EHR-Conditioned Patient Self-Report Generation with Factual Constraints
Journal of Clinical Practice and Health Sciences · 10.09.2026 · 6 (2) · 174-184
DOI: 10.71465/jcphs82 ↗Kurz gefasst
Fachartikel in der Zeitschrift Journal of Clinical Practice and Health Sciences, online erschienen am 10.09.2026. Verfasst von Seo-Hyun Yang und 2 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
The digitization of modern healthcare systems has resulted in an unprecedented accumulation of clinical data, predominantly stored in Electronic Health Records. While these records are invaluable for clinical decision-making and continuous patient monitoring, they remain highly inaccessible to patients due to the prevalence of complex medical jargon, dense abbreviations, and unstructured narrative formats. Transforming these clinical documents into patient-friendly self-reports is a critical step toward enhancing patient health literacy, empowering shared decision-making, and fostering better adherence to treatment protocols. However, leveraging advanced natural language generation models for this task introduces a significant risk of hallucination, where the model may generate medically inaccurate or unsupported information. Such factual deviations in the healthcare domain can lead to severe consequences, including patient anxiety or incorrect self-management. This paper comprehensively investigates the problem of Electronic Health Record-conditioned patient self-report generation with a stringent focus on factual constraints. We explore the architectural modifications necessary to bind natural language generation processes strictly to the underlying clinical source texts. By integrating entity-level factual constraints and semantic verification mechanisms directly into the generation pipeline, the proposed frameworks ensure that all generated patient-facing text remains strictly faithful to the original clinical documentation. Through extensive methodological analysis and simulated experimental validations, this study demonstrates how natural language processing can bridge the communication gap between healthcare providers and patients without compromising clinical accuracy.
Zitieren
APA
Yang, S., Yoo, S., & Song, M. (2026). EHR-Conditioned Patient Self-Report Generation with Factual Constraints. Journal of Clinical Practice and Health Sciences, 6(2), 174-184. https://doi.org/10.71465/jcphs82
Vancouver
Yang S, Yoo S, Song M. EHR-Conditioned Patient Self-Report Generation with Factual Constraints. Journal of Clinical Practice and Health Sciences. 2026;6(2):174-184. doi:10.71465/jcphs82
BibTeX
@article{Yang2026,
author = {Yang, Seo-Hyun and Yoo, Soo-Min and Song, Min-Jae},
title = {EHR-Conditioned Patient Self-Report Generation with Factual Constraints},
journal = {Journal of Clinical Practice and Health Sciences},
year = {2026},
volume = {6},
number = {2},
pages = {174--184},
doi = {10.71465/jcphs82}
}
EUVIMED