Indian Pediatrics 2001; 38: 596-602
Early Predictors of Mortality in Very Low Birth Weight Neonates
Tarun Gera and S. Ramji
From the Department of Pediatrics, Maulana Azad
Medical College, New Delhi 110 002, India
Objective: To evaluate early predictors of mortality in very low birth weight neonates. Setting: Teaching hospital, Design: Case control study. Methods: Hospital born very low birth weight newborns (500-1500 g) enrolled for study and followed up till death or 28 days. Infants’ birth data and data on physiologic alterations, investigation and interventions in the first 24 hours of life and CRIB score were analyzed for their ability to predict neonatal mortality. Results: 115 subjects were enrolled into the study of which 47 died in the neonatal period. The factors significantly associated with early neonatal mortality included birth weight, gestation, low Apgar scroes, need for assisted ventilation at birth, need for supplemental oxygen and mechanical ventilation in the first 24 hours, presence of shock, hypoxia and acidosis ( p <0.05). The factors associated with late neonatal mortaility were birth weight and gestation only. Multivariate analysis of these factors showed that besides low birth weight, shock, need for mechanical ventilation, acidosis and high alveolar-arterial oxygen gradients were significant predictors of neonatal mortality. When compared with the CRIB score, birth weight<1200g proved to be an equally good predictor of mortality risk. Conclusions: VLBW neonates with disturbed cardio-pulmonary physiology during the first 24 hours of life, esepcially those in need of mechanical ventilation, are at an increased risk of early neonatal mortality.
Very low birth weight (VLBW) babies constitute approximately 4-7%(1) of all live births but appropriate a major share of effort, time and resources for their care. Despite this the mortality amongst this subgroup of neonate is high, contributing as much as 30% to the early neonatal mortality(1). Survival of these babies is inversely associated with their birth weight, gestation and illness severity(2). Efforts to define illness severity to predict neonatal mortality have been non-unanimous; an important confounder being the diverse management protocols. The possible exist-ence of a geographical bias in the pre-existing prognostic scoring systems, like the CRIB(3) and the SNAP(4), has not been addressed satisfactorily, this assumes a special signi-ficance in a country like India, where almost 30% of the neonates are intrauterine growth retarded (IUGR), unlike in developed countries wherein low birth weight rates are significantly lower. This study was therefore conducted to determine the factors predictive of mortality within the first 12-24 hours of life in the Indian setting.
Subjects and Methods
The study included all consecutive intramural live born infants, delivered between March to December 1998, with birth weight of 500-1500 g. A case control design in a prospectively assembled cohort was used to study risk factors for neonatal mortality.
All neonates received standardized neo-natal care. A pediatrician, trained in resusci-tation, was present at all deliveries. Antenatal steroids were rarely used and surfactant was not used in any of the babies. In the Neonatal Intensive Care Unit (NICU) careful attention was given to fluids, nutritional, metabolic, hematologic and environmental parameters. Respiratory support included oxygenation and mechanical ventilation (SLE 2000, Babylog 8000). Standard indications for ventilation were used (respiratory failure, recurrent apnea, shock, etc.). Babies with birth weight <750g and those with multiple malformations were not mechanically ventilated. Exception-ally, a neonate in need of mechanical ventila-tion could not be provided ventilation due to non-availability of a ventilator bed. The cost of care of all neonates was borne by the state. The data was prospectively recorded in all infants at birth and during the first 24 hours of admission to the NICU on a standardized proforma, as per the details given below:
All the patients were followed up till the time of death or the first 28 days of life which ever was earlier.
A univariate analysis was used to identify the significant predictors of early and late neonatal mortality. Continuous data (normally distributed, e.g., heart rate, birth weight, gestation) were analyzed by Student’s t test. Non parametric continuous variables (e.g., Apgar scores, A-aDO2) were analyzed using the Kruskal Wallis test. All proportions were analyzed using the Chi square test. To adjust for confounders, all variables found to be significant on univariate analysis were analyzed using a forward stepwise logistic regression analytic process. Variables were considered significant if they had a p <0.05.
