Analysis of Social Anxiety Factors using Ordinal Logistic Regression in Australia 2019-2022

Abstract

Social Anxiety Disorder (SAD) is a prevalent mental health condition that is characterised by intense fear of social interactions and negative evaluation. It’s a global prevalence that has increased in recent years, where previous studies focused on limited factors or used binary classifications, resulting in an overlook of the gradual severity of anxiety. This study aims to determine the main characteristics of SAD, identify the associations between anxiety levels and explanatory variables and estimate the likelihood of individuals experiencing varying levels of social anxiety. This study utilised a Social Anxiety Dataset comprising 11,000 individuals in Australia between 2019 and 2022, employing Cross-Tabulation Analysis (CTA), the Chi Square Test, and Ordinal Logistic Regression (OLR) models to achieve the research objectives. The first objective used a CTA to reveal main characteristics with higher anxiety levels associated with shorter sleep duration, lower physical activity, abnormal breathing rate, more frequent therapy frequency, poor diet quality, high stress, sweating, dizziness, and family history of anxiety. These associations have been assessed using the Chi-Square Test, with the results showing that all explanatory variables are significantly associated with anxiety levels (p-value is less than 0.05), except for gender. Meanwhile, the OLR models are developed to estimate the likelihood of individuals experiencing different anxiety levels. After comparing several model specifications, Model 2, which excludes gender, performs best according to the statistical metric. Model 2 records the lowest Akaike Information Criterion (AIC) with a value of 33174.83, the lowest corrected Akaike Information Criterion (AICc) at 33175.10, the lowest Bayesian Information Criterion (BIC) at 33452.44, deviance of 33098.83 and McFadden’s Pseudo of 0.2465 indicating that Model 2 demonstrates a strong overall fit.  The model 2 revealed that risk factors increase the likelihood of SAD severity with an odds ratio greater than one, including excessive caffeine intake at 2.6585, excessive alcohol consumption at 1.1279, elevated stress levels at 1.8898, higher sweating levels at 1.0529, smoking at 1.1133, dizziness at 1.086, family history of anxiety at 1.439, medication use at 1.1549, and being a freelancer at 1.2327. The findings highlight the importance of lifestyle, physiological, and behavioural factors in shaping anxiety severity and demonstrate the value of ordinal logistic regression in mental health research.