Skip to Main Content
Frequently Asked Questions
Submit an ETD
Global Search Box
Need Help?
Keyword Search
Participating Institutions
Advanced Search
School Logo
Files
File List
SaraConroyFinal.pdf (1.07 MB)
ETD Abstract Container
Abstract Header
A Novel Approach for Modeling Time to Event Data in Maternal Child Health
Author Info
Conroy, Sara A
ORCID® Identifier
http://orcid.org/0000-0001-9155-9456
Permalink:
http://rave.ohiolink.edu/etdc/view?acc_num=osu1574348181518311
Abstract Details
Year and Degree
2019, Doctor of Philosophy, Ohio State University, Public Health.
Abstract
Background
: One of the most important and commonly studied factors in maternal child health is gestational age at delivery. Infants born too soon are more likely to experience negative health outcomes such as admittance to the neonatal intensive care unit, cerebral palsy, developmental delays, and death. Even with decades of research, we still do not understand why some infants are born too soon or how to optimize the length of the gestational period. We propose looking at the data from a new perspective through a novel application of an established statistical method, threshold regression. Our overarching hypothesis is that application of threshold regression will have utility in advancing our understanding of the etiology of gestational age at delivery. We will evaluate this hypothesis through the following specific aims: 1) To examine the utility of threshold regression to assess the association of a well established risk factor on time to delivery when using gestational age for the time scale; 2) To investigate how threshold regression parameter estimates are impacted in situations with measurement error in the time scale, in this case gestational age.
Methods
: For Aim 1, we assessed the impact of smoking on gestational age at delivery from a large prospective cohort of pregnant women. We compared the overall conclusions from threshold regression and two common statistical methods, logistic regression and Cox proportional hazards regression. For Aim 2, we investigated how measurement error in estimated gestational age may impact the parameter estimates of threshold regression by using four different estimates of gestational age at delivery.
Results
: Logistic regression provided evidence that the odds of birth before 37 weeks completed gestation were higher for smokers compared to non-smokers. Similarly, there was evidence from the Cox proportional hazards regression that the spontaneous live birth rate for was higher for smokers compared to non-smokers until about 37 weeks completed gestation. However, from 37 weeks until 44 weeks gestation there was no evidence of a difference in the birth rate. Threshold regression provided evidence that the latent fetal growth/development process has a greater rate of change for smokers compared to non-smokers, which corresponded to an estimated gestational age at delivery about two days sooner for smokers compared to non-smokers. Additionally, in a low-risk, prospective cohort of pregnant women, potential measurement error from gestational age estimation methods did not have a large impact on threshold regression estimates.
Conclusions
: Current research typically models the odds ratio or the hazard ratio when studying factors associated with gestational age at delivery. Threshold regression offers a new perspective for studying time to delivery by focusing on the latent fetal growth/development process. By learning more about factors related to the latent fetal growth/development process separately from the birth threshold, we can better understand the impact of interventions on the different parts of the process. Threshold regression is a flexible survival model and may help elucidate effects in other processes; for example, time to maternal or infant death after delivery or time to pregnancy loss. Increased use of this model may help facilitate deeper understanding of the fetal growth/development process and identify new interventions to improve both maternal and infant health outcomes.
Committee
Courtney D. Lynch, PhD, MPH (Advisor)
Michael L. Pennell, PhD (Advisor)
Erinn M. Hade, PhD (Committee Member)
William C. Miller, MD, PhD, MPH (Committee Member)
Pages
156 p.
Subject Headings
Epidemiology
Keywords
survival analysis, preterm birth, threshold regression, time to delivery, gestational age
Recommended Citations
Refworks
EndNote
RIS
Mendeley
Citations
Conroy, S. A. (2019).
A Novel Approach for Modeling Time to Event Data in Maternal Child Health
[Doctoral dissertation, Ohio State University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=osu1574348181518311
APA Style (7th edition)
Conroy, Sara.
A Novel Approach for Modeling Time to Event Data in Maternal Child Health.
2019. Ohio State University, Doctoral dissertation.
OhioLINK Electronic Theses and Dissertations Center
, http://rave.ohiolink.edu/etdc/view?acc_num=osu1574348181518311.
MLA Style (8th edition)
Conroy, Sara. "A Novel Approach for Modeling Time to Event Data in Maternal Child Health." Doctoral dissertation, Ohio State University, 2019. http://rave.ohiolink.edu/etdc/view?acc_num=osu1574348181518311
Chicago Manual of Style (17th edition)
Abstract Footer
Document number:
osu1574348181518311
Download Count:
313
Copyright Info
© 2019, some rights reserved.
A Novel Approach for Modeling Time to Event Data in Maternal Child Health by Sara A Conroy is licensed under a Creative Commons Attribution-NonCommercial-NoDerivs 3.0 Unported License. Based on a work at etd.ohiolink.edu.
This open access ETD is published by The Ohio State University and OhioLINK.