Anxiety disorders are a growing concern, but reliable tools for accurate diagnosis and monitoring are limited. Traditional methods of assessment often fail to integrate personalized, real-time data. This study aims to develop a predictive system for anxiety disorders using a hybrid machine learning approach on the Internet of Medical Things (IoMT) framework. Survey data based on GAD-7 (Generalized Anxiety Disorder-7), PHQ-9 (Patient Health Questionnaire-9), and UCLA-8 (University of California, Los Angeles Loneliness Scale-8), along with demographic factors like occupation, economic status, and body mass index (BMI), were collected. The system, built on the IoMT framework and hosted on a Flask cloud server, processes user data submitted via a mobile app for anxiety detection. It uses seven machine learning algorithms— Logistic Regression (LR), Naive Bayes (NB), Extreme Gradient Boosting (XGBoost/ XGB), Random Forest (RF), Support Vector Machine (SVM), Linear Discriminant Analysis (LDA), CatBoost (CatB), Decision Tree (DT), and Neural Network (NN)—to predict anxiety levels and stages. Model performance is evaluated using metrics like accuracy, precision, recall, F1-score, ROC-AUC (Receiver Operating Characteristic - Area Under the Curve), and Jaccard Score. LR and LDA achieved 97% accuracy, while SVM and CatB scored 95% and 94%, respectively. The system provides reliable real-time anxiety detection and mental health monitoring, with potential for optimization.