Setting
We conducted this study in a tertiary hospital, Paropakar Maternity and Women’s Hospital, located in Kathmandu, Nepal. This government-funded hospital has 415 beds, with 407 staff equipped to provide comprehensive obstetric and gynecological services. The hospital has three delivery units, as well as a Kangaroo Mother Care unit, special newborn care unit and neonatal intensive care unit for the care of small and sick babies born there. Around 22,000 deliveries take place annually, with an incidence of preterm birth and low birth weight at 9 and 11 %, respectively in 2012, and a neonatal mortality rate of 9 per thousand live births [
11].
As part of the larger study evaluating the impact of neonatal resuscitation protocol implementation, the study received approval from the Hospital’s Institutional Review Committee, the Nepal Health Research Council (Reg. No. 37/2012) and the Ethical Review Board of Uppsala University (dnr 2012/267). The study was registered as clinical trial, ISRCTN 97846009 [
10]. Written consent was taken from all women who participated in the study.
Data collection
A surveillance system was set up to collect socio-demographic, obstetric and postpartum information from the women in the case and referent populations at the admission, delivery and postnatal units. A surveillance team member at the admission unit collected information from the women who were admitted to the hospital for delivery. The team randomly selected 20 % of women admitted to the hospital using a lottery technique. The surveillance team members at the delivery and postnatal units followed the referent women until discharge and followed up on the birth outcome through telephone interview conducted 28 days after delivery. The surveillance team at the delivery and postnatal units also collected information on the case population, i.e. all neonatal deaths that occurred in the hospital. Information about the case and referent populations was taken from the women’s individual client journals, including demographic characteristics, obstetric history, intrapartum clinical progress and outcomes, and neonatal information. For certain socio-economic information, short interviews were completed with the women from the case and referent populations.
Variables
Neonatal mortality: Death of an infant from the time of birth until 28 days.
Preterm birth: Babies born before 37 completed weeks of gestation, estimated by the date of the mother’s last menstrual period or based on clinical examination of the newborn.
Term birth: Babies who were born at, or after, 37 completed weeks of gestation, estimated by the mother’s last menstrual period or based on clinical examination of the newborn.
Small for gestational age (SGA): Babies whose birth weight was less than the 10th percentile according to the appropriate gestational age and sex-specific reference population standards [
8].
Appropriate for gestational age (AGA): Babies whose birth weight was greater than or equal to the 10th percentile according to the appropriate gestational age and sex-specific reference population standards [
8].
Low Birth Weight (LBW): Babies who weighed less than 2500 grams at the time of birth.
Wealth quintile: The wealth index is a measure of socioeconomic position, used in nationally representative health surveys (Demographic Health Surveys) to compare the socio-economic inequalities [
12,
13]. During the interviews with mothers, data was collected on ownership of durable assets (e.g. car, refigerator, bicycle, radio, television), housing characteristics (e.g. number of rooms, dwelling floor and roof materials, toilet facilities) and access to services (e.g. electricity supply, drinking water source). Using the scores from the first principal component analysis, a wealth index (asset index) was contructed. Based on the value of the index, individuals were sorted and established to create cut-off values for percentiles within the population. These quintiles were then ranked from bottom to top as poorest, poorer, middle, richer and richest [
14].
Ethnicity: The group within the social hierarchical system of Nepal to which the women’s family belongs [
15].
Parity: Number of times a woman has given birth after the age of viability, i.e. 22 weeks, including both live and still births [
16].
Antenatal care attendance: The number of antenatal care visits that a woman went to in order to receive antenatal care from a skilled health worker.
Obstetric complication during pregnancy: Any complication that a woman had during the pregnancy period [
17], including the following:
Antepartum hemorrhage: Excessive vaginal bleeding occurring before the onset of labor.
Hypertensive disorder during pregnancy: Classified by maternal diastolic blood pressure greater than or equal to 90 mmHg in two different recordings, at least 4 h apart.
Multiple pregnancies: When a woman was pregnant with more than one fetus.
Medical disorder: When a woman had any of the following: diabetes mellitus, severe anemia (Hb <7 gm/L), epilepsy, or other serious medical condition during pregnancy.
Obstetric complication during delivery: Any complication that a woman had during the intrapartum period [
17], including the following:
Hypertensive disorder: Classified by maternal diastolic blood pressure greater than or equal to 90 mmHg in two separate recordings.
Mal-presentation: Presentation of the fetus in any position besides vertex, i.e. with the top of the head appearing first.
Prolonged labor: When cervical dilation did not move beyond 4 cm after 8 h of regular contractions, or if cervical dilation was to the right of the alert line on the partogram.
Prolapsed cord: Characterized by the presence of the umbilical cord in the birth canal below the fetal presenting part, or at the vagina following the rupture of membranes.
Data analysis
The demographic, social and obstetric characteristics of the case and referent populations were compared using a Pearson’s chi-square test, Wilcoxon rank-sum t-test or Fisher’s exact test to assess whether there was a difference (p < 0.05) between the two groups.
For comparison of the demographic, social and obstetric characteristics of the case and referent populations, categorical variables were created. Maternal age was categorized into 5-year intervals including ≤20, 21–25, 26–30, and >30 years; maternal education was categorized as less than 6 years of education or greater than or equal to 6 years of education; ethnicity was classified into six groups as Brahmin/Chhetri (hill and terai), relatively advantaged Janajatis (Newar, Gurung, Thakali), disadvantaged Janajatis, Dalit (hill and terai), non-Dalit terai, and Muslims; wealth was classified into five population quintiles: poorest, poorer, middle, richer and richest; parity was classified into three groups: primiparous, multiparous (1–2) and multiparous (3 or more); antenatal care attendance was classified as having attended any antenatal care or none; obstetric complications during pregnancy were classified as having any or having none; obstetric complications during the intrapartum period were also classified as having any versus none; the number of babies was categorized as multiple pregnancy or not (i.e., singleton); the sex of the baby as male or female; mode of delivery as vaginal, instrumental or cesarean section; gestational age at birth was classified as term versus preterm; and size according to gestational age was classified as SGA or AGA. We also created a binary variable grouping babies who were born both preterm and SGA versus those who were neither.
Univariate logistic regression analysis was done to test the association between neonatal death and demographic, social and obstetric characteristics of the women and babies that showed differences (p < 0.01) between the case and referent populations. Three different multivariable models were created to assess the level of association of neonatal mortality with preterm and/or SGA after adjusting for maternal age, maternal educational status, antenatal care attendance, wealth status, complication during the intrapartum period, mode of delivery, parity and multiple pregnancy. The first multivariable model assessed the level of association between neonatal mortality and preterm birth compared to term; the second multivariable model assessed the level of association between neonatal mortality and being born SGA compared to babies born AGA; and, the third model assessed the level of association between neonatal mortality and being born both preterm and SGA compared to being born only preterm, or only SGA, or neither.
We used the multiple imputation method to deal with data missing at random from the case or referent populations for the demographic, social, and/or obstetric variables [
18].