BACKGROUND: Electrocardiographic (ECG) characteristics were analyzed in postoperative cardiac surgery patients in an attempt to predict development of new-onset postoperative atrial fibrillation (AF). METHODS: Nineteen ECG characteristics were analyzed using computer-based algorithms. The parameters were retrospectively analyzed from ECG signals recorded in postoperative cardiac surgery patients while they were in the cardiovascular intensive care unit (CVICU) at our institution. ECG data from 99 patients (of whom 43 developed postoperative AF) were analyzed. A bootstrap variable selection procedure was applied to select the most important ECG parameters, and a multivariable logistic regression model was developed to classify patients who did and did not develop AF. RESULTS: Premature atrial activity (PAC) was greater in AF patients (P < 0.01). Certain heart rate variability (HRV) and turbulence parameters also differed in patients who did and did not develop AF. In contrast, P-wave morphology was similar in patients with and without AF. Receiver operating curve (ROC) analysis applied to the model produced a C-statistic of 0.904. The model thus correctly classified AF patients with more than a 90% sensitivity and a 70% specificity. CONCLUSION: Among the 19 ECG parameters analyzed, PAC activity, frequency-domain HRV, and heart rate turbulence parameters were the best discriminators for postoperative AF.
BACKGROUND: Electrocardiographic (ECG) characteristics were analyzed in postoperative cardiac surgery patients in an attempt to predict development of new-onset postoperative atrial fibrillation (AF). METHODS: Nineteen ECG characteristics were analyzed using computer-based algorithms. The parameters were retrospectively analyzed from ECG signals recorded in postoperative cardiac surgery patients while they were in the cardiovascular intensive care unit (CVICU) at our institution. ECG data from 99 patients (of whom 43 developed postoperative AF) were analyzed. A bootstrap variable selection procedure was applied to select the most important ECG parameters, and a multivariable logistic regression model was developed to classify patients who did and did not develop AF. RESULTS: Premature atrial activity (PAC) was greater in AFpatients (P < 0.01). Certain heart rate variability (HRV) and turbulence parameters also differed in patients who did and did not develop AF. In contrast, P-wave morphology was similar in patients with and without AF. Receiver operating curve (ROC) analysis applied to the model produced a C-statistic of 0.904. The model thus correctly classified AFpatients with more than a 90% sensitivity and a 70% specificity. CONCLUSION: Among the 19 ECG parameters analyzed, PAC activity, frequency-domain HRV, and heart rate turbulence parameters were the best discriminators for postoperative AF.