Variations in data collection can influence outcome measures of BMI measuring programmes.

BACKGROUND: The World Health Organization (WHO) promotes the surveillance of obesity prevalence through standardized and harmonized surveillance systems. However, variations in data collection between countries, or between coordinating regions in countries can affect outcome measures. METHODS: Multilevel analysis of 2007/08 National Child Measurement Programme (NCMP) data estimating the relationship between BMI z-score and data collection variations within coordinating regions whilst adjusting for individual-level and school-level variables. The 2007/08 NCMP collected height and weight measurements for 478,381 Reception year pupils (4-5-year-olds) and 496,297 year 6 pupils (10-11-year-olds) from 17,279 primary schools in 152 data collection coordinating regions in England. RESULTS: Data collection variables accounted for 29.7% of the regional variation in BMI z-score for Reception year pupils but only 5.3% for the older Year 6 pupils. Digit preference in the rounding of weight measurements had the greatest impact of all the data collection variables, explaining 26.4% of the regional variation in BMI z-score for Reception year pupils and 4.0% for Year 6 pupils. CONCLUSIONS: Although variations in data collection may have a small effect on individual measurements their impact can be magnified when scaled up to regional or national figures. All measurement programmes must regularly identify and minimize variations in data collection to improve accuracy of outcome measures. These factors include those identified within this study: participation and opt out rates, the time in the year the measurements are taken and the recording of measurements to the correct decimal place.