Planning for a Crisis: Predicting Anxiety in a Population During COVID-19 Using Machine Learning

Area of Science:

• Public Health
• Computational Psychiatry
• Epidemiology

Background:

• The COVID-19 pandemic has significantly impacted global population mental health.
• Understanding anxiety prevalence and predictors is crucial for public health interventions.
• Statistics Canada surveyed Canadians to assess mental health challenges during the pandemic.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting anxiety symptoms.
• To identify key factors contributing to anxiety among Canadian adults.
• To inform the development of virtual care strategies during health crises.

Main Methods:

• Utilized a dataset of 45,989 Canadian survey respondents.
• Applied and compared eight machine learning algorithms: Logistic Regression, Random Forest, Naive Bayes, K Nearest Neighbours, Adaptive Boost, Multi-layer Perceptron, XGBoost, and LightGBM.
• Evaluated model performance using Area Under the Curve (AUC) scores.

Main Results:

• LightGBM demonstrated the highest predictive performance with an AUC of 87.45%.
• The most influential features for predicting anxiety were 'perception of mental health compared to before physical distancing', 'perceived life stress', and 'perceived mental health'.
• A noted limitation is the sample's representativeness of the broader Canadian population.

Conclusions:

• Machine learning models can effectively predict anxiety symptoms in a surveyed Canadian population.
• Perceptual factors related to mental health and life stress are significant predictors of anxiety.
• Findings highlight the need to consider these factors in planning virtual mental health care during crises.