Aims: Takotsubo syndrome (TTS) is associated with a substantial rate of adverse events. We sought to design a machine-learning (ML) based model to predict the risk of in-hospital death and to perform a clustering of TTS patients to identify different risk profiles. Methods and results: A Ridge Logistic Regression-based ML model for predicting in-hospital death was developed on 3482 TTS patients from the International Takotsubo Registry, randomly split in a train and an internal validation cohort (75% and 25% of the sample size, respectively) and evaluated in an external validation cohort (1037 patients). 31 clinically relevant variables were included in the prediction model. Model performance represented the primary endpoint and was assessed according to area under the receiver-operating characteristic curve (AUC), Sensitivity and Specificity. As secondary endpoint, a K-Medoids clustering algorithm was designed to stratify patients into phenotypic groups based on the ten most relevant features emerging from the main model. The overall incidence of in-hospital death was 5.2%. The InterTAK-ML model showed an AUC of 0.89 (0.85-0.92), Sensitivity 0.85 (0.78-0.95) and Specificity 0.76 (0.74-0.79) in the internal validation cohort and an AUC of 0.82 (0.73-0.91), a sensitivity of 0.74 (0.61-0.87) and a specificity of 0.79 (0.77-0.81) in the external cohort for in-hospital death prediction. By exploiting the 10 variables showing the highest feature importance, TTS patients were clustered into six groups associated with different risks of in-hospital death (28.8% vs 15.5% vs 5.4% vs 0.8% vs 0.5%) which were consistent also in the external cohort. Conclusion: A ML-based approach for the identification of TTS patients at risk of adverse short-term prognosis is feasible and effective. The InterTAK-ML model showed unprecedented discriminative capability for the prediction of in-hospital death. This article is protected by copyright. All rights reserved.
Machine-learning based prediction of in-hospital death for patients with takotsubo syndrome: the InterTAK-ML model / De Filippo, Ovidio; Cammann, Victoria L; Pancotti, Corrado; Di Vece, Davide; Silverio, Angelo; Schweiger, Victor; Niederseer, David; Szawan, Konrad A; Würdinger, Michael; Koleva, Iva; Dusi, Veronica; Bellino, Michele; Vecchione, Carmine; Parodi, Guido; Bossone, Eduardo; Gili, Sebastiano; Neuhaus, Michael; Franke, Jennifer; Meder, Benjamin; Jaguszewski, Miłosz; Noutsias, Michel; Knorr, Maike; Jansen, Thomas; Dichtl, Wolfgang; von Lewinski, Dirk; Burgdorf, Christof; Kherad, Behrouz; Tschöpe, Carsten; Sarcon, Annahita; Shinbane, Jerold; Rajan, Lawrence; Michels, Guido; Pfister, Roman; Cuneo, Alessandro; Jacobshagen, Claudius; Karakas, Mahir; Koenig, Wolfgang; Pott, Alexander; Meyer, Philippe; Roffi, Marco; Banning, Adrian; Wolfrum, Mathias; Cuculi, Florim; Kobza, Richard; Fischer, Thomas A; Vasankari, Tuija; Airaksinen, K E Juhani; Napp, L Christian; Dworakowski, Rafal; Maccarthy, Philip; Kaiser, Christoph; Osswald, Stefan; Galiuto, Leonarda; Chan, Christina; Bridgman, Paul; Beug, Daniel; Delmas, Clément; Lairez, Olivier; Gilyarova, Ekaterina; Shilova, Alexandra; Gilyarov, Mikhail; El-Battrawy, Ibrahim; Akin, Ibrahim; Poledniková, Karolina; Toušek, Petr; Winchester, David E; Massoomi, Michael; Galuszka, Jan; Ukena, Christian; Poglajen, Gregor; Carrilho-Ferreira, Pedro; Hauck, Christian; Paolini, Carla; Bilato, Claudio; Kobayashi, Yoshio; Kato, Ken; Ishibashi, Iwao; Himi, Toshiharu; Din, Jehangir; Al-Shammari, Ali; Prasad, Abhiram; Rihal, Charanjit S; Liu, Kan; Schulze, P Christian; Bianco, Matteo; Jörg, Lucas; Rickli, Hans; Pestana, Gonçalo; Nguyen, Thanh H; Böhm, Michael; Maier, Lars S; Pinto, Fausto J; Widimský, Petr; Felix, Stephan B; Braun-Dullaeus, Ruediger C; Rottbauer, Wolfgang; Hasenfuß, Gerd; Pieske, Burkert M; Schunkert, Heribert; Budnik, Monika; Opolski, Grzegorz; Thiele, Holger; Bauersachs, Johann; Horowitz, John D; Di Mario, Carlo; Bruno, Francesco; Kong, William; Dalakoti, Mayank; Imori, Yoichi; Münzel, Thomas; Crea, Filippo; Lüscher, Thomas F; Bax, Jeroen J; Ruschitzka, Frank; de Ferrari, Gaetano Maria; Fariselli, Piero; Ghadri, Jelena R; Citro, Rodolfo; D'Ascenzo, Fabrizio; Templin, Christian. - In: EUROPEAN JOURNAL OF HEART FAILURE. - ISSN 1388-9842. - (2023). [10.1093/eurheartjsupp/suac121.714]
Machine-learning based prediction of in-hospital death for patients with takotsubo syndrome: the InterTAK-ML model
Bossone, Eduardo;
2023
Abstract
Aims: Takotsubo syndrome (TTS) is associated with a substantial rate of adverse events. We sought to design a machine-learning (ML) based model to predict the risk of in-hospital death and to perform a clustering of TTS patients to identify different risk profiles. Methods and results: A Ridge Logistic Regression-based ML model for predicting in-hospital death was developed on 3482 TTS patients from the International Takotsubo Registry, randomly split in a train and an internal validation cohort (75% and 25% of the sample size, respectively) and evaluated in an external validation cohort (1037 patients). 31 clinically relevant variables were included in the prediction model. Model performance represented the primary endpoint and was assessed according to area under the receiver-operating characteristic curve (AUC), Sensitivity and Specificity. As secondary endpoint, a K-Medoids clustering algorithm was designed to stratify patients into phenotypic groups based on the ten most relevant features emerging from the main model. The overall incidence of in-hospital death was 5.2%. The InterTAK-ML model showed an AUC of 0.89 (0.85-0.92), Sensitivity 0.85 (0.78-0.95) and Specificity 0.76 (0.74-0.79) in the internal validation cohort and an AUC of 0.82 (0.73-0.91), a sensitivity of 0.74 (0.61-0.87) and a specificity of 0.79 (0.77-0.81) in the external cohort for in-hospital death prediction. By exploiting the 10 variables showing the highest feature importance, TTS patients were clustered into six groups associated with different risks of in-hospital death (28.8% vs 15.5% vs 5.4% vs 0.8% vs 0.5%) which were consistent also in the external cohort. Conclusion: A ML-based approach for the identification of TTS patients at risk of adverse short-term prognosis is feasible and effective. The InterTAK-ML model showed unprecedented discriminative capability for the prediction of in-hospital death. This article is protected by copyright. All rights reserved.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
