Prediction of pacing using HRV

Pacemaker use due to conduction abnormalities is a common complication following surgical aortic valve replacement (AVR). Heart rate variability (HRV) is associated with sinus node dysfunction and significant dysrhythmias. However, its predictive value for postoperative electrical pacing requirements after AVR remains unclear. This retrospective study reviewed pre-registered electrical records from 194 adult patients who underwent isolated AVR. HRV parameters in both time and frequency domains were obtained prior to anesthesia induction and before initiating cardiopulmonary bypass. Tree-based machine learning (ML) models, including RandomForest, LightGBM, and ExtraTrees, were developed using HRV parameters and clinical variables to predict postoperative pacing needs. The incidence of temporary electrical pacing postoperatively was 35.1% (34.8% in the training set and 35.9% in the test set). The RandomForest model incorporating both HRV and clinical features achieved an area under the receiver operating characteristic curve of 0.731 (95%CI, 0.681–0.781) and an area under the precision-recall curve of 0.687 (95%CI, 0.619–0.746). In conclusion, ML models leveraging HRV demonstrated potential for predicting postoperative pacemaker requirements following isolated surgical AVR. Accurate prediction of significant conduction disturbances through HRV-based ML algorithms may enable timely interventions and improved management for at-risk patients in clinical practice.