Smartphone addiction and temporomandibular disorders among university students: A machine learning based multiple...

Temporomandibular disorders (TMD) involve complex interactions among behavioral and musculoskeletal factors, and machine learning (ML) has increasingly been used to model these multidimensional relationships. This study aimed to investigate the relationship between smartphone addiction and TMD symptoms by evaluating the possible contributions of clinical variables such as head posture, cervical dysfunction, and masseter muscle sensitivity to this relationship and by examining the multivariate structure using ML methods. A single-blind, cross-sectional study was conducted among university students in Turkey. Participants completed several assessments: Smartphone Addiction Scale-Short Version (SAS-SV), Fonseca Anamnestic Index for TMD severity, Neck Disability Index, Numerical Rating Scale, forward head posture, and masseter muscle pressure pain threshold using algometry. Data were analyzed using Pearson correlation, logistic regression, and various ML classification algorithms (e.g., extreme gradient boosting [XGBoost], support vector machines). Shapley additive explanations analysis was applied to determine variable importance in ML models. In this study, XGBoost demonstrated the most successful performance in determining the relationships between smartphone addiction and TMD symptoms, while the support vector machines model showed the most successful performance in classifying participants. The findings suggest that this ML based approach, taking into account postural abnormalities and masseter muscle sensitivity, could contribute to the early identification of individuals at risk for TMD and the planning of targeted intervention strategies. Research findings have revealed complex relationships between SAS-SV and TMD symptoms among university students. In analyses conducted using ML algorithms, the support vector machines model stood out with the highest classification accuracy (area under the curve [AUC] = 0.94) and performance indicators. Shapley additive explanations analyses showed that smartphone usage time, masseter muscle sensitivity, and Fonseca Anamnestic Index score were the most decisive variables on SAS-SV. In TMD prediction, the highest predictive power was explained by temporomandibular joint pain. Ensemble models such as random forest and gradient boosting yielded more successful results with lower error rates and higher R2 values compared to classical regression methods. The findings indicate that smartphone use indirectly affects TMD through postural disorders and muscle sensitivity.