We read with interest the article by Leha et al.,¹ developing and validating the TRIM risk scores for predicting the risk of 30-day mortality following transcatheter aortic valve implantation (TAVI) using machine learning (ML) models. We commend the authors for developing two models based on TAVI pre-procedural (TRIMpre) and post-procedural (TRIMpost) variables; however, we would like to raise some concerns and discuss potential methodological challenges that might have influenced the results.