The prevalence of preterm births has shown a growing trend in many countries, including developed ones. Studies in Brazil have shown that the Information System on Live Births (SINASC, in Portuguese), until 2010, underestimated the prevalence of preterm births, when compared with studies based on primary data. Starting in 2011, gestational age at birth has been calculated in SINASC according to the last menstrual period (LMP), when available. This study sought to evaluate the accuracy of the gestational age assessment using LMP, compared with two other estimates, and correlate it with birth weight. This is a population study with data from SINASC available from Brazilian Health Informatics Department between 2011 and 2015. Definitions of preterm birth, low birth weight and birth asphyxia were taken from the literature. Adequacy of birth weigh to gestational age was calculated based on Fenton and Intergrowth-21 curves. We compared weight means according to the presence or lack of preterm birth. gestational age assessment was based on LMP in 58.5% and 41.5% used another method. We found that the preterm proportion was 12% in the LMP group and 8.4% in the other method group, while low birth weight was 6.5% and 8.4%, respectively. Mean weight of preterm infants was higher in the LMP group. Use of LMP as a gestational age estimator overestimated the proportion of weight equal to or higher than 2,500g among preterm infants, which does not seem compatible with the expected distribution for this group. LMP favored "correction" of prematurity for the parameters that are comparable to those of primary data studies, though the distortions we found between gestational age and birth weigh may indicate that there are still problems with this estimator.