Study designs for estimating association between time-varying exposures during pregnancy and preterm birth: a simulation study Open Access

Liu, Anran (2015)

Permanent URL: https://etd.library.emory.edu/concern/etds/h128nf53w?locale=en
Published

Abstract

Preterm birth, defined as birth occurring before 37 weeks of gestation, is the leading cause of perinatal morbidity and mortality, and long-term neurological disabilities. An increasing number of studies have investigated the association between environmental exposures during pregnancy and preterm birth. However, the results are inconsistent across studies and across exposure windows because ambient air pollution levels and temperature have strong seasonal patterns and there is no standard analytic method to study time-varying exposures. The purpose of our simulation study is to examine the performances of 4 commonly used study designs, including logistic regression, case-crossover design, time-series analyses, and discrete time survival model. We first simulated the outcome of gestational age using a discrete time survival model with true relative risk for the exposures of interest equal to 1.000, 1.002, 1.004, and 1.010 for the exposure of interest. Then we used the 4 methods to estimate the risk. We compared the root mean square error (RMSE), power, coverage of 95% confidence interval, bias, and average standard error from the first 3 designs to those from discrete time survival model. We found that logistic regression is as good as the discrete time survival model when examining time invariant exposure windows; but it would overestimate the risks when examining the time varying exposure windows. The longer the exposure window, larger bias is associated with logistic regression. Case-crossover design and time-series analyses were used to examine the 1-week lag exposure and we found that case-crossover design would introduce large bias, large average standard errors, large RMSEs, small power and small 95% coverage. We also found that time-series analyses with and without stratification give similar results to the discrete time survival model.

Table of Contents

Contents

1 Introduction-------------------------------------------1

2 Methods-----------------------------------------------5

2.1 Data Generation--------------------------------------5

2.2 Model for the 4 approaches---------------------------7

2.2.1 Logistic Regression---------------------------------7

2.2.2 Discrete Time Survival Analyses---------------------8

2.2.3 Case-crossover Design-----------------------------8

2.2.4 Time-series Analyses-------------------------------9

2.3 Simulation Evaluation--------------------------------10

3 Results-----------------------------------------------11

4 Discussions-------------------------------------------14

Reference----------------------------------------------16

Appendices---------------------------------------------17

Tables--------------------------------------------------17

Sample R code------------------------------------------21

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
Department
Subfield / Discipline
Degree
Submission
Language
  • English
Research field
Keyword
Committee Chair / Thesis Advisor
Committee Members
Partnering Agencies
Last modified

Primary PDF

Supplemental Files