Abstract
The use of Twitter data for health-related research has been increasing over time. While the organic nature of the data offer new opportunities, the limited understanding of how and by whom the data are generated poses a challenge for advancing health-related research. Individual-level data linkage could shed light into the data generation mechanism.
This paper investigates whether consent to link survey data with Twitter public data is associated with sociodemographic and Twitter use pattern factors and whether consenters and non-consenters differ on health-related outcomes.
Data from three health related surveys that use probability samples of the target population were used: 1) A college population web survey in KU Leuven University, 2) An adult population web survey of the US population, and 3) A population face-to-face survey in the Kingdom of Saudi Arabia (KSA). In all surveys, respondents reported whether they have a Twitter account, and Twitter users were asked to provide consent for linking their survey responses to their public Twitter data.
Consent rate estimates from the two web surveys in Belgium and the US were 24% and 27% respectively. The face-to-face survey in KSA yielded a higher consent rate of 45%. In general, respondents' sociodemographic characteristics were not significantly associated with consent to link. However, more use of social media and reporting sensitive information in the survey were found to be significantly correlated with higher consent. Consenters and non-consenters were not found to be statistically different on any of the health related measures.
Very few differences were found between those who consented to link their survey data with their Twitter public data and those who did not. Modifiable design variables need to be investigated to maximize consent while maintaining balance between consenters and non-consenters.