Leveraging machine learning and EMA data to identify factors that predict children's energy intake

The National Institute of Child Health and Human Development awarded Pennington Biomedical Research Center $316,936 on September 28, 2025, under the Child Health and Human Development Extramural Research program (CFDA 93.865) to develop and test machine learning algorithms that identify contextual factors predicting children's energy intake.

The research trains and tests a series of machine learning algorithms on ecological momentary assessment (EMA) and dietary data to evaluate their ability to identify factors that predict children's subsequent energy intake, including location, food preparation style, people present, and presence of media devices. In Aims 1 and 2, algorithms are trained and tested across all children to evaluate their predictive ability. An exploratory aim evaluates the ability to build personalized, individual-specific prediction models. The work leverages EMA and dietary data from the Mothers and Their Children's Health (MATCH) Study, which enrolled 202 mothers and their 8- to 12-year-old children. The research combines hands-on experience in data analytics, coursework, and guided mentorship to develop the applicant's skills in this cross-disciplinary line of work.

Performance occurs at Pennington Biomedical Research Center in Baton Rouge, Louisiana. The period of performance extends from September 28, 2025, through May 31, 2028. This is a project grant, a form of direct assistance funding.

Abstract

The collection of real time data through technology-based approaches, such as ecological momentary assessment (EMA), presents a rich source of information for understanding concurrent predictors of children's eating behaviors. However, advanced statistical programming methods are needed to transition from real-time assessment to real-time modification of eating behaviors. In order to complete the foundational step for this transition, the proposed study aims to train and test a series of machine learning algorithms that identify and describe the effects of concurrent contextual factors, including location, food preparation style, people present, and presence of media devices, on children's subsequent energy intake. In aims 1 and 2, algorithms will be trained and tested on data across all children to evaluate their ability to identify factors that predict children's energy intake. By training and testing individual-specific algorithms in the exploratory aim, this research will evaluate the ability to build personalized prediction models. To achieve the proposed aims, this research will leverage EMA and dietary data from the mothers and their children's health (match) study (R01 HL119255; PI: dunton), which used a longitudinal, observational, dyadic, case-crossover design, and enrolled 202 mothers and their 8 to 12 year-old children. By matching mothers' and children's EMA responses to the time- stamped 24-hour dietary recall data, the proposed research will identify predictors of children's energy intake. This proposal represents an innovative approach to studying children's eating behaviors and brings together cross-disciplinary training to pursue this line of work. With a strategic training plan that combines hands-on experience in data analytics, coursework, and guided mentorship, the applicant will test the proposed aims and develop skills and expertise critical for founding an independent research career in real-time interventions using machine learning techniques. The applicant will work closely with her primary sponsor, DR. Corby martin (an expert in mobile health interventions), her CO-sponsors DR. Genevieve dunton (an expert in EMA approaches), DR. Sujoy ghosh (an expert in computational biology and machine learning), and DR. John apolzan (an expert in clinical nutrition trials). This sponsoring team, in combination with the resources available through the pennington biomedical research center, represent the ideal conditions to accomplish the applicant's research and training goals, facilitate professional development, and increase the applicant's autonomy as an independent researcher.