You are here

  • Home
  • Archive
  • Volume 7, Issue 10
  • Comparing human papillomavirus vaccine concerns on Twitter: a cross-sectional study of users in Australia, Canada and the UK

Article Text

Article menu

Download PDFPDF

XML
Communication
Research
Comparing human papillomavirus vaccine concerns on Twitter: a cross-sectional study of users in Australia, Canada and the UK
  1. Gilla K Shapiro1,2,
  2. Didi Surian3,
  3. Adam G Dunn3,
  4. Ryan Perry2,
  5. Margaret Kelaher4
  1. 1 Department of Psychology, McGill University, Montreal, Canada
  2. 2 Centre for Health Policy, School of Population and Global Health, University of Melbourne, Melbourne, Victoria, Australia
  3. 3 Centre for Health Informatics, Australian Institute of Health Innovation, Macquarie University, Sydney, New South Wales, Australia
  4. 4 Head Evaluation and Implementation Science, Centre for Health Policy, School of Population and Global Health, University of Melbourne, Melbourne, Victoria, Australia
  1. Correspondence to Gilla K Shapiro; gilla.shapiro{at}mail.mcgill.ca

Abstract

Objective Opposition to human papillomavirus (HPV) vaccination is common on social media and has the potential to impact vaccine coverage. This study aims to conduct an international comparison of the proportions of tweets about HPV vaccines that express concerns, the types of concerns expressed and the social connections among users posting about HPV vaccines in Australia, Canada and the UK.

Design Using a cross-sectional design, an international comparison of English language tweets about HPV vaccines and social connections among Twitter users posting about HPV vaccines between January 2014 and April 2016 was conducted. The Health Belief Model, one of the most widely used theories in health psychology, was used as the basis for coding the types of HPV vaccine concerns expressed on Twitter.

Setting The content of tweets and the social connections between users who posted tweets about HPV vaccines from Australia, Canada and the UK.

Population 16 789 Twitter users who posted 43 852 tweets about HPV vaccines.

Main outcome measures The proportions of tweets expressing concern, the type of concern expressed and the proportions of local and international social connections between users.

Results Tweets expressing concerns about HPV vaccines made up 14.9% of tweets in Canada, 19.4% in Australia and 22.6% in the UK. The types of concerns expressed were similar across the three countries, with concerns related to ‘perceived barriers’ being the most common. Users expressing concerns about HPV vaccines in each of the three countries had a relatively high proportion of international followers also expressing concerns.

Conclusions The proportions and types of HPV vaccine concerns expressed on Twitter were similar across the three countries. Twitter users who mostly expressed concerns about HPV vaccines were better connected to international users who shared their concerns compared with users who did not express concerns about HPV vaccines.

  • Human papillomavirus
  • Vaccination
  • Twitter
  • Health Belief Model
  • Social media.

This is an Open Access article distributed in accordance with the Creative Commons Attribution Non Commercial (CC BY-NC 4.0) license, which permits others to distribute, remix, adapt, build upon this work non-commercially, and license their derivative works on different terms, provided the original work is properly cited and the use is non-commercial. See: http://creativecommons.org/licenses/by-nc/4.0/

https://doi.org/10.1136/bmjopen-2017-016869

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    Strengths and limitations of this study

    • This study conducted an international comparison investigating how human papillomavirus (HPV) vaccination concerns are expressed on Twitter.

    • Machine learning methods were used to identify and classify the proportion and types of concerns expressed in thousands of tweets.

    • The analysis of social connections among Twitter users posting about HPV vaccines in three English-speaking countries (Australia, Canada and the UK) revealed the potential for concerns to spread internationally.

    • While Twitter is used by a substantial number of people, this study is not designed to allow for generalisation to the general population.

    • This study used follower networks to examine social connections, but further research could use other social interactions to measure the spread and impact of negative attitudes about HPV vaccines.

