Is the SNAP-II score useful for predicting mortality in mechanically ventilated in neonates within the first 12 hours of admission?

This prospective observational cohort study was performed in the 12-bed level III NICU at Hat Yai Hospital between January 2010 and December 2010. Infants who were intubated before NICU admission were included in the study. Infants with an additional major congenital anomaly or cyanotic heart disease, or who died within 12 hours following NICU admission, were excluded. The study including prospectively abstracting of patient information from clinical charts was approved by the Hat Yai Hospital Ethics Committee.

The basic characteristics of the patients were noted, and the worst readings of six physiological parameters (lowest blood pressure, lowest temperature, lowest PaO₂/FiO₂ ratio, lowest serum pH, urine output, and seizures) in the first 12 hours following NICU admission for SNAP-II were abstracted from clinical and laboratory records, using the variable definitions reported in the original SNAP-II report [4].

During the study period, 626 infants were admitted to the NICU, of whom 241 required mechanical positive pressure ventilation. Of these 241 infants, 46 (19.1%) were initially intubated before NICU admission. Table 1 presents the basic characteristics of the study population. The mean gestational age and birth weight were 35.4 ± 4.5 weeks and 2,401 ± 1,022 g, respectively, and 33 infants (72%) were male. Eight (18%) were referred to our institution. Respiratory distress syndrome (RDS) was the most common etiology of respiratory disorder in the ventilated infants in our study (19 of 46, 41%), and this group had a mortality rate of 13%. The second most common cause of respiratory disorder in our study was meconium aspiration syndrome (13 of 46, 28%), partly complicated by persistent pulmonary hypertension of the newborn (PPHN) (3 of 13, 23%). In univariate analysis, comparing with infants who survived, the infants who died had lower OI, lower SNAP-II, lower birth weight, less length of stay. In multivariate analysis, there was no detectable factor found associated with mortality.

Table 1

Characteristics of the 46 study infants

The study population had a mean SNAP-II of 22.8 ± 14.2. The mean SNAP-II was significantly higher in nonsurvivors than in survivors (36.5 ± 11.5 vs. 22.3 ± 13.6, P < 0.01, respectively). Infants classified as having ‘severe illness’ (a SNAP-II of ≥32) had the greatest mortality risk with HR of 9.7 (95% confidence interval (CI) 1.7 to 56.5). Every one point increase in the SNAP score increased the HR of mortality by 1.08 (95% CI 1.02 to 1.14, P < 0.01). On the other hand, the median oxygen index (OI) score was significantly higher in nonsurvivors than in survivors (25.3 (range: 11.6-124.2) vs. 6.0 (range: 1.8-40.8), P < 0.01, respectively). Every one point increase in the OI increased the HR of mortality by 1.04 (95% CI 1.02 to 1.06, P < 0.01).

The AUC of ROC for the SNAP-II, and OI was 0.79 (95% CI 0.66 to 0.92, P = 0.01), and 0.85 (95% CI 0.76 to 0.98, P = 0.01), respectively. The goodness-of-fit by the Hosmer–Lemeshow test showed that the SNAP-II test and OI were well calibrated in our study to assess deviation between the observed and expected numbers of death cases (P = 0.34, and P = 0.22, respectively). A SNAP-II cut-off of 32 was derived from the ROC analysis for prediction of mortality. The sensitivity, specificity, and positive and negative predictive values (NPV) of SNAP-II ≥32 in predicting mortality were 75.0%, 76.3%, 40.0%, and 93.5%, respectively.

Currently, the treatment modalities of NICUs for infants with respiratory failure may influence mortality. Hence, early identification of infants with respiratory disorders that present a high mortality risk may be helpful in guiding clinical decision-making and resource allocation. In the present study, the SNAP-II was studied to determine its usefulness in predicting mortality in mechanically ventilated infants in our NICU. We found that the SNAP-II offered an acceptable discriminative power with an AUC of 0.79, and had a significantly good calibration for both expected and observed mortalities.

An association between mortality and higher SNAP-II in sick infants has been previously reported and the authors of the current study have previously reported the SNAP-II accurately predicted mortality in infants with PPHN [2, 9]. Those results were similar to those in the present study, because in both studies the discrimination levels were acceptable (AUC 0.72 vs. 0.79, respectively). Our results were consistent with those in the study by Skarsgard et al. who showed the discrimination ability of SNAP-II to predict mortality in congenital diaphragmatic hernia had an AUC of 0.76 [10]. Ma et al. [2] reported the epidemiology of respiratory distress associated the severity illness in late preterm or term infants. Their study had a similar study population to ours, and found that a higher SNAP-II was significantly correlated with death, with an odds ratio (OR) from multivariate logistic regression analysis of 1.07 (95% CI 1.04 to 1.10), nearly the same as the HR from the present study of 1.08 (95% CI 1.02 to 1.14). Unfortunately, Ma et al. did not analyze the discriminative ability of the SNAP-II. In the present study, the evaluation of the SNAP-II with a cut-off value of 32 in the ROC had acceptable levels of sensitivity and specificity as a predictor of mortality in mechanically ventilated infants. This cut-off point was less than the cut-off point in infants with PPHN that we found, probably because the infants with PPHN were more severely ill and had a higher mortality rate [9].

The results of our study confirm that the SNAP-II and OI can be useful predictors of mortality in infants with respiratory failure. Our results also confirm that OI can be as useful as SNAP-II in predicting mortality in sick infants [2, 11]. Therefore, we suggest using SNAP-II and OI together in assessing infants with respiratory failure within 12 hours of admission, because both are relatively simple and inexpensive, and results can be obtained within only a few minutes. In addition, birth weight was a also factor associated with mortality. However, the multivariable analysis was inconclusive, probably because of the small sample size. Morbidity prediction with the SNAP-II has been previously examined in subgroups of illnesses such as chronic lung disease or intraventricular hemorrhage [12].

Our study has some limitations. The most significant is that our mortality data was based on short term outcomes only, i.e., death or hospital discharge. Ultimately, tools used for outcome prediction should reflect longer-term outcomes of neonatal respiratory failure including growth and potential neurodevelopmental effects of the illness being assessed. Moreover, the sample size in this study was small. Although the SNAP-II has been shown to be a useful predictor of mortality outcome for a number of neonatal respiratory diseases, a larger sample size is required to validate the SNAP-II’s ability to predict mortality in infants with respiratory failure.

In conclusion, the SNAP-II has good discriminatory power in predicting mortality in infants with respiratory failure who need mechanical positive pressure ventilation after delivery. The SNAP-II score can be applied with an adequate degree of accuracy for severity assessment in clinical practice.