IADR Abstract Archives
Artificial Intelligence Algorithms for Understanding the Determinants of Oral Health
Objectives: Oral health plays a critical role in individuals’ overall well-being and quality of life. Better identifying individuals at higher risk for oral health issues, without having them undergo a dental examination, is of great interest. The aim of this study was to use artificial intelligence by developing machine learning predictive models to forecast the likelihood of permanent tooth loss, as an indicator of overall oral health, based on various behavioral and lifestyle factors.
Methods: Data on the Behavioral Risk Factors, obtained from the 2022 CDC’s Behavioral Risk Factor Surveillance System (BRFSS), was used in this study. Different factors from a diverse group of respondents were collected, including age, gender, education, income, smoking history, chewing tobacco use, e-cigarette use, alcohol consumption, physical activity, sleep patterns, general health status, and dental care visits. After cleaning and refining the dataset, a total of 293,398 individuals were evaluated. Five different machine-learning techniques were employed to predict tooth loss, including K-nearest neighbor, logistic regression, decision trees, random forests, and extreme gradient boosting trees.
Results: Our findings showed that although age and routine dental care were the strongest predictors for tooth loss, socioeconomic conditions also played a significant role, indicating their importance in predicting this condition. Indeed, models incorporating socioeconomic characteristics outperformed those relying solely on clinical dental indicators. According to our findings, the best performing machine-learning algorithm was the extreme gradient boosting trees which exhibited the highest predictive performance in determining tooth loss (AUC = 81.2%).
Conclusions: This study highlights the ability of machine-learning algorithms in predicting tooth loss risk based on Behavioral Risk Factors. The findings suggest that incorporating socioeconomic factors into predictive models can enhance their accuracy and effectiveness. These models have the potential to find practical application in clinical settings for identifying individuals at risk of tooth loss, enabling healthcare professionals to prioritize preventive interventions.
Methods: Data on the Behavioral Risk Factors, obtained from the 2022 CDC’s Behavioral Risk Factor Surveillance System (BRFSS), was used in this study. Different factors from a diverse group of respondents were collected, including age, gender, education, income, smoking history, chewing tobacco use, e-cigarette use, alcohol consumption, physical activity, sleep patterns, general health status, and dental care visits. After cleaning and refining the dataset, a total of 293,398 individuals were evaluated. Five different machine-learning techniques were employed to predict tooth loss, including K-nearest neighbor, logistic regression, decision trees, random forests, and extreme gradient boosting trees.
Results: Our findings showed that although age and routine dental care were the strongest predictors for tooth loss, socioeconomic conditions also played a significant role, indicating their importance in predicting this condition. Indeed, models incorporating socioeconomic characteristics outperformed those relying solely on clinical dental indicators. According to our findings, the best performing machine-learning algorithm was the extreme gradient boosting trees which exhibited the highest predictive performance in determining tooth loss (AUC = 81.2%).
Conclusions: This study highlights the ability of machine-learning algorithms in predicting tooth loss risk based on Behavioral Risk Factors. The findings suggest that incorporating socioeconomic factors into predictive models can enhance their accuracy and effectiveness. These models have the potential to find practical application in clinical settings for identifying individuals at risk of tooth loss, enabling healthcare professionals to prioritize preventive interventions.