Abstract 13727: Prediction of Pacemaker Implantation in Patients With Postoperative Heart Block: A Data Analytics Approach

Introduction: Current guidelines recommend observation for AV node recovery until postoperative days (POD) 7 to 10 for patients with complete heart block (CHB) after congenital cardiac surgery prior to permanent pacemaker (PPM) placement. We created a predictive tool to identify patients at high risk of requiring PPM before POD 7 to aid in surgical decision making. Methods: Postoperative CHB cases from August 2009 to January 2019 were identified. We collected demographic, anatomic, and surgical details. Random forest, logistic regression, and decision tree models were trained and evaluated using 5-fold cross validation, then prospectively validated on data from January 2019 to April 2021. Separate models were developed for all patients on POD 0 and those without recovery on POD 4. Results: Of 101 cases of CHB (22 in validation cohort), 60 (59%) required PPM. Median days [IQR] to AVN recovery was 3 [1-7] and PPM was 10 [7-13]. A decision tree model was used for its performance and interpretability, and included: (1) aortic valve replacement, subaortic stenosis repair, or Konno incision; (2) ventricular L-loop; (3) atrioventricular valve replacement; (4) use of preoperative antiarrhythmic agent (in POD 0 model only; figure 1). The POD 4 model (table 1) had a positive predictive value (PPV) of 0.94 [95% CI 0.81-0.98], which was stable in prospective validation (PPV 1.0). Conclusions: Our decision tool identified patients requiring PPM with high specificity and PPV. On POD 4, patients with left ventricular outflow tract repair, AV valve replacement, or ventricular L-Looping could be considered for PPM placement before POD 7 to reduce risks of temporary pacing and improve care efficiency.