A "Model-on-Demand" Methodology For Energy Intake Estimation to Improve Gestational Weight Control Interventions

Abstract

Energy intake underreporting is a frequent concern in weight control interventions. In prior work, a series of estimation approaches were developed to better understand the issue of underreporting of energy intake; among these is an approach based on semi-physical identification principles that adjusts energy intake self-reports by obtaining a functional relationship for the extent of underreporting. In this paper, this global modeling approach is extended, and for comparison purposes, a local modeling approach based on the concept of Model-on-Demand (MoD) is developed. The local approach displays comparable performance, but involves reduced engineering e ort and demands less a priori information. Cross-validation is utilized to evaluate both approaches, which in practice serves as the basis for selecting parsimonious yet accurate models. The effectiveness of the enhanced global and MoD local estimation methods is evaluated with data obtained from Healthy Mom Zone, a novel gestational weight intervention study focused on the needs of obese and overweight women.

Keywords: Estimation; Model-on-Demand; Semi-physical Identification; Weight Interventions.

Figures

Fig.1.

Weight predictions from the energy balance model per (1) using a representative participant from the Healthy Mom Zone Study shows evidence of significant underreporting of energy intake.

Fig. 2.

Block diagram of the correction model.

Fig. 3.

Estimate comparison on validation data set with the two approaches for Participant A (an intervention participant) and B (a control participant) from HMZ Study. Model structure C was used for semi-physical approach, while one input case for MoD approach.