Social Phenotyping for Cardiovascular Risk Stratification in Electronic Health Registries.

Abstract

Evaluation of social influences on cardiovascular care requires a comprehensive analysis encompassing economic, societal, and environmental factors. The increased utilization of electronic health registries provides a foundation for social phenotyping, yet standardization in methodology remains lacking. This review aimed to elucidate the primary approaches to social phenotyping for cardiovascular risk stratification through electronic health registries. Social phenotyping in the context of cardiovascular risk stratification within electronic health registries can be separated into four principal approaches: place-based metrics, questionnaires, ICD Z-coding, and natural language processing. These methodologies vary in their complexity, advantages and limitations, and intended outcomes. Place-based metrics often rely on geospatial data to infer socioeconomic influences, while questionnaires may directly gather individual-level behavioral and social factors. Z-coding, a relatively new approach, can capture data directly related to social determinant of health domains in the clinical context. Natural language processing has been increasingly utilized to extract social influences from unstructured clinical narratives-offering nuanced insights for risk prediction models. Each method plays an important role in our understanding and approach to using social determinants data for improving population cardiovascular health. These four principal approaches to social phenotyping contribute to a more structured approach to social determinant of health research via electronic health registries, with a focus on cardiovascular risk stratification. Social phenotyping related research should prioritize refining predictive models for cardiovascular diseases and advancing health equity by integrating applied implementation science into public health strategies.