Identifying Factors Associated with Periodontal Disease Using Machine Learning

OBJECTIVE: This study aimed to identify combinations of chronic conditions associated with the presence and severity of periodontal disease (PD) after accounting for a series of demographic and behavioral characteristics in a nationally representative sample of US adults. MATERIALS AND METHODS: A cross-sectional study of the 2013–2014 National Health and Nutrition Examination Survey (n = 4555). Outcome measure: PD using clinical attachment loss (measured as none, mild, moderate, or severe). The main independent variables were self-reported chronic conditions, while other covariates included demographic and behavioral variables. Classification and regression tree analysis was used to identify combinations of specific chronic conditions associated with PD and PD with higher severity. Random forest was used to identify the most important variables associated with the presence and severity of PD. RESULTS: The prevalence of PD was 77% among the study population. The percentage of those with PD was higher among younger and middle-aged (< 61 years old) than older (> 61 years old) adults. Age and education level were the two most important predictors for the presence and severity of PD. Other significant factors included alcohol use, type of medical insurance, sex, and non-white race. Accounting for only chronic conditions, hypertension and diabetes were the two chronic conditions associated with the presence and severity of PD. CONCLUSIONS: Sociodemographic and behavioral factors emerged as more strongly associated with the presence and severity of PD than chronic conditions. Accounting for the co-occurrence for sociodemographic and behavioral factors will be informative for identifying people vulnerable to the development of PD.