Completeness and representativeness of body mass index in children's electronic general practice records: Linked cross-sectional study in an ethnically-diverse urban population in the United Kingdom.

OBJECTIVE To assess completeness and accuracy of children's body mass index (BMI) recorded in general practice electronic health records (GP-EHRs). METHODS We linked National Child Measurement Programme (NCMP) records from 29 839 5-year-olds and 26 660 11-year-olds attending state schools in inner London to GP-EHRs (95% linked; 49.1% girls). We estimated adjusted odds (aOR) of at least one GP-BMI record by sex, ethnic background, area-level deprivation, weight-status and long-term conditions. We examined within-child BMI differences and compared obesity prevalence from these sources. RESULTS 10.5% (2964/28330) and 26.0% (6598/25365) of 5- and 11-year-olds respectively had at least one GP-BMI record. Underweight (aOR;95% CI:1.71;1.34,2.19), obesity (1.45;1.27,1.65), South Asian background (1.55;1.38,1.74), presence of a long-term condition (8.15;7.31,9.10), and residence in deprived areas (Wald statistic 38.73; P-value<0.0001) were independently associated with at least one GP-BMI record. NCMP-BMI and GP-BMI differed by +0.45(95% Limits of Agreement -1.60,+2.51) and + 0.16(-2.86,+3.18) in 5- and 11-year-olds, respectively. The prevalence of obesity based on GP-BMI was 18.2%(16.1,20.5) and 35.9%(33.9,38.0) in 5- and 11-year-olds respectively, compared to 12.9%(12.5,13.3) and 26.9%(26.4,27.4) based on NCMP-BMI. CONCLUSION Child BMI is not comprehensively recorded in urban general practice. Linkage to school measurement records is feasible and enables assessment of health outcomes of obesity.