Opportunities and Challenges in Using National EHR Networks for AI in Learning Health Systems

Abstract

ABSTRACT Background National electronic health record (EHR) networks can support learning health systems (LHSs) by enabling large‐scale data aggregation, monitoring, and benchmarking, but their capacity to produce trustworthy and locally deployable machine learning and artificial intelligence (ML/AI) models remains uncertain. We characterized major US national EHR networks and examined barriers to ML/AI development and deployment across the LHS cycle. Methods We conducted an environmental scan combining PubMed searches with reviews of network websites, governance documents, and federal and vendor white papers through September 2025. Eligible networks aggregated patient‐level EHR data nationally. We abstracted scale, settings, data domains, harmonization approaches, and access models, identified ML/AI studies, and mapped barriers onto a seven‐step LHS‐AI cycle from data capture to implementation and monitoring. Results We identified 23 national EHR networks spanning federal and academic consortia, vendor‐led consortia, commercial aggregators, and practice‐based research networks, covering fewer than 1 million to more than 200 million patients. Most used common data models to standardize inputs. We identified 34 ML/AI studies, but only a small subset was prospectively evaluated or integrated into clinical workflows. Common barriers included heterogeneous data capture, privacy and linkage constraints, residual variation despite harmonization, limited representativeness, need for local recalibration, and sociotechnical challenges related to implementation and evaluation. Conclusion National EHR networks offer critical infrastructure for LHSs but currently function primarily as research rather than ML/AI platforms. Addressing data, implementation, and evaluation barriers to enable ML/AI development and subsequent local deployment is essential to realizing their potential.