Adaptive Bayesian Networks for quantitative risk assessment of foreign body injuries in children

Abstract

Injuries due to foreign body (FB) aspiration/ingestion/insertion represent a common public health issue in paediatric patients, which causes significant morbidity and mortality. The aim of this study is to present a Bayesian Network (BN) model for the identification of risk factors for FB injuries in children and provide their quantitative assessment. Combining a priori knowledge and observed data, a BN learning algorithm was used to generate the pattern of the relationships between possible causal factors of FB injuries. Finally, the BN was used for making inference on scenarios of interest, providing, for instance, the risk that an accident caused by a spherical object swallowed by a male child aged five while playing leads to hospitalization. BNs as a tool for quantitative risk assessment may assist in determining the hazard of consumer products giving an insight into their most influential specific features on the risk of experiencing severe injuries.