Modeling of Indicators of Nutritional Status of Pregnant Women from a Multilevel Approach

A strategy for the construction of anthropometric indicators employed for the evaluation of the pregnant women's nutritional state is presented. The references of weight by pregnancy week, according to height and the body mass index at the beginning of pregnancy, are constructed by means of multilevel models for repeated measurements. The maximum likelihood estimations are compared to those obtained by the bootstrap method in order to verify the consistency of the estimated values and the fitting of the possible data imbalance effect, caused by missing observations. The obtained results evidence not only the power of the multilevel models for the construction of reference patterns, but they also permit estimate ranks of gain in weight recommended for pregnant women representative of the population under study.