Towards a data-driven characterization of behavioral changes induced by the seasonal flu

Area of Science:

• Public Health
• Behavioral Science
• Data Science

Background:

• Self-initiated behavioral changes are crucial for managing infectious disease outbreaks, such as seasonal influenza.
• Understanding the factors influencing these changes can inform public health interventions and communication strategies.

Purpose of the Study:

• To identify the primary factors driving self-initiated behavioral changes during seasonal flu epidemics.
• To utilize machine learning to predict the extent of behavioral changes based on individual characteristics and perceptions.

Main Methods:

• A questionnaire was administered via the Influweb participatory surveillance platform in Italy during the 2017-18 and 2018-19 flu seasons.
• Data from 599 surveys (434 users) included socio-demographics, flu concerns, illness history, and implemented behavioral changes.
• Machine learning algorithms, specifically Gradient Boosted Trees, were employed to classify behavioral change levels and assess feature importance.

Main Results:

• The study categorized behavioral changes into no (26%), moderate (36%), and significant (38%) levels.
• Gradient Boosted Trees achieved 66% accuracy in classifying individuals with significant behavioral changes.
• Key predictors for significant behavioral changes included the intensity and recency of past illnesses, and perceived susceptibility and severity of infection.

Conclusions:

• Individual characteristics, particularly past illness experiences and perceptions of infection risk, significantly influence self-initiated behavioral changes during flu seasons.
• Machine learning models can effectively predict behavioral responses to infectious diseases, highlighting the importance of data-driven approaches.
• Findings contribute empirical evidence to the data-driven characterization of behavioral changes in response to infectious diseases.