Risk prediction using a neonatal therapeutic intervention scoring system in VLBW and ELBW preterm infants


Nihal Oygur, MD, Department of Pediatrics, Division of Neonatology, Akdeniz University Medical School, 07070-Antalya, Turkey. Email: nihaloygur@akdeniz.edu.tr


Background:  The Neonatal Therapeutic Intervention Scoring System (NTISS) is a therapy-based severity-of-illness index, The aim of the present study was to evaluate whether: (i) NTISS can predict the severity of illness with the same accuracy both in very low-birthweight (VLBW) and extremely low-birthweight (ELBW) infants, using all parameters; and (ii) the performance of NTISS can be increased by using only the significant variables.

Methods:  All inborns <1500 g, and all outborns <1500 g transferred in the first 12 h of postnatal life, were included. NTISS using 63 variables was assessed for all infants at the 24th hour. Predictive performance for the overall variables was assessed using area under the curve (AUC) for group 1 (500–1499 g), 2 (1000–1499 g) and 3 (500–999 g). Variables with good prediction were identified for each group and a second AUC was assessed using only sensitive variables. Receiver operating characteristic (ROC) curve area for all variables was compared with the ROC area for sensitive variables.

Results:  A total of 364 preterm infants fulfilled the eligibility criteria. The AUC of groups 1, 2 and 3 with all variables were 0.851; 0.834 and 0.749, respectively. The number of parameters with good prediction was 33 in group 1, 30 in group 2 and 18 in group 3. The AUC for sensitive variables was 0.848 in group 1; 0.821 in group 2 and 0.823 in group 3. When compared, increase in the description of outcome was significant only for group 3 patients (P = 0.02).

Conclusion:  NTISS using all parameters seems to be less predictive in ELBW infants. It is probably related to the use of some interventions, done as a routine procedure in almost all ELBW preterm infants, therefore NTISS may be modified according to birthweight in order to obtain a more sensitive prediction.