J Clin Exp Dent. 2026 Jul 29;18(8):e982-e989. doi: 10.4317/jced.64297. eCollection 2026 Jul.
ABSTRACT
BACKGROUND: The aging global population is characterized by a high prevalence of multimorbidity and polypharmacy. This study aimed to investigate the independent and combined associations of these factors on the severity and progression of periodontal disease in older adults and to develop a predictive machine learning model for clinically meaningful disease progression.
MATERIALS AND METHODS: This retrospective cohort study utilized de-identified electronic health records from patients aged 65 and older with periodontitis. The primary independent variables were polypharmacy (categorized as 0-4, 5-9, or 10+ medications) and a multimorbidity index (a sum of self-reported systemic conditions). The outcome variables were baseline mean clinical attachment level (MEAN_CAL) and the annualized rate of MEAN_CAL progression. Multivariable linear regression models were used to assess associations, and a Random Forest Classifier was developed to predict clinically relevant disease progression (CAL 1mm).
RESULTS: The study included 5,645 patients for cross-sectional analysis and a sub-cohort of 2,130 for longitudinal analysis. Both polypharmacy (5-9 medications: 0.18 mm; 10+ medications: 0.29 mm) and each additional systemic disease (0.09 mm) were significantly associated with higher baseline MEAN_CAL (p < 0.001). Longitudinally, hyperpolypharmacy (10+ medications) and each additional comorbidity were associated with an annual CAL loss of 0.08 mm (p = 0.002) and 0.03 mm (p = 0.001), respectively. The machine learning model predicted disease progression with an AUC-ROC of 0.81, 94% specificity, and 61% recall. Baseline MEAN_CAL, smoking, age, bleeding on probing, diabetes, and the number of medications were the most important predictors.
CONCLUSIONS: Multimorbidity and polypharmacy are significant and clinically relevant independent predictors of more severe periodontal disease and faster progression in older adults. Predictive modeling can help identify high-risk individuals, emphasizing the need for integrated medical-dental care for this population.
PMID:42542801 | PMC:PMC13428309 | DOI:10.4317/jced.64297