Contribution of social determinants to symptoms of generalized anxiety disorder

Symptoms of anxiety are known to be triggered by a range of life context factors including early life trauma, poor sleep quality, infrequent exercise, unemployment and social isolation. Machine learning techniques offer a powerful method for analyzing these factors in combination, enabling the evaluation of aggregate predictive associations rather than causal pathways and the identification of their relative association with anxiety symptoms. However, most studies examining these factors have either been small-scale or included only a small number of factors. Here we applied multiple machine learning approaches (Random Forest, Gradient Boosting, Naïve Bayes, Information Gain, and SHAP) to a cross-sectional data sample of 4,186 individuals to reveal how a broad range of lifestyle and life context factors are associated with the experience of anxiety symptoms, as measured by the Generalized Anxiety Disorder-7 screening questionnaire (GAD-7). The results showed that, in combination, early life trauma, poor sleep quality, infrequent exercise, unemployment, and deterioration of social bonds were substantially associated with anxiety symptoms, particularly for older age groups, with frequency of a good night's sleep having an outsized impact. For older ages, this was followed by employment status and experience of interpersonal trauma, as well as frequency of in-person socializing. For younger ages (18-34), employment status was less important with interpersonal trauma being a more significant factor. Specifically, poor sleep, rarely socializing in person, not being able to work or being unemployed, bullying by peers, or neglect/abuse by a parent or caregiver had the largest associations with anxiety symptoms. These findings have implications for how we approach both prevention and treatment of anxiety.