NIH VideoCast - Learning About U.S. Child Obesity from Multiple Longitudinal Surveys

Skip Navigation

Learning About U.S. Child Obesity from Multiple Longitudinal Surveys

Loading... Loading video...

158 Views

Air date:	Friday, June 8, 2012, 1:00:00 PM Time displayed is Eastern Time, Washington DC Local
Views:	Total views: 158 (1 Live, 157 On-demand)
Category:	BSSR - Behavioral and Social Sciences
Runtime:	01:05:12
Closed captioning:	Check box to display CC outside of the video Note: You can drag the captioning window around and resize it
Description:	BSSR Lecture Series A model of child weight-status evolving from birth to early adolescence is first described. This evolution is modeled from children’s having probabilities of transitioning into and out of overweight and obesity between different ages, typically 1 to 3 years apart. To estimate components of this model for the U.S. child population, we draw on four nationally-representative longitudinal surveys with observations of children through the late-2000s, and use the National Health and Nutrition Examination Surveys as the standard of comparison. The model is first used to evaluate the ages at which Hispanic and black children’s higher obesity prevalence by early adolescence emerges most strongly. We find that Hispanic child obesity emerges earlier than black child obesity. Half of the 50 greater prevalence of obesity among Hispanic than non-Hispanic white children in 8th grade has emerged already by kindergarten. In contrast, less than a quarter of the 50 greater prevalence of obesity among black than white children in 8th grade emerges by kindergarten, and as much as half is generated between the 5th and 8th grades. To understand better the sources of obesity disparities emerging by kindergarten, we estimate two sets of multivariate regressions. Each set of regressions uses two nationally-representative surveys, with one survey in common across the two sets. The first set of regressions compares the obesity of black and white children and the second set compares the obesity of children of immigrant mothers (the majority of whom are Hispanic) with children of U.S.-born mothers. Mother’s pre-pregnancy weight status emerges as an important but not decisive factor in both sets of regressions across all three surveys. In particular, immigrant mothers’ overall lower obesity than that for U.S.-born mothers is not sufficient to prevent the children of immigrants from becoming more obese than the children of natives by kindergarten. The main innovations in statistical methods that we developed to generate these results are described, including the application of Markovian demographic models, pooled-survey model-fit diagnostics, and cross-survey multiple imputation for variables present in one survey but not the other. Finally, longitudinal survey data evaluation results are described. We focus on the problems of parental reporting of the height and weight components of the standard Body Mass Index (BMI) used in tracking the time series and socio-demographic group variations in U.S. child obesity. A simulation method is used to learn that parental under-reporting of their children’s rapidly changing height induces much larger biases in the estimation of obesity prevalence than does parental misreporting of their children’s weight. Michael Rendall is a Professor of Sociology and Associate Director of the Population Research Center at the University of Maryland, College Park and adjunct senior social scientist at the RAND Corporation. Between 2007 and 2011 Dr. Rendall served as Director of the RAND Population Research Center and Director of the Postdoctoral Program in Population Studies. His research has included analyses of fertility, family structure, and international and internal migration; the development of new statistical methods for combining survey and population data, and for the simulation of cohort and population dynamics. His current research is supported by R01HD061967 (PI: Rendall) in response to a 2008 request for applications jointly supported by OBSSR, NICHD, and NHLBI to propose “Innovative Computational and Statistical Methodologies for the Design and Analysis of Multilevel Studies on Childhood Obesity.” Professor Rendall has a Ph.D. in Sociology and A.M. in Economics from Brown University.

NLM Title:	Learning about U.S. child obesity from multiple longitudinal surveys / Michael S. Rendall.
Author:	Rendall, Michael S. National Institutes of Health (U.S.)
Publisher:
Abstract:	(CIT): A model of child weight-status evolving from birth to early adolescence is first described. This evolution is modeled from children"s having probabilities of transitioning into and out of overweight and obesity between different ages, typically 1 to 3 years apart. To estimate components of this model for the U.S. child population, we draw on four nationally-representative longitudinal surveys with observations of children through the late-2000s, and use the National Health and Nutrition Examination Surveys as the standard of comparison. The model is first used to evaluate the ages at which Hispanic and black children"s higher obesity prevalence by early adolescence emerges most strongly. We find that Hispanic child obesity emerges earlier than black child obesity. Half of the 50 greater prevalence of obesity among Hispanic than non-Hispanic white children in 8th grade has emerged already by kindergarten. In contrast, less than a quarter of the 50 greater prevalence of obesity among black than white children in 8th grade emerges by kindergarten, and as much as half is generated between the 5th and 8th grades. To understand better the sources of obesity disparities emerging by kindergarten, we estimate two sets of multivariate regressions. Each set of regressions uses two nationally-representative surveys, with one survey in common across the two sets. The first set of regressions compares the obesity of black and white children and the second set compares the obesity of children of immigrant mothers (the majority of whom are Hispanic) with children of U.S.-born mothers. Mother"s pre-pregnancy weight status emerges as an important but not decisive factor in both sets of regressions across all three surveys. In particular, immigrant mothers" overall lower obesity than that for U.S.-born mothers is not sufficient to prevent the children of immigrants from becoming more obese than the children of natives by kindergarten. The main innovations in statistical methods that we developed to generate these results are described, including the application of Markovian demographic models, pooled-survey model-fit diagnostics, and cross-survey multiple imputation for variables present in one survey but not the other. Finally, longitudinal survey data evaluation results are described. We focus on the problems of parental reporting of the height and weight components of the standard Body Mass Index (BMI) used in tracking the time series and socio-demographic group variations in U.S. child obesity. A simulation method is used to learn that parental under-reporting of their children"s rapidly changing height induces much larger biases in the estimation of obesity prevalence than does parental misreporting of their children"s weight.
Subjects:	Child Longitudinal Studies Obesity--epidemiology United States--epidemiology
Publication Types:	Lecture Webcast

NLM Classification:	WD 210
NLM ID:	101589390
CIT Live ID:	11336
Permanent link:	https://videocast.nih.gov/watch=11336