Explainable Artificial Intelligence for Clinical Decision Support Systems
Year: 2025
Type: article
Abstract: The growing integration of artificial intelligence into clinical decision support systems promises improvements in diagnostic accuracy, treatment personalization, and workflow efficiency, yet adoption remains constrained by the opacity of complex models, which undermines clinician trust and raises concerns about safety, accountability, and fairness. Explainable artificial intelligence has emerged as a critical enabler for bridging the gap between model performance and clinical interpretability. This paper presents a systems-level analysis of explainability within clinical decision support systems, moving beyond algorithmic description to examine architectural trade-offs, deployment infrastructure, governance frameworks, robustness under distributional shift, sustainability across institutional boundaries, and fairness implications. We argue that explainability must be treated not as a post-hoc overlay but as a multi-layered system property that spans data pipelines, model architecture, user interface design, and regulatory compliance. Through a structured examination of design paradigms, ranging from inherently interpretable models to post-hoc explanation engines, we highlight fundamental tensions between fidelity, latency, stakeholder needs, and evolving clinical contexts. We discuss how federated learning, edge deployment, and model monitoring architectures influence explainability guarantees, and we explore the policy landscape shaped by emerging regulatory instruments that elevate transparency to a legal requirement. Case illustrations from acute care and chronic disease management illuminate the concrete consequences of poor explainability and the systemic barriers to achieving it. The paper concludes by outlining a research agenda that prioritizes context-aware explanations, continuous verification pipelines, and multi-stakeholder governance mechanisms as preconditions for trustworthy, sustainable clinical decision support. Our analysis underscores that explainable artificial intelligence in medicine is not solely a technical problem; it is a socio-technical infrastructure challenge demanding alignment across engineering, clinical practice, and policy.
Authors: Hravastava, Tohit, Hamilton, Niklas R.
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Keywords: explainable AI, clinical decision support systems, clinical interpretability, model opacity, clinician trust, distribution shift, federated learning, edge deployment, model monitoring, regulatory compliance, transparency, multi-stakeholder governance, acute care, chronic disease management, socio-technical infrastructure
Open Access status: diamond
OpenAlex ID: W7220088448
