Modelling predictive gender- and gestation-specific weight reference centiles for preterm infants using a population-based cohort study

Watkins, W. John ORCID: https://orcid.org/0000-0003-3262-6588, Farewell, Daniel ORCID: https://orcid.org/0000-0002-8871-1653, Banerjee, Sujoy, Nasef, Hesham, James, Anitha and Chakraborty, Mallinath 2020. Modelling predictive gender- and gestation-specific weight reference centiles for preterm infants using a population-based cohort study. Scientific Reports 10 , 4302. 10.1038/s41598-020-60895-6

Preview

PDF - Published Version Available under License Creative Commons Attribution. Download (1MB) | Preview

Abstract

We aimed to model longitudinal data to create predictive growth charts for weight in preterm infants from birth till discharge, that took into account the differing growth rates post-birth when compared to in-utero growth and therefore was more representative of the data than the UK1990 reference charts. Data from birth until discharge (or death), was collected and rigorously cleaned for all infants born at <32 weeks of gestation over a 4-year period. Means and standard deviations from the UK1990 reference charts were used to compute standard deviation scores (SDS) for our cohort. 2/3rd of the data was randomly selected and used to create gestation and gender-specific predictive weight centile lines through novel application of mixed modelling methods. The remaining 1/3rd of the data was used to test model fit by comparing expected vs actual weights for the new model with those predicted by the UK1990 model. Data from 1,510 preterm infants was analysed. 1067 of these were used to produce the predictive model. Weekly SDS were significantly lower than predicted throughout hospital stay for all gestation groups when compared with UK1990 data. The test data (n = 539) fitted the new centile lines substantially better than those modelled by the UK1990 centile lines. Mixed modelling of longitudinal data produced new predictive references for weight centiles of preterm infants. A large population-based prospective study is needed to produce representative longitudinal reference growth charts using these methods.

Item Type:	Article
Date Type:	Publication
Status:	Published
Schools:	Biosciences Medicine
Publisher:	Nature Publishing Group
ISSN:	2045-2322
Date of First Compliant Deposit:	5 March 2020
Date of Acceptance:	17 February 2020
Last Modified:	25 Aug 2024 01:06
URI:	https://orca.cardiff.ac.uk/id/eprint/130155

Actions (repository staff only)

Edit Item