Impact of Clustering Algorithm Choice on Dietary Pattern Identification in Periconceptional Nutrition Research Restricted; Files Only

Abstract

Objective: This study compared dietary patterns derived using multiple clustering algorithms and evaluated differences in preeclampsia risk across the patterns identified by these algorithms.

Methods: Data was analyzed from 8,259 participants enrolled in the Nulliparous Pregnancy Outcomes Study: Monitoring Mothers-to-Be (nuMoM2b). Periconceptional diet was assessed using a modified Block FFQ and summarized into 38 food groups. Dietary patterns were identified using K-means, K-medoids, Hierarchical clustering, and Gaussian Mixture Modelling, with a two-cluster solution selected across methods. Cluster membership was then used as the exposure in Augmented Inverse Probability Weighting analyses performed separately for each clustering algorithm to estimate the marginally standardized risk difference for preeclampsia, while adjusting for a wide range of confounders.

Results: Dietary patterns differed in their defining food groups across algorithms, with only a small group of foods consistently distinguishing clusters. Agreement in cluster assignment was also low across algorithms (ARI range: -0.005 to 0.386), indicating inconsistent grouping of participants. Adjusted risk differences for preeclampsia comparing cluster 2 vs cluster 1 varied by algorithm, with K-means indicating lower risk (-1.73; 95% CI: -3.96, 0.50), Hierarchical clustering indicating higher risk (1.67; 95% CI: -0.44, 3.78), and K-medoids (-0.46; 95% CI: -2.26, 1.35) and GMM (0.93; 95% CI: -0.44, 2.31) showing minimal differences. 

Conclusions: Dietary pattern characterization and their associations with preeclampsia were sensitive to clustering algorithm choice, leading to inconsistent findings across algorithms. Careful consideration is therefore needed when applying clustering algorithms in nutritional epidemiology to derive dietary patterns.

Table of Contents

INTRODUCTION ... 1

MATERIALS AND METHODS ... 2

RESULTS ... 7

DISCUSSION ... 13

CONCLUSIONS ... 16

REFERENCES ... 16

About this Master's Thesis

Rights statement

• Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.

School	Rollins School of Public Health
Department	Epidemiology
Subfield / Discipline	Epidemiology - MPH & MSPH
Degree	M.P.H.
Submission	Master's Thesis
Language	English
Research Field	Health Sciences, Epidemiology
Keyword	Clustering algorithms Nutritional epidemiology
Committee Chair / Thesis Advisor	Ashley Naimi, PhD, Emory University

Last modified

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	File download under embargo until 21 May 2028	2026-04-23 15:41:50 -0400	File download under embargo until 21 May 2028

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions