Background There are exist several different scoring systems for predicting neonatal morbidity and mortality: They have been infrequently compared, especially out of the countries where they have been designed.

Objectives To compare neonatal mortality prediction models Clinical Risk Index for Babies (CRIB), Transport Risk Index of Physiologic Stability Score (TRIPS) and Score for Neonatal Acute Physiology (SNAP-II) in the ventilated newborn in ICU of Voronezh Regional Children’s Hospital, Russia.

Methods Data were collected on 221 ventilated newborns admitted in sequence to intensive care unit. The mean birth weight was 2272 (4400–760) g; gestational age 32 (27–41) weeks; male 139 (62,9%), female 82 (37,1%); 1 min Apgar score less than 4 was in 68 (30,8%) and 5 min Apgar score less than 6 was in 69 (31,2 %) babies. The mortality rate was 1,4%. Discrimination was quantified as the area under the curve (AUC) for the receiver operating characteristic curves (ROC). The calibration of the model with the best discrimination was assessed using the standardized mortality ratio.

Results Discrimination of all the scales was good: AUC for CRIB 0,821+−0,03; TRIPS 0,797+−0,03; SNAP-II - 0,738+−0,04. The standardized mortality ratios for the CRIB was 1,45. Using logistic regression method we constructed a model for mortality prediction with the independent variables birth weight and CRIB (AUC  = 0.879).

Conclusion Discrimination for these illness severity scores is good. Published models for severity of illness under predict hospital mortality in the ventilated newborn in Russia and need recalibration.