Explainable AI for Geriatric Clinical Risk Prediction: A Governance-Aware SHAP and LIME Framework
Clinical risk prediction for older adults increasingly relies on machine learning architectures whose opacity conflicts with the interpretability demands of geriatric clinical practice. Existing explainable artificial intelligence implementations in clinical settings are technically sound but governance-incomplete: they produce attribution outputs without specifying the accountability structures, equity evaluation protocols, or use-restriction mechanisms required for responsible clinical deployment. The proposed Governance-Aware Explainable Clinical AI (GXAI) framework — a structured design specification integrating Shapley Additive Explanations, Local Interpretable Model-Agnostic Explanations, an equity evaluation protocol, and a four-layer governance stack for geriatric risk stratification. The framework synthesises evidence from clinical AI, health equity research, and enterprise risk management to produce an actionable blueprint for health systems deploying transparent, auditable AI in geriatric care contexts. A geriatric-specific feature engineering taxonomy encompassing frailty trajectory, polypharmacy burden, functional status, cognitive indicators, and social determinant domains is proposed alongside a dual-attribution interface architecture differentiated by clinical user role. The governance stack positions XAI outputs as primary accountability instruments, with escalation thresholds, use-restriction protocols, and equity root-cause analysis as mandatory governance functions rather than optional enhancements. The GXAI framework addresses a critical implementation gap: the disconnect between technically capable clinical AI systems and the institutional governance requirements for safe, equitable geriatric deployment. Prospective evaluation using stepped-wedge cluster randomized designs and equity-stratified outcome monitoring is recommended to validate the framework's clinical and governance effectiveness. Index Terms — explainable artificial intelligence; geriatric risk prediction; SHAP; LIME; algorithmic fairness; clinical governance; frailty; machine learning framework
Authors
- Uche Carine Ekemezie
Publication Details
- Journal
- INTERNATIONAL JOURNAL OF HEALTH AND PHARMACEUTICAL RESEARCH
- Published
- 2026-11-17
- DOI
- https://doi.org/10.56201/ijhpr.v9.no5.2024.pg106.147
- Primary Topic
- Explainable Artificial Intelligence (XAI)
- Type
- article
- Field-Weighted Citation Impact
- 0.00