A Hybrid RAG and Rule-Based System for Explainable Healthcare Claims Adjudication

Abstract

This study addresses the critical challenge of automating healthcare claims adjudication within the complex regulatory landscape of dental and medical insurance policies. Traditional rule-based systems, while reliable for straightforward cases, often falter when confronted with contextual policy interpretations, whereas purely AI-driven models risk generating unsubstantiated decisions, undermining trust and compliance. To bridge this gap, we propose a hybrid adjudication framework that synergistically integrates deterministic rule-based logic with retrieval-augmented generation (RAG) and large language model (LLM) reasoning. The system leverages a synthetic corpus that emulates Medicaid policy manuals from multiple anonymized U.S. states (State A, State B, State C and State D), using semantic indexing and similarity-based retrieval to anchor AI reasoning in structured policy text. The adjudication pipeline first applies rule-based filters to resolve clear-cut cases, subsequently invoking RAG-enhanced LLM inference for complex scenarios requiring interpretive judgment. Empirical evaluation across diverse dental and medical claims demonstrates that this hybrid approach achieves superior accuracy, transparency, and policy alignment compared to standalone methods. Notably, the system generates structured, auditable explanations with precise policy citations, enhancing interpretability and regulatory compliance. These findings suggest that hybrid RAG and rule-based architectures offer a robust, scalable solution for modernizing healthcare claims processing, balancing the rigor of deterministic rules with the flexibility of AI-driven reasoning.Explainability is not merely desirable but legally mandated in healthcare claims, where denial decisions must include specific policy justifications accessible to beneficiaries, providers, and regulators. By grounding every adjudication decision in explicit policy text with citations, the system bridges the critical gap between AI capabilities and healthcare regulatory requirements, enabling both sophisticated automated reasoning and complete auditability.