Thesis
Predictors of complete edentulism in older adults in the United States using the 2020 Behavioral Risk Factor Surveillance System (BRFSS) data
University of Iowa
Master of Science (MS), University of Iowa
Spring 2022
DOI: 10.17077/etd.006435
Abstract
The proportion of adults aged 65 and older in the United States continues to grow. This increase in population growth will affect the oral health status and oral health care in the coming decades. With the increasing aging population, it is expected that there will be an increase in the need for various forms of care, including chronic oral diseases. Many older adults are keeping more natural teeth due to the benefits enjoyed from improved oral health over the decades; however, due to the persisting health inequities, many still face significant barriers in accessing oral health care. Tooth loss, whether partial or complete, is an incapacitating condition that marks the endpoint of several disease processes that affect an individual's oral health. It is a chronic oral disease of global public health importance due to its high prevalence and attributable disability. In 2010, about 158 million individuals were completely edentulous globally. The loss of natural teeth places an enormous burden on an individual's physical and psychological well-being, which can significantly affect the quality of life of a person. In 2019, the years of healthy life lost due to disability (YLDs) due to edentulism and severe tooth loss was 9.62 million (95% UI 6.15 - 14.2) YLDs. Also, the Institute of Health Metrics and Evaluation describes edentulism and tooth loss as the “22nd ranked Level 4 cause of disability” worldwide.
The older age group and economically disadvantaged are most affected by complete edentulism and its consequences, with an estimated prevalence of over 10% worldwide in individuals aged 50 years and above. Even with the decline in the prevalence of edentulism in the United States, significant disparities still occur in some population groups. Addressing these disparities is very important because reducing the number of people who completely lose their natural teeth is a vital objective of Healthy People 2030. Thus, investigating the prevalence and the predisposing risk factors of edentulism will provide us with an indication of where we are with population oral health, identify predictors and gauge the response of the oral health care system to this condition. One way to investigate the prevalence of edentulism nationwide would be to analyze information gathered from national surveys. National surveys allow for the extensive use of properly documented information from a large sample to examine multiple predictors and relationships.
The Behavioral Risk Factor Surveillance System (BRFSS) data set, a telephone survey that gathers health-related information from U.S. residents across the nation about health-related risk behaviours, chronic health conditions, and the use of preventive services, meets this criterion and will be used to evaluate the prevalence and risk factors for edentulism in this thesis. This thesis utilized different algorithms mining techniques, namely Naive Bayes, K-Nearest Neighbors, Random Forest, AdaBoost, Gradient Boosting, Logistic Regression and Ensemble, to identify predictors for complete edentulism in older adults in the United States using the 2020 BRFSS data set.
Machine learning (ML) is increasingly used in analyzing big data, including health care data. It simply refers to using artificial intelligence to develop predictive models. It can improve our understanding of characterizing and predicting risk factors for different health outcomes, including oral health. ML techniques have also been better adapted to situations where convectional statistical methods have not yielded the best performance.
Analysis showed that health access factors – (last dental visit), socioeconomic factors – (level of education, marital status and employment status), health status and health behavior factors – (general health status and smoking), chronic disease factors – (heart attack and coronary heart disease (CHD) or myocardial infarction (MI)) and the presence of a disability – (difficulty walking and seeing) were strong predictors of complete tooth loss in older adults.
Overall, ML techniques can improve the population's oral health and health outcomes by taking advantage of the enormous data. Furthermore, the information can enhance oral health policy and decision-making in care providers by better specifying the intervention points. Also, data from population-based studies such as the BRFSS could help develop systems that could better integrate dental and medical care.
Details
- Title: Subtitle
- Predictors of complete edentulism in older adults in the United States using the 2020 Behavioral Risk Factor Surveillance System (BRFSS) data
- Creators
- Abimbola M. Oladayo
- Contributors
- Leonardo Marchini (Advisor)Azeez Butali (Committee Member)Dan Caplan (Committee Member)Erliang Zeng (Committee Member)
- Resource Type
- Thesis
- Degree Awarded
- Master of Science (MS), University of Iowa
- Degree in
- Dental Public Health
- Date degree season
- Spring 2022
- DOI
- 10.17077/etd.006435
- Publisher
- University of Iowa
- Number of pages
- xi, 116 pages
- Copyright
- Copyright 2022 Abimbola M. Oladayo
- Language
- English
- Description illustrations
- Charts, tables, graphs
- Description bibliographic
- Includes bibliographical references (pages 85-95).
- Public Abstract (ETD)
Over the years, the prevalence of tooth loss in older adults in the United States (US) has declined. However, significant disparities still exist among some population groups. Tooth loss is an incapacitating condition that marks the endpoint of several disease processes that affect an individual's oral health. This thesis aimed to identify the predictions of completed edentulism in older US adults using different machine learning algorithms.
The data used form this thesis was from the 2020 cycle of the Behavioral Risk Factor Surveillance System (BRFSS). Six machine learning algorithms were applied to the data set, and predictive performances were examined by AUC under ROC (area under the receiver operating characteristic curve).
The Adaptive boosting (AdaBoost) and Gradient Boosting showed the highest performance with the highest AUC score in the prediction of edentulism (AUC = 84.9%), respectively. In agreement with previous research, our analysis showed that health access factors – (last dental visit), socioeconomic factors – (level of education, marital status and employment status), health status and health behavior factors – (general health status and smoking), chronic disease factors – (heart attack and coronary heart disease (CHD) or myocardial infarction (MI)) and the presence of a disability – (difficulty walking and seeing) were strong predictors of complete tooth loss in older adults. Also, other predictors were identified.
ML techniques can improve the population's oral health and health outcomes by taking advantage of the enormous data. It will be interesting to see the outcome when used on longitudinal data. Furthermore, the information can enhance oral health policy and decision-making in care providers by better specifying the intervention points to prevent tooth loss in the older population.
- Academic Unit
- Preventive and Community Dentistry; Craniofacial Anomalies Research Center
- Record Identifier
- 9984271451702771
Metrics
1 File views/ downloads
156 Record Views