Abstract
Objective. To validate a method of estimating non-response bias to a school-based survey based on body mass index (BMI). Methods. A population survey of Australian students in Kindergarten and Grades 2, 4, 6, 8 and 10 (N=5 407) was conducted in 2004, and height and weight measured. Teachers matched the height and weight of all students who were selected to participate, but who were not measured (non-participating students) and of half of the students whose height and weight had been measured based on comparisons with the remaining half of measured students. Among those students whose height and weight had been both measured and estimated based on matching with the most similar student, percent agreement between the BMI categories based on measured and estimated values was calculated to determine the validity of the estimated BMI value. The prevalence of overweight and obesity among participating and non-participating students was compared to determine the magnitude of non-response bias. Results. For more than 88% of classes, percent agreement between BMI categories based on estimated and measured height and weight was >60%. The differences in the prevalence of overweight and obesity between participating and non-participating students were all less than one percentage point. Conclusions. Most teachers were able to validly match students based on their height and weight, allowing estimation of and adjustment for non-response bias among non-participating students.
Acknowledgements
This study was funded by the NSW Department of Health.
Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.