Beyond accuracy: The translational case for explainable poststroke anxiety prediction

This letter discusses the translational significance of the explainable machine learning framework developed by Lyu et al. for predicting poststroke anxiety. The selection of logistic regression over more complex algorithms illustrates the clinical value of prioritizing generalization, calibration, interpretability, and bedside feasibility when discrimination is comparable. We clarify that SHAP was used to interpret the fitted model and inform its subsequent simplification, rather than as a formal feature-selection procedure. Although the reported predictive performance is encouraging, the small test partition and absence of genuinely external validation limit conclusions regarding transportability to independent populations. We also identify inconsistencies across the workflow figure, Methods, Results, and figure legends that merit correction. Addressing these issues would strengthen an otherwise valuable contribution to clinically oriented, explainable prediction research in stroke care.