Risk Prediction Model for Permanent Pacemaker Implantation after Transcatheter Aortic Valve Replacement

Vejpongsa, Pimprapa; Zhang, Xu; Bhise, Viraj; Kitkungvan, Danai; Shivamurthy, Poojita; Anderson, H. Vernon; Balan, Prakash; Nguyen, Tom C.; Estrera, Anthony L.; Dougherty, Anne H.; Smalling, Richard W.; Dhoble, Abhijeet

doi:10.6084/m9.figshare.6182219.v3

Supplemental material.docx (31.94 kB)

Risk Prediction Model for Permanent Pacemaker Implantation after Transcatheter Aortic Valve Replacement

Version 3 2020-10-23, 20:10

Version 2 2018-05-11, 20:44

Version 1 2018-04-25, 14:06

journal contribution

posted on 2020-10-23, 20:10 authored by Pimprapa Vejpongsa, Xu Zhang, Viraj Bhise, Danai Kitkungvan, Poojita Shivamurthy, H. Vernon Anderson, Prakash Balan, Tom C. Nguyen, Anthony L. Estrera, Anne H. Dougherty, Richard W. Smalling, Abhijeet Dhoble

Background: Atrioventricular conduction disturbance requiring permanent pacemaker (PPM) implantation is the most common complication after transcatheter aortic valve replacement (TAVR). Improved risk stratification for potential need for post-procedure PPM implant prior to the TAVR procedure is warranted. The aim of this study was to develop and validate a risk-prediction model for PPM implantation after TAVR.

Methods: This PPM risk assessment model was developed using the 2012&2013 National Inpatient Sample (NIS). A logistic regression model was built to identify the predictors of PPM placement. The performance of the model was validated using the NIS 2014 dataset.

Results: Of 18,400 patients in the development cohort, 1,825 (9.9%) patients required PPM implantation after TAVR. After multivariate analysis, final predictive covariates of PPM implantation included left or right bundle branch block, bradycardia, 2nd-degree AV block and transfemoral approach. The estimated regression coefficients associated with these predictors were used to develop a scoring system. The proposed scoring system showed good discrimination in both development and validation cohorts, with c-statistics of 0.754 (95% CI: 0.726–0.782) and 0.746 (95% CI: 0.721–0.772) respectively. Calibration analysis indicated a good agreement between the observed rate of PPM and predicted risks of PPM by the risk score.

Conclusions: This PPM risk prediction model derived using the NIS database is a simple tool that can estimate individual risk of PPM prior to TAVR procedure. The model displayed good discrimination and calibration indices. This risk score can provide valuable information for patients’ counseling and also help identify high-risk patients who need close monitoring immediately after the TAVR procedure.