Bleeding incidence in children admitted to hospital : Combining AI and
Manual Chart Review
Abstract
This study addresses the lack of information about bleeding incidences,
location and risk factors in children admitted to hospital. The primary
objective of this study was to determine the incidence of bleeding in
children admitted to hospital. Methods: In a retrospective
observational cohort study, the first admittance of 13,842 children
(<18 years old) to Odense University Hospital from 2015-2020
was analyzed. Bleeding episodes and the anatomical location of bleeding
were identified in electronic health record (EHR) text using a
combination of artificial intelligence and manual validation. The study
determined cumulative bleeding incidence during admission with
Kaplan-Meier analysis and the anatomical locations of bleedings in
frequencies and numbers. Subgroup analyses included spontaneous and
major bleeding, and bleeding during intensive care unit (ICU) stay.
Results: Overall, 1869 children bled at admission and 1523
during admission. The most frequent locations of bleeding were
cutaneous, internal, and gastrointestinal, comprising 70% of episodes.
The cumulative incidence of bleeding during admission was 29.1% and
19.0% for any bleeding and spontaneous bleeding, respectively.
Spontaneous and major bleeding occurred in 794 (6.6%) and 174 (1.5%)
of children, and six bleedings contributed to death (all central nervous
system bleeds). Bleedings occurred in 36% admissions complicated with
ICU stay. Independent risk factors for bleeding were prematurity,
hematological cancer, infection, congenital anomalies, anticoagulants,
administration and ICU stay. Conclusions: The study found a
high incidence of bleeding in admitted children, and fatal bleeding
episodes were registered. Thus, bleeding risk shall be considered during
admission of children.