Machine learning versus logistic regression for permanent pacemaker implantation prediction after transcatheter aortic valve replacement—a systematic review and meta-analysis

Background: Transcatheter aortic valve replacement (TAVR) has improved clinical management of aortic valve disease. Postoperative permanent pacemaker implantation (PPI) remains the main complication associated with TAVR. We have conducted a study-level meta-analysis which compares logistic regression (LR) models with machine learning (ML) models to predict the association between peri-procedural risk factors and PPI in patients who underwent TAVR. Methods: We included three studies reporting PPI after TAVR as primary outcome comparing ML and LR models’ predictive ability. Model performance was primarily assessed in terms of discrimination ability for PPI prediction. The retrieved estimates of C-statistic for ML and LR models were summarized into a weighted average. A Bayesian estimation framework was utilized to calculate the meta-analytic estimates. Pooled C-statistics (PCS) and 95% credibility interval (CI) were calculated, and the overall accuracy was presented as net benefit. As a sensitivity analysis, the main comparison was performed excluding derivation models and models with extremely reduced accuracy levels. Results: No significant difference was noticed in the predictive accuracy between ML and LR models (ML PCS =0.72, 0.63–0.81 vs. LR PCS =0.64, 0.50–0.78, net benefit: +8.0%, P=0.34). The sensitive analysis showed no significant difference excluding the derivation model from Tsushima and 1 year model from Agasthi (ML PCS =0.72, 0.51–0.92, vs. LR PCS =0.64, 0.43–0.88, net benefit: +8.0%, P=0.60); and also no significant difference excluding derivation model from Tsushima and the 30 days model from Agasthi (ML PCS =0.74, 0.57–0.90 vs. LR PCS =0.64, 0.42–0.88, net benefit: +10.0%, P=0.28). Conclusions: Despite showing a higher net benefit, ML models do not perform significantly better than LR models in predicting the association between peri-procedural risk factors and PPI after TAVR.