Objective To determine the independent risk factors for early onset neonatal sepsis (EONS) in a setting where the policy is to use intrapartum antibiotic prophylaxis (IAP) for known risk factors.
Design Prospective cohort study.
Setting Level III neonatal unit in a developing country.
Patients Consecutive mother–infant dyads (gestation ≤34 weeks) with no major neonatal malformations.
Interventions Thirteen putative maternal and neonatal risk factors and use of IAP were assessed. Neonates were followed until 72 h of life for signs of EONS. Blood cultures were drawn on clinical suspicion of EONS and/or prior to starting prophylactic antibiotics for high risk asymptomatic neonates.
Main outcome Culture-proven EONS (onset at <72 h).
Results 601 mother–infant dyads were enrolled (mean (SD) gestation=31.8 (2) weeks; mean (SD) birth weight 1559.4 (452) g). The best fitted multivariable logistic regression model had six independent risk factors (adjusted OR (95% CI)): vaginal examinations ≥3 (9.5 (3 to 31)), clinical chorioamnionitis (8.8 (2 to 43)), birth weight <1500 g (2.8 (2 to 5)), male sex (2.7 (2 to 5)), gestation <30 weeks (2 (1 to 4)) and no IAP (2 (1.04 to 4)). Regression coefficients were converted into scores of 6, 6, 3, 3, 2 and 2, respectively. Internal prediction accuracy was 86.5% and c statistic was 0.75 (95% CI 0.70 to 0.81, p<0.001).
Conclusions Vaginal examinations ≥3, clinical chorioamnionitis, birth weight <1500 g, male sex, gestation <30 weeks and no intrapartum antibiotics were independent risk factors for EONS among preterm infants of ≤34 weeks’ gestation.
Statistics from Altmetric.com
Early onset neonatal sepsis (EONS) is defined as onset of sepsis within 72 h of life.1 Several maternal and peripartum risk factors have been described in association with EONS. These include recto-vaginal carriage of group B streptococcus (GBS), prematurity, prolonged rupture of membranes, preterm premature rupture of membranes, spontaneous preterm onset of labour, prolonged duration of labour, repeated per vaginal examinations, chorioamnionitis, maternal systemic infections, foul smelling liquor and birth asphyxia.2,–,7
Clinical diagnosis of neonatal sepsis is difficult because of the non-specific nature of signs and symptoms; hence, antibiotics are often started empirically based on the presence of perinatal risk factors. Despite the fact that the risk factors often occur together and in various combinations, there have been few comprehensive multivariable analyses to identify independent risk factors for EONS. There have been even fewer attempts at devising a prognostic score from multivariable models. The clinician is often faced with a dilemma as regards starting antibiotics in preterm babies of ≤34 weeks’ gestation. He or she has to balance the need for not missing any cases of EONS with not over-treating with antibiotics. Therefore, a prognostic score could be a valuable tool for the clinician.
What is already known on this topic
There are several maternal and perinatal risk factors for early onset neonatal sepsis (EONS).
Intrapartum antibiotic prophylaxis (IAP) is recommended in the presence of maternal recto-vaginal carriage of group B streptococcus (GBS) and/or risk factors for EONS.
What this study adds
Many of the well-recognised risk factors for EONS are not associated with EONS when a policy of IAP is instituted for high risk mothers.
Clinical chorioamnionitis, repeated per vaginal examinations, male sex, birth weight <1500 g, gestation <30 weeks and non-exposure to IAP are independent risk factors for EONS.
Almost all previous studies are from developed countries, which have high rates of vaginal colonisation with GBS.4,–,6 The significance of the risk factors and their inter-play in countries with low carriage rates of GBS (such as India) is not known.8 The incidence of GBS bacteraemia in India has been reported to be 0.17 per 1000 live births, whereas it is 3.6 per 1000 live births in the UK.9 The impact of the risk factors in situations where intrapartum antibiotic prophylaxis (IAP) is administered for known risk factors is also not well elucidated. The use of IAP may blunt the effect of some of the known maternal risk factors on the occurrence of EONS.
Keeping the above limitations in mind, we designed a prospective study to evaluate perinatal risk factors for EONS in preterm infants admitted to a neonatal unit in India which has a policy of administering IAP for selected perinatal risk factors; and by multivariable analysis to develop a prognostic model.
