Tooth decay is a highly prevalent chronic disease in childhood. To understand its etiology and time-course profile, many longitudinal studies of dental caries have been conducted. Despite being able to inform caries prevention and treatment, survival analyses of tooth decay data from longitudinal studies are challenging because of the complications including interval censoring, intra-oral correlation, unknown tooth emergence time, and time-varying risk factors and/or interventions among others. The proposed research will develop two suites of survival analysis methods that account for several such complications simultaneously.
Aim 1 is to develop a joint analysis method for correlated interval-censored age-to-caries data of primary teeth and longitudinal data of a time-varying covariate.
Aim 2 is to develop a time-to-caries analysis method that copes with interval censored tooth emergence times by treating non-merged tooth as a state in semi-Markov multistate models. This method also accounts for interval-censored times to caries and the correlation between teeth within a mouth. The two suites of new survival analysis methods both include statistical inference and outcome prediction approaches. Simulation studies will be conducted to evaluate the finite sample performance of the new methodology. The final methods will be programmed into an R package to be disseminated through the Comprehensive R Archive Network. The developed methods will be applied to the Detroit Dental Health Project (DDHP), a three-wave cohort study on oral health of young African American children living in downtown Detroit, Michigan, to illustrate their utility and discover subject matter knowledge, particularly the effects of some important risks factors (e.g., sugar intake and dietary calcium) and an educational intervention implemented since Wave 2 on primary tooth decay. The completion of the project will lead to novel survival analysis methods that can delineate the tooth decay profile and predict future caries probabilities based on complex time-to-event data in caries research, and the methods will come with user-friendly implementing software for free public use.
The proposed research will develop statistical methods for analyzing complex lifetime data from longitudinal studies of childhood caries and predicting tooth-level caries risk for individuals. The success of the project will advance our knowledge of childhood dental caries in low-income African Americans, which may then help their prevention and treatment in these underserved populations.
