Oral Presentation Joint Annual Scientific Meeting of the Nutrition Society of NZ and the Nutrition Society of Australia

The dietary modelling tool: Using online constraint optimisation techniques to streamline food-based dietary prescriptions WITHDRAWN (284)

Yasmine C Probst 1 , Emma Sullivan 1 , Evan Morrison 1 , Hoa K Dam 1
  1. University of Wollongong, Wollongong, NSW, Australia

Background/Aims: Standardised background diets for participants of food-based randomised trials are vital. For this,  dietary modelling based on the percentage contributions from macronutrients or particular food group and their serving sizes is employed. Modelling accuracy is critical to assuring reliability and validity of the research and can significantly impact on the translation to practice. This study, describes the development of the dietary modelling tool and its relative validity compared to manual and semi-automated approaches to modelling.

Methods: Constraint optimisation techniques were used to determine whether non-linear constraints are best suited to the development of the automated modelling tool. Using standard practice as a reference, dietary models were produced and compared (manual and semi-automated).

Results: The Dietary Modelling Tool was produced online using constraint optimisation, incorporating estimated energy requirement calculations and based on Australian food guidance systems. Output for all tools was the number of food group servings required by a person of given energy requirements, to meet the macronutrient criteria of a hypothetical study. Percentage differences between tools revealed similar results across modelling tools. Polyunsaturated fatty acids and monounsaturated fatty acids showed the most variation (up to 200%) across tools, though not clinically significant (equating to 2 tsp difference).

Conclusion: Automated modelling tools can streamline the modelling process for food-based trials, however, appropriate constraints must be entered to achieve desired results. Relatively, similar results were found with less automation, suggesting interchangeability of tools though implementation should reflect the requirements and capacity of study.

Funding source: University of Wollongong ECR grant.