Modeling Spatiotemporal Pattern of Depressive Symptoms Caused by COVID-19 Using Social Media Data Mining

Area of Science:

• Computational social science
• Public health informatics
• Mental health analytics

Background:

• The COVID-19 pandemic caused widespread stress due to infection fears, job losses, and educational disruptions.
• Traditional psychological surveys are time-consuming and prone to biases, limiting real-time depression analysis.
• There is a need for scalable, objective methods to monitor population-level stress during public health crises.

Purpose of the Study:

• To propose and evaluate the CorExQ9 algorithm for detecting COVID-19-related stress symptoms.
• To analyze stress symptom spatiotemporal patterns across the United States.
• To understand the relationship between public news, risk perception, and stress levels.

Main Methods:

• Integration of the Correlation Explanation (CorEx) learning algorithm with the Patient Health Questionnaire (PHQ) lexicon.
• Application of the CorExQ9 algorithm to social media data for spatiotemporal analysis in the US.
• Minimizing human intervention and ambiguity in social media data mining for topic detection.

Main Results:

• Demonstrated a strong correlation between reported stress symptoms and increased COVID-19 case counts in major US cities (e.g., Chicago, New York).
• Observed that public risk perception is highly sensitive to COVID-19 news and media messaging.
• Identified a shift in public concerns from infection fear (Jan-Mar) to financial worries (April) regarding long-term COVID-19 impacts.

Conclusions:

• The CorExQ9 algorithm offers a robust method for real-time, large-scale analysis of public stress during pandemics.
• Spatiotemporal stress patterns are linked to disease prevalence and public health communication.
• Understanding evolving public concerns is crucial for effective public health interventions and communication strategies.