{"id":"46087","type":"article","url":"https://hartvaat.nl/2026/04/10/zuinig-machine-learningmodel-voorspelt-1-jaarssterfte-en-zinloosheid-na-tavi/","title":"Zuinig machine-learningmodel voorspelt 1-jaarssterfte en zinloosheid na TAVI","title_en":"","category":"algemeen","category_label":"Algemeen","professions":["cardioloog"],"tags":[],"journal":"Heart","doi":"http://heart.bmj.com/cgi/content/short/112/9/488?rss=1","source_url":"https://doi.org/http://heart.bmj.com/cgi/content/short/112/9/488?rss=1","authors":["Bharucha","A. H.","Brown","S.","Theofilatos","K.","Singh","B.","Galusko","V.","Rashid","H.","Abdelgawad","H.","Dworakowski","R.","Papachristidis","A.","Patterson","T.","Redwood","S. R.","Prendergast","B.","Byrne","J. A.","MacCarthy","P. A.","OGallagher","K.","Eskandari","M."],"significance":6,"published":"2026-07-23","source_date":"2026-04-10","image":"https://hartvaat.nl/global/img/5e97691e9888.webp","kennis":["https://hartvaat.nl/kennis/kleplijden/tavi-transcatheter-aortaklepimplantatie/","https://hartvaat.nl/kennis/kleplijden/esc-richtlijn-kleplijden-2021/"],"congress":"","summary_en":"Existing risk scores inadequately predict long-term mortality after transcatheter aortic valve replacement (TAVR), hampering decisions about procedural futility. The investigators developed a machine-learning model using only preprocedural variables to predict 1-year all-cause mortality. The model was trained on 1,025 TAVR patients with 52 clinical and echocardiographic variables; via an evolutionary algorithm it was reduced to 13 variables while preserving performance and simplicity. The final model was externally validated in an independent cohort of 270 patients and compared with EuroSCORE II, FRANCE-2 and TAVI2-SCORE. A simple, preprocedural-data-based model can thus identify patients with high 1-year mortality — useful to support decision-making around TAVR.","created":"2026-07-03T20:27:17Z","updated":"2026-07-04T19:09:44Z","licence":"Citeer vrij, met bronvermelding en een link naar hartvaat.nl (de url van het record). Samenvattingen zijn redactioneel werk van HartVaat; de oorspronkelijke publicaties blijven van hun uitgevers (doi). Geen medisch advies.","body_markdown":"Bestaande risicoscores voorspellen de langetermijnsterfte na transkatheter-aortaklepimplantatie (TAVI) onvoldoende, wat de besluitvorming over zinloze ingrepen ('futility') bemoeilijkt. De onderzoekers ontwikkelden een machine-learningmodel dat uitsluitend preprocedurele variabelen gebruikt om de 1-jaarssterfte te voorspellen. Het model werd getraind op 1.025 TAVI-patiënten met 52 klinische en echografische variabelen; via een evolutionair algoritme werd het teruggebracht tot 13 variabelen, met behoud van prestaties en eenvoud. Het uiteindelijke model werd extern gevalideerd in een onafhankelijk cohort van 270 patiënten en vergeleken met de EuroSCORE II-, FRANCE-2- en TAVI2-SCORE. Een eenvoudig, op preprocedurele gegevens gebaseerd model kan zo de patiënten identificeren bij wie de 1-jaarssterfte hoog is — bruikbaar om de besluitvorming rond TAVI te ondersteunen.","abstract_original":"Background\n

Current risk scores inadequately predict long-term mortality after transcatheter aortic valve replacement (TAVR), limiting their ability to guide decisions around procedural futility. We aimed to develop and externally validate a machine learning (ML) model using only preprocedural variables to predict 1-year all-cause mortality.

\n\nMethods\n

An ML model was trained on a retrospective cohort of 1025 TAVR patients using 52 clinical and echocardiographic variables. Feature selection and model tuning were performed via a multiobjective evolutionary algorithm to optimise predictive performance and model simplicity. The final model used 13 preprocedural variables and was externally validated in an independent cohort of 270 patients. Performance was compared with European System for Cardiac Operative Risk Evaluation II (EuroSCORE II), FRANCE-2 and TAVI2-SCORE using the area under the curve (AUC), calibration and net reclassification improvem"}