MACHINE LEARNING FOR EFFECTIVE IDENTITY MATCHING IN HEALTHCARE CUSTOMER DATA PLATFORMS

Abstract

Accurate patient identity resolution across disparate healthcare systems is a significant challenge due to the absence of a universal patient identifier. This fragmentation hampers cohesive patient care and poses obstacles to effective data exchange. Beyond healthcare, industries employing Customer Data Platforms (CDPs) face analogous issues in unifying customer identities from varied data sources. This paper proposes a hybrid AI-driven patient identity matching model that integrates probabilistic record linkage (PRL) with a deep learning-based transformer network. Designed to enhance the accuracy of linking fragmented patient records across multiple healthcare systems, this model also offers valuable insights for improving customer identity resolution within CDPs. Experimental results demonstrate the model’s effectiveness in improving precision and recall in identity matching, laying the foundation for enhanced interoperability in healthcare data exchange and more robust customer identity management in CDPs.