Predicting childhood overweight and obesity using maternal and early life healthcare and administrative data in Wales, UK

Ziauddeen, Nida, Fraser, Simon D.S., Stannard, Seb, Berrington, Ann, Chiovoloni, Roberta, Akbari, Ashley, Owen, Rhiannon K., Paranjothy, Shantini and Alwan, Nisreen A. (2025) Predicting childhood overweight and obesity using maternal and early life healthcare and administrative data in Wales, UK. Journal of Epidemiology & Community Health, 79, A74, [P76]. (doi:10.1136/jech-2025-SSMabstracts.152).

Abstract

Background: in Wales, 11.4% of children aged 4–5 years live with obesity, with the issue being more pronounced for those living in poorer areas. Weight status tracks from childhood to adolescence to adulthood, and obesity is linked to developing multiple long-term conditions (MLTCs). There is no early identification system during pregnancy or in early life to detect those at high risk of childhood obesity. We have previously developed childhood obesity prediction models using healthcare data in Hampshire, externally validated in Bradford. Pregnancy predictors included maternal body mass index (BMI), smoking, age and ethnicity, and early life predictors included birthweight, child sex and weight at 1 or 2 years. In this study, we aimed to predict childhood obesity using healthcare and wider demographic, socioeconomic and area-level data in Wales as part of the MELD-B project examining the role of early life factors in relation to later MLTCs.

Methods: the Secure Anonymised Information Linkage (SAIL) Databank in Wales contains routinely-collected individual-level anonymised data from health records and administrative data (census and birth registrations). We linked singleton births between 15thMarch 2010 and 28th March 2012 (to enable inclusion of Census 2011 data) to mother’s pregnancy records. Age- and sex-adjusted BMI at 4–5 years was used to define outcome of overweight and obesity using the UK clinical cut-off of ≥91st centile. Backward stepwise logistic regression models with multivariable fractional polynomials were used to develop the models in two stages – first using clinical factors and then adding census and area-level factors.

Results: data was available on 53,815 children at 4–5 years. Variables retained in the first stage included maternal factors: age, BMI, smoking, parity, ethnic group, marital status, anaemia, venous thromboembolism, and child factors: birthweight, gestational age at birth, gender and breastfeeding at birth. Additional variables retained in the model on adding census and area-level factors included: unpaid carer, maternal educational attainment, type of area of residence (urban/rural) of mother during pregnancy, and Welsh Index of Multiple Deprivation of child’s residence. Discrimination (Area Under the Curve) improved from 0.66 with clinical predictors to 0.67 with wider factors.

Conclusion: clinical factors retained were largely consistent with existing literature. Additional insights were provided by the inclusion of census and area-level characteristics though the increase in model discrimination was marginal. Childhood obesity can act as a mediator on the pathway to MLTCs, and risk identification tools may be beneficial to target early prevention.

This record has no associated files available for download.

Contributors

Author: Nida Ziauddeen

Author: Simon D.S. Fraser

Author: Seb Stannard

Author: Ann Berrington

Author: Nisreen A. Alwan

Download statistics