Abstract
BACKGROUND: Low to moderate agreement between self-reported and directly measured anthropometry is shown in studies for adults and children. However, this issue needs further evaluation during puberty, a period marked by several transitions. We examined the correspondence of BMI status based on self-reported versus measured anthropometric data among Estonian adolescents with a specific focus on gender and age differences. METHODS: Self-reported height and weight were determined in a national representative sample of Estonian schoolchildren collected within the framework of the HBSC (health behaviour of school-aged children) survey. Self-reported and directly measured height and weight were collected from 3379 students (1071 aged 11, 1133 aged 13 and 1175 aged 15 years). The standardized HBSC questionnaire was used for collecting self-reported data; direct anthropometric measures were taken after the HBSC questionnaires were completed. The accuracy of the self-reported values by age and gender groups were determined by comparing mean differences, Bland-Altman plots with limits of agreement, Kappa statistics, and by estimation of the sensitivity and positive predictive value for detecting overweight. RESULTS: Mean self-reported weight, height and body mass index (BMI) values were significantly lower than corresponding values obtained using direct measurements. Mean differences between self-reported and directly measured weight, height and BMI were largest among 11-year-olds and smallest among students aged 15 years. Underestimation of overweight prevalence (includes obese) showed a graded trend which decreased in older age groups; the difference was greater among girls than boys in all age groups. The mean underestimation of overweight prevalence based on self-reported anthropometry was 3.6 percentage points. The positive predictive value was 72.3 % for boys and 63.4 % for girls. CONCLUSION: A distinct age-related pattern in underestimation of weight, height and prevalence of overweight was found; the bias decreased with increasing age. The mean underestimation of overweight prevalence based on self-reports was small, 3.6 %. Self-reported height and weight remain the method of choice in large surveys for practical and logistical reasons.