A total of 115 subjects were enrolled into the study. Of these 47 died; 36 (76.5%) in the early neonatal period and the rest (23.5%) in the late neonatal period. The causes of early neonatal period mortality included septicemia (30.6%), hyaline membrane disease (30.6%), intraventricular hemorrhage (22.2%), pulmo-nary hemorrhage (8.3%), extreme prematurity (5.5%) and birth asphyxia (2.8%). The causes of late neonatal mortality included sepsis (8 deaths), milk aspiration (2 deaths) and congenital malformations (1 death). The sepsis in all the infants were nosocomial. Of the 25 infants ventilated only 5 (20%) developed ventilator related sepsis or pneumonia.
Risk Factors for Early Neonatal Mortality
Maternal factors such as age, parity, number of antenatal visits, medical and obstetric complications were comparable between neonates who survived and those who died. Neonatal factors such as sex and intrauterine growth status were comparable between the survivor and mortality groups (males 44.3% and SFD 73.4% in survivors; males 44.0% and SFD 75.0% in mortality group). Infants who died had a significantly lower birth weight, gestation and Apgar scores than the survivors (Table 1). The distribution of Apgar scores across different gestational ages was comparable. The odds for mortality amongst VLBW needing resuscitation at the time of birth was 3.15 (95% CI = 1.22, 8.19). However there were no differences with regard to the duration of resuscitation in the two groups.
Physiologic parameters such as heart rate, respiratory rate, temperature, incidence of apnea, and the urine output were comparable in the two groups. Shock was found to be significantly associated with neonatal morta-lity with an odds ratio of 19.3 (95% CI = 3.6, 136.5) (Table I).
It was observed that a singnificantly larger proportion of neonates who died required supplemental oxygenation in the first 24 hours (Table I). The need for supplemental oxygen predicted a 5.9 (95% CI = 2.3, 15.9) times higher risk of early neonatal mortality compared to those who did not require it. It was also observed that a significantly larger number of neonates who died required mechanical ventilation in the first 24 hours as compared to those who survived (OR = 18.5, 95% CI = 5.4, 67.8) (Table I). The need for mechanical ventilation is a possible proxy for illness severity.
The infants who died had a lower mean PO2 and a higher PCO2, reflecting a more severe lung disease amongst those who died. The infants who died were more acidotic (a combination of both respiratory and meta-bolic, as reflected in the higher mean PCO2 and greater base excess) as compared to survivors, suggesting a significant alteration in pulmonary physiology of the neonates who died. This is further supported by the significantly higher A-aDO2 in the group that died. There were no significant differences in the two groups with regard to the other laboratory parameters. These observations suggest that neonates who died had a significantly greater pulmonary illness, i.e., mechanical ventilation needs being a possible proxy to illness severity. The greater illness severity in the mortality group is supported by a significantly higher CRIB score in the mortality group [median (range) = 4 (0-18)] as compared to the survivors [median (range) = 1 (0.5)].
Risk Factors for Late Neonatal Mortality
The only variables which were signifi-cantly associated with late neonatal death were birth weight and gestation. It was observed that the babies who died had a significantly lower mean (S.D.) birth weight [1148(200)g] and gestation [30.9(3.1) weeks] as compared to the birth weight [1352(156) g] and gestation [34.2(2.7) weeks] of the infants who survived
Table I Significant Factors Associated with Early Neonatal Mortality
* p <0.001, ** p <0.05, ***p = 0.07.
Multivariate Analysis of Risk Factors
Multivariate analysis was used only for risk factors associated with early neonatal mortality (Table II). The variables selected for multivariate analysis included birth weight, gestation, Apgar scores at 1, 5 and 10 minutes, need for resuscitation at birth, shock, need for supplemental oxygen and mechani-cal ventilation, pH and the alveolar arterial oxygen gradient. The initial modeling did not include mechanical ventilation as its effect was expected to be dependent on the skills of the care providers. The independent pre-dictors of mortality were shock, lower birth weight, low pH and large A-aDO2. Since shock, acidosis and hypoxia were indications of ventilation, another model was used wherein shock, pH and A-aDO2 were replaced by mechanical ventilation, the other variables being the same. It was observed that the birth weight and mechanical ventilation were the only significant risk factors for mortality. However, the odds of mortality due to ventilation was far greater than the combined odds of shock, low pH and large AaDO2, suggesting that iatrogenesis could possibly be a confounder in ventilated babies.
