Abstract

Background: Illness severity scores are increasingly used for risk adjustment in clinical research and quality assessment. Recently, a simplified version of the score for neonatal acute physiology (SNAPPE-II) and a revised clinical risk index for babies (CRIB-II) score have been published.

Aim: To compare the discriminatory ability and goodness of fit of CRIB, CRIB-II, and SNAPPE-II in a cohort of neonates < 1500 g birth weight (VLBWI).

Methods: Data from 720 VLBWI, admitted to 12 neonatal units in Lombardy (Northern Italy) participating in a regional network, were analysed. The discriminatory ability of the scores was assessed measuring the area under the receiver operating characteristic curve (AUC). Outcome measure was in-hospital death.

Results: CRIB and CRIB-II showed greater discrimination than SNAPPE-II (AUC 0.90 and 0.91 v 0.84, p < 0.0004), partly because of the poor quality of some of the data required for the SNAPPE-II calculation—for example, urine output—but also because of the relative weight given to some items. In addition to each score, several variables significantly influenced survival in logistic regression models. Antenatal steroid prophylaxis, singleton birth, absence of congenital anomalies, and gestational age were independent predictors of survival for all scores, in addition to caesarean section and not being small for gestation (for SNAPPE-II) and a five minute Apgar score of ⩾ 7 (for SNAPPE-II and CRIB).

Conclusions: CRIB and CRIB-II had greater discriminatory ability than SNAPPE-II. Risk adjustment using all scores is imperfect, and other perinatal factors significantly influence VLBWI survival. CRIB-II seems to be less confounded by these factors.