Postoperative New-Onset Heart Block in Noncardiac Surgery: Model Development, Validation, and Long-Term Prognostic...

Heart block is common in cardiac surgery but understudied in noncardiac settings despite its poor prognosis.

This study aims to investigate epidemiology, develop predictive models, and assess the prognostic value of new-onset heart block after noncardiac surgery.

This retrospective study analyzed 281,497 patients from 2008 to 2019. The primary outcome was new-onset heart block within 30 days. We used least absolute shrinkage and selection operator for variable selection and machine learning algorithms (eg, Nnet, LGBM, SVM, and XGB) for prediction. Piecewise Cox regression was used to analyze 1-year survival in elderly patients.

New-onset heart block occurred in 1,000 of 281,497 patients (0.36%; 95% CI: 0.33%-0.38%). Subtypes included 192 of 281,497 atrioventricular (AV) blocks (0.07%; 95% CI: 0.06%-0.08%); 51 of 281,497 left bundle branch blocks (0.02%; 95% CI: 0.01%-0.02%); and 757 of 281,497 right bundle branch blocks (RBBBs) (0.27%; 95% CI: 0.25%-0.29%). The XGB model achieved a validation area under the curve of 0.804 using 12 predictors. Median follow-up for elderly patients was 162 days (Q1-Q3: 85-256) for nonsurvivors and 365 days (Q1-Q3: 365-365) for survivors. AV block predicted poor early-phase prognosis (adjusted HR: 6.24; 95% CI: 1.83-21.23), whereas RBBB was associated with poor late-phase prognosis (adjusted HR: 2.67; 95% CI: 1.26-5.69).

New-onset heart block affects 0.36% of noncardiac surgery patients. Machine learning effectively predicts this complication. Given the significant mortality risks associated with AV block and RBBB, enhanced perioperative surveillance is crucial.