Incidence, prevalence, and trajectories of dental caries experience of young adults

Tooth decay remains a public health issue across all age groups, both in the United States and globally. As a chronic and cumulative disease, tooth decay develops over time and can occur at any stage of life if there are susceptible tooth surfaces. There are very few long-term prospective studies that have attempted to describe patterns of tooth decay over a substantial portion of the life course, probably because of the difficulty following participants for a long time.

This dissertation has three parts. The first part focused on the incidence of tooth decay and associated risk factors from age 17 to 23. The second study focused on the prevalence of tooth decay at age 23 and factors from age 9 to 23 that are associated with the prevalence. The third study focused on trajectories of tooth decay from childhood (age 9) to young adulthood (age 23) and factors from age 9 to 23 that are associated with the trajectory group each individual belongs to. All the data were from the Iowa Fluoride Study. The dental exams were performed at ages 9, 13, 17, and 23. The predictor variables used in the dissertation were sex, mother’s education, family income, composite SES, tooth decay experience at ages 9, 13, and 17, and the ages 5-9, 9-13, 13-17, and 17-23 cumulative exposure variables relating to fluoride, diet, and behaviors. The analysis of the first part of the dissertation was performed using traditional statistical methods while the analysis of the second and third parts of this dissertation was performed using machine learning techniques.

In the first part of this dissertation, 40% of the participants had an increment in the number of decayed tooth surfaces from age 17 to 23. The average number of newly decayed tooth surfaces from age 17 to 23 was 1.70. Further analysis showed that higher composite SES and higher combined daily fluoride intake were associated with a lower number of newly decayed tooth surfaces from age 17 to 23. Higher frequency of milk intake and higher amount of sugar-sweetened beverages intake were associated with a lower number of newly decayed tooth surfaces from age 17 to 23. Higher 100% juice intake and age 17 dental caries count were associated with a higher number of newly decayed tooth surfaces from age 17 to 23.

The second study found a high prevalence of cavities at age 23 (69.1%). The average number of tooth surfaces with cavities at age 23 was 4.75. After performing the prediction of the risk of tooth cavity at age 23 using machine learning techniques, the LASSO regression was the best performing model (compared to gradient boosting machine (GBM), generalized linear model (negative binomial (NegGLM)), and extreme gradient boosting model (XGBOOST)). The accuracy of LASSO in correctly predicting tooth cavities at age 23 was 83.7%. The model was precise in correctly classifying those who truly had had tooth decay at age 23 (85.9%) and the sensitivity was 93.1%. Having had tooth decay at age 13 and 17, as well as the drinking of sugar-sweetened beverages intake at age 13 and age 17, were important risk factors positively associated with the prediction of tooth cavity at age 23.

The third study found three trajectory groups, with 70.5%, 21.1%, and 8.4% of participants in low, medium, and high caries trajectory groups, respectively. There were steeper increases in the number of decayed tooth surfaces of the three trajectory groups between age 13 and 17, with less steep but also strongly positive slopes from age 17 to 23, suggesting that the period from age 13 to 17 is the highest risk period. The machine learning model XGBOOST was most accurate (85%) compared to LASSO and GBM in predicting the trajectory group an individual belongs to. The model was 96%, 81%, and 80% precise in correctly classifying individuals into high, medium, and low caries trajectory groups, respectively. The model was also 97.5%, 82.7%, and 75.5% sensitive in detecting individuals that are wrongly classified into low, medium, and high caries trajectory groups, respectively. Sex (female) was the most important variable. The other 12 most important variables were related to the mother’s educational level, socioeconomic status, fluoride intake, milk intake, sugar-free beverages intake, sugar-sweetened beverages intake, and 100% juice intake.

In conclusion, the main findings from the three studies suggest that age 13 to 17 (early teenage years) may be the most critical period in the process of developing tooth decay from age 9 to 23. Dietary factors (sweetened sugar beverages intake, milk, and 100% juice) appear to be the most prominent & consistent modifiable factors associated with tooth decay across the three studies. Also, non-modifiable factors that were associated with tooth decay across the three studies were previous tooth decay experience and socioeconomic status. This dissertation demonstrates the great potential of machine learning as a tool for predicting the risk and pattern of tooth decay, as well as the identification of predictors of tooth decay, not just in young adults, but potentially also across other age groups. Finally, this dissertation suggests the need for more studies across different ages, races/ethnicity, and socioeconomic backgrounds, with a wider array of clinical and behavioral variables, to better understand the patterns of and risk and protective factors for tooth decay.