Skip to main content
NUTRITION 2026 Main banner

(PTFS01-02-26) Understanding Engagement With an AI-Based Nutrition Education Chatbot in a Food Is Medicine Program: A Mixed-Methods Evaluation

Saturday, July 25, 2026
10:04 AM - 10:08 AM ET
Location: Prince George's Hall B-E (The Hub) - Poster Theater 1 / Poster Board Number: PT1-2

    Presenting Author(s)

  • Lorena Macias-Navarro, M.Sc.

    Doctoral candidate
    UTHealth Houston School of Public Health
    Magnolia, Texas, United States

    Disclosure(s): No relevant financial relationship(s) with ineligible companies to disclose.

    • Co-Author(s)

  • NR

    Nalini Ranjit, PhD

    UTHealth Houston School of Public Health
    Austin, Texas, United States

  • NH

    Natalia I. Heredia

    UTHealth Houston School of Public Health, United States

  • SM

    Sahiti Myneni

    UTHealth Houston School of Public Health, United States

  • SS

    Shreela V. Sharma

    UTHealth Houston School of Public Health, United States

Disclosure(s):
Lorena Macias-Navarro, M.Sc.: No relevant financial relationship(s) with ineligible companies to disclose.
Objectives: This study evaluates a bilingual AI-based chatbot providing personalized nutrition education to high-risk pregnant individuals within a Food is Medicine randomized controlled trial. The purpose is to quantitatively assess engagement patterns and qualitatively explore barriers and motivators influencing user experiences.

Methods: This study uses a two-phase sequential explanatory mixed-methods design. The quantitative phase will involve a secondary analysis of engagement data (n=106) to model chatbot use and intensity using binomial logistic and zero-inflated negative binomial (ZINB) regressions. Predictors include baseline cooking skills and frequency. The qualitative phase will consist of semi-structured telephone interviews (n≈20) stratified by usage status to explore contextual factors.

Results: Preliminary quantitative analysis (n=106) indicates that 31% (n=33) of participants engaged with the chatbot. While no statistically significant differences in race or ethnicity were observed between users and non-users, chatbot users were significantly older on average (Mean=34.55, SD=5.33) compared to non-users (Mean=31.78, SD=6.37; p=0.032). Additionally, educational background significantly differed between the groups (p=0.034). Ongoing qualitative research aims to identify facilitators and barriers to digital adoption to help contextualize these initial demographic differences.

Conclusions: Preliminary findings suggest that age and educational attainment may influence engagement with digital nutrition education among high-risk pregnant individuals. Final integrated findings could potentially inform the equitable design and scalability of nutrition education and technology-enabled support within clinical and community-based FIM settings.

Funding Sources: American Heart Association and Walmart Foundation