Fig. 5: Partial dependence plots illustrating the influence of individual features on sepsis prediction.
From: Streamlined machine learning model for early sepsis risk prediction in burn patients
Each plot shows how the value of a single feature (x-axis) impacts its contribution to the model’s output (y-axis, SHAP value) for every patient. Higher SHAP values push the prediction towards a higher risk of sepsis. The features from the final Random Forest model are: a Burned Body Surface Area, b Age, c full-thickness burns (Burn Depth 3), d deep partial-thickness burns (Burn Depth 2b), e Inhalation Injury, and f Hypertension.
