| Abstract: | This paper extends the use of the multistate proportional hazards model proposed by R. Kay for transitions, reverse transitions, and repeated transitions. A simple method of testing the equality of vectors of parameters for transitions and repeated transitions is also shown in addition to estimates for the underlying cumulative hazards for different types of transition. The multistate survival models applied to contraceptive use data collected by the International Center for Diarrheal Disease Research, Bangladesh, for the period January 1978 to December 1982 provide useful findings relating to interstate transitions. The author stresses the need to not treat separately states of contraceptive use given their link to the acceptance of the first method, transitions from one method to another, first method discontinuation, periods of non-use, the use of subsequent methods, loss to follow-up, the use of irreversible methods, and subsequent periods of non-use. Religion is an important indicator for the acceptance of an irreversible method as a first or subsequent method; compared to women of other religions, Muslims were less likely to accept an irreversible method as a first or subsequent method. Religion is not, however, associated with the acceptance or subsequent use of a reversible method or with discontinuation of a reversible method. Higher age in the study area is associated with both lower acceptance and longer continuation of a method. Women with a larger number of living sons tend to accept their first method and subsequent methods at an higher rate than women with fewer or no sons. As for educational status, women with higher education proved to be serious users of contraception, although women with less or no education tend to accept an irreversible method at an higher rate than the women with an higher level of education. Further, the desire for more children is a strong predictor for non-use of a method even if contraception is used to space births. The score test suggested in the paper for testing the equality of parameters in models 1-4 reveals that there is no significant difference in the parameters of the models. This paper shows that the factors which affect the acceptance of any reversible method of contraception are different from those for an irreversible method. Moreover, the factors which affect a transition differ from those for a reverse transition in the case of adopting reversible methods. |
