Study design and setting
We analyzed data from the Nagpur, India site of the Maternal and Newborn Health Registry (MNHR) collected between January 1, 2010 and December 31, 2013. The MNHR is a multi-country prospective, population-based observational study established by the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) Global Network for Women's and Children's Health Research to track pregnancy and birth outcomes. The MNHR study site in Nagpur, India, covers a catchment area of 20 rural primary health centers with an estimated total population of approximately 536,000 people in 2010.
Details of the MNHR study design and implementation have been published previously . Briefly, registry administrators identified eligible pregnant women; 99.9 % consented to participate. Registry administrators visited participants within one week after delivery and collected information about the pregnancy, delivery and maternal and neonatal outcomes through interviews with family, birth attendants and review of available medical records. They again visited participants at six weeks postpartum and collected information about maternal and neonatal adverse conditions from delivery until six weeks of age. If the infant died, a basic death audit was conducted. Data were transmitted weekly to the data coordinating center at RTI International.
Inclusion and exclusion criteria
We included all live born singleton infants with a birth weight of 1.5 kg and above. Babies weighing less than 1.5 kg were excluded because morbidity and mortality in this group is more likely related to their prematurity rather than just infection alone. Reported stillbirths, miscarriages and abortions were excluded. Infants were also excluded if birth weight information was missing or taken after seven days of age, or if there was no follow-up information at six weeks of age.
Assessment of exposure variables
Information regarding maternal age and education, parity, and hemoglobin was collected at enrollment. Information on antenatal care seeking, pregnancy complications, delivery mode and location, maternal and neonatal treatment, and breastfeeding was collected at the first postpartum visit. This information was collected by medical officers from parents, birth attendants and review of available medical records. Delivery location was categorized as home/other or as referral government center (e.g. district or other government hospitals), referral private center (e.g. private clinics, nursing homes), or first level facility (public health center or sub-center) based on facility codes. Most referral government and private centers had the capability to perform C-sections, while first level facilities did not. Presence of prolonged or obstructed labor/failure to progress was indicated as a single variable in the MNHR and determined according to adapted WHO definitions. Birth weight was usually obtained on the day of birth using study scales. Low birth weight was defined as 1.5- < 2.5 kg. Initiation of breastfeeding within one hour of delivery was based on maternal report. Gestational age was not included due to concerns about the accuracy of the last reported menstrual period and lack of ultrasound availability to confirm gestational age. No data on income were collected.
Receipt of antenatal corticosteroids was identified through the MNHR and through data collected for the Antenatal Corticosteroid Trial (ACT), an 18-month cluster-randomized trial conducted by the Global Network in six countries to identify women at risk of preterm birth and promote appropriate use of antenatal corticosteroids . In Nagpur, the trial was implemented in ten intervention clusters within the MNHR study area between June 11, 2012-December 11, 2013. Mothers could have also received corticosteroids outside the ACT study at their delivery facility, as antenatal corticosteroids were also available in some tertiary care centers.
Case Definition of Possible Serious Bacterial Infection (PSBI)
We defined PSBI based on a modified version of the Young Infant Clinical Signs Study criteria [3, 16]. The presence or absence of PSBI signs and symptoms in infants 0-6 weeks of age was recorded by registry administrators retrospectively within one week of delivery and again at six weeks of age based on interviews with family, birth attendants and/or available medical records. Live born infants met the case definition for PSBI if they fulfilled one of the following criteria: 1) Presence of breathing problems/difficulty breathing, feeding problems/stopped suckling or feeding, high fever (T > 38C or estimated by touch), hypothermia (T < 35C or estimated by touch), convulsions, or bleeding/pus-like discharge from the umbilicus, 2) Cause of death coded as “infection”, or 3) Presence of free text entries under sections on symptoms or cause of death with any symptom listed in criteria 1, infection, sepsis, possible sepsis, meningitis, and/or pneumonia. There was no clinical confirmation of symptoms, information on timing of symptoms, nor did we have the ability to collect laboratory or microbiological data.
Poisson regression models with robust variance estimation were used to assess whether maternal, peripartum and neonatal characteristics were associated with PSBI and to estimate risk ratios (RR) between groups [17, 18]. Generalized estimating equations with an independent working correlation structure and robust parameter covariance estimates were used to adjust for clustering by primary health center. We first estimated unadjusted risk ratios between PSBI and individual risk factors potentially associated with PSBI based on published literature and biological plausibility that were collected reliably in the MNHR. We then included all potential risk factors in a multivariable model for the purpose of estimating adjusted risk ratios. We included maternal, peripartum and neonatal factors, as well as year of delivery as a categorical variable to account for trends over time. Data were analyzed using STATA (version 14) .
We performed a sensitivity analysis to assess the relationship between the risk factors and two modified outcome variables classifying PSBI in a more restrictive manner. We first excluded those with breathing problems as a single symptom (Modified PSBI Outcome Variable 1), and then excluded those with breathing or feeding problems as a single symptom (Modified PSBI Outcome Variable 2). We did this to create a more specific outcome variable since breathing problems were present in 50 % of PSBI cases and could indicate a spectrum of symptoms, including transient respiratory distress following delivery, fast breathing (RR > 60), which is considered the least severe of the clinical signs that predict severe illness, or a more concerning sign such as severe chest indrawing [3, 20]. Feeding problems as a single symptom was excluded in Modified PSBI Outcome Variable 2 because it could also represent varying levels of difficulty, such as poor latch at birth, or a more serious sign such as inability to feed.
The study was approved by the ethics committees of Lata Medical Research Foundation, Partners Human Research Committee, and RTI International. Prior to MNHR study initiation, sensitization meetings were conducted for local community approval. Informed consent was obtained from all study participants. A data monitoring committee appointed by the Eunice Kennedy Shriver National Institute of Child Health and Human Development (NICHD) reviews the MNHR annually. This trial is registered at ClinicalTrials.gov (NCT01073475).