Predicting the risk of childhood overweight and obesity at 4–5 years using population-level pregnancy and early-life healthcare data

Abstract

Abstract Background Nearly a third of children in the UK are overweight, with the prevalence in the most deprived areas more than twice that in the least deprived. The aim was to develop a risk identification model for childhood overweight/obesity applied during pregnancy and early life using routinely collected population-level healthcare data. Methods A population-based anonymised linked cohort of maternal antenatal records (January 2003 to September 2013) and birth/early-life data for their children with linked body mass index (BMI) measurements at 4–5 years (n = 29,060 children) in Hampshire, UK was used. Childhood age- and sex-adjusted BMI at 4–5 years, measured between September 2007 and November 2018, using a clinical cut-off of ≥ 91st centile for overweight/obesity. Logistic regression models together with multivariable fractional polynomials were used to select model predictors and to identify transformations of continuous predictors that best predict the outcome. Results Fifteen percent of children had a BMI ≥ 91st centile. Models were developed in stages, incorporating data collected at first antenatal booking appointment, later pregnancy/birth, and early-life predictors (1 and 2 years). The area under the curve (AUC) was lowest (0.64) for the model only incorporating maternal predictors from early pregnancy and highest for the model incorporating all factors up to weight at 2 years for predicting outcome at 4–5 years (0.83). The models were well calibrated. The prediction models identify 21% (at booking) to 24% (at ~ 2 years) of children as being at high risk of overweight or obese by the age of 4–5 years (as defined by a ≥ 20% risk score). Early pregnancy predictors included maternal BMI, smoking status, maternal age, and ethnicity. Early-life predictors included birthweight, baby’s sex, and weight at 1 or 2 years of age. Conclusions Although predictive ability was lower for the early pregnancy models, maternal predictors remained consistent across the models; thus, high-risk groups could be identified at an early stage with more precise estimation as the child grows. A tool based on these models can be used to quantify clustering of risk for childhood obesity as early as the first trimester of pregnancy, and can strengthen the long-term preventive element of antenatal and early years care.

Keywords

• Pregnancy
• Early life
• Overweight
• Obesity
• Prediction