A Framework for Personalised Diet-Health Behavior Change Theory

Thesis Abstract

This study addresses the persistent gap between generic dietary guidelines and individual dietary adherence by developing a framework for personalised diet-health behavior change that integrates nutrition science with behavior change theories and digital health data. The aim is to formulate a robust, testable model that explains how personalized dietary recommendations influence sustained health behaviors and outcomes across diverse populations. Specific objectives are to (i) synthesize and integrate constructs from Self-Determination Theory, Health Belief Model, and the Transtheoretical Model into a unified decision-support framework; (ii) identify individual difference factors (genetic, metabolic, psychosocial, and environmental) that moderate diet adherence and health outcomes; (iii) develop a dynamic, adaptable algorithm for tailoring dietary messages and goals using multi-omics and ecological momentary assessment (EMA) data; (iv) examine the mediating and moderating relationships among motivation, self-efficacy, perceived barriers, goal setting, and actual dietary intake; and (v) validate the framework against health outcomes such as weight, glycemic control, and lipid profiles in a real-world setting. A mixed-methods design is employed. The qualitative strand uses thematic analysis of semi-structured interviews with 40 adults across varied BMI ranges and health statuses to elicit perceived facilitators and barriers to personalized diet change, and to refine the theoretical constructs for integration. The quantitative strand adopts a prospective cohort of 600 adults aged 18–65 from three urban regions, with stratified sampling to ensure representation across sex, ethnicity, and socioeconomic status. Data collection combines validated instruments, including the Diet Adaptation and Motivation Scale, New Dietary Restraint Scale, and EMA-based dietary intake logs collected over 12 weeks, alongside biomarker data (fasting glucose, HbA1c, LDL-C, HDL-C, triglycerides) at baseline and 6 months. Anthropometric measures (weight, waist circumference) are recorded at baseline, 3 months, and 6 months. The analytic plan includes structural equation modeling to test the integrated framework’s pathways, multilevel modeling to handle repeated EMA data, and machine learning approaches (random forests and gradient boosting) to optimize personalization rules based on multi-omic and psychosocial inputs. Mediation analyses will examine whether motivation and self-efficacy transmit effects from personalized recommendations to intake change, while moderation analyses will assess the influence of genetic risk scores and environmental context on adherence. Expected findings indicate that a unified theory-driven personalization framework, incorporating motivation and perceived barriers with adaptive messaging, will significantly predict adherence to dietary recommendations and improvements in metabolic biomarkers over 6 months. It is anticipated that higher autonomous motivation, greater self-efficacy, and timely feedback will mediate diet change, while genetic and environmental moderators will refine individual response profiles. The study also expects that EMA-informed, real-time tailoring of dietary goals will produce greater adherence and more favorable health outcomes compared to static dietary advice. The contribution to knowledge lies in offering a concrete, testable framework that merges theoretical constructs with operationalized personalization algorithms, validated in a representative population, and applicable to digital health platforms. The framework advances understanding of how to translate nutrition science into actionable, individualized behavior-change interventions and provides a blueprint for scalable implementation in clinical and community settings. The study concludes that integrating established behavior theories with personalisation driven by real-time data and multi-omics context yields superior adherence and health improvements relative to non-personalized approaches. Recommendations include embedding the framework into primary care and digital health interventions, prioritizing user-centered design to enhance engagement, and pursuing longitudinal trials to assess long-term sustainability. Limitations acknowledged include potential bias in self-reported intake, rapid evolution of digital health tools, and generalizability across diverse cultural settings, with future work suggested to extend the model to pediatric and geriatric populations and to explore cost-effectiveness analyses.