Résumé
OBJECTIVE: To evaluate the accuracy of prediction of birth weights from body weights of newborns till six days after birth. DESIGN: Prospective follow-up. SETTING: Four villages near Hyderabad. METHODS: Weights of 47 newborns were recorded daily from the day of birth for seven days. The birth weights were regressed on the weights of the babies taken on the 2nd day to the 7th day. Specificity and sensitivity of the predicted birth weights to arrive at the prevalence of low birth weight (LBW) were computed. RESULTS: The co-efficient of determination (R-square) for between the days measurements decreased from 95% on the second day to 86% on seventh day with an increase in the standard error of the estimate from 84 g to 154 g. Based on the "predicted birth weights", the prevalence of LBW in the community was arrived at and compared with the actual observation. The sensitivity and specificity of these regression equations was high and ranged from 0.95 to 0.85 and 0.96 to 0.93, respectively. CONCLUSIONS: In situations where the birth weight cannot be recorded, weight of the baby taken within the first week after birth may be reliably utilized to assess the "birth weight", particularly in relation to categorization as LBW. This methodology can serve as a tool to monitor various developmental programs aimed at improving birth weights.