RESHAPE Réunion scientifique - 8 Avril 2022 - 12h30

- Vincent LEQUERTIER: Prédiction de la durée de séjour à l'aide de données hospitalières standardisées provenant de soins aigus et d'urgence, grâce à un "réseau neuronal profond" (deep neural network)

Estimating length of stay (LOS) is important for organizing care activities. In particular, predicting it accurately contributes to better resources allocation and a better preparation of patients’ discharge. In this talk, a method for predicting LOS using a Feed-Forward Neural Network (FFNN) with standardized administrative data from acute and emergency care is presented. Performances of the prediction were compared to random forest and logistic regression with the accuracy, Cohen’s kappa and a Bland-Altman plot, through a 5-folds cross-validation