Skip to main content
. 2026 Jun 11;13:1830474. doi: 10.3389/fmed.2026.1830474

Figure 5.

Panel A shows a scatter plot of SHAP values by TRVmax with points colored by FE status. Panel B is a scatter plot of SHAP values by FE, colored by TRVmax. Panel C is a scatter plot of SHAP values by NIV, colored by TRVmax. All plots visualize model interpretability for different variables using SHAP values.

SHAP dependence plots illustrating the effects and interactions of key predictors on the model output. (A) Dependence of the SHAP value on TRVmax, with points colored by FE status (Yes/No), showing how the contribution of TRVmax varies with frequent exacerbation history. (B) SHAP values stratified by FE (Yes/No) and colored by TRVmax, indicating the joint influence of frequent exacerbation history and maximum tricuspid regurgitation velocity. (C) SHAP values stratified by NIV (Yes/No) and colored by TRVmax, demonstrating the interaction between non-invasive ventilation and maximum tricuspid regurgitation velocity on the predicted risk. Positive SHAP values indicate increased predicted risk, whereas negative values indicate decreased predicted risk. TRVmax, maximum tricuspid regurgitation velocity; FE, frequent exacerbation history (≥2 hospitalizations due to acute exacerbations within the past year); NIV, non-invasive ventilation; SHAP, SHapley Additive exPlanations.