The determinants of the duration of contraceptive use in China: a multilevel multinomial discrete-hazards modeling approach.

Often in demography, individuals may change state over time for a variety of reasons. Competing-risks hazards models have been developed to model such situations. This paper describes the extension of the discrete-time competing-risks hazards model to a multilevel framework that allows for data at different levels of aggregation. The model is illustrated with data from the 1988 Chinese National Survey of Fertility and Contraceptive Prevalence, which collected complete contraceptive histories. Women may stop using a method of contraception for a number of reasons; this paper describes how one can control for correlations between the outcomes of repeated spells of contraceptive use.