Study design and settings
The data for the study came from selected slums in Dhaka (North & South) and Gazipur City Corporations, where icddr,b has been maintaining a Health and Demographic Surveillance System (HDSS) since 2015 for over 120,000 people. A birth cohort who were born in the study sites from 2016 to 2018 were used in this study. These births were followed for their survival until the neonatal period (< 29 days). The main focus of the surveillance system is to record the major demographic events such as pregnancy outcome, deaths, migration, and health indicators for maternal and child health. Female Field Workers visited each household every three months to record if any pregnancy outcome, death, migration and maternity and child health related event occurred in the surveillance site. The data contains the date of conception, date of delivery, mode of delivery, number of antenatal visits, litter size, sex of the baby, and data on the mother’s date of birth, education, and occupation; the death data contains information on the date of death and cause of death.
The study area is in proximity where people of middle- and high-income groups live and many garment factories; this is an opportunity for the slum dwellers to get easy access to work from home to these places for livelihood. The people living in urban slums are at risk groups as the environment in the slum is favourable for disease transmission with the overcrowded living conditions and limited access to public health infrastructure [21]. In these slums, most households possessed one bedroom (82%) with a mean dwelling area of 119 sq. ft. About 95% of households used pipe water for drinking, while only one-third had access to sanitary latrine flush to sewerage/septic tank. Sharing of water sources (92%), latrines (90%), and cooking places (60%) were very common in these slums.
Participants
During the study period (2016–2018), 8,421 conception had been recorded where 6,989 were recorded as live births that used as birth cohort for this study. Of these live births, 265 died during the neonatal period.
Variables
Exposures
Mother’s age at birth was calculated by subtracting mother’s date of birth from her child’s date of birth, converted in years and categorized (< 18, 18–24, and 25 or more years). Gestation age was calculated by subtracting delivery date and date of conception. Preterm birth was defined as a livebirth those born between 28 and 36 weeks of gestation and further categorized as very/moderate preterm (28 to 33 weeks) and late preterm (34 to 36 weeks) births; those born at 37 or more weeks of gestation were classified as term birth. Sex of the child (boy and girl), mother’s years of schooling (0, 1–4 and 5 or more years), mother’s working status (working and not working), litter size (singleton and multiple), antenatal care visits during pregnancy period (0, 1–3, and 4 or more). Later the mother’s occupation was converted into mother’s working status who were economically active and earn for their family.
The standard clinical and surgical definition of caesarean delivered babies were followed, whereas, for the vaginal delivery, those babies delivered through birth canal with or without instrumental and medicinal support. The vaginal delivery could either be at home or at facility. For vaginal delivered, the neonatal survivals were checked for those delivered at home (4.3% died) and those at facility (5.3% died) and found no significant difference in survival of these two groups. So, in the analyses the mode of delivery was categorised into two (vaginal and caesarean).
Outcome variable
Our main outcome, neonatal deaths were accounted to those deaths within 28 days after births (0 to 28 days of life) and was dichotomized (death or alive).
Quality of data
The urban HDSS running since 2015, the standard protocol was followed to collect data with 17 trained female Field Workers with 3 supervisors. The data was validated by the data-management team and any inconsistency reported to the supervisors are checked by the Field Workers consult available records, as well as through field visit, if needed. Field supervisors also visited 2–3% of the households for ensure the data quality [22]. Therefore, no sampling error as each household is covered, however there could be reporting bias. To minimise reporting error, female Field Workers were adequately trained to collect the data particularly for date of event (conception, pregnancy outcome, and death), however, reported conception date is usually been criticised for accuracy. For ascertaining conception, the female Field Worker asked each eligible married woman (15–49 years) during their routine data collection about whether they had been menstruating or not; if not, then they asked about their last mensuration period to ascertain the conception status. Once the conception was confirmed, the woman was followed for subsequent pregnancy outcomes. The death data contains information on the date of death and cause of death; these deaths were collected from the informed household member. The interviewer then wrote down a brief description of the cause of death; however, cause of death data has limitations as it is not collected through structured questionnaire. Subsequently, the death data were coded as the cause by a trained medical-assistant under the guidance of a physician.
Statistical analyses
Both bivariate and multivariate analyses were performed. Chi-square tests were used for a general association of nominal variables. For multivariate analyses, multiple logistic regression analysis was performed to examine the effects of caesarean delivery on neonatal mortality; odds ratios and 95% confidence intervals were calculated. For the adjusted regression model, age of mother, sex of children, mother’s education, mother’s working status, litter size, mode of delivery, number of antenatal visits, and categories of gestation age (preterm birth) were treated as independent variables. All the analyses were performed in the STATA 16.1 windows version (Stata.Corp, TX) and the whole manuscript was reported under the CONSORT guidelines (See the supplementary file).
Patient and public involvement
No patients were directly involved in setting the research question, outcome measures nor involved in the design of the study. They were not involved in interpretation of the results, however there is plan to disseminate the results among mother/women those attending health care service centre. Their written consent/assent has been taken before the data collection.