Early Detection of Anxiety Disorders in Single Mothers Through Explainable Machine Learning Models

Abstract

The objective of this study was to develop and evaluate explainable machine learning models for the early detection of anxiety disorders among single mothers living in Tehran, with an emphasis on identifying key psychosocial and socioeconomic predictors of anxiety risk. This cross-sectional study was conducted among single mothers residing in Tehran who were primary caregivers of at least one child under 18 years of age. Participants were recruited from community health centers and social support organizations. Data were collected using standardized self-report measures assessing anxiety symptoms, perceived stress, sleep quality, depressive symptoms, social support, and relevant sociodemographic characteristics. Following data preprocessing, multiple supervised machine learning algorithms, including regularized logistic regression, support vector machines, and tree-based ensemble models, were trained and evaluated using stratified cross-validation. Model performance was assessed using accuracy, sensitivity, specificity, and area under the receiver operating characteristic curve. Explainable machine learning techniques were applied to interpret model predictions and determine the relative importance of predictors at both global and individual levels. Inferential analyses indicated that ensemble-based machine learning models achieved the highest predictive performance, with gradient boosting models demonstrating superior accuracy and discrimination. Sensitivity and specificity values suggested strong capability in correctly identifying both high-risk and low-risk individuals. Explainable analyses revealed that perceived stress and sleep quality were the strongest predictors of anxiety risk, followed by social support and depressive symptoms, while socioeconomic factors such as income showed smaller but meaningful contributions through interaction effects. Explainable machine learning models provide an accurate and interpretable approach for early detection of anxiety disorders among single mothers, enabling timely identification of high-risk individuals while highlighting the psychosocial and structural factors underlying anxiety vulnerability. These findings support the integration of transparent data-driven screening tools into community and primary care settings.