Early neonatal mortality in twin pregnancy: Findings from 60 low- and middle-income countries

Background Around the world, the incidence of multiple pregnancies reaches its peak in the Central African countries and often represents an increased risk of death for women and children because of higher rates of obstetrical complications and poor management skills in those countries. We sought to assess the association between twins and early neonatal mortality compared with singleton pregnancies. We also assessed the role of skilled birth attendant and mode of delivery on early neonatal mortality in twin pregnancies. Methods We conducted a secondary analysis of individual level data from 60 nationally-representative Demographic and Health Surveys including 521 867 singleton and 14 312 twin births. We investigated the occurrence of deaths within the first week of life in twins compared to singletons and the effect of place and attendance at birth; also, the role of caesarean sections against vaginal births was examined, globally and after countries stratification per caesarean sections rates. A multi-level logistic regression was used accounting for homogeneity within country, and homogeneity within twin pairs. Results Early neonatal mortality among twins was significantly higher when compared to singleton neonates (adjusted odds ratio (aOR) 7.6; 95% confidence interval (CI) = 7.0-8.3) in these 60 countries. Early neonatal mortality was also higher among twins than singletons when adjusting for birth weight in a subgroup analysis of those countries with data on birth weight (n = 20; less than 20% of missing values) (aOR = 2.8; 95% CI = 2.2-3.5). For countries with high rates (> 15%) of caesarean sections (CS), twins delivered vaginally in health facility had a statistically significant (aOR = 4.8; 95% CI = 2.4-9.4) increased risk of early neonatal mortality compared to twins delivered through caesarean sections. Home twin births without SBA was associated with increased mortality compared with delivering at home with SBA (aOR = 1.3; 95% CI = 1.0-1.8) and with vaginal birth in health facility (aOR = 1.7; 95% CI = 1.4-2.0). Conclusions Institutional deliveries and increased access of caesarian sections may be considered for twin pregnancies in low- and middleincome countries to decrease early adverse neonatal outcomes. We conducted an analysis based on publicly available data sets from the Demographic and Health Surveys (DHS) [14]. We included data from 60 countries which represented the latest country DHS over the last 15 years with available data on mortality and caesarean sections in single and multiple pregnancies (Table 1). List of countries and years under study DHS are cross-sectional nationally-representative household surveys that provide data for a wide range of maternal and infant health and nutrition indicators [14]. With more than 300 surveys in 90 countries, the DHS program is considered the best available way of obtaining cross-sectional information on health indicators in developing countries. In these surveys, women are interviewed about their reproductive history with survival of their offspring as well as their personal and household socioeconomic characteristics. Standard DHS surveys have large sample sizes (usually between 5 000 and 30 000 households) and typically are conducted about every 5 years, to allow comparisons over time. They are conducted by trained personnel using a standardized questionnaire and strict methods for sampling and data collection. The figures obtained from the DHS refers to births that occurred up to 5 years previous to the data of the survey [14]. We included all singleton and twin births over the five years preceding from the most recent standard country-survey within the last fifteen years. We excluded triplets and higher order multiple births as well as all neonatal deaths occurring after the first week of life. We merged country data sets into one cross-sectional database. Our main outcome was death during the first week of life (days 0-6, “early neonatal deaths”). We explored the association between the main outcome and the type of pregnancy regarding the number of foetuses (singleton vs twin pregnancies). Additionally, we examined the association between mode of delivery (caesarean section vs vaginal delivery) and early neonatal mortality separately for singleton and twins for those births taking place in health facilities. Subgroup analysis was performed after stratifying countries according to overall caesarean section rates as low (<5%), medium (5%-15%), and high (>15%) [15]. We also investigated the result of place and attendance at birth on early mortality for the twins under study. Place and attendance at birth were based on women reports and were categorized as follows: births at home without skilled birth attendance (SBA), births at home with SBA and births in health facilities. In addition, the association between early neonatal mortality in twins and low birth weight (LBW) was studied, with LBW defined as weight at birth less than 2500 g. In the logistic regression, we adjusted for the following co-variates: a wealth index [16] derived from an index of household assets, the number of antenatal visits, mother’s education, maternal age at birth of child, parity and previous birth interval categorized in “less than 18 months”, “18-23 months”, “18-35 months” and “more than 35 months”. After the exclusion of triplets and higher order of multiple pregnancies and the exclusion of late neonatal deaths, 536 179 births, 521 867 singletons and 14 312 twins, were eligible for this analysis (Figure 1). Study participants flowchart. We initially tabulated the distribution of livebirths and early neonatal deaths, singletons and twin births by country and by World Health Organization (WHO) region. WHO classifies the 194 Member States in six regions: African Region (AFR, n = 47), Region of the Americas (AMR, n = 35), South-East Asia Region (SEAR, n = 11), European Region (EUR, n = 53), Eastern Mediterranean Region (EMR, n = 21), and Western Pacific Region (WPR, n = 27). We also tabulated the distribution of caesarean sections as well as place and attendance at delivery for all singleton and twin pregnancies by country, WHO region and the total. We conducted a descriptive analysis of selected maternal and delivery characteristics in singleton and twin pregnancies. A logistic regression was performed to calculate unadjusted and adjusted odds ratio for the association between early neonatal mortality and type of pregnancy (singleton and twin). The pooled OR for early neonatal mortality among twins vs singletons was adjusted for the following confounders: the presence of at least one antenatal care visit, mode of delivery (vaginal/caesarean section), household wealth (as a proxy for socio-economic status), and other birth-related confounding variables like birth spacing [17,18]. The pooled OR for early newborn mortality was adjusted for these confounders. We used a random effect model to control for unobserved factors at primary sampling unit and country levels [19]. The model also accounted for the clustering of twins within mothers, which is often overlooked and can affect the precision of estimates [20]. In consideration of the likely confounding effect of birthweight in the association of early neonatal death with singleton/twin pregnancies we conducted a subgroup analysis using logistic regression on the pooled data set of 20 countries having less than 20% of missing data on weight at birth. These are Albania 2008/09, Armenia 2010, Azerbaijan 2006, Bolivia 2008, Congo (Brazzaville) 2011/12, Dominican Republic 2013, Gabon 2012, Guyana 2009, Honduras 2011/12, Indonesia 2012, Jordan 2012, Kyrgyz Republic 2012, Maldives 2009, Moldova 2005, Peru 2012, Philippines 2013, Sao Tome and Principe 2008/09, Swaziland 2006/07, Taijikistan 2012, Ukraine 2007. Data on the type of birth (vaginal or CS) for health facility births was available for 9 732 twins (99.0%) and for 315 635 singletons (99.0%). To explore the association between early neonatal mortality and place and attendance at birth for the twin pregnancies population, 14 096 newborns (98.5%) with data on the exposure variables remained. Throughout the analysis, p-values of <0.05 were considered significant. Statistical analysis was performed with STATA 13.1 SE (StataCorp LP, College Station TX, USA) [21]. This study used existing data obtained from ORC Macro through formal request mechanisms. No additional ethical review for the secondary analysis was required since each country and the institutional review board of ORC Macro (Calverton, MD, USA) approved the DHS data collection procedures.