Derivation of Dietary Patterns dependent on Diabetes status using ordinal Supervised Robust Profile Clustering: Results from Hispanic Community Health Study/Study of Latinos

2025-08-11

AI Summary

Hispanic/Latino populations in the U.S. bear a disproportionately high diabetes burden, yet dietary pattern analysis is confounded by racial and geographic heterogeneity. Method: We propose and extend ordinal supervised Robust Profile Clustering (ordinal sRPC)—a novel method for identifying diet patterns under ordinal disease states (normoglycemia, prediabetes, diagnosed diabetes)—and apply it for the first time to baseline dietary data from this population. Contribution/Results: In both simulation and empirical analyses, ordinal sRPC outperforms conventional supervised latent class models in discriminative accuracy. Key findings reveal monotonically increasing intakes of fruits, snacks, and refined grains across diabetes severity levels, suggesting potential risk associations. This work establishes an interpretable, robust framework for ordinal dietary phenotyping in heterogeneous populations and provides a methodological foundation and empirical evidence for precision nutrition interventions.

Technology Category

• Machine Learning: Learning Preferences or Rankings
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning
• Search and Optimization: Mixed Discrete/Continuous Search

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• User Modeling, Personalization and Recommendation: Explainable and interpretable methods for personalization
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

The burden of diabetes has disproportionately impacted Hispanic/Latino residents in the United States, with diet recognized as a major modifiable risk factor. Outcome-dependent dietary patterns provide insight into what foods may be associated with the increased severity and progression of diabetes. However, the ethnic and geographical heterogeneity of US Hispanic/Latino adults makes it difficult to identify and distinguish differences within their diet as risk increases. Supervised robust profile clustering (sRPC) is a flexible joint model that can identify dietary patterns associated with diabetes, while partitioning out those defined by their ethnicity and geography. However, sRPC has only been applied to binary outcomes. We extend the existing model to develop the ordinal sRPC. Using baseline dietary data (2008-2011) from the Hispanic Community Health Study/Study of Latinos, we illustrate the utility of our model to identify dietary patterns associated with the three-levels of diabetes status (i.e. normal, pre-diabetes, diabetes). Simulation studies confirmed that ordinal sRPC improved identification and characterization of these patterns compared to a standard supervised latent class model. Results indicated that participants who had greater consumption of fruits, snack foods, and refined grain breads may be more likely to be associated with an increasing severity of diabetes status.

Problem

• Identify dietary patterns linked to diabetes severity levels
• Address ethnic and geographic heterogeneity in Hispanic/Latino diets
• Extend binary sRPC model to ordinal diabetes status outcomes

Innovation

• Extends sRPC to ordinal diabetes status levels
• Identifies dietary patterns linked to diabetes progression
• Accounts for ethnic and geographic dietary heterogeneity