Oral health problems have been a major public health concern profoundly affecting people’s general health and quality of life. Given that oral health data is composed of several measurable dimensions including clinical measurements, socio-behavioral factors, genetic predispositions, self-reported assessments, and quality of life measures, strategies for analyzing multidimensional data are neither computationally straightforward nor efficient. Researchers face major challenges to identify tools that circumvent the processes of manually probing the data.
The purpose of this dissertation is to provide applications of the proposed methodology on oral health-related data that go beyond identifying risk factors from a single dimension, and to describe large-scale datasets in a natural intuitive manner. The three specific applications focus on the utilization of 1) classification regression tree (CART) to understand the multidimensional factors associated with untreated decay in childhood, 2) network analyses and network plots to describe connectedness of concurrent co-morbid conditions for pediatric patients with autism receiving dental treatments under general anesthesia, and 3) random forests in addition to conventional adjusted main effects analyses to identify potential environmental risk factors and interactive effects for periodontitis.
Compared to findings from the previous literature, the use of these innovative applications demonstrates overlapping findings as well as novel discoveries to the oral health knowledge. The results of this research not only illustrate that these data mining techniques can be used to improve the delivery of information into knowledge, but also provide new avenues for future decision making and planning for oral health-care management.
Available at: http://works.bepress.com/hsin-fang-li/23/