Materials and methods
This prospective cohort study was conducted over a year in a level III neonatal unit of a tertiary care teaching hospital in north-west India. The hospital caters for a middle class and lower middle class population. All mother–infant dyads were screened for possible enrolment in the study. Inborn babies with a gestational age of 34 weeks or less were included. Gestation was determined by the best estimate from last menstrual period, first trimester ultrasonogram and early urine pregnancy test, and cross-checked with the New Ballard Score. Subjects with life-threatening congenital malformations and those who were discharged from hospital prior to 72 h of life were excluded. Babies were enrolled after informed written parental consent was obtained. The study had the approval of the Institute Ethics Committee.
Demographic data and information relating to putative perinatal risk factors and intrapartum antibiotics were recorded at enrolment. The putative risk factors were defined as follows:
Prolonged rupture of membranes: rupture of membranes for longer than 18 h.
Prolonged duration of labour: duration of labour longer than 18 h.
Preterm premature rupture of membranes: pre-labour rupture of membranes with preterm delivery.
Spontaneous preterm onset of labour: onset of labour before term without medical intervention.
Clinical chorioamnionitis: intrapartum fever with two or more of the following: fetal tachycardia, uterine tenderness, malodorous vaginal discharge or maternal blood leucocytosis (>15 000 leucocytes/μl).10
Intrapartum maternal fever: temperature >37.8°C in the period from onset of labour to delivery.
Frequent per vaginal examinations: three or more per vaginal examinations in labour.
Maternal systemic infections: maternal sepsis, urinary tract infection or diarrhoea in a 7-day period prior to delivery.
Foul smelling liquor: abnormal smell of liquor as perceived by the attending obstetrician or frankly purulent liquor.
Birth asphyxia: 5 min Apgar score ≤3.
Enrolled subjects were followed up for clinical signs of sepsis, recorded at least every 6 h, for a 72 h period after birth or until death, whichever was earlier. Blood culture and other appropriate cultures were performed within the 72 h period if subjects developed clinical signs of sepsis. In the event of an unexpected death, where investigations for sepsis could not be carried out before death, the patient was excluded. If investigations were carried out prior to such a death, the patient remained in the study.
As per unit policy, only those babies were started on prophylactic antibiotics who had a high perinatal risk of sepsis, even if they were asymptomatic.11 Babies at high perinatal risk underwent a blood culture within 30 min of birth from a peripheral vein, prior to starting antibiotics. If such a patient became symptomatic within 72 h of life due to suspected sepsis, a fresh blood culture was performed.
Two 20 ml blood culture bottles (brain heart infusion broth and bile broth) were simultaneously inoculated with 0.5–1 ml blood. The culture was considered positive if both culture bottles grew the same organism. If there was growth in only one culture bottle, or if it grew known non-pathogens such as aerobic spore bearers, it was designated contaminated after discussion with the microbiologists.
Definite sepsis was the key outcome in our study. It was defined as either (A) clinical signs of sepsis within 72 h of birth with a positive blood culture or (B) blood culture positive at 30 min after birth in an asymptomatic baby who was started on prophylactic antibiotics.
Sample size was calculated for one key putative risk factor—prolonged rupture of membranes. Yancey’s data suggest that 40% of babies with prolonged rupture of membranes have EONS. We calculated that 150 subjects with prolonged rupture of membranes would be required to identify a 40% incidence of EONS with the width of 95% CI being 15%.4 As our unit data showed that 25% of preterm deliveries were preceded by prolonged rupture of membranes, we planned to recruit 600 consecutive subjects.
The baseline variables were described using descriptive statistics. To determine the association between each risk factor and EONS, OR and 95% CIs were calculated. Factors that had a p value of <0.1 on univariate analysis were included in backward stepwise multivariable logistic regression models to determine the independent risk factors for EONS. Before variables were included in the model, they were tested for multi-collinearity. Interaction terms were also tested by introducing them into the model and evaluating them for statistical significance. To adjudge the model with the best fit, we used Hosmer-Lemeshow goodness-of-fit. The p value of the model, predictive accuracy and c statistic were also calculated.
