Advancing Healthcare Administrative Automation: A Multi-Agent Framework for Prior Authorization and FHIR Interoperability

Abstract

Prior Authorization (PA) remains one of the most administratively burdensome processes in United States healthcare, delaying medically necessary care and generating billions of dollars in annual administrative costs. The CMS Interoperability and Prior Authorization Final Rule (CMS-0057-F), published in January 2024, requires impacted payers to expose standardized HL7 HL7® FHIR® application programming interfaces by January 1, 2027 and enforce shorter decision timelines beginning January 1, 2026: 72 hours for urgent requests and seven calendar days for standard requests. This regulatory shift creates a strong opportunity for agentic artificial intelligence systems capable of autonomously executing multi-step administrative workflows, invoking external tools, synthesizing heterogeneous clinical documents and generating structured, auditable outputs. We present a general-purpose agentic framework for healthcare administrative automation built around HL7® FHIR®-native data exchange, the Model Context Protocol (MCP) for secure tool orchestration and deterministic structured outputs designed for regulatory traceability. The framework comprises three cooperating agents: an Intake Agent for patient and coverage extraction, a Clinical Review Agent for multi-format document synthesis and a Policy Evaluation Agent for evidence-to-criterion mapping with confidence scoring. Each agent is grounded by typed FHIR tooling and governed by strict output schemas. We describe design principles for accuracy, latency, auditability and HIPAA compliance; map system components to CMS-0057-F requirements; and demonstrate the framework on a representative prior authorization case for lumbar spinal fusion. The results show how the multi-agent system assembles clinical evidence, evaluates policy criteria with calibrated confidence and produces structured outputs aligned with CMS-0057-F transparency requirements.