Sociodemographic Factors Influenced Response to the 2015 National Nutrition Survey on Preschool Children: Results From Linkage With the Comprehensive Survey of Living Conditions

• Published in: Journal of epidemiology Vol. 30; no. 2; pp. 74 - 83
• Main Authors: Okubo, Hitomi, Yokoyama, Tetsuji
• Format: Journal Article
• Language: English
• Published: Japan Japan Epidemiological Association 05.02.2020
• Subjects: Age composition; Bias; Body weight; Children; Estimates; Households; Human nutrition; japan; Living conditions; multiple imputation; national surveys; Nutrition; Nutrition surveys; Original; Overweight; Polls & surveys; Preschool children; Regression analysis; Response rates; sociodemographic factors; Sociodemographics; sociodemographic factors; response rates; multiple imputation; national surveys; Japan
• ISSN: 0917-5040; 1349-9092; 1349-9092
• DOI: 10.2188/jea.JE20180176

The National Nutrition Survey on Preschool Children, Japan (NNSPC) provides fundamental information for policy making for child nutrition. However, the response rate and background characteristics of subjects are unclear. Here, we examined response rate and sociodemographic factors related with response to the survey and evaluated the magnitude of bias due to selective response in the survey estimates of the NNSPC. This study was based on two national surveys conducted in 2015: the NNSPC and the Comprehensive Survey of Living Conditions (CSLC). Because potential survey participants of the NNSPC were children aged <6 years and their households that answered the CSLC, we examined response rates and respondent characteristics by linking the data of the NNSPC and CSLC. Multiple logistic regression analysis was used to identify sociodemographic factors associated with response. Potential bias caused by non-response in the survey estimates was examined after considering missingness through multiple imputation. Among the 5,343 children who participated in the CSLC, 3,426 children responded to the NNSPC (response rate = 64.1%). Variables associated with response were living in a smaller city, a large number of children, three-generation family structure, older maternal age, and a non-working mother. The prevalence of overweight was underestimated by 20%, but the bias for almost all variables examined was small. Response to the survey varied by sociodemographic characteristics. Some biases, mostly small, were seen in survey estimates of the 2015 NNSPC. Further insight into the effect of selective response is important to assess associations between variables more precisely.