Predicting contraceptive use in Bangladesh: a logistic regression analysis

Abstract

Family planning knowledge, attitude and practice surveys typically assess respondents' reproductive attitudes and intentions to use contraception. Longitudinal observation of individual respondents nevertheless shows that such questions are not strongly predictive of subsequent behaviour. This study examines 3 years' data which show that a set of such responses to questions are nevertheless substantially superior in predicting behaviour than any single indicator. Thus statistical techniques which bring into account the apparent multidimensionality of contraceptive motivation can greatly improve upon the estimation of future practice of family planning in a population. PIP:

Using a sample of 2111 rural women in Matlab thana Bangladesh, this study tests whether cross-sectional survey responses reliably predict subsequent contraceptive behavior. The Family Planning Health Services Project collected the data in 3 rounds 1977-1979. Results show that the stated desire for additional children is apparently a somewhat better predictor of subsequent use than the stated intention to use contraception, but even this variable is weakly predictive; among users in 1977 who wanted no more children, contraceptive practice declined slightly with successive rounds. The data suggest that current behavior in round 1 is a better predictor than attitudes. The increasing use of contraception from 1977-1979 was most pronounced among women with many living children. The effect of husband's educational attainment declined in importance with time, perhaps because intensive initiation and follow-up over 2 years substituted for the effect of husband's education. Women who were from land owner households and other occupational types were early to accept contraception, while those from laborer class households were relatively slow. By 1979, however, differences in use levels between occupational classes were not pronounced, although families of fishermen still lagged behind. Multivariate analysis shows that the 1977 characteristics of respondents taken together are strongly predictive of contraceptive use in 1979. Only the husband's occupation has no effect. Current use is most predictive, followed by family size, and intentions. Younger women are more likely to use contraception once the number of living children is allowed for. Thus, the activities of the Matlab Family Planning Health Services Project can be reasonably predictive of contraceptive behavior when considered as a set of variables