Determinants of linear growth retardation in Brazilian preschool children: a multilevel approach

This study focuses on the determinants of linear growth deficit in under-five children using multilevel modeling. The sample included 3,746 preschoolers from 15 Brazilian municipalities (ten from the State of Bahia and five from São Paulo), with a three-tier analysis (municipalities, households, and children). Municipal data were obtained from the 1991 National Census. Individual and household data were collected from 1999 to 2001 using structured questionnaires and measuring children's weight and height. Analysis used the MLwiN software. In the final multilevel model, low HDI and prenatal care at the municipal level, poor environmental and economic conditions and maternal schooling at the household level, and low birth weight, incomplete immunization, prior history of malnutrition, and no report of breastfeeding at the individual level were strongly associated with linear growth deficit. Total variability of deficit explained by the model was 27.4%, of which 51.3%, 33%, and 15.7% were explained, respectively, by individual, household, and municipal variables.