An Explainable Artificial Intelligence-Driven Blockchain Architecture for Secure Healthcare Data Exchange and Clinical Decision Support

Abstract

The digitization of healthcare records has improved clinical efficiency but has simultaneously introduced significant challenges around data security, interoperability, patient consent management, and the transparency of AI-assisted clinical decisions. This paper proposes a Hybrid Explainable Artificial Intelligence-Driven Blockchain Architecture that integrates a permissioned blockchain network for secure, auditable healthcare data exchange with an ensemble machine learning clinical decision support model whose predictions are made transparent through an integrated explainability layer combining SHAP, LIME, and attention-based rationale. The architecture employs a permissioned Hyperledger Fabric network in which hospitals, insurers, and regulatory auditors participate as distinct peer types, governed by smart contracts that enforce patient consent, fine-grained access control, and an immutable audit trail, while bulk clinical data is stored off-chain in an encrypted vault referenced by on-chain hashes to preserve scalability. The clinical decision support module was evaluated on a multi-institutional patient dataset comprising 186,420 records across five chronic disease prediction tasks. Experimental results show that the proposed model achieves 97.3% accuracy, 95.8% precision, 94.6% recall, and a 95.2% F1-score, outperforming the strongest baseline (deep neural network) by 4.4 percentage points in F1-score, while blockchain performance evaluation shows sustained throughput of up to 858 transactions per second at 24 peer nodes with average transaction latency remaining below 420 milliseconds.