Latent Class Analysis for PTSD Subtype Discovery Open Access

Abstract

Background: Exposure to trauma presents a public health concern worldwide. Common traumatic events include child abuse, military combat, personal assaults, and car accidents. It has been shown that more than half of the female and male population experience at least one traumatic event in their lifetime. Although more than half experience trauma, only a small percentage develop post-traumatic stress disorder (PTSD). Moreover, little is understood about the underlying neurobiological mechanism contributing to this extreme heterogeneity in PTSD.

 

Objective: To discover PTSD subtypes in our cohort study for explaining the extreme symptomatologic heterogeneity.

 

Methods:  A cluster analysis was performed on the Grady Trauma Project study. The cluster analysis involved a Dirichlet Process (DP)-based Latent Class Analysis (LCA) to discover PTSD subtypes. We performed a non-parametric Bayesian technique, DP, in conjunction with the LCA to non-empirically discover subtypes of PTSD.

 

Results: The clustering analysis revealed 4 distinct subtypes of patients with resulting groups of 23 patients, 15 patients, 9 patients and 31 patients in each of the groups, respectively. Likewise, the three clinically-defined symptom (intrusive, avoidance/numbnesss, and hypearousal) categories characterize the PTSD subtypes into clear, separable clusters – cluster 1 with moderate- to high- intrusive symptom-present patients, low- to high- avoidance/numbness symptom-present patients, low- hyperarousal patients, cluster 2 with moderate- to high- symptom-present patients, cluster 3 with high symptom-present patients, and cluster 4 with symptom-absent to low symptom-present patients.   

 

Conclusions: Our research reveals discovery of PTSD subtypes as a benchmark for explaining the inherent symptomatologic heterogeneity in PTSD symptom profiles. However, this raises important questions regarding the association between the underlying neurobiological mechanism and behavioral difference. We aim to further explain the symptomatologic heterogeneity through brain network analyses to discover important brain connectivity patterns that influence the onset of the various symptoms of PTSD. 

Table of Contents

Table of Contents

Introduction 1-5

Methods 5-6

Data description 5

Grady Trauma Project 5-6

Clinical data 6

Analysis 7-9

Stage I Analysis: Bayesian-based Cluster Modeling 7

Dirichlet Process 7-8

Latent class analysis (LCA) model 8-9

Results 9-14

Clustering results 9-14

Discussion 14-20

PTSD subgroup identification 15

Further research 15-20

Stage I Further research 15-16

Stage II Further research 16-20

Limitations and Conclusion 20-21

References 22-25

Appendix I: Modified PTSD Symptom Scale (MPSS) 26

Appendix II: Proposed two-stage approach (Future Research) 27

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	Biostatistics
Subfield / Discipline	Biostatistics - MPH & MSPH
Degree	M.S.P.H.
Submission	Master's Thesis
Language	English
Research Field	Statistics
Keyword	PTSD, clustering, Dirichlet Process, Latent Class Analysis
Committee Chair / Thesis Advisor	Dr. Suprateek Kundu, Emory University
Committee Members	Dr. Benjamin Risk, Emory University

Last modified

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	Latent Class Analysis for PTSD Subtype Discovery ()	2019-04-09 17:21:30 -0400	Press to Select an action Download

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions