Risk factors for possible serious bacterial infection in a rural cohort of young infants in central India

Serious bacterial infection, such as meningitis, pneumonia and sepsis, is a significant cause of morbidity and mortality in neonates. Worldwide, there were an estimated 6.9 million cases in 2012 and each year serious bacterial infection is responsible for one-fifth of neonatal deaths [1, 2]. Diagnosis is challenging even in high-resource settings since signs and symptoms in neonates are often non-specific. In resource-limited settings where access to laboratory facilities is limited, diagnosis of possible serious bacterial infection (PSBI) has relied on validated clinical algorithms to identify those who would benefit from antibiotic therapy [3]. PSBI has an estimated 10 percent case fatality risk; therefore early diagnosis and treatment could save hundreds of thousands of lives [2, 4]. India bears a high burden of disease, accounting for one-quarter of all neonatal deaths, with more than 25 percent attributable to serious bacterial infections [5].

Risk factors for neonatal infection in resource-limited settings have mostly been studied in children admitted to facilities and have examined the outcome of clinical or culture-confirmed sepsis. These studies have shown that prematurity, low birth weight, prolonged rupture of membranes and maternal fever are important risk factors for neonatal sepsis, similar to resource-rich settings [6–8]. Delivery at health care facilities and skilled birth attendance has been advocated to reduce neonatal mortality, prompting the implementation of policies such as India’s conditional cash transfer scheme (Janani Suraksha Yojana) in 2005 to incentivize deliveries at facilities [9, 10]. The rates of facility delivery vary widely across India, ranging from 32–62 % in 2007–08 [11], with some areas reaching facility delivery rates greater than 90 % from 2010–2013 [12]. While facility-based delivery has the potential to improve newborn outcomes, poor infection control practices intrapartum and postpartum may contribute to the development of neonatal infections [13]. Although much has been published on PSBI in India over the last twenty years, the effect of the recent rapid change to high rates of facility delivery in India on risk factors for PSBI in the community remains less clear.

Population-based data are needed to address the full burden of PSBI, since many neonates with signs of infection, even those born in facilities, may develop signs of infection at home after discharge and do not return to a facility for evaluation. To our knowledge, there are no published population-based data on risk factors for PSBI in resource-limited settings where the overwhelming majority of births (>90 %) occur in facilities [2]. Identification of babies at highest risk for PSBI is important to further reduce neonatal mortality from PSBI. Our objective was to describe the burden of and risk factors for PSBI in infants from 0–6 weeks of age in a prospective population-based pregnancy registry in central India in the era of facility deliveries.

Our results show that PSBI occurred in 6 % of babies born to a population-based cohort of 37,379 pregnant women in rural central India, where 96 % of deliveries occurred in facilities. This is similar to the 5.1 % rate of PSBI found in a large community-based study in Ghana where 68 % of deliveries occurred in a facility [21]. It is also consistent with the pooled estimate of PSBI incidence in South Asia (7.2 %, 95 % CI 5.0–9.3) developed by Seale et al. using data from multiple studies [2]. However our estimate was lower than the 14.4 % and 9.2 % rates found in community-based studies in Bangladesh [22] and Nepal [23], respectively, where the majority of deliveries (>80 %) occurred at home.

In this population-based study with a majority of facility deliveries, many of the factors associated with PSBI were similar to those previously reported in babies admitted to facilities [6–8]. Factors that were associated with an increased risk of PSBI in our population included low birth weight, nulliparity, parity > 2, first antenatal care visit after the first trimester or no antenatal care, lack of early initiation of breastfeeding, antenatal corticosteroids, and male sex.

Within low-resource settings, low birth weight is an established risk factor for neonatal mortality and has been shown to be a risk factor for neonatal sepsis in facility-based studies [1, 6–8, 24]. Nulliparity has also been shown to be associated with early onset sepsis in a large South African cohort (RR 1.8; 95 % CI 1.4–2.3) [7]. In that cohort, nulliparity was associated with a number of traditional sepsis risk factors, such as prolonged labor and greater number of vaginal examinations. Appropriate antenatal care has been associated with improved neonatal outcomes [9]. A facility-based study in Uganda found that no antenatal care was associated with a 3-times higher odds of culture-confirmed sepsis compared to mothers who did receive antenatal care (odds ratio 3.21, 95 % CI 1.24–8.33) [25].

