Cardiovascular Disease Events and Life Expectancy Lost Attributable to Machine Learning-Derived Dietary Networks:...

Area of Science:

• Nutritional epidemiology
• Computational biology
• Public health

Background:

• Artificial intelligence and machine learning (ML) offer novel methods for analyzing complex dietary patterns.
• Traditional methods often miss conditional food-food dependencies crucial for understanding health outcomes.
• ML can uncover co-consumption patterns linked to cardiovascular disease (CVD) and mortality.

Purpose of the Study:

• To apply ML network analysis to identify dietary communities in Canadian adults.
• To investigate associations between these dietary communities and CVD risk, mortality, and life expectancy (LE).

Main Methods:

• Utilized data from the Canadian Community Health Survey-Nutrition (2004, 2015).
• Employed semiparametric Gaussian copula graphical models for food group dependency analysis.
• Applied the Louvain algorithm for unsupervised community detection and Cox models for health outcome associations.

Main Results:

• Identified three dietary communities: Vegetable-Rich (VR), High-Sugary Beverage and Low Fruit (HSBLF), and High-Fat Breakfast (HFB).
• Higher VR scores correlated with significantly lower all-cause mortality and CVD risk.
• Higher HSBLF scores were associated with increased mortality, particularly in males.

Conclusions:

• ML-based network analysis successfully identified distinct dietary communities.
• These communities show differential associations with mortality and CVD, highlighting their relevance for chronic disease prevention.
• Advanced computational methods are valuable for characterizing dietary patterns and informing public health strategies.