Machine learning score using only echocardiographic data for prediction of in-hospital outcomes in ICCU patients - 23/12/23
, A. Coisne 1, K. Hamzi 2, S. Toupin 3, C. Bouleti 4, C. Fauvel 5, J.B. Brette 6, D. Montaigne 1, R. Rossanaly Vasram 7, A. Trimaille 8, G. Lemesle 9, G. Schurtz 10, J.-G. Dillinger 2, P. Henry 2, T. Pezel 11
Bienvenue sur EM-consulte, la référence des professionnels de santé.
Article gratuit.
Connectez-vous pour en bénéficier!
Résumé |
Introduction |
Several studies have shown the strong prognostic value of parameters measured by transthoracic echocardiography (TTE) at the admission of patients hospitalised in intensive care cardiac unit (ICCU) for acute cardiovascular event. Machine learning (ML) methods is an innovative way to evaluate accurately the benefit of a score integrating these TTE parameters.
Objective |
To investigate the feasibility and accuracy of a ML-score using TTE data only to predict major adverse cardiovascular events (MACE) in consecutive patients admitted to ICCU.
Method |
In April 2021, all consecutive patients admitted in 39 French ICCU were included and a complete TTE was performed at admission including 20 traditional TTE parameters. Several ML-models (LASSO, Random Forest and XGBoost) were then trained on 70% of patients and evaluated on the other 30% as internal validation. The primary composite outcome was MACE defined by all-cause mortality, cardiogenic shock (requiring medical or mechanical haemodynamic support) and resuscitated cardiac arrest.
Results |
Of 1,499 consecutive patients screened (63±15 years, 70% male), MACE occurred in 67 patients (4.5%) including: 27 (1.8%) deaths, 38 (2.5%) cardiogenic shocks, 14 (0.9%) resuscitated cardiac arrests. To build the ML-model, 5 TTE parameters were selected as being the most important in predicting MACE: LVOT VTI, E/e’ ratio, sPAP, TAPSE, and LVEF. The XGBoost model showed the best performance compared with the other ML-models (AUROC for XGBoost: 0.83, Random-forest: 0.81, LASSO regression 0.79 and to a regression model: 0.76). Using the XGBoost model as our final ML-score, SHAP values for the parameters were: 0.047 for LVOT VTI, 0.041 for E/e’ ratio, 0.028 for SPAP, 0.021 for TAPSE and 0.01 for LVEF. Our ML-score exhibited a higher AUC compared with any existing scores (Fig. 1), and an incremental prognostic value for predicting MACE over and above clinical or biological data (Chi2 59.72 P<0.001; C-Index 0.8 [0.71;0.89], P=0.012) (Fig. 2).
Conclusion |
The ML-score including only TTE parameters exhibited a higher prognostic value to predict MACE compared with any existing traditional method or traditional scores.
Le texte complet de cet article est disponible en PDF.Plan
Vol 117 - N° 1S
P. S66 - janvier 2024 Retour au numéro
Article précédent
- Optimal cut-off thresholds of LGE extent to predict death using machine-learning in a large cohort of ICM patients
- A. Unger, S. Toupin, T. Pezel, P. Garot, T. Unterseeh, S. Champagne, T. Goncalves, T. Ah-Sing, L. Hamzi, J.-G. Dillinger, P. Henry, V. Bousson, F. Sanguineti, T. Hovasse., J. Garot
- Left atrioventricular coupling index assessed using cardiac CT as a prognostic marker of cardiovascular death
- T. Pezel, J.-G. Dillinger, S. Toupin, S. Miraillesx, D. Logeart, A. Cohen Solal, A. Unger, E.S. Canuti, F. Beauvais, A. Lafont, T. Gonçalves, A. Léquipar, E. Gall, A. Boutigny, T. Ah-Sing, L. Hamzi, J. Lima, V. Bousson, P. Henry
Bienvenue sur EM-consulte, la référence des professionnels de santé.