The finding of a lower rate of PSBI among C-section deliveries in the multivariable model was unexpected. Differences in reasons for not breastfeeding early may have contributed to this finding. In the babies who did receive early breastfeeding, risk of PSBI did not differ by delivery type. In the babies who were not breastfed early, the risk of PSBI was higher in vaginal deliveries compared to C-sections. This suggests that the difference in PSBI rates seen between vaginal and C-section deliveries may have been driven by the population of babies that was not breastfed early and their differences in baseline risk. The small percentage of babies delivered vaginally who were not breastfed early (4 %) may have been sicker, requiring medical attention that prevented breastfeeding immediately after birth. In contrast, 51 % of babies delivered by C-section did not have early breastfeeding, which was more likely due to maternal post-operative recovery rather than an indicator of the health of the baby. Nevertheless, babies delivered by C-section who were not breastfed early had a higher risk of PSBI compared to babies delivered by C-section who were. Breastfeeding has been associated with decreased rates of neonatal sepsis and other neonatal infections [26, 27]. One meta-analysis found that initiation of breastfeeding within 24 h of birth was associated with a lower neonatal mortality rate (RR 0.55, 95 % CI 0.36–0.84) [28]. As more births are occurring in facilities and the percentage of C-sections are rising [29], improving rates of early initiation of breastfeeding following C-section is a potentially important area for intervention.

Our finding of the association between antenatal corticosteroids and PSBI deserves further exploration. Antenatal corticosteroids have been shown to reduce neonatal morbidity and mortality from pre-term birth in hospitals [30]. However, a recent cluster-randomized trial of antenatal corticosteroid use in six low- and middle-income countries showed an increased 28-day neonatal mortality rate in the group who received corticosteroids (RR 1.12, 95 % CI 1.02–1.22) [15]. This may have been due to the administration of corticosteroids to women thought to be at risk for preterm birth who ultimately delivered late preterm or term infants. On secondary analysis of these data, there was no significant increase in risk of PSBI in the intervention clusters who received antenatal corticosteroids when compared to the control clusters [31]. When the composite outcome of PSBI and death was analyzed, risk was increased in the intervention clusters (RR 1.17, 95 % CI 1.04–1.30), which was driven by the difference among infants with a birth weight greater than or equal to the 25^th percentile. Within our study population of babies with birth weight ≥ 1.5 kg, corticosteroids may have resulted in a compromised immune system and increased risk for infection in the neonate if the corticosteroids were administered to the mother after week 34 of pregnancy. It is also possible that our PSBI outcome variable misclassified cases because it was based only on reported symptoms. Therefore, further study of antenatal corticosteroid use and its relationship with neonatal infection as assessed by clinical and laboratory criteria is urgently required as antenatal corticosteroids are being scaled up in low-resource settings [32].

This study has important strengths. We determined rates of PSBI up to 42 days of age in a large, population-based cohort of mothers and babies over a four-year period. Trained study staff prospectively collected information on PSBI, risk factors and outcomes, reducing the risk of selection bias when babies admitted to facilities with PSBI are studied. Our study also has several limitations. First, PSBI was defined on the basis of clinical signs and symptoms, not laboratory data or microbiological confirmation of infection. However, because laboratory diagnosis of infection in young infants is difficult in resource-limited settings, identification of PSBI has traditionally relied on versions of the definition used in this study [2]. Secondly, we did not have the ability to determine the exact timing of PSBI symptoms, nor did we require clinical confirmation of the symptoms. Questions regarding the symptoms of breathing or feeding problems, which are the least objective measures, may have been interpreted differently by respondents and/or study staff. We addressed this issue by excluding the babies with breathing or feeding problems as a single symptom in our sensitivity analysis and obtained results similar to those from the primary analysis. Third, the MNHR was not specifically designed to assess rates of PSBI and therefore does not include all factors that may be associated with PSBI, such as the duration of rupture of membranes, maternal fever, etc. We also did not include gestational age in our analysis and are thus unable to separate the effect of birth weight from that of prematurity. We chose not to include estimates of gestational age based on last menstrual period dating due to the poor reliability of this dating in our population and the inability to confirm dates with ultrasound. Lastly, because the study was completed at a single site in India, which consists of many diverse states, the findings may not be representative of the entire country.

Since neonatal infections are the third most common cause of neonatal deaths worldwide and the leading cause of death in the late neonatal period (7–27 days) [1], achieving significant reductions in neonatal mortality requires improvements in the identification and treatment of neonatal infections. Our study in a large rural population-based cohort around Nagpur, India, with a high rate of facility delivery, showed that many of the known risk factors for PSBI such as low birth weight and late or no antenatal care, were still present. Our data suggest that infants born to nulliparous women, women who present late to antenatal care, those receiving antenatal corticosteroids and infants who are low birth weight could be targeted for extra antenatal and post-natal visits for early recognition of signs and symptoms of PSBI and prompt referral. Promoting early breastfeeding following C-section may also be a potential opportunity for intervention.