Menu
Menu
Menu
Menu
Background: First childbirth in a woman’s life is one of the most important events in her life. It marks a turnaround when she might have to drop roles of career building and education, for motherhood and parenthood. The timing of the commencement of these roles affects the child bearing behavior of women as they progress in their reproductive ages. Prevalent early first childbirth in Nigeria has been reported as the main cause of high population growth and high fertility, mortality and morbidity among women, but little has been documented on the progression into first birth as well as factors affecting it in Nigeria. This paper modelled timing of first birth among women in Nigeria and determined socio-demographic and other factors affecting its timing. Methods: We hypothesized that background characteristics of a woman will influence her progression into having first birth. We developed and fitted a survival analysis model to understand the timing of first birth among women in Nigeria using a national representative 2013 NDHS data. Women with no children were right censored as of the date of the survey. The Kaplan Meier survival function was used to estimate the probabilities of first birth not occurring until certain ages of women while Cox proportional hazard regression was used to model the timing of first births at 5 % significance level. Results: About 75.7 % of the respondents had given birth in the Northern region of Nigerian compared with 63.8 % in the South. Half (50.1 %) of the first childbirth occurred within the 15-19 years age bracket and 38.1 % within 20-29 years. The overall median survival time to first birth was 20 years (North 19, South 22), 27 years among women with higher education and 18 years for those with no formal education. The adjusted hazard of first birth was higher in the Northern region of Nigeria than in the South (aHR = 1.24, 95 % CI: 1.20-1.27), and higher in rural areas than in urban areas (aHR = 1.15, 95 % CI: 1.12-1.19). Also, hazard of earlier first birth tripled among women with no education (aHR = 3.36, 95 % CI: 3.17-3.55) compared to women with higher education. The significant factors affecting age at first birth are education, place and zone of residence, age at first marriage, religion, ethnicity and use of contraceptives. Conclusions: This study showed that progression into early first birth is most affected by the education standing of women as well as age at first marriage. Delay of first childbirths as a strategy for fertility reduction and maternal health improvement can be achieved if women are empowered early in life with quality education. Stakeholders should therefore, give adequate attention to educating the girl child. Adverse socio-cultural norms of betrothing and marrying young girls should be abrogated, while health education and promotion of need to delay child bearing must be intensified especially among rural dwellers and also in Northern Nigeria.
Data from the 2013 Nigeria Demographic and Health Survey (NDHS) [1], a cross-sectional national representative data, was used for this study. The National Population Commission (Nigeria) and ICF International, United States gave access and authorized the use of the data. The survey used clusters as the primary sampling unit based on the EAs from the 2006 census frame and sampled respondents using a stratified three-stage cluster design consisting of 904 clusters, 372 in urban areas and 532 in rural areas across the six zones, 36 states, and the Federal Capital Territory, Abuja. A total of 39,902 women aged 15–49 years were identified as eligible for individual interviews, and 98 % of them were successfully interviewed. We extracted information on the women’s background characteristics, sexual and reproductive history and knowledge, source and use of contraception. The dependent variable in this study was age at first child birth while region and geographical zones of residence, education, religion, residence and ethnicity were the independent variables. Included also as independent variables are responses on; if the woman ever smoked, whether she had terminated a pregnancy or not and whether she has ever used something to prevent pregnancy. We collapsed the six zones into two regions: The North Central, North East and North West constituted the “North” while South East, South South and South West formed the “South”. We used women’s current educational attainment as a proxy for education as of the time of first child birth. This is justified, because the educational status does not change for persons who had none or primary education throughout their lifetime since primary education is mostly attained at age 12. “Ever smoked” was used as a proxy for peer pressure. We used survival analysis to model the determinants of age at first birth. Survival analysis is analysis of history of events which uses statistical procedures to deal with analysis of time duration, until one or more events of interest happen. It is usual in a follow up study, such as the current study, for some participants not to have experienced the event of interest at the end of the study or some participants were “lost to follow up” or some might have withdrawn during the study. Bias may be introduced if these categories of participants were excluded in further analysis as they could possess unique characteristics that could be useful in answering the research question. In such cases, the length of time the participants stayed in the study would be recorded as their study time and marked as “censored”. Two quantitative terms are important in survival analysis. They are the survivor function S(t) and hazard function h(t). In relation to the present study, the survivor function gives the probability that a woman “survives” longer than some specified time t without a birth, while the hazard function gives the instantaneous potential per unit time to have a first childbirth after time t, given that the individual had not had a first childbirth up to time t. Survival and hazard function are mathematically denoted by and In contrast to the survivor function (S(t)) which describes the probability of not failing before time t, hazard function (h(t)) addresses the failure rate at time t among those individuals who are alive at time t. Also two variables are compulsory in survival analysis; they are survival time and the censoring index. The “survival time” or “follow-up time”, is assumed to be a discrete random variable that takes on only positive integer. In this study, the population at risk are all women involved in the study since they are all likely to give birth one time or the other. The “survival time” for age at first childbirth is the age of the women at first birth while the survival time for those with no birth as of the time of the survey was their current age at the time of the survey. Thus their censoring index were coded “1” and “0” respectively. The usual logistic regression techniques therefore, become unsuitable in a follow up study such as the current study where the follow up time could be determined and used in explaining the event of interest. Kaplan-Meier method, developed for scenarios where survival time is measured on a continuous scale whereby only intervals containing an event contribute to the estimate, was used to compute the survival estimates. The Kaplan-Meier estimates of S(t) were obtained from equation (3) where n j is the number of subject observed at time tj and dj is the number of subject that experienced the event of interest at time tj. We applied the Cox-proportional Hazard model to the age at first birth. The model assumed that proportion of hazards are constant from time to time. In proportional hazard model, the effect of a unit increase in covariate is multiplicative with respect to hazard rate. The Cox model gives an expression for the hazard at time t for an individual with a given specification of a set of independent variables denoted by X to predict individuals’ hazard. The model assumes the relationship for one covariate where ho(t) is the baseline hazard function, xi are the covariates and β i are the coefficients. We also determined Cox regression estimates for all levels of each of the covariates. In which case, the hazard at time t for a subject in group is assumed to be The coefficients are assumed to be the same, regardless of group, but the baseline hazard can be group specific. The sign of the coefficient indicates how a covariate affects the hazard rate. The hazard ratio (HR), expressed as the exponentials of the coefficients, implies more exposure to event of interest if >1, HR < 1 means low exposure while HR = 1 has no effect on the exposure. The statistical significance of the coefficient indicates whether these changes in the expected duration will be statistically significant or not. In the stratified Cox analysis, we tested whether the proportional-hazards assumption was violated using the significance of the hazard ratios, the log likelihood tests and the Wald chi square statistics. The significant variables in the independent Cox regression were plugged into the multiple Cox regression so as to control for the effects of other variables. We fitted four models. The first model involved the two most significant independent variables at the bivariate logistic level, Model II involved the other socio-cultural characteristics of the women (place and regions of residence, ethnicity and religion). In model III we added education and age at marriage to the socio-cultural factors while model IV is the full model. The log likelihood tests and the Wald chi square statistics were used to select the best model. The data was weighted to adjust for differences in population in each state and FCT. Statistical significance was determined at p-value = 0.05. We used the Stata (version 13) statistical analysis software for all the analysis.
N/A
