Distributions of Postpartum Amenorrhea: Some New Evidence

Breastfeeding and the amenorrhea related to it are important determinants of fertility levels. Breastfeeding practices vary widely across and within countries, and there is evidence of change in the extent of breastfeeding in several areas. It is therefore important in demographic research on fertility to measure precisely the length of amenorrhea. Because of the large costs involved in data collection, most data on amenorrhea have been obtained from retrospective studies. In such studies when women have been asked when they resumed menstruation after the birth of their last child, the resulting data have usually been of poor quality, commonly with gross heaping on multiples of six months. Models of amenorrhea have been developed to adjust available data and for use in fertility research. Thus far there have been three major efforts to describe lengths of amenorrhea: (1) Barrett's (1969) modified Pascal distribution, (2) Lesthaeghe and Page's (1980) logit model, and (3) Potter and Kobrin's (1981) mixed geometric negative binomial model. Realistic models are needed to adjust the distorted data prevalent in most studies. A major problem, however, in the evaluation of the available models has been the lack of data with which to study them. Data from large prospective studies of breastfeeding women are needed both to examine the true distribution of amenorrhea and to evaluate the ability of models to produce realistic distributions. The aim of this paper is to examine distributions of the length of postpartum amenorrhea in large prospective studies of breastfeeding women for these two purposes.