Validation and refinement of an Australian customised birthweight model using routinely collected data

Background: Published customised birthweight models designed to account for individual constitutional variation have not been validated in an independent population to verify the results. Aims: To validate our previously reported customised birthweight model with additional data from the same hospital and to revise this model using a larger, more refined dataset. Methods: With the accumulation of further data, a set of coefficients was derived based on the 12-year dataset. Using shrinkage statistics, records between July 2005 and December 2008 were used to validate the model. Stepwise multiple regression using a more refined dataset of births between January 1997 and December 2008 was used to derive updated coefficients. Performance of the model was assessed using individualised birthweight ratios and the absolute difference between customised and actual birthweight. Results: Previous coefficients were validated, with shrinkage of less than 1%, indicating that the model is stable over time. An updated set of coefficients based on a dataset of 61 630 births, including refined ethnicity categories and the addition of a smoking term, is presented, which resulted in improved model statistics (primarily an improved multiple correlation coefficient of 0.51). Conclusion: The customised birthweight model appears to be stable over time in the same hospital. Initial comparisons to literature indicate that models from different geographic locations may lead to similar coefficients; but, there remains a need to formally assess this aspect of birthweight models. The updated coefficients differ slightly from those previously published and are considered superior because of refinement in the dataset.