Abstract 1460 Poster Session III, Monday, 5/3 (poster 62)

BACKGROUND: SNAP-II is a parsimonious (6-item) physiologic measure of illness severity which has not been previously used to assess NICU mortality. Since variations in NICU mortality among Canadian NICUs has not been previously described, we used SNAP-II to model and assess mortality variations among 17 NICUs (80% of NICU beds in Canada) in the Canadian NICU Network.

METHOD: Prospectively collected data from all 19,510 admissions to participating Canadian NICUs were analyzed. Patients moribund on admission (n=72) were excluded. Transfers between hospitals and re-admissions were accounted for. Crude mortality rates were calculated. Logistic regression was used to model and assess variations in mortality after accounting for population risks (sex, gestation, multiple births, antenatal steroids, inborn/outborn status, surgery including ECMO), small-for-gestation (SGA), 5-min Apgar score and illness severity (SNAP-II).

RESULTS: Crude mortality rates varied from 0.6% to 11% among the 17 NICUs. Variables predictive of mortality were gestation, day one SNAP-II, SGA, 5-min Apgar, antenatal steroids, and surgery (including ECMO). Multivariate analysis revealed that mortality rates were significantly lower (OR 0.09-0.57) in 6 NICUs compared (p<0.05) to the reference (median) NICU, and significantly higher (OR 1.99-2.02) in 2 NICUs. SNAP-II was highly predictive of mortality and had the highest Odds Ratio among all the factors considered.

CONCLUSIONS: SNAP-II is highly predictive of NICU mortality. Significant variations in mortality rates exist among Canadian NICUs. This may be attributable to differences in population risks which are not captured by the present analysis but may also reflect differences in outcomes resulting from practice differences. Further analysis of practice differences may lead to improved outcomes in NICUs.

Funded by the Medical Research Council of Canada