Explainability and Deployment
Subgroup performance and fairness will be audited across sex, age bins, and hospital type. SHAP will be used for individual predictions and global feature importance. Final models will be deployed as Shiny web-based clinical risk calculato…
1 sources - 5 claims
Subgroup performance and fairness will be audited across sex, age bins, and hospital type. SHAP will be used for individual predictions and global feature importance. Final models will be deployed as Shiny web-based clinical risk calculators with local SHAP waterfall explanations. SHAP will help assess whether models over-rely on patient demographics.