Modeling Age at first pregnancy among teenage Women in Nigeria: A Survival Analysis Approach

Abstract

Cox PH model has been the most popular model known in literature for survival time data when the baseline hazard function is unspecified. Discrete time model is used when only the interval in which the event of failure occurs is known or the event itself occurs in discrete time scale. In addition to the observed factors collected on individuals, there often exist unobservable covariates in the data at cluster level which if not accounted for during analysis, may lead to loss of information and biased estimates. This paper uses Cox model and discrete-time logit model to investigate the effect of some covariates (risk factors) on the hazard of age at first pregnancy for teenagers. The analysis explicitly account for the unobserved heterogeneity (frailty) at the level of ethnicity which was included as a random effect in the models. Comparison was made between the models with and without random effects to know which one performs best. Teenage pregnancies were viewed within the broader socio-economic and socio-cultural environment in which the adolescents operate. The results of the analysis reveal that the teenage pregnancy depend on geopolitical zone, location of residence, level of educational attainment, marital status, religion, circumcision, sex in exchange of gift, the age of first sexual initiation, HIV/AIDS status and the use of contraceptives.