Analysis of the Conversations on Twitter regarding HPV Vaccine

Abstract

Background and objectives Infection with human papillomavirus (HPV) is the main cause of cervical cancer, and vaccination is an effective method to prevent HPV infection. In Japan, adverse reactions were reported in some HPV-vaccinated people in March 2013, and while Japan’s Ministry of Health, Labor, and Welfare withdrew active recommendation of the vaccine in June 2013, the social movement to refuse vaccination has continued. The Ministry of Health, Labor, and Welfare (MHLW) has devised a plan to accurately disseminate information that promotes vaccination, but less than 1% of the eligible population was vaccinated, and the number has not increased. Besides, inaccurate information about health information can disseminate rapidly on social networks. Social networking services (SNS), mainly used by young people, can be used by the public to obtain medical information. However, according to the World Health Organization (WHO), SNSs are prone to spreading inauthentic and misleading information when it comes to information related to health and medical care. “Infodemic” is defined as a situation in which unidentified and false information is widely disseminated on SNS, causing WHO to issue international alerts. This study aimed to organize information about HPV vaccination disseminated on SNS in Japan. Methods We extracted 208 tweets with the keyword “HPV vaccine” posted in Japan between April 1, 2014, and September 30, 2017. The extracted tweets included data points such as ID, language, posting date and time, and latitude and longitude. The location information of the senders was obtained from the latitude and longitude, and the tweets were organized by prefecture, city, town, village, and ID. Then, we evaluated whether the information at the URLs was accurate by examining retweets, likes, and the number of comments on the tweet. Python version 3.7.7 was used to extract the tweets. Results The results of classification of the tweets by prefecture are as follows: the Hokkaido prefecture accounted for four tweets; the northeast, six tweets; southern Kanto, 123 tweets; northern Kanto-Koshin, six tweets; Hokuriku, five tweets; Tokai, 35 tweets; Kinki, 10 tweets; Chugoku 4 tweets; Shikoku, three tweets; and Kyushu, nine tweets. A total of 93 users posted tweets; four users posted five or more tweets; 14 users posted 2–4 tweets, and 75 users posted one tweet. In particular, 66 tweets in Kanagawa prefecture, 14 tweets in Shizuoka prefecture, and two tweets in Tokyo were posted from the same ID. Regarding the type of tweet, there were 109 tweets, 65 retweets, and 34 replies. There were 137 tweets with and 71 tweets without URLs. When organized by the linked URL, 50 posts linked to a blog, 46 posts linked to a news item, seven posts linked to Facebook, five posts linked to a government agency homepage, four posts linked to YouTube, three posts linked to the home page of the City Council rep, two posts linked to a medical site, and 20 posts linked to other sources that could not be categorized. In terms of the authenticity of the posts, 25 tweets were judged as “accurate,” 14 were judged to be “inaccurate,” and 16 were judged as “unknown.” We classified the posts as follows; “accurate” for those that contained accurate information and “inaccurate” for those that contained inaccurate information. Discussion The distribution of tweets and the uneven distribution of the users suggest that few people spread information about the HPV Vaccine on Twitter in Japan. Regarding the content, more than half of the tweets could not be judged as accurate or inaccurate because the verification results regarding adverse reactions of the HPV vaccine were not published at the time of sending, and in news and blog articles, personal opinions were stated rather than authentic medical information. In this study, we clarified the characteristics of tweets regarding HPV vaccination in Japan and the status of transmission. In the future, it will be necessary to change the keywords and time periods for which tweets need to be extracted, and the data set used for the analysis will need to be compared and examined.