Illness severity scores are widely used in neonatal care. These scoring systems are useful for Group predictions (Comparing study groups for similarity of risks, Auditing the severity of illness in different units, Comparing performance of different units, Reviewing if infants are treated appropriately for risk, Determining trends in results over time, Comparing rates of complications) and Individual predictions (Giving prognostic information, Stratifying infants in trials, Determining individual treatment) .
Usually the scores are created in one of the two ways. “Medical” scores are derived by expert panel using clinical knowledge to select the variables to be included in the score and their relative weights. “Statistical” scores are created when collected data is used in identifying which variables have strong association with the outcome of interest and their relative weights. There is evidence that statistical scores outperform purely medical scores. However, clinical knowledge should contribute to the choice of variables included in the final model .
To be practically useful such a score should rely on physiology rather than on diagnosis or therapy. It should be easy to calculate in clinical and research settings, be reliable, have sufficient range to distinguish among the broad spectrum of cases seen in NICUs, be applicable early (at admission), have ability to predict mortality, specific morbidities or cost of various categories of neonates .
Various researchers have tried to derive and validate the scores that fulfill the above said criteria to as large an extent as possible. Certain well established of these scores include APACHE and TRISS scores in adults and PRISM and PeRF scores in pediatric intensive care. Neonates are a different set of patients as they have difference in physiology, nature of diseases, associated congenital malformations and different need of care. Thus, these scores were felt insufficient for neonates in general. International Neonatal Network proposed a system of predicting mortality in preterm neonates and small for gestational age babies in 1993. Since then various scoring systems have been proposed for neonates [3,16-19].
Many risk factors have been identified to assess prognosis and to compare the outcome in different nurseries. The traditional risk factors include birth weight, gestational age, respiratory distress at birth, and male sex. That low birth weight is a major determinant of neonatal mortality has been recognized since 1930, when the Finnish pediatrician Yllpo argued that infants born weighing 2500 g or less were at substantially increased risk of death. More recent evidence, however, indicates that birth weight (even in conjunction with other demographic markers such as sex and race) is insufficient to explain large variations in neonatal mortality among neonatal intensive care units [20-22].
The birth weight and gestational age are two markers that have been used widely to classify the neonatal patients in such groups. On univariate analysis in our study, the gestational age and admission weight were significant factors in predicting factors for overall mortality. The relative risk of increase in mortality was three fold for preterm neonates while that was two-fold for low birth weight neonates undergoing surgery. However, on multivariate analysis only the gestational age was significant predictor of overall mortality. This is because of deficient reserves and immunity in preterm neonates. The admission weight could not achieve significance in predicting mortality. This could also point to the fact that the mortality is independent of weight of neonate if the organ systems are mature and proper care is provided to the neonates. The reason could be that the previous studies have been done in VLBW neonates with immature body systems and that has more profound effects on the survival or that in our cohort we included neonates with higher birth weight (only 9 VLBW neonates and 92 low birth weight neonates) with not so much different functional systems, and hence less difference in survival.
The age of the neonate at admission/ surgery have had insignificant effect on mortality as in previous studies. The effect of gender on mortality has been different in studies with most denying any significant association between gender and mortality. Our results also suggested no association between sex of the neonate and mortality.
Among neonatal physiologic parameters, we studied the vitals, respiratory distress and urine output as predictors of mortality. The heart rate signifies hydration status and cardiac function. It may be deranged in any stress, fever, hypothermia, CNS depression secondary to hypoxia or otherwise. Bradycardia was seen in 6 patients at admission of whom 4 (66.7%) died; tachycardia was seen in 18 patients of whom 12 (66.7%) died. The heart rate at presentation was insignificant for early mortality but tachycardia was found to be a significant predictor for overall mortality. It was found that the patients who presented in shock and tachycardia could survive early post-operative by intensive care, ionotropes and in many cases ventilatory support. But the effect of shock persisted and increased overall mortality. The mean blood pressure was found to be significant by univariate analysis but on sequential multivariate analysis could not predict mortality separate from tachycardia. The effect of shock on mortality has been demonstrated by various authors in literature and is also included in the SNAP score.
The respiratory rate (RR) is an indicator of respiratory distress and CNS depression. It may increase in cases of malformed, premature lungs, external compression, increased oxygen demand (like sepsis), pneumonia, etc. It may decrease in cerebral hypoxia or respiratory failure. The respiratory distress was found to be significant in predicting early as well as overall mortality by univariate analysis in accordance with English literature. Two (1.2%) babies were received gasping. Both the babies died. The presence of tachypnea increased the risk of mortality, in the present study, by a factor of 4. This was found to be the most important factor in predicting mortality. The respiratory distress has been included in CRIB scoring system as minimum and maximum FiO2 needed to keep saturation above 85%. Similarly, the SNAP score includes respiratory system immaturity and abnormality as RR, pO2, pO2/FiO2, pCO2 and oxygenation index. Respiratory distress has also found its place in other proposed scoring systems like NICHD score, Berlin score and NMPI score. In 1990, Tarnow Mordi et al found increased sensitivity with inclusion of FiO2 and pH in assessing prognosis of VLBW neonate. The importance of respiratory decompensation as strong predictor of neonatal mortality was identified as early as 1962 by Waterston et al in neonates with esophageal atresia [10,23].
