.png)
186 – Current Topics in Phone Surveys
A Bayesian Prediction Model of Severe Intra-ventricular Hemorrhage in Very Pre-term Infants
Michael Anderson
The University of Oklahoma
Suzanne Dubnicka
Kansas State University
Shahab Noori
Children's Hospital Los Angeles
Nearly 20% very preterm infants (gestational age < 28 weeks) will experience severe intra-ventricular hemorrhage (IVH stage 3 or 4)(Stoll 2010). Prophylactic use of indomethacin has been shown to reduce the risk of severe IVH but this intervention's side effects require its judicious use on only those infants at the greatest risk of severe IVH. Current research suggests infants with increased cerebral oxygenation may be at greater risk of severe IVH (Noori 2013) and neonate cerebral sensors can easily and continuously measure this biometric resulting in high dimensional data sets. In this work we employ a Bayesian prediction model on the cerebral tissue oxygenation index (c-TOI) measures of 22 very preterm infants (5 experienced IVH) continuously monitored for 72 hours and evaluate the model's performance via leave-on-out-cross validation at 5, 10, and 12 hours. By constructing conditional densities of this biometric at each time point for both groups of infants we obtain conditional group assignment probabilities that are sequentially updated at each time point. At 12 hours, this model has 80\% sensitivity and 82\% specificity.