The chronic mental condition schizophrenia affects someone’s thoughts as well as their emotions. We look into a cross-country dataset that includes rates of bipolar disorders, eating disorders, anxiety disorders, drug and alcohol use disorders, and depression in various countries and years. Six machine learning models—three regressors (MLP Regressor, Random Forest Regressor, Decision Tree Regressor) and three classifiers (Naïve Bayes, Logistic Regression, Decision Tree Classifier)—were used and tuned with grid search and five-fold cross-validation. Decision Tree Classifier had 99.73% accuracy (AUC = 0.99), and Random Forest Regressor provided R2 = 0.99876 (MSE = 2 × 10−⁶). While outcomes appear promising, these have to be tested using patient-level information to deploy clinically. This work fills the gap between binary classification and severity estimation and offers a scalable framework for deployment in low-resource environments. Addition with neuroimaging would potentially enhance clinical utility in the future.