Discovering Long COVID Symptom Patterns: Association Rule Mining and Sentiment Analysis in Social Media Tweets

Area of Science:

• Medical Informatics
• Public Health
• Social Media Analytics

Background:

• The COVID-19 pandemic has led to long COVID syndrome, a condition with persistent, debilitating symptoms impacting daily life.
• Breathlessness, fatigue, and brain fog are frequently reported long-term effects of COVID-19 infection.

Purpose of the Study:

• To analyze patterns and behaviors of long COVID symptoms discussed on Twitter.
• To enhance the understanding of long COVID through social media data mining.

Main Methods:

• Collected and analyzed 30,327 tweets related to long COVID from May 2020 to December 2021.
• Employed association rule mining and the Apriori algorithm to identify frequent symptoms and their relationships.
• Utilized high confidence level detection (10% minimum confidence, 0.01% minimum support, positive lift) to find significant symptom associations.

Main Results:

• Brain fog (25.8%), fatigue (17.4%), and breathing/lung issues (15.7%) were the most frequently reported symptoms.
• Other reported symptoms include heart issues, flu symptoms, depression, and general pains.
• Identified 57 meaningful relationship rules between symptoms, with lung/breathing problems and loss of taste strongly associated with loss of smell (77% confidence).

Conclusions:

• Active social media discussions provide valuable insights into long COVID.
• Natural language processing and association rule mining can effectively reveal symptom patterns and relationships.
• This research highlights the potential of social media data for understanding long COVID syndrome.