Birthweight by gestational age in preterm babies according to a Gaussian mixture model

Objective: To provide a statistically sound criterion for identifying implausibly large birthweights for gestational age. Design: Review of ISTAT 1990-1994 national newborn records. Setting: Italy Population: Forty-two thousand and twenty-nine single first and second liveborn preterm babies. Methods: Two-component Gaussian mixture models are used to describe the birthweight distributions stratified by gestational age. Implausibly large babies are identified through model-based probabilistic clustering. Main outcome measures: Gestational age misclassification and weight-for-gestational age centile curves Results: Gestational age appears under-estimated by about six weeks in 12.3% of the cases. Large babies are equally present in males and females, but are more frequent in second-borns than in first-borns, even when parity-specific models are fitted. Conclusions: The approach allows for a quantification of the gestational age under-estimate error and for data correction through model-based clustering. Correct birthweight distributions and growth curves are also provided.