A Novel Framework Predicting Anxiety in Chronic Disease Using Boosting Algorithm and Feature Selection Techniques

Introduction: Chronic diseases represent a major public health challenge and contribute to a substantial number of deaths. Individuals living with chronic illnesses frequently experience anxiety due to the long-term impact on their overall well-being. The development of a predictive model to accurately identify anxiety in these patients is of great importance for healthcare providers. This study aimed to create a predictive model focused on anxiety among patients with chronic diseases within the Moroccan population. Methods and materials: In this study, a comparative analysis was performed using five different machine learning algorithms. The researchers utilized a cross-sectional dataset comprising 938 patients with chronic diseases. These patients were under monitoring at the Hassan II University Hospital Center in Fez, Morocco, from October 2019 to December 2020. Anxiety levels were evaluated using the validated Moroccan version of the Hospital Anxiety and Depression Scale (HADS). Results: To assess the models’ performance, several metrics were calculated, including accuracy, AUC, precision, recall, and F1-measure. The CatBoost algorithm achieved the highest accuracy of 0.7 in the evaluation. It also obtained an AUC of 0.68, a precision of 0.7, a recall of 0.55, and an F1-measure of 0.6. Based on its accuracy performance, the CatBoost algorithm is considered the optimal model in this study. Conclusion: This study created a CatBoost model to predict anxiety in patients with chronic diseases, offering valuable insights for anxiety prevention. Early psychological support and intervention are beneficial for high-risk patients, aiding in anxiety management.