Differences in bleeding outcome capture between electronic health record review using natural language processing and ICD-10 coding in hospitalised children.

Abstract

NLP applied to paediatric EHRs identified substantially more clinically documented bleeding events than ICD-10 coding. Demonstrates considerable loss of clinically assessed bleeding information when relying on administrative datasets alone. Shows that narrative-based ascertainment captures the full spectrum of documented bleeding events not transferred to structured coding. Provides a scalable method for documentation-based bleeding surveillance in hospitalised children. Supports use of NLP to improve outcome ascertainment in research, safety monitoring, and quality registries.