Predicting Low Back Pain Following Lumbar Interbody Fusion: A Comparison of Multivariate Model Selection Criteria 公开

Abstract

Developing an adequate predictive regression model and indentifying appropriate functional and interaction relationships between its predictors is perhaps one of the most challenging tasks in regression analysis. Specifically, linear models applied to a high-dimensional dataset, such as those commonly found in clinical settings, are usually over-parameterized, unclear for interpretation, and have high variances for estimated parameters. Traditionally used hypothesis testing frameworks for selection criteria have proven to be inconclusive and unreliable, especially in medical settings. The goal of this thesis was to overcome these challenges by integrating clinical experience with statistical theory and develop empirical predictive models using different variable selection criteria.

This thesis also contributes to research in health-related quality of life (HRQOL) outcomes for the assessment of lumbar interbody fusion (IBF) efficacy. Research regarding statistical methods in this field is preliminary and has been limited to univariate analysis. In this thesis, pre-operative HRQOL measures and patient predictors associated with 3-month post-operative low back pain (LBP) improvement were determined and multivariate linear regression models were selected for the data from the Georgia Spine Patient Outcomes Registry.

The results were very sensitive to the selection criteria used, as well as noise variables included in the full dataset. The findings of this thesis show that further work in the analysis of lumbar IBF HRQOL outcomes needs to be done to transform the knowledge and teaching base from theory and person experience to one of statistical evidence.

Table of Contents

1 Introduction

1.1 Low Back Pain ........................................................................ 1
1.2 Lumbar Interbody Fusion ......................................................... 3
1.3 Evaluation of Efficacy .............................................................. 3
1.4 Study Motivation & Goals.......................................................... 8
2 Statistical Methods
2.1 Multivariate Linear Regression .................................................. 11
2.2 Model Selection ...................................................................... 12
2.3 Stepwise Selection .................................................................. 13
2.4 Mallows' Cp Criterion ............................................................... 15
2.5 Model Validation ...................................................................... 18

3 Study Materials
3.1 Data Source ............................................................................ 20
3.2 Predictor Variables ................................................................... 21
3.3 Outcome Variables ................................................................... 21
3.4 Sample Selection ..................................................................... 23

4 Analysis & Results

4.1 Univariate Analysis ................................................................... 24
4.2 Stepwise Selection .................................................................... 26
4.3 Mallows' Cp Criterion ................................................................. 28
4.4 Model Validation ....................................................................... 30
5 Discussion ................................................................................ 31

References ................................................................................... 33

A HRQOL Instruments
A.1 Numeric Rating Scale (NRS) ........................................................ 40
A.2 Oswestry Disability Index (ODI) ................................................... 40
A.3 SF-36v2^TM (SF-36) ..................................................................... 41

B Proofs
B.1 Proof of Equation (2.7) ................................................................ 42
B.2 Lemma B.2 ................................................................................ 43
B.3 Proof of Equation (2.8) ................................................................ 44

C Supplemental Materials
C.1 Results of Univariate Analysis ....................................................... 46

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 and Bioinformatics
Subfield / Discipline	Biostatistics - MPH & MSPH
Degree	MSPH
Submission	Master's Thesis
Language	English
Research Field	Statistics Health Sciences, Public Health Health Sciences, Medicine and Surgery
关键词	low back pain multivariate predictive model HRQOL outcomes model selection criteria Mallows' Cp lumbar interbody fusion
Committee Chair / Thesis Advisor	Weiss, Paul, Emory University
Committee Members	Cotsonis, George A, Emory University
Partnering Agencies	Hospital or other health care provider

最新修改

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	Predicting Low Back Pain Following Lumbar Interbody Fusion: A Comparison of Multivariate Model Selection Criteria ()	2018-08-28 11:15:30 -0400	Press to Select an action Download

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions