Time to first birth and its predictors among reproductive-age women in Ethiopia: inverse Weibull gamma shared frailty model

Background: High maternal and child death with high fertility rate have been reported in Ethiopia. Extreme age at first birth is linked with both maternal and child morbidity and mortality. However, literatures showed there were limited studies on the timing of the first birth and its predictors in the area so far. Therefore, determining the time to first birth and its predictors will help to design strategies to improve maternal and child survival. Methods: A community-based cross-sectional study was conducted among reproductive-age women in Ethiopia using the Ethiopian demographic health survey, 2016 data. Stratified two-stage cluster sampling technique was used for sampling. The Kaplan–Meier method was used to estimate time to first birth. Inverse Weibull gamma shared frailty model applied to model the data at 95% confidence interval (CI), adjusted hazard ratio (AHR) and median hazard ratio (MHR) were reported as effect size. Proportional hazard assumption checked using Schoenfeld residual test. Information Criteria were applied to select a parsimonious model. Stratified analysis performed for the interaction terms and statistical significance was declared at p value < 0.05. Results: The overall median age at first birth was found to be 20 years (IQR, 16–24 years). The independent predictors of time to first birth were: married 15–17 years (AHR = 2.33, 95% CI 2.08–2.63), secondary education level (AHR = 0.84, 95% CI 0.78–0.96), higher education level (AHR = 0.75, 95% CI 0.65–0.85), intercourse before 15 years in the married stratum (AHR = 23.81, 95% CI 22.22–25.64), intercourse 15–17 years in married stratum (AHR = 5.56, 95% CI 5.26–5.88), spousal age difference (AHR = 1.11, 95% CI 1.05–1.16),and use of contraceptives (AHR = 0.91, 95% CI 0.86–0.97). The median increase in the hazard of early childbirth in a cluster with higher early childbirth is 16% (MHR = 1.16, 95% CI 1.13–1.20) than low risk clusters adjusting for other factors. Conclusion: In this study, first birth was found to be at an early age. Early age at first marriage, at first sexual intercourse and their interaction, high spousal age difference, being Muslim were found to increase early motherhood. Conversely, living in the most urban region, secondary and higher women education were identified to delay the first birth. Investing on women education and protecting them from early marriage is required to optimize time to first birth. The contextual differences in time to first birth are an important finding which requires more study and interventions. Community based Cross sectional survey was conducted from January 18, 2016 to June 27, 2016 among reproductive-age women in Ethiopia [44]. The study was conducted in Ethiopia one of the Sub-Saharan African country where the maternal mortality ratio 412 per 100,000 live births, skilled delivery coverage 28%,the median age at first marriage 17.1 years and the median age at first sexual intercourse 16.6 years, the contraceptive prevalence among married 36%, sexually active unmarried women 58% [44]. The estimated population in 2016 was 102 million with a fertility rate of 4.46 and the second largest population in Africa. The majority (78%) of women lived in rural [44]. The study was conducted from January 18 to June 27, 2016. The study included all reproductive age-women (15–49 years) found in the selected clusters at least one night before data collection period January 18, 2016 to June 27, 2016. Taking reproductive age-women (15–49 years) of Ethiopian in place of source population, reproductive age women living in selected clusters as study population and reproductive age-women (15–49 years) found in 2016, Ethiopian demographic health survey (EDHS) enumeration areas at least one night before data collection as per Sample population [44]. Women declared infecund were excluded. Access to media Respondents were asked how often they read a newspaper, listened to the radio, or watched television. Those who had exposure to one of them at least once a week are considered being regularly exposed to media [44, 45]. Time to first birth refers to the age of a mother in years when she gave birth to the first child after puberty [1, 2, 38]. Censored Those women who did not gave birth until the 2016 EDHS data collection end date. Event/Uncensored mothers who gave first birth until 2016 EDHS data collection end date. Declared infecund married or in union women for 5 + years, had no children in the past 5 years and never used contraception [45]. Time to event/waiting time it is the time in years from puberty to age at first birth. Beginning time women at puberty (10 years from her birth date). The 2016 EDHS sample was selected using stratified two-stage cluster sampling design and census enumeration areas (EAs) were the sampling units for the first stage and the detail published [46]. A total of 18,008 households were selected for the sample, of which 17,067 were occupied. Of the occupied households, 16,650 were successfully interviewed, yielding a response rate of 98%. In the