Background The National Child Measurement Programme (NCMP) assesses the weight status of 4–5 and 10–11 year-old children attending state schools. We linked pseudonymised NCMP records to pseudonymised unique household identifiers derived from general practice electronic health records to investigate the likelihood of overweight or obesity in a younger household child member based on the oldest child’s weight status, sex and ethnic background.

Methods We assigned an encrypted Unique Property Reference Number using a validated address-matching algorithm to addresses of patients registered in mid-2018 with all 159 GP practices in four inner east London boroughs. We identified 36,789 children (18,714 five-year-olds [9,326 girls] and 18,075 11-year-olds [8,989 girls]) with NCMP records living in 16,434 households, representing 33.8% of 48,602 registered households with two or more children who would have been eligible for NCMP measurements in the preceding five school years. We identified the oldest child in the household with a NCMP record as the reference child.

We used binary logistic regression to estimate the likelihood (Odds Ratio [OR]; 95% confidence intervals [CI]) of at least one younger household member with NCMP-recorded overweight or obesity (BMI centile ≥91st) according to the reference child’s weight status (reference category: healthy weight), sex (male), and ethnic background (White) after adjustment for household size and composition, and borough of residence (Stata/MP 15).

Results The number of children with a NCMP record per household ranged from 2–8 (median 2, IQR 2–2) with 13,170 (80.1%) including only two. 50.0% of 16,434 reference children were girls, with 18.2%, 33.5% 18.6% from White, South Asian or Black ethnic backgrounds respectively. 5.4% (n=890) of reference children were severely obese (≥120% of the 95th centile) and 24.0% (n=3,936) had a younger household member with NCMP-recorded overweight or obesity.

After mutual adjustment, reference children who were obese (OR 3.50; 95% CI: 3.08,3.96), severely obese (4.29; 3.69,4.99) or from Black ethnic backgrounds (1.52; 1.35,1.72) were more likely, and those who were underweight (0.42; 0.28,0.64) less likely, to share a household with an overweight or obese younger child. No difference was identified in relation to the reference child’s sex (1.03; 0.95,1.12).

Conclusion Our findings suggest that, over a five-year period in a geographically-defined and ethnically-diverse population with high childhood obesity prevalence, linked NCMP and GP records can be used to identify households with a high burden of childhood obesity. Further analyses of household social, environmental and health characteristics is underway to identify potentially modifiable factors at the household level.