Abstract 1413 Poster Session III, Monday, 5/3 (poster 61)

Two measures of illness severity specific to neonatal intensive care, SNAP (Score for Neonatal Acute Physiology), SNAP-II, and their perinatal extensions (PE) were validated by their use in important morbidity for smaller babies. The SNAP was developed for babies of all birth weights and had been validated as a predictor of mortality, nursing workload, therapeutic intensity, and length of stay. To complement previous studies, the current study was aimed at validating the four scores (SNAP, SNAP-PE, SNAP-II, SNAP-II-PE) on the first day of admission in predicting the occurrence of death, BPD (bronchopulmonary dysplasia), grade III and IV IVH (intra-ventricular hemorrhage), and PVL (periventricular leucomalacia) among infants less than 33 weeks gestational age.

Admissions to participating NICUs of the Canadian NICU Network from January 1996 to October 1997 were recorded. The exclusion criteria included gestational age greater than 32 weeks, age of admission beyond three days, length of stay less than 24 hours, and moribund on admission. If readmissions/transfers were within the participating NICUs, outcomes were linked through all admissions for an infant. The resultant sample was 4907 infants/admissions.

Predictive performance of the four scores were assessed by the area under the Receiver Operator Characteristic curve (AUC) and their standard errors (SE), using the techniques of Hanley and McNeil. (Table)

Table 1 No caption available
Full size table

The results showed comparable performance of SNAP and SNAP-II in predicting death, BPD, IVH, and PVL. PE had better performance than SNAP and SNAP-II. Overall, the four scales performed well for death, BPD, and IVH, but poor for PVL. The findings supported the use of day-1 score to control for initial illness severity in making comparison between hospitals for outcomes of death, BPD, and IVH among infants less than 33 weeks gestational age. The scores performed poorly for PVL, which suggests PVL is not associated with initial illness severity or that the physiological variables included in the SNAP were not indicative of the physiological derangement associated with PVL. Further investigation is needed to better define physiologic predictors for PVL.

Funded by the Medical Research Council of Canada