Dissertation
Bayesian modeling of dental caries progression over time in the oral cavity space
University of Iowa
Doctor of Philosophy (PhD), University of Iowa
Summer 2023
DOI: 10.25820/etd.006989
Abstract
Dental caries is a prevalent disease with significant public health implications. In this dissertation, we investigate the longitudinal progression of dental caries and associated risk factors, using data from the Iowa Fluoride Study (IFS). During dental examinations, clinicians classify tooth surfaces as healthy/sound, non-cavitated caries, or cavitated caries, and our goal is to model the transitions among these states.
Two modeling strategies are proposed to capture transitions among caries disease states: a multi-state Markov model that characterizes progression, and a joint competing-risks time-to-event model that accounts for transitions from a healthy state to any caries state, and progressions and regressions from the non-cavitated state. By focusing on tooth surface-level transitions, we address correlation of sites within people over time, and within the spatial domain of the oral cavity. We employ Bayesian hierarchical modeling methods, to address the multiple levels of correlation. In each modeling framework, a person-level random effect is included, as is a tooth-level spatial effect. Two parameterizations of the spatial effect are considered: the Besag-York-Mollie (BYM) model, and the associated reparameterized BYM2 model. The latter allows for direct modeling of the proportion of variance coming from the spatial component. Two different computational programs, Stan and INLA (Carpenter, Gelman, Hoffman, et al., 2017; Rue, Martino, & Chopin, 2009), are considered to fit these Bayesian models.
This research demonstrates the feasibility and importance of modeling dental caries at the tooth surface level while considering repeated measures and spatial correlation. We highlight the need for alternative approximate techniques due to the computational demands of fitting Bayesian spatial models and for more efficient use in clinical research. The outcomes and approaches presented in this dissertation have implications for dental care, treatment planning, and preventive strategies, contributing to the advancement of statistical modeling in the dental caries research field.
Details
- Title: Subtitle
- Bayesian modeling of dental caries progression over time in the oral cavity space
- Creators
- Carissa Comnick
- Contributors
- Brian Smith (Advisor)Xian Jin Xie (Advisor)Steven Levy (Committee Member)Jacob Oleson (Committee Member)
- Resource Type
- Dissertation
- Degree Awarded
- Doctor of Philosophy (PhD), University of Iowa
- Degree in
- Biostatistics
- Date degree season
- Summer 2023
- Publisher
- University of Iowa
- DOI
- 10.25820/etd.006989
- Number of pages
- x, 96 pages
- Copyright
- Copyright 2023 Carissa Comnick
- Language
- English
- Date submitted
- 07/22/2023
- Description illustrations
- illustrations, graphs, tables
- Description bibliographic
- Includes bibliographical references (pages 90-96).
- Public Abstract (ETD)
- Tooth decay is a common disease that affects many people and has impacts on public health. In this dissertation, we want to understand how tooth decay progresses over time and what factors are associated with it, using data from the Iowa Fluoride Study (IFS) to investigate. When dentists examined teeth during this study, they recorded tooth surface conditions as healthy, having early signs of decay, or having more advanced decay. Our goal was to create models that can predict how teeth transition among these different conditions.
We used two different models to look at these questions. One model looked at the overall progression of tooth decay, while the other model considered the risks of moving from a healthy state to any decay state, as well as the chance of tooth surfaces with early decay getting either worse or better. By focusing on specific tooth surfaces and following how they change over time, we took into account the fact that people’s teeth can change and that different areas of the mouth might be affected differently.
This research shows that it is possible to create models that predict tooth decay at the individual tooth surface level. This is important because it allows us to consider the measurements taken over time and the spatial patterns within the mouth. We also found that there is a need for more efficient techniques to fit these models because they can be computationally demanding. This focus on computational speed will make the models more useful in clinical research and dental care.
The findings and approaches we presented in this study have implications for how we provide dental care, plan treatments, and develop preventive strategies. By advancing statistical modeling in the field of tooth decay research, we hope to improve oral health.
- Academic Unit
- Biostatistics
- Record Identifier
- 9984454435702771
Metrics
1 File views/ downloads
23 Record Views