PROJECT SUMMARY
A decade ago, the U.S. Dietary Guidelines Advisory Committee recommended dietary pattern approaches
to examine relationships between diet and health outcomes. Meanwhile, longitudinal dietary data have
become increasingly available. However, methods are underdeveloped for characterizing dynamic diet-quality
variations and remain rudimentary for validating longitudinal diet-quality patterns, thus, leading to unclear
evidence for assessing diet-health relationships and formulating dietary guidelines. A noticeable gap exists
between the dietary pattern literature and the fast-growing statistical learning field with explosive growth of
artificial intelligence algorithms. We propose to develop “iPAT:Intelligent Diet Quality Pattern Analysis for
Harmonized MA-National Trials”. iPAT will leverage original and newly harmonized dietary data generated
from 7 studies funded by NIDDK, NHLBI, and NIMH: 4 longitudinal randomized controlled trials (RCT) in
Massachusetts (MA), and 3 large-scale longitudinal multi-site national studies, an RCT and one observational
study (OS) from the Women’s Health Initiative (WHI), and one OS from the Coronary Artery Risk Development
in Young Adults (CARDIA) study. We aim to harness over 20 newly-harmonized dietary datasets from these
highly-comparable longitudinal studies that span up to 35 years and cross 50 clinical and health community
centers to: 1) innovate by adapting our new visualization-aided trajectory pattern-recognition and validation
algorithm to an intelligent and streamlined pattern analysis tool (iPAT) for longitudinal dietary data; 2) enable
a new multi-view and comprehensive understanding of diet-quality trajectory patterns for multiple chronic
disease outcomes that may not be discoverable from individual studies at different levels of granularity; and 3)
create an accessible and expandable harmonized dietary database and open-access iPAT tool for diet-related
studies. Our harmonized-data-driven approach will increase the likelihood of successfully addressing complex
and subtle questions with large-scale dietary data, including but not limited to the cultural, age, gender and
geographic variation in diet quality patterns and how diet quality may vary with context and time. Our iPAT
approach will be built upon PI Fang’s behavioral trajectory pattern-recognition method which has been
validated and replicated in five NIDA/NCI/NHLBI-funded longitudinal OS and RCTs. Developing this evidence-
based iPAT tool will contribute to the infrastructure for diet-related studies, advance pattern-recognition methods,
help scientific communities and the public to compare individual dietary behavior with local and national diet-
quality patterns and associated dietary health risks. Our work will also help grow more valid evidence for dietary
guidelines. More broadly, this iPAT project will contribute to creating a platform that supports harmonized data
management, near-real-time pattern analyses and adaptive interventions.