Scores for each independent predictor were assigned in proportion to the magnitude of the β coefficient derived from the logistic regression model, such that the item scores were converted into whole numbers. The sum of the item-wise scores gave the total prognostic score. A receiver operator characteristic (ROC) curve was generated for the prognostic score. The score was subdivided into risk categories and likelihood ratios were determined for each category. Analysis was carried out using SPSS v 10.0.
Overall, 728 babies of ≤34 weeks’ gestation were delivered during the study period. Of these, 127 were excluded because they were discharged before 72 h, had major malformations or did not give consent or for other logistical reasons, leaving 601 babies included in the trial (figure 1). Thirty babies died before 72 h of life and all of them had been investigated for sepsis prior to death.
The mean (SD) gestational age of the study population was 31.8 (2) weeks and the mean (SD) birth weight was 1559.4 (452) g. Thirty nine (6.5%) subjects were large for gestational age, 129 (21.5%) were small for gestational age and the remainder were appropriate. Of 601 babies, 493 were singletons, 89 were born of twin pregnancies, 15 of triplet pregnancies and four were quadruplets. Seventy mothers had received a full course of antenatal steroids. Eighty five subjects were diagnosed as having EONS on the basis of positive blood cultures; none had GBS infection. Gram positive organisms accounted for 18 (Staphylococcus aureus 16, CONS 2) and the remainder were Gram negative Enterobacteriaceae. The prevalence of various risk factors in the study population is described in table 1. Intrapartum antibiotics were administered to 137 (22.8%) mothers. Among mothers who had received IAP, the median duration from the first dose of antibiotics until delivery was 12 h (first, third quartiles: 6, 36). The decision to start antibiotics was at the discretion of the treating obstetrician.
On univariate analysis, gestation <30 weeks, birth weight <1500 g, per vaginal examinations ≥3 times in labour, male sex and “no IAP” were significantly associated with EONS (table 1). Small for gestational age and clinical chorioamnionitis showed a trend towards achieving significance. No other factors were significantly associated with EONS.
There was no multi-collinearity between the above factors, with all coefficients being less than 0.8. Specifically, the coefficient for gestation <30 weeks and birth weight <1500 g was −0.49. The seven factors that were either statistically significant or showed a trend towards significance on univariate analysis were included for stepwise backward logistic regression (table 2). The most parsimonious logistic regression model with the best fit was the one that had the following six risk factors: gestation <30 weeks, birth weight <1500 g, per vaginal examinations ≥3, male sex, clinical chorioamnionitis and no intrapartum antibiotics. The p value of our model was <0.001 and the internal prediction accuracy was 86.5%. The model fitted the data well (Hosmer Lemeshow goodness-of-fit, p=0.95). The c statistic of the model was 0.75 (95% CI 0.70 to 0.81, p<0.001).
Interaction terms “gestation <30 weeks×birth weight <1500 g”, “birth weight <1500 g×male sex” and “gestation <30 weeks×male sex” were forced into the model. In this model none of the interaction terms achieved statistical significance, while the six risk factors described above retained their significance. “Gestation <30 weeks×birth weight <1500 g” had a regression coefficient of −1.16 (adjusted OR 0.31 (95% CI 0.07 to 1.34), p=0.12), “birth weight <1500 g×male sex” had a regression coefficient of −0.84 (adjusted OR 0.43 (95% CI 0.09 to 1.98); p=0.28) and “gestation <30 weeks×male sex” had a regression coefficient of −0.26 (adjusted OR 0.77 (95% CI 0.23 to 2.56); p=0.67).
The scores that we allotted to the six risk factors are given in table 2. The total score was 22. We constructed a receiver operator characteristics (ROC) curve using the total score as the predictor variable and culture positive EONS as the dichotomous outcome. The area under the ROC curve was 0.76 (95% CI 0.70 to 0.81, p<0.001).
Subjects were categorised according to the prognostic score into groups with the following scores: 0, 2, 3–4, 5–6, 7–8 and 9 and above (table 3). The probability of developing EONS increased as one ascended the order of risk groups (χ2 test for linear trends, p<0.00001). Similarly, the likelihood ratio of developing EONS increased.
We analysed the factors that were associated with the obstetric decision to give IAP (table 4). Preterm premature rupture of membranes, prolonged rupture of membranes, prolonged duration of labour and frequent per vaginal examinations were significantly associated with this decision.