Table II Unweighted Logistic Regression Analysis of Risk Factors for Early Neonatal Mortality
The factors found to be significantly associated with mortality on multivariate analysis were then compared, both indivi-dually and in combination with the CRIB score. A CRIB >1 had a sensitivity of 71.7% and a specificity of 83.4% in predicting neonatal mortality. It was observed that birth weight <1200 g, need for ventilation and shock together had a specificity of 100% but a very low sensitivity (17%) in predicting mortality. Similar results were obtained for shock (specificity = 93%, sensitivity = 25.5%) and need for ventilation (specificity = 95.5%, sensitivity = 46.8%). However, a birth weight <1200 g alone compare well with CRIB score in predicting mortality (sensitivity = 63.8% and specificity = 88.2%).
Survival of VLBW babies admitted to NICUs in India is generally lower than in the developed world with wide variations in the performances in the different NICUs within the country itself. This situation has been attributed to the intrinsic limitations of a poor infrastructure and limited resources. The purpose of having an outcome assessment score or a prediction model is not aimed at limiting treatment or to withdraw treatment from those with a poor prognosis but to identify the patients who are at a higher mortality risk and to optimize the health resource distribution. Furthermore, an agree-ment on a minimum care data set of clinical and physiological information could improve neonatal audit and help identify effective treatment policies.
A number of antenatal and intrapartum factors have been reported to be significantly associated with perinatal and neonatal mortality(7-11). In the present study none of these factors were observed to be significantly associated with neonatal mortality in VLBW babies.
Various authors have elucidated the significant role of prematurity and low birth weight in contributing to neonatal mortality (9,12,14). Studies have, however, been inconsistent with regard to the relative contri-bution of the two and the role of intrauterine growth retardation in VLBW babies. The present study observed that the relative contri-bution of birth weight to mortality was much greater than gestation amongst VLBW babies.
Various investigators, including us, have found low Apgar scores to be significantly associated with neonatal mortality(8,15). On multivariate analysis however, Apgar scores or need for resuscitation were not significant. This is probably attributable to the fact that the more premature group of infants who died had a greater difficulty in making the transi-tion from an intrauterine to an extrauterine environment and required ventilatory support even beyond the delivery room and, therefore, continued ventilatory support served as a better indicator of mortality.
The presence of RDS has been included among the four traditional factors influencing mortality, besides birth weight, gestation age and sex. Almost 30% of the deaths were attributed to RDS and all these infants required ventilatory support, indicating a more severe pulmonary pathology in the babies who died. However the intercentre variation in the definition and the diagnosis of disease limits its use for predictive purposes. This prompted us to use a more objective criterion to define the respiratory illness in the neonates. The infants who died had significantly lower PO2, higher PCO2, greater need for supplemental oxygen and a higher A-aDO2; all pointers to the existence for a pulmonary pathology, probably a RDS like illness. On multivariate analysis these factors, when replaced by need for mechanical ventilation, suggested that mechanical ventila-tion served as a more useful proxy for the cumulative effect of all these variables. This observation assumes great significance in view of the fact that none of the existing scor-ing systems include therapeutic interventions such as need for ventilation amongst the predictors of mortality. Appropriate inter-vention may significantly blunt the predictive power of physiologic and laboratory para-meters. In this context the need for such therapeutic measures by itself can serve as a reasonably good mortality predictor.
An important observation made in the present study was that the risk of mortality in ventilated babies was far greater than the combined odds of shock, low pH and high A-aDO2. This highlights several points. Firstly, mechanical ventilation of very ill and tiny neonates requires well trained medical and nursing personnel, and improved infra-structure. The lack of these in the Indian context is a factor governing high mortality in the Indian NICUs compared to those of the developed nations.
In conclusion it can be said that VLBW neonates, especially those with a birth weight less than 1200 grams, who are critically ill due to disturbances in the pulmonary and circulatory physiology have a very high risk of mortality. Interventions such as the need for ventilation may increase the burden of mortality contributed to by severe illness. It is therefore imperative that NICUs caring for large number of VLBW infants must maintain high standards of neonatal care with a continous audit of the care provided in order to salvage more of these VLBW infants.
Contributors: TG conducted the study and drafted the manuscript. SR prepared the study design, supervised the study and conducted the statistical analysis. He will act as guarantor for the paper.