Abstract Background and Aims Atrial fibrillation (AF) is a prevalent complication after cardiac surgery, worsening patient outcomes. Considering the established role of Ca2+-handling abnormalities in AF pathogenesis, this study aimed to evaluate if integrating cytosolic Ca2+-handling measurements with clinical risk factors enhances the risk prediction of post-operative AF. Methods Clinical data from 558 patients undergoing cardiac surgery without pre-existing AF from two centres were analysed. From 94 of these patients, atrial cardiomyocytes were isolated from collected right atrial appendages and Ca2+ handling (L-type Ca2+ current, intracellular Ca2+ concentration) was assessed using patch-clamp. The predictive performance of combining both clinical and single-cell Ca2+ handling parameters was tested using sequential feature selection and logistic regression models. Results Single-cell Ca2+-handling parameters through cluster analysis correlated with post-operative AF development and several cardiac diseases. Integration of Ca2+-handling parameters into a new post-operative AF risk prediction model improved its predictive accuracy by increasing the areas under the receiver operating characteristic (ROC) curves from 0.69 to 0.71 in the training and 0.76 to 0.79 in the validation cohort. Systolic Ca2+ level, along with clinical parameters such as age, left atrial dilatation, valvular heart disease, impaired renal function, and serum magnesium, was identified as an independent risk factor for post-operative AF. Additionally, a predictive score for AF occurrence at discharge and during rehabilitation has been developed, with area under the curve (AUC) values of 0.84 and 0.71, respectively. Incorporating the occurrence of AF during the immediate post-operative period as an additional predictor significantly enhanced the prediction of AF at discharge, achieving an AUC value of 0.94. Conclusions Integrating cellular Ca2+ handling signature with clinical predictors improves the prediction of post-operative AF, highlighting the potential of incorporating functional cellular data into clinical risk models.