Background: In majority of previous studies prediction of TAVI outcomes has been traditionally performed using Cox regression analysis. In this study we investigated whether a decision tree algorithm (DTA) could be used to predict one-year mortality and complications following TAVI.  

Patients & Methods: A total of 738 TAVI patients (male = 403, median age = 81 (77-84), baseline Euro score II (ESII = 3.90; 2.26-6.67)) treated in our department from 2019 until 2021 were included. The primary endpoint was one-year mortality, secondary endpoints were TAVI-related complications as assessed by a dedicated DTA.

Results: Within a median follow up of 361 (IQR: 144-537 days) days 73 patients died. DTA revealed that baseline ESII score is the single independent predictor of mortality and TAVI-related complications. DTA identified 3 risk categories of mortality based on baseline ESII cut-off values of 1.99 and 3.90, namely 1) low mortality risk = 0.7%, if baseline ESII was < 1.99, 
2) with moderate risk = 6.7%, if ESII = 1.99-3.90 and 3) high risk = 15.6%, if ESII >3.90 (Figure 1). DTA identified 3 risk categories of complications based on ESII cut-off values of 1.61 and 7.45, namely 1) low risk = 9.6%, if baseline ESII < 1.61, 2) moderate risk = 21.2%, if ESII = 1.61-7.45 and 3) high risk = 31.3%, if ESII >7.45 (Figure 2).


Figure 1



Figure 2

Conclusion: Using just a single metric and our novel DTA–analysis, based on a prospectively gathered significant amount of patient data, clinicians will be able to provide targeted personalized information and consultation to TAVI candidates prior to the intervention.