Mining Twitter to Assess the Determinants of Health Behavior toward Human Papillomavirus Vaccination in the United States

In light of low human papillomavirus (HPV) vaccination coverage in the United States (US), there is a need to understand factors that affect people's health behaviour in order to promote vaccination uptake. The rapid growth of social media, which has transformed the health communication landscape, can be used as a tool for understanding factors that affect individuals' decision-making processes. This study sought to test the feasibility of using Twitter data to assess determinants of consumers' health behaviour toward HPV vaccination.

The study is informed by the Integrated Behavior Model (IBM), a theory holding that individuals' intention is the most important determinant of their health behaviours (i.e., HPV vaccination uptake, in the present case), while behaviour intention is subsequently determined by attitude (e.g., feelings about the behaviour), perceived norms (e.g., the social pressure one feels to perform the behaviour), and personal agency (e.g., self-efficacy).

The researchers used 3 Twitter datasets collected independently using Twitter application programming interface with HPV-related keywords. The 3 datasets covered overlapping date ranges, spanning from January 2014 to April 2018. The researchers collected 2,846,495 tweets, analysed 335,681 geocoded (to US states) tweets, and then built a rule-based model that classified each tweet into either promotional information or consumers' discussions. Through topic modeling, they identified 122 high-quality topics and subsequently explored the associations between the topics learned from consumers' discussions and the responses of HPV-related questions in the Health Information National Trends Survey (HINTS).

The most discussed consumer topics were "cervical cancer screening" and "defunding of Planned Parenthood", which account for 24.92% of all consumers' tweets; among promotional tweets, the most popular topic is to increase awareness of "HPV causes cancer". The study found that 112 of the 122 high-quality topics are relevant to consumers' discussions and can be mapped to 6 different IBM constructs: (1) "feelings about behavior" (97 topics), (2) "behavioral beliefs" (92 topics), (3) "normative beliefs - other's behavior" (36 topics), (4) "knowledge" (23 topics), (5) "normative beliefs - other's expectation" (7 topics), and (6) "environmental constrains" (2 topics).

Eighty-seven (of 122) consumer topics are correlated with promotional information, suggesting that promotional health information on Twitter has an impact on consumers' discussions. According to the researchers, these strong correlations might indicate that coordinated national efforts and promotion strategies on raising public awareness of HPV have been relatively successful.

The researchers also examined the alignment between their Twitter findings and the results obtained from HINTS. Thirty-five topics can be mapped to HINTS questions by keywords, 112 topics can be mapped to IBM constructs, and 45 topics have statistically significant correlations with HINTS responses in terms of geographic distributions. Most of these topics are related to people's "attitude" and "perceived norm."

Reflecting on the process, the researchers suggest that their hybrid use of computational tools (e.g., machine-learning-based classification and topic modeling approaches) and qualitative methods (e.g., guided by behavioural theory and manual coding of topics) enables analysis and understanding of consumers' needs based on their discussions. Furthermore, mining Twitter to assess consumers' health behaviour determinants can obtain results comparable to surveys, like HINTS. These results may motivate development of further ways of using social media data (e.g., to understand factors that are precursors to adopting a health behavioural change) and leveraging social media platforms (e.g., to design creative and tailored intervention strategies).