Unique Determinants of Relations Between Diet Quality, Socioemotional Development, and Academic Performance in U.S. Schoolchildren

• In Event: Poster Session 06
  In Poster Session: PS 06 Topic - Education, Schooling

Abstract

Objective:
Socioemotional development (SED), and academic performance (AP) have multifaceted etiologies. In analyses evaluating main associations between diet quality (DQ) and these variables among U.S. schoolchildren, we wished to identify additional, influential covariates reflecting the child’s socio-ecological context to include in adjusted models to control for confounding.

Methods:
Cross-sectional relations of DQ to SED and DQ to AP were assessed using nationally representative data from 7,755 U.S. fifth-graders with survey-weighted, multiple linear regression models. SED (dependent variable [DV]) were measured using three Self-Description Questionnaire scales: internalizing and externalizing problem behaviors [IPB, EPB] and peer relations competency. Standardized reading, math, and science test scores represented AP (DV). DQ (independent variable [IV]) was calculated from a brief food frequency questionnaire. Twelve child-level, 21 parent/household, and four school level variables were tested as potential covariates, selected given variable availability, existing literature, and theory. Model building began with inclusion of all significant covariates based on bivariate analyses, and final models were arrived at after extensive post-estimation processing and testing that minimized multicollinearity, optimized model contributions through Wald tests, and considered regression diagnostics as well as pragmatics of data collection.

Results:
After adjustment, small but significant relations (range: β=-0.006 to β=0.008 for SED and β=0.065 to β=0.109 for AP) were observed reflecting more positive SED and AP with higher diet quality. Final models for each of the SED and AP dependent variables were all different and controlled for different factors in addition to DQ (models ranged between 4-13 covariates). While some variables appeared in nearly all models, there were also unique variables specific to particular SED and AP outcomes. Also, when variables of all levels were considered altogether, important covariates that remained were not all at the child level or demographic in nature.

Conclusions:
Controlling confounding in relations between DQ and SED and DQ and AP may benefit from the consideration and inclusion of more interdisciplinary characteristics beyond sociodemographic factors.

Author

• Dorothy Chiu, UC Berkeley School of Public Health
  Presenting Author