Reliable Prediction of Postoperative Atrial Fibrillation from High-resolution ECG-based Assessment of Cardiac Autonomic Derangement and altered Heart Rhythm Dynamics
Jurij M. Kalisnik, MD, PhD1, Viktor Avbelj, PhD2, Jon Vratanar3, Tilen Tumpaj, MD4, Janez Zibert, PhD5.
1Department of Cardiac Surgery – Cardiovascular Center, Klinikum Nuernberg - Paracelsus Medical University, Nuremberg, Germany, 2Department of Communications and Computer Networks, Jožef Stefan Institute, Ljubljana, Slovenia, 3School of Medicine, University of Ljubljana, Ljubljana, Slovenia, 4Department of Cardiovascular Surgery, University Medical Centre Ljubljana, Ljubljana, Slovenia, 5Faculty of Health Sciences, University of Ljubljana, Ljubljana, Slovenia.
OBJECTIVE: We and others have shown that patients developing postoperative AF (poAF) after cardiac surgery, have severe cardiac autonomic derangement and altered heart rhythm dynamics already preoperatively. In the way towards ECG based online (po)AF prediction we present novel approach further improving the performance of poAF prediction.
METHODS: 179 consecutive patients scheduled for cardiac surgery were enrolled in our study prospectively. Eight were excluded, 4 died, 17 had high ectopical activity not amenable to analysis, so 150 represented the final study sample. High-resolution 20-minute ECG recordings were obtained one day before surgery to determine RR, PQ, and QT intervals as well as linear (Time and Frequency domain) and nonlinear Heart Rate Variability parameters such as Fractal Dimension (FD) and Detrended Fluctuation Analysis (DFA). Statistical analyses were performed and p ≤ 0.05 was considered significant. Relevant predictors of AF were determined by using logistic (stepwise) regression modeling. To estimate the classification performance, the final model was built by using 9 most powerful predictors of AF in logistic regression and evaluated with leave-one-out cross validation approach. The prediction of AF was measured with AUC in ROC analysis.
RESULTS: 31 patients developed poAF after operation (poAF Group) and 119 did not. The two groups were similar, except for more Arterial Hypertension, higher Age, Euroscore II and Leukocyte Count on the second postoperative day in poAF group. PQ intervals were shorter in poAF group (156 ± 23 vs. 173 ± 31 ms; p=0,011). Among nonlinear parameters was DFAα1 lower in poAF group (0,95 ± 0,36 vs. 1,11 ± 0,30; p=0,032). Nine the most relevant predictors of poAF were: Preoperative Heart Rate, Use of Inotropes, DFA α2, FD (High), Ejection Fraction, CPB and Aortic Crossclamp TIme, Age and PQ interval. The AUC of poAF prediction on train data was 88.2%, while leave-one-out cross validation approach produced the AUC of 79.6%.
CONCLUSIONS: Determination of Cardiac Autonomic Modulation and Heart Rhythm Dynamics through digital ECG offers a platform for a true prediction of poAF now in the good-to-excellent range.
Back to 2016 Annual Meeting Posters