interviewed households, 16,583 eligible women were identified for individual interviews. Interviews were completed with 15,683 women, yielding a response rate of 95% [44]. After the exclusion of primary infertile (57 women) from the data, the effective sample size became 15,626 (Fig. 1). Sampling procedure of time to first birth and its predictors among reproductive age women in Ethiopia, 2016 EDHS The dependent variable in the current study is time to first birth in years when a woman gave her first childbirth until data collection period. The independent variables included, socio-demographic and reproductive health related factors (Age at first sexual intercourse, age at first marriage, Ever married, Spousal age difference); socio-economic and information related factors (respondent’s education, respondent’s occupation, Husband’s education, Husband occupation, Wealth index and Mass media exposure); Community level factors (region and residence) and Use of contraception as an immediate factor [46]. For this study secondary data from the 2016 EDHS was used. The data set downloaded from the website https://dhsprogram.com after approval letter for use had obtained from the measure DHS. Variables were extracted from the EDHS 2016 individual women’s data set using a data extraction tool. Dependent variable, time to first birth measured in years was taken from age at first birth for mothers at least gave their first birth and the current age of respondent for event censored women. For the purpose of analysis those women gave birth event coded 1 (success) and those who did not give birth 0 (censored). Independent variables age at first sexual intercourse and age at first marriage classified in to three categories; less than 15, 15–17 and 18 and above years, the highest age category taken as reference. Ever married coded as married and not married. Spousal age difference categorized as less than 5 years and 5 and above years. Respondent’s and husband education categorized into (no education, primary, secondary and higher education) and no education taken as reference. Respondent’s and husband occupation coded as not working, agriculture and non-agriculture with non-agriculture reference. Wealth index was classified as (poorest, poor, middle, richer and richest) by taking poorest as comparison group. Mass media exposure (yes/no), and use of contraception (yes/no). The regions were classified into six categories because there socio-cultural and economic similarities and geographical relations of the regions. These are northern regions (Amhara and Tigray), Oromia, Southern Nations, Nationalities and Peoples (SNNP), eastern pastoralist referring to the pastoralist dominant Afar and Somali regions, western region semi pastoralist representing Gambella and Benishangul-Gumuz, and most urban regions representing (Addis Ababa and Dire Dawa city administrations and Harari),while residence classified as urban and rural [46]. After all, questionnaires were finalized in English; they were translated into local languages (Amarigna, Tigrigna, and Oromiffa) and pretested at Bisheftu. Computer-assisted personal interview data collection system was carried out to collect data by trained EDHS data collectors and mobile version CSPro software was used for entering and capturing the data [44]. For this study the same source population used for both those who gave birth or not to make comparable. The data collectors and study participants were blind to the study hypothesis since the analysis considered later. Data extraction checklist was prepared and data extracted using Stata version 14.0. After the data were extracted, cleaned and weighted descriptive measures such as median, percentiles, graphs and frequency tables were used to characterize the study population. We estimated time to first birth using the Kaplan–Meier (K–M) method and compared across categorical predictor variables using log rank test. Schoenfeld residual test was applied to check the proportional hazard assumption. Since our data were correlated at cluster level, shared frailty model were modeled by taking enumeration areas/clusters as a random effect for predictors of time to first birth among reproductive-age women in Ethiopia assuming time to first birth to be constant in the same clusters. The efficient model was selected by the smallest AIC value. Model adequacy was checked using Akaike Information Criteria (AIC), Cox-Snell residuals and R2 type statistic. Stratified analysis and chi-square test were applied for interaction terms. Finally adjusted hazard ratio (AHR) and adjusted time ratio (ATR) as a measure of effect size reported at 5% significant level and p value < 0.05. Stata 14.0/SE software for data extraction, cleaning and analysis was used. The associations within group members are measured by Kendall’s, which is given by where τ ɛ [0, 1]. For Gamma frailty distribution (θ > 0) where τ ɛ (0, 1/2). For Inverse Gaussian frailty distribution (θ > 0). The median hazard ratio (MHR) was used to compare between high and low risk clusters of time to first childbirth. MHR = e2θ∗Φ-1∗34 where θ = variance of frailty, Φ−1 = inverse normal distributions.