Estimating Steatosis Prevalence in Overweight and Obese Children: Comparison of Bayesian Small Area and Direct Methods

Authors

1 Associate Professor in Biostatistics, Inpatient’s Safety Research Center, Department of Biostatistics and Epidemiology, Urmia University of Medical Sciences, Urmia, Iran.

2 Associate Professor of Pediatric Gastroenterology, Maternal and Childhood Obesity Research Center, Urmia University of Medical Sciences, Urmia, Iran.

3 Associate Professor in Biostatistics, Department of Health and Preventive Medicine, Urmia University of Medical Sciences, Urmia, Iran.

4 Associate Professor in Health Education and Promotion, Solid Tumor Research Center, Department of Health and Preventive Medicine, Urmia University of Medical Sciences, Urmia, Iran.

Abstract

Background
Often, there is no access to sufficient sample size to estimate the prevalence using the method of direct estimator in all areas. The aim of this study was to compare small area’s Bayesian method and direct method in estimating the prevalence of steatosis in obese and overweight children.
Materials and Methods: In this cross-sectional study, was conducted on 150 overweight and obese children aged 2 to 15 years referred to the Children's digestive clinic of Urmia University of Medical Sciences- Iran, in 2013. After Body mass index (BMI) calculation, children with overweight and obese were assessed in terms of primary tests of obesity screening. Then children with steatosis confirmed by abdominal Ultrasonography, were referred to the laboratory for doing further tests. Steatosis prevalence was estimated by direct and Bayesian method and their efficiency were evaluated using mean-square error Jackknife method. The study data was analyzed using the open BUGS3.1.2 and R2.15.2 software.
Results: The findings indicated that estimation of steatosis prevalence in children using Bayesian and direct methods were between 0.3098 to 0.493, and 0.355 to 0.560 respectively, in Health Districts; 0.3098 to 0.502, and 0.355 to 0.550 in Education Districts; 0.321 to 0.582, and 0.357 to 0.615 in age groups; 0.313 to 0.429, and 0.383 to 0.536 in sex groups. In general, according to the results, mean-square error of Bayesian estimation was smaller than direct estimation (P<0.05).
Conclusion
The study suggests that estimation of prevalence using Bayesian estimation method via the logistic mixed model was more efficient and better than direct estimation method.

Keywords