425 Mortality Risk Scores for Very-Lowbirthweight Infants (VLBWI) Admitted to a Single Neonatal Unit (NICU) in a 15 Year Period

Abstract

Background: CRIB score was designed to assess VLBWI initial disease severity and mortality risk. Recently, its validity has been questioned and other scales used. Aim: To assess the validity of CRIB to predict VLBWI pre-discharge mortality cared in a single NICU during 15 years, and to compare it with other scores. Methods: CRIB score was prospectively calculated within 24 hours of life in 1,048 inborn VLBWI admitted to a single NICU from 1994 to 2008. Scores validated by two networks Neocosur -birthweight, gestational age (GA), gender, prenatal steroid use and 1-min Apgar score-, and EuroNeoNet (adds C-Section) were calculated by regression models using the step-wise selection method. Calibration was measured by Hosmer-Lemeshow goodnessof- fit and discrimination by area under ROC curve. Comparisons between models were based on nonparametric tests over ROC curves. P< 0.05. Results: Mean birthweight and GA were 1,120+262 g and 29+3 wks. Overall neonatal and pre-discharge mortality rates were 12.4% and 15.6%. The model to predict pre-discharge mortality that combined birthweight and GA (ROC: 0.862) adjusted significantly better than those based on birthweight (0.85) or GA (0.83). CRIB- adjusted model (ROC: 0.84) didn't improve pre-discharge mortality prediction compared to birthweight and GA model. Both, ROC curves for Neocosur (0.868) and EuroNeoNet (0.87) scores were significantly higher than for CRIB. Conclusions: Nowadays, CRIB score has a lower predictive ability for pre-discharge mortality than birthweight and GA combined or than the Neocosur and EuroNeoNet scores. Reasons for this should be explored. Performed partly by SANCO's EuroNeoStat II grant (2008/1311).