The documentation of hypothermia was also found to be statistically significant in predicting overall mortality by univariate analysis. But hypothermia at presentation lost its statistical significance for overall mortality by multivariate analysis. This suggests that the effect of hypothermia wanes in long term and good neonatal care to keep baby warm can prevent late mortality.
Most of the patients included in the study were out-born, among which most were home deliveries. Thus accurate documentation of APGAR score was not available. Although the previous studies have shown APGAR score to be significant factor in neonatal mortality, we could not evaluate the factor for the same.
The decreased urine output in immediate post-operative period is suggestive of inadequate hydration status, presence of shock, poor renal perfusion and multi-organ dysfunction syndrome in sepsis. The urine output was shown to be a significant factor in predicting overall mortality by univariate analysis. However, the urine output could not achieve significance by multivariate analysis. This can be because oliguria seen in cases with shock/ multi-organ dysfunction syndrome affects mortality while that due to hypovolemic shock can be improved with adequate resuscitation.
The presence of associated life threatening anomalies is significant predictor as these anomalies (mainly cardiac anomalies) can cause death unrelated to the anomaly for which the neonate has been operated. Such decompensation is precipitated by anesthesia and surgical stress. The exact role cannot be defined as the presence could not be exactly quantified.
The hematocrit and total leukocyte count at presentation could not achieve statistical significance. This is expected as the neonate has normal hematocrit even in presence of severe anomalies and maternal anemia. Hematocrit has been included only in SNAP scoring and is removed from SNAP II derived for purpose of simplifying the SNAP score. Lack of statistical significance for TLC may be explained by the delay in the host immune response to antigen thus no significant difference in TLC at presentation. TLC is not included in any score other than SNAP. The platelet count achieved significance by univariate analysis in predicting overall mortality, but the significance was not demonstrated by multivariate analysis.
The renal function at presentation in terms of serum values of blood urea, serum creatinine and serum electrolyte values did not demonstrated any significant difference in cohorts that died and survived. This may be explained by the fact that neonatal blood levels may be within normal limits till 48 hours of age due to feto-placental exchange of factors till birth and slow build up of these in blood over time. They thus are not good predictors of mortality in first 48 hours of life.
Hypoglycemia is considered a very important factor in neonatal care. Hypoglycemia is one of the prognostic markers included in SNAP and is also recognized as an important cause of neonatal mortality in nurseries. We documented significance in predicting mortality by univariate analysis but the association was not found by multivariate logistic regression analysis.
This is understandable as all the neonates undergoing surgical intervention were given continuous infusion of intravenous fluids containing glucose for a considerable period and thus preventing chances of hypoglycemia while in hospital stay. Also, the preterm and VLBW neonates, who are prone to repeated hypoglycemia, were very few in numbers.
Metabolic acidosis as documented by arterial blood gas analysis (pH, base excess, and serum bicarbonate levels) were found to be significant factors in predicting early and overall mortality. The metabolic acidosis occurs in neonates by tissue hypo-perfusion, ischemia resulting in anaerobic metabolism with lacto-acidosis.
The liver function tests as documented by serum bilirubin levels did not statistically differentiate between the neonates who died from those that survived. This is expected as the liver is immature in almost all neonates and the serum bilirubin levels were increased in both the cohorts.
C-reactive protein is a marker for neonatal sepsis. The exact levels could not be documented due to unavailability of standard laboratory. The raised levels were shown to be significant in predicting overall mortality by univariate analysis that lost significance in multivariate analysis. This is against expected as neonatal sepsis is the major cause of mortality in our study population and also shown by other studies previously. This can be explained by the fact that levels of CRP may not have risen at the admission and sepsis is acquired later. The finding may also be biased due to unavailability of standard laboratory in our hospital.
The overall mortality is predicted by shock, respiratory distress and gestational age. The effect of premature respiratory and renal systems was not seen to influence the survival, because the majorities were either full term or borderline preterm.
The improvement requires good antenatal care so as to reduce the chances of pre-term deliveries and better supportive management of these babies. The shock and respiratory distress point directly or indirectly to neonatal sepsis as the most important cause for mortality. The effect of sepsis on respiratory, cardiac and renal function was evident and related directly to the pre-mortality incidents. The importance of hand hygiene and maintenance of strict asepsis in nursery cannot be overemphasized. It was observed, if proper antibiotic if started in proper dosage and at proper time (before onset of irreversible damage), the neonate could be salvaged in most circumstances. Impact of proper supportive care could also be identified in the study group.
We, in the present study, could not derive enough statistical relation to derive any scoring system universally applicable to all surgical neonates. We believe that the variation in derangement of physiology and predisposition to sepsis depends on surgical condition and hence deriving at any reliable scoring system for all surgical neonates may not be possible. Although a larger study may disprove the belief and further studies are needed.