Table 1.
Predictors and auxiliary variables.
Fig 1.
Flowchart of patient selection and data segregation.
Table 2.
Basic characteristics.
Table 3.
Test performance of predictive models.
Fig 2.
SHAP summary plots of the 20 most influential predictors (A) The predictive model for death or bronchopulmonary dysplasia (B) The predictive model for death or severe bronchopulmonary dysplasia (C) The predictive model for death before discharge.
Fig 3.
k-means clustering of the Shapley additive explanations values of the test set, visualized by t-distributed stochastic neighbor embedding.
On the left are clusters based on the predictive model for death or bronchopulmonary dysplasia. In the middle, clusters are based on the predictive model for death or severe bronchopulmonary dysplasia. On the right, clusters are based on the predictive model for death before discharge.