Page 63 - IJPS-11-4

International Journal of
Population Studies Early marriage and infant mortality in SSA

to derive women’s participation in household decision- the same model. We applied weighting factors provided
making from responses to questions about who made by the DHS program to account for the survey’s complex
decisions on visits to relatives, major purchases, husband’s nature ensuring the national representativeness of the
earnings, and the respondent’s health care and earnings. data. We applied the sample size weight for the pooled
The resultant composite scores were divided into (i) less data using the formula 1/(C [n /n ]), in which C denotes
p
c
empowerment and (ii) more empowerment. Other the number of countries involved in the analysis and n
c
selected covariates were considered based on the reviewed and n indicate the sample size for each studied country
p
literature. These variables included birth order, place of and the pooled data, respectively. Multivariable analyses
delivery, religion, wealth index, parity, place of residence, of the pooled data were performed using the sample size
and number of antenatal visits. Antenatal care visits and weight. Measures of association were presented as hazard
place of delivery were included in the model based on the ratios (HRs) with a 95% confidence interval (CI) and at
assumption that child brides are likely to have limited access a significance level of p < 0.05. The normative and largest
to these life-saving measures for their infants due to their groups were chosen as the reference categories during
relatively low status in their families and communities. We the multivariable analysis. All statistical analyses were
also controlled for the country of residence to compare performed using Stata (version 16.0; StataCorp, USA).
the estimated effects of child marriage on infant mortality
across countries. 2.4. Patient and public involvement
2.3. Statistical analysis No patients were involved in the design or dissemination
of this study.
We performed three levels of statistical analyses. At
the univariate level, the study samples were distributed 3. Results
according to the selected countries and the key 3.1. Infant mortality rate
independent variables. The bivariate analysis assessed
the distribution according to the sample characteristics. Table 1 and Figure 1 present the infant mortality rates and
Cox proportional hazard models were fitted at the the percentages of infant deaths occurring in the neonatal
multivariable level to determine the influence of child period in the selected countries. Table 1 clarifies that the
marriage on infant mortality while adjusting for the neonatal mortality rate was found to be more than 30 per
selected control variables. We employed the Cox 1000 live births in 13 of the 28 selected countries. Moreover,
proportional hazards model because it is appropriate the infant mortality rate was higher than 30 per 1000 live
for analyzing survival data and handling censored births in 27 of the 28 selected countries. At least, 45% of the
observations. Censoring occurs when the value of an total infant mortality occurred during the neonatal period
observation is not comprehensively identified. Some in all 28 selected countries. The infant mortality rate was
children in our study sample were not completely highest in Guinea (73.5 deaths/1000 live births) and lowest
exposed to mortality risk at the time of the survey. The in Ghana (25.3 deaths/1000 live births).
probability of infant death was regarded as the hazard in
using the Cox proportional hazards model. 3.2. Level of women’s empowerment
A total of seven Cox proportional hazard models were Figure 2 illustrates the women’s empowerment levels
fitted for the outcome variable. Model 1 represented an registered by the study participants in the selected countries.
unadjusted model examining the relationship between Women’s empowerment levels were low across all selected
age at marriage (as a continuous variable) and infant countries except Burundi, the Democratic Republic of the
mortality. Model 2 had a similar objective but also adjusted Congo, Rwanda, Uganda, and Zimbabwe, where women
for a proxy measure of women’s empowerment (i.e., were moderately (or intermediately) empowered.
involvement in household decision-making). Model 3 was 3.3. Age at first marriage
unadjusted with age at marriage as a categorical variable,
whereas Model 4 included women’s empowerment. Figure 3 presents the ages of the respondents in the
Model 5 incorporated maternal characteristics, selected countries at their first marriage. A high percentage
and Model 6 considered additional variables (i.e., of women married at age 18+ in Burundi, Gabon, the
characteristics of children). Model 7 represented the Democratic Republic of the Congo, Kenya, Lesotho, and
full model incorporating all the independent variables, Namibia. Meanwhile, the proportion of women who
including the country of residence. Multiple models married before age 15 or at ages 15 – 17 was 50% or more
were fitted to confirm rigorous analyses and to ensure in Burkina Faso, Guinea, Liberia, Mali, Mozambique,
that highly correlated predictors were not included in Niger, Tanzania, and Zambia.

Volume 11 Issue 4 (2025) 57 https://doi.org/10.36922/ijps.2411