Background In England, 1 in 3 children aged 10–11 years live with overweight or obesity, with the prevalence in the most deprived areas being more than twice as that in the least deprived. It is important to identify children at risk of becoming overweight or obese in the future to apply early prevention interventions. We aimed to develop and internally validate prediction models of childhood overweight and obesity at age 10–11 years (Year 6) using weight and height measurements at age 4–5 years (Year R) as well as antenatal and birth data in Hampshire.

Methods A population-based anonymised linked cohort of maternal antenatal and delivery records for all births taking place at University Hospital Southampton, between 2003 to 2018 and child health records including information on postnatal growth, type of feeding and childhood body mass index (BMI) up to the age of 14 years. Childhood age- and sex- adjusted BMI at 10–11 years was used to define the outcome of overweight and obesity (≥91st centile) in the models. 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. Models were developed in stages, incorporating data collected at 4–5 years and then first antenatal booking appointment, birth and early life predictors. Predictive accuracy was evaluated by assessing model discrimination and calibration.

Results Childhood BMI was available for 6566 children between 4–5 years (14.6% overweight/obese) and 10–11 years (26.1% overweight/obese) with 10.8% overweight/obese at both ages. One-fifth of normal weight children at 4–5 years became overweight or obese by 10–11 years, 30.3% of overweight children at 4–5 years were obese by 10–11 years and 68% of obese children remained obese. The area under the curve (AUC) was 0.82 for the model only incorporating BMI at 4–5 years and child gender. AUC increased to 0.84 on incorporating maternal predictors (BMI, smoking, age, educational attainment, ethnicity, parity, and employment status) as measured/reported at the booking appointment. Variables from birth and early life were not retained in the model.

Conclusion This prediction modelling can be applied at 4–5 years to identify the risk for later childhood overweight or obesity at 10–11 years, with improved prediction with the inclusion of pregnancy data. These prediction models demonstrate that routinely collected healthcare data can be used to target early preventive interventions.