The strengths of this study were that it was a large, prospective cohort study on an unselected population of preterm infants, in which we evaluated a comprehensive repertoire of putative risk factors by multivariable analysis, without any alteration of the on-going antibiotic policies of the obstetric unit. The definition of EONS was restricted to definite sepsis proven by blood culture.
We found six independent risk factors for EONS in our study. Repeated per vaginal examinations had the highest adjusted OR. This risk factor had been reported earlier by Seaward et al.7 Repeated examinations increase the chances of introducing bacteria from the external environment into the birth canal. Clinical chorioamnionitis had the second highest adjusted OR, although on univariate analysis, clinical chorioamnionitis had fallen short of achieving statistical significance. Chorioamnionitis is the end result of several other important risk factors, and thereby acts as a composite surrogate for them. It is more than just a risk factor, it is an indicator that the host inflammatory response to an infection is already established.10
Birth weight <1500 g and gestation <30 weeks emerged as independent risk factors. The adjusted ORs of these factors were lower than the unadjusted ORs. This could be due to the fact that although birth weight and gestation are basic biological determinants of neonatal outcomes, they are interlinked with several risk factors for EONS. The more premature and lighter the baby, the poorer is its immune system.4 12 Unexplained prematurity may also serve as a surrogate for a number of related risk factors (eg, silent chorioamnionitis).13 Male sex has been described earlier as a risk factor for EONS.14
The non-usage of intrapartum antibiotics emerged as an independent risk factor. This may imply that the selective use of intrapartum antibiotics for known risk factors was protective in our population. Many risk factors traditionally associated with EONS did not emerge as significant in our study: preterm premature rupture of membranes, prolonged rupture of membranes, prolonged labour, foul smelling liquor, unclean per vaginal examinations, maternal septicaemia or other systemic infections. One of the reasons behind this phenomenon could be that the obstetric decision to administer IAP was associated with some of these risk factors. Thus, the antibiotics may have prevented the vertical transmission of infection and stopped the occurrence of EONS. Other reasons could be the low genital carriage of GBS in this study population and the difference in intrapartum antibiotic policies.
Our data (table 4) show that not all mothers with risk factors received intrapartum prophylaxis. Several obstetricians were involved in managing the mothers, resulting in varying policies. Often, referred mothers arrive in our hospital at the penultimate stage prior to delivery without having received intrapartum prophylaxis for known risk factors.
If, for some reason adequate IAP is not administered despite the presence of these risk factors, they may still indicate that antibiotics should be administered to the neonate. There is not enough evidence as to how many doses of IAP constitute adequate protection.
It may also be noted that in our study population, there was a paucity of subjects (<5% each) with very prolonged rupture of membranes (≥72 h), very prolonged labour (≥22 h), foul smelling liquor and maternal septicaemia or other systemic infections. Since such subjects were under-represented, they may not have been reflected in our prognostic score. A limitation of our study was that the clinical chorioamnionitis was not confirmed histologically.
The study closest to ours in design and size was that by Yancey et al, who prospectively included 823 mothers admitted to a tertiary care obstetrics unit and performed a multivariable analysis of risk factors.4 In contrast to our study, they excluded mothers who had received antibiotics, only 141 (17.1%) of births were preterm deliveries and the study was carried out in a population with high GBS carriage. Other studies were either retrospective15,–,18 or restricted to either very low birthweight babies,14 babies with a single risk factor16 17 or those with GBS infection.5 Multivariable analysis has been attempted in a limited number of studies4 5 15 and some studies included both early and late onset sepsis.5 12 18
The results from our study may not be generalisable to settings where the rate of developing EONS is lower than ours, where intrapartum antibiotic policy is different or where GBS is the dominant organism. It may be noted that in the USA, the overall rate of developing EONS in very low birthweight babies is only 1.5–1.9%, whereas in our study population it was as high as 14%.19 Our study can help the paediatrician decide which babies are at high risk of developing EONS because the model takes into account the influence of intrapartum antibiotics given to the mother. However, without adequate validation the prognostic score cannot yet be recommended for regular use.
Competing interests None.
Ethics approval This study was conducted with the approval of the Institute Ethics Committee of the Postgraduate Institute of Medical Education and Research, Chandigarh, India.
Patient consent Parental consent obtained.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.