Data Created (NOT) for AI: Ethnographic Understanding of EMR creation

Abstract

This study explores the implications of data origin in electronic medical records (EMRs) within the context of increasing digitalization and the use of big data analytics, artificial intelligence, and machine learning in healthcare. While EMRs are promoted globally to enhance efficiency, precision, and data quality, our qualitative analysis using interviews and vignettes revealed that similar medical documentation can originate from diverse contextual practices. Notably, the data origin of unmarried working women differed from that of married couples visiting for obstetric or gynecological concerns, and women with adverse obstetric histories or accompanied by their mothers exhibited distinct documentation patterns not reflected in EMRs. These findings highlight that visually similar records may embody different meanings, emphasizing the need to consider documentation practices when analyzing EMR data. We conclude that understanding the “data origin story” is crucial for designing and applying data-intensive healthcare models responsibly and contextually.