Identification of hematuria with a natural language processing model and validation of hematuria diagnosecodes
Rasmus Søgaard Hansen a), Rasmus Bank Lynggaard a), Martin Sundahl Laursen a), Freja Maack Lykke a), Pernille Just Vinholt a) b)
a) Dept. of Clinical Biochemistry, Odense University Hospital, Odense, Denmark
b) Dept. of Clinical Research, University of Southern Denmark, Odense, Denmark
Highlights
• The accuracy of hematuria ICD-10 codes in this study was low.
• Natural Language Processing (NLP) can be useful for accurate identification.
• An NLP pipeline can with high accuracy assist in identifying hematuria episodes.
• An NLP augmented chart review reduces the likelihood of overlooking relevant cases.