: New-onset atrial fibrillation (NOAF) is a common arrhythmia in patients on an intensive care unit (ICU). NOAF in this setting is associated with adverse short and long-term outcomes. Little is known about modifiable risk factors for NOAF. Developing a model to identify patients at high risk of developing NOAF is vital for future studies investigating prevention strategies to allow stratification, sample enrichment and efficient study design. Methods: We will use data from 7 general ICUs in the UK and USA to develop a predictive model for NOAF. We will assess whether including longitudinal, dynamic predictors, along with the use of machine learning approaches improves model predictive ability. Generalisability and Implications: This work will be the largest assessment of predictors of NOAF in patients treated on a general ICU. It will be the first to analyse international data to ensure worldwide applicability. A detailed understanding of modifiable risk factors underpins future work to prevent NOAF in these patients.