Abstract Introduction Post-operative atrial fibrillation (POAF) represents the most common arrhythmia in the post-operative setting, with a peak incidence from day 0 to 5 after surgery and it represents a main cause of morbidity, mortality, length of stay, thromboembolic events and stroke. Cardiothoracic surgery has the highest rates of POAF, while data about other surgeries are contrasting amongst various studies. Aim of this study was to detect POAF onset in the 28 days after surgery and to better assess its predictors, especially the role of inflammation. Methods This is a retrospective single center cohort study of 53.387 patients undergoing surgery from January 2016 to January 2020. Patients were classified in four groups according to types of surgery performed: (I) orthopedic surgery, (II) not thoracic nor abdominal surgery, (III) abdominal and esophageal surgery and (IV) lung and cardiovascular surgery. Kaplan–Meier estimates were used to draw the cumulative incidence curves by surgery groups; finally, they were compared with a log-rank test. Furthermore, multivariable Cox proportional hazards (PH) models of prognostic factors were used. Confounders were selected according to a review of the literature, statistical relevance, and consensus opinion by an expert group of physicians and methodologists. After fitting the model, the PH assumption was examined on the basis of Schoenfeld residual. Results The primary endpoint of AF onset occurred in 570 patients (1.1%) with a mean incidence after surgery of 3.4±2.6 days. 90 patients died (0.17%) after an average of 13.7±8.4 days. The Kaplan-Meier analysis showed the lowest 28-day event-free survival in group IV and the highest in group I (log-ranks test p=0.0001). In patients who developed AF, levels of C-Reactive Protein (CRP) were higher than the others (mean 0,70+0.03 log10 mg/dL versus mean 0,40+0.01 log10 mg/dl; p <0.0001) with higher levels in group III and group IV. In the univariable Cox regression, CRP was a strong predictor of AF (HR per 1 unit increase in log-scale, 2.64; 95% C.I,1.74–4.0; p<0.0001). This was confirmed at the multivariable analysis, adjusting for confounding factors like age, gender, length of stay in hospital and group of surgery (adjusted HR per 1 mg/dL increase in log-scale, 1.81; 95% CI,1.18–2.79; p = 0.007). Other strong predictors of POAF were age (HR per 1 year increase, 1.06; 95% CI, 1.04–1.08; p< 0.0001) and surgery of group III and IV (HR, 23.62; 95% CI, 5.65–98.73; p< 0.0001 and HR,6.26; 95% CI, 1.48–26.49; p 0.013, respectively).The PH assumption was not violated (p=0.12). Conclusions POAF represents a frequent complication of surgery and major burden for healthcare. Inflammation may represent a major driver in its pathophysiology, especially in non-cardiac surgery, in which manipulation of cardiac tissue is avoided. This may explain the poor response to antiarrhythmic drugs and its self-limiting nature, which expires when post-operative inflammation turns off.