Optimizing electronic health records to support artificial intelligence

Abstract

Electronic health records (EHRs) provide the most important data sources for artificial intelligence (AI). Gaining access to quality data suitable for advanced analytics continues to be challenging. This rapid review documents the current state of available data; identifies foundational AI data/information needs; and explores the benefits of adopting new and emerging technologies to design and implement next-generation EHRs. Opportunities to optimize EHRs for AI purposes are identified. This review was informed by expert knowledge and shared experiences supported by the literature, including technical standards. Main findings include poor ecosystem-wide infrastructures due to the lack of adopting the right set of standards, and current data and knowledge governance no longer fit for purpose. While many jurisdictions are continuing the use of legacy systems, some forward-looking national health systems and health-care facilities are adopting transformational strategies by adopting a strong data and digital focus to transition to new-generation systems. New foundational-level national infrastructures with strong leadership and governance are essential to enhance the governance and quality of available data, from collection at source throughout the entire data supply chain. Secure and ubiquitous access to high-quality EHR data at scale will foster the evolution of more intelligent and trustworthy AI. Key characteristics of next-generation EHRs supported by currently available technologies and standards that are able to meet digital era demands are provided in this paper. We conclude that the use of generative AI in clinical settings can only be reliably achieved when EHRs are optimized throughout the entire global digital health ecosystem.