Modeling Usual Nutrient Intake Distribution: A Comprehensive Comparative Study Using Hierarchical Models.

Area of Science:

• Nutritional epidemiology
• Statistical modeling
• Public health nutrition

Background:

• Accurate estimation of usual dietary intake distribution is vital for assessing population nutritional risk and informing public health policy.
• Traditional dietary assessment methods using short-term recalls have significant within-individual variability.
• The choice of statistical model impacts the accuracy of usual intake distribution estimation.

Purpose of the Study:

• To compare various statistical models for estimating usual dietary intake distribution.
• To evaluate the impact of model selection on the estimation of nutrient inadequacy prevalence.
• To identify the most suitable modeling approaches for nutritional epidemiology.

Main Methods:

• Utilized repeated 24-hour recalls (n=5800, INCA3, France) with appropriate weighting.
• Applied diverse statistical models including crossed and nested random effects models, and regression-based approaches.
• Compared descriptive statistics and nutrient inadequacy prevalence across different models.

Main Results:

• Descriptive statistics of usual intake were similar across models, but nutrient inadequacy prevalence varied significantly.
• Hierarchical models, especially nested random effects models, showed better fit and lower errors.
• Prevalence estimates differed notably between hierarchical and simpler classical models, particularly for nutrients with high variability.

Conclusions:

• Model selection critically influences the estimated prevalence of nutritional inadequacy.
• Hierarchical modeling approaches accurately account for within-individual variation, yielding more reliable estimates.
• Rigorous statistical methodology is essential for accurate nutritional risk assessment and public health interventions.