Fig. 4 | Scientific Reports

Fig. 4

From: Development and external validation of an interpretable machine learning model for the prediction of intubation in the intensive care unit

Fig. 4

SHAP value impact on model output. Illustrates the relationship between feature values and their corresponding SHAP values in Catboost with 5 features. In this figure, red points represent relatively higher numerical feature values, while blue points correspond to relatively lower ones. Grey points represent the missing values. The horizontal axis denotes the assigned SHAP value for patients belonging to each category based on this particular feature. A positive SHAP value signifies an increased risk of intubation, whereas a negative value implies a decreased risk. Features were ranked according to the sum of absolute SHAP values across all samples analyzed. Notably, GCS (Glasgow Coma Scale) emerges as the most influential predictor for intubation in our final model, where a lower GCS score indicates a higher risk of intubation.

Back to article page