[Discontinuation of contraceptive methods in Northeast Brazil, 1986-1991]

This paper examines the dynamics of contraceptive discontinuation in the Northeast Region of Brazil. A hierarchical discrete-time competing risks hazards model is used to derive cumulative rates of discontinuation in the first 12 months of use, by causes, according to a set of socio-demographic variables. The hierarchical approach is used not only to obtain corrected parameter estimates and standard errors but also to have a better understanding of the contraceptive discontinuation process. Despite the large number of socio-demographic variables in the model, the study shows that there is a substantial amount of unexplained variation at the level of individual women.