Background: In 1995, large differences were identified in rates of grade 3-4 intraventricular/periventricular haemorrhage (major IVH) among neonatal intensive care units (NICUs) in the Australian and New Zealand Neonatal Network.
Aims: To develop a predictive model for major IVH in order to allow risk adjustment for the variation in rates of major IVH among NICUs.
Methods: Rates of IVH were determined in 5712 infants of 24-30 weeks gestation born from 1995 to 1997. Significant antenatal and perinatal variables for major IVH in 1995 and 1996 were identified by univariate and multivariate analysis. A predictive model was developed and then validated on 1997 data.
Results: Rates of all grades of IVH fell from 1995 to 1997 (30.4 to 24.3%) but wide interunit variation remained. Seven antenatal and perinatal characteristics had significant association with major IVH: fetal distress, intrauterine growth restriction (protective), antenatal corticosteroids (protective), gestational age, 1 minute Apgar <4, male gender, and transfer after birth. A predictive model based on the last five of these variables was developed using data from 1995 and 1996 which gave an area under the receiver operator characteristic (ROC) curve of 0.76. This model was then validated on the 1997 dataset where an identical ROC curve resulted.
Conclusions: Antenatal and perinatal factors are important in the pathogenesis of major IVH. The predictive model developed from these factors can be used to adjust for confounders in interunit outcome comparison.