Menu
Menu
Menu
Menu
Background: Sub-Saharan African (SSA) newborns are ten times more likely to die in the first month than a neonate born in a high-income country. The objective of this study was to examine the relationship between educational attainment and neonatal mortality (NM) among women with cesarean section (CS) deliveries in SSA countries. Methods: Using data from recent demographic and health surveys from 33 countries in SSA, we applied propensity score matching to estimate the effect of education attainment on post-CS neonatal mortality using a propensity-matched cohort where being educated was defined as completing at least primary school education Results: The number of reported CS births ranged from 186 in Niger to 1695 in Kenya. The odds of neonatal mortality between uneducated and educated women ranged from as low as 2.31 in Senegal to 35.5 in Zimbabwe, with a pooled overall risk for NM from all of the countries of OR 2.54 (95% CI: 1.72–3.74) and aOR 1.7 (95% CI: 1.12–2.57). From the 17,220 respondents, we successfully matched 11,162 educated respondents with 2146 uneducated respondents. Uneducated women had a 6% risk compared to a 2.9% risk among educated women for neonatal mortality, with an overall risk of 3.4%; babies from uneducated women were twice as likely to die compared to babies from educated women, RR 2.1 (95% CI, 1.69–2.52). Conclusion: Neonates from uneducated women were twice as likely to die following CS delivery than neonates from educated women. This evidence suggests that a means of achieving Sustainable Development Goal target 3.2 to lower newborn and child mortality is ensuring that everyone has access to high-quality care with efforts made at ensuring education for all and improving socio-economic conditions.
This was a secondary data analysis of recent cross-sectional demographic and health survey (DHS) datasets from 33 low- and middle-income countries (LMIC). The DHS surveys are nationally representative household surveys conducted in LMIC, and constitute the largest worldwide effort used for obtaining health and demographic data from developing countries, with reliable quality assurance mechanisms and rigorous survey methods. In this study, we used data from 33 recent DHS surveys conducted between 2010 and 2021 in SSA available as of April 2022. A three-stage stratified cluster sampling with households as the sampling unit is used in these DHS surveys [11]. Within each sample household, all women and men meeting the eligibility criteria are interviewed. These women aged 15–49 years, and men, aged 15–59 years, were interviewed using both men’s and women’s questionnaires. The survey questionnaires are standardized and used across all countries involved in these surveys, with some modifications to suit each country’s needs. The surveys are not self-weighting. Hence, sampling weights are calculated to account for unequal selection probabilities and for non-responses. With these weights, survey findings represent target populations. Data are available for households, including for women, men, and children within these households. These surveys are conducted every five years by ICF International in collaboration with respective national institutions of the countries. In addition, financial support is provided by the US Agency for International Development (USAID). The methods and details of data collection procedures have previously been published [11]. Ethical approval for the survey was granted by the respective countries’ national research ethics committees or equivalent bodies. Informed consent was obtained and participation was voluntary. Access to de-identified datasets was granted by Measure DHS, ICF International USA. We included cesarean delivery data from the following 33 countries in SSA: Angola, Benin, Burkina Faso, Burundi, Cameroon, Chad, Comoros, Congo, Congo DR, Cote d’Ivoire, Ethiopia, Gabon, Gambia, Ghana, Guinea, Kenya, Lesotho, Liberia, Malawi, Mali, Mozambique, Namibia, Niger, Nigeria, Rwanda, Senegal, Sierra Leone, South Africa, Tanzania, Togo, Uganda, Zambia, and Zimbabwe. The outcome variable was neonatal mortality, defined as death occurring less than or equal to 28 days of life and coded as 1 and otherwise 0. The exposure variable was maternal education, categorized as no formal education (1) or educated (0), whereby the respondent had at least completed primary education. Primary education is generally the first phase of formal learning for school-age children, the duration of which is variable in different countries but six years on the average (i.e., primary one to six). Covariates included the age of the mother in completed years (≤24, 25–34, and 35–49), the mother’s place of residence as rural or urban, wealth index quintiles (poorest, poorer, middle, richer, and richest), the number of antenatal care visits during pregnancy from none, one to four visits, and five or more visits, and media access (a dichotomized variable representing any form of access to the media or no access (magazine, television, or radio)). The wealth grouping in the DHS is a proxy measure of the long-term standard of living based on ownership of certain goods and social facilities by individual households [12]. Briefly, each household had an index of economic status constructed using a principal components analysis based on several household variables: the number of rooms per house, or ownership of car, motorcycle, bicycle, fridge, television, and telephone including any kind of heating device [ibid]. The unit of analysis was any delivery within the last five years preceding the survey. Our analytical approach included descriptive statistics, a meta-analysis, and a propensity score matching analysis. Using basic descriptive statistics, we summarized variable proportions as absolute numbers and percentages, mean, and standard deviations, as applicable. Country weights were used for descriptive statistics in this study. We examined and calculated NM rates at country-level before describing and analyzing women who received CS. For each country, we generated odds ratios (OR) for the relationship between educational attainment and neonatal mortality. To calculate pooled OR across nations, we utilized the DerSimonian–Laird technique (random-effects model) [13]. Cochran’s Q test was used to assess the results’ homogeneity. The metric I2 denotes percentage variation among heterogeneous investigations [14]. Negative I2 values were adjusted to zero (no heterogeneity) to produce an I2 range of 0 to 100%, with greater values indicating increasing heterogeneity [15,16]. Since babies born to uneducated women were more likely to be at a higher risk of neonatal mortality due to individual and contextual factors, we used propensity score matching to ensure that the uneducated and educated groups in this study were comparable in terms of important covariates. A standardized difference of 10% or more was suggestive of imbalance. The propensity score approach was used to minimize potential biases in factors that might influence assignment and outcome. To construct this balanced sample, birth-specific propensity scores were estimated from a logistic regression model, which included covariates examined in this study. To create matches and evaluate the quality of matching, simple nearest-neighbor matching with one neighbor (and no replacement) was used. After matching, we examined the quality of matching and gauged comparability of the matched groups using a graph inclusive stata command called ‘pstest’. Having obtained a balanced matched sample, we conducted a pre- and post-match descriptive analysis comparing between-group differences in baseline characteristics for births between uneducated vs. educated women. Finally, we estimated the effect (average treatment effect ATE) of uneducated women on NM outcomes among CS-born babies using ATE, which measures the impact of no-education and whether CS will result in NM. We calculated the absolute difference in the probability of NM among uneducated and educated women in the propensity score-matched cohort. All analyses were completed using Stata version 14 for Windows (StataCorp, College Station, TX, USA). The null hypothesis was tested against a two-sided alternative hypothesis and the statistical significance level was set at p < 0.05.
N/A
