Associations of employment status with social media addiction, anxiety, and parenting self-efficacy among mothers of infants: a cross-sectional study using machine learning.

BACKGROUND: Differences in social media addiction, anxiety, and parenting self-efficacy according to maternal employment status have important implications for maternal and child public health. However, evidence comparing these psychosocial outcomes between employed and non-employed mothers of infants remains limited, hindering the development of targeted maternal and child health interventions. This study compared social media addiction, anxiety, and parenting self-efficacy between employed and non-employed mothers and evaluated the performance of machine learning models for classifying employment status. METHODS: This descriptive cross-sectional study included 462 mothers with children aged 0-24 months (231 employed and 231 non-employed). Data were collected between November 2023 and February 2024 using the Social Media Addiction Scale, the Perceived Maternal Parenting Self-Efficacy Scale, and the Generalized Anxiety Disorder-7 Scale. Group differences were assessed using independent-samples t-tests, followed by analysis of covariance (ANCOVA) adjusted for age, education level, and socioeconomic status, and several supervised machine learning algorithms were used to classify employment status. RESULTS: Non-employed mothers had significantly higher social media addiction and parenting self-efficacy scores than employed mothers (both p < 0.001), whereas anxiety scores did not differ significantly (p = 0.655). Among the machine learning models, XGBoost showed the best classification performance, and SHAP analysis identified parenting self-efficacy as the most important feature contributing to employment status classification. CONCLUSION: Employment status was associated with differences in social media addiction and parenting self-efficacy, but not anxiety, among mothers of infants. Machine learning approaches may complement conventional statistical analyses by identifying variables that contribute most to employment status classification. Given the cross-sectional design, these findings should be interpreted as associations rather than evidence of causal relationships.