    Introduction

    Human papillomavirus (HPV) is a prevalent sexually transmitted infection that can cause cancers and anogenital warts.1–5 Since 2006, three prophylactic vaccines have been developed to protect adolescents from HPV-associated health problems.6 Research has demonstrated that these vaccines are safe and effective in reducing HPV related infections, genital warts and pre-cancers.7–12 As a result, at least 65 countries have implemented HPV vaccination programmes for women in their national immunisation schedules.11

    There is notable variation between countries’ HPV vaccine programmes and coverage rates. Australia’s school-based vaccination programme targets girls (since 2007) and boys (since 2013) aged 12–13 years.13 According to Australia’s National HPV vaccination programme register, 85.6% of women and 77% of men received the HPV vaccine (2015 data).14 15 In Canada, all provinces and territories introduced school-based vaccination programmes for girls aged 9–13 years (2007–2010), and six provinces also include boys in HPV vaccine programmes (since 2013).16 According to national parental surveys, 72.3% of women (2013 data) and less than 3% of men received the HPV vaccine (2014 data).17–19 Lastly, the UK only provides a school-based vaccination programme for girls aged 12–13 years (since 2008). According to Public Health England, 89.5% of women in the UK received the HPV vaccine (2015 data).20 HPV vaccine coverage rates are lower than other child or adolescent vaccines in these countries’ national immunisation programmes21–23; suboptimal coverage hinders cancer prevention efforts.24

    The media has the potential to dramatically impact vaccine coverage through influencing parental awareness, perception and attitudes.25–29 Unconfirmed reports of adverse events associated with the HPV vaccine published in the media dramatically affected female HPV vaccine coverage in Japan and Colombia.11 30 31 Many individuals use the internet and social media to access health information; however, these sources have been described as a risky platform that can rapidly amplify unbalanced, distorted or inaccurate information about vaccines.25 32–34 For example, a study by Betsch et al found that even 5–10 min of access to vaccine-critical websites negatively influenced individuals’ risk perception and intentions to be vaccinated.35 Similarly, Nan and Madden report that, compared with a control group, participants who were exposed to negative online blogs about HPV perceived the vaccine as less safe, held more negative attitudes and reported a reduced intention to receive the vaccine.36

    Previous research has evaluated the public discourse concerning HPV vaccination in newspapers,37–40 online news,41 comments to online news articles,42 Facebook,43 blogs or online forums36 44 45 and YouTube videos.46 47 Twitter is a microblogging service, established in 2006, that has over 313 million users active monthly. Twitter is an important source of information regarding HPV and vaccine hesitancy,32 48 49 and several studies have examined the representation of HPV vaccines on Twitter.33 43 50–54 Though many of these studies analyse a limited number of HPV-related tweets, a few have used data mining and machine learning techniques to analyse a large number of tweets.51 52 54 55 However, no research has conducted an international comparison to evaluate and compare how vaccination concerns are expressed across countries. Furthermore, no research has examined the domestic and international network connectedness of HPV vaccine concern expression.

    The aim of this study was to explore the proportion of HPV vaccine concern on Twitter, examine the type of concern expressed in Australia, Canada and the UK and investigate differences in the ways Twitter users connected locally and internationally.

    Methods

    Study overview

    Tweets related to HPV vaccines were collected from January 2014 to April 2016 in Australia, Canada and the UK. These countries were selected because they are English-speaking countries, share a similar history and commonwealth membership and their similarity in administering the HPV vaccination in schools. Data captured included information about users’ locations, the text of the tweets and information about social connections. To enable the classification of a large number of tweets, two stages of machine learning classifiers were constructed from a sample of tweets that were manually coded by two investigators.

    Study data

    Using a similar approach to previous studies that examined large number of tweets in communities of Twitter users posting about HPV vaccines,54 56 the Twitter Search Application Programming Interface (API) was used to collect tweets in the English language about HPV vaccines from January 2014 to April 2016. The search terms were ‘Gardasil’, ‘Cervarix’, ‘hpv AND vaccin*’ and ‘cervical AND vaccin*’. Information extracted from each tweet included the unique tweet identifier, tweet text, creation time, the identifier of the user posting the tweet and geographical coordinates (if available). Without any restrictions applied to the locations of users, the entire dataset included 358 194 tweets (including retweets) by 129 286 users.

    A gazetteer was used to transform the text provided by users into coordinates, and any users with self-reported locations that were located in coordinates in Australia, Canada or the UK were included in this analysis (online supplementary material, Section 1).

    Supplemental material

    Other data that were used in the analyses included the set of social connections formed among the users who were included in the analyses. For each user, the Twitter Search API was used to collect the set of all follower relationships in which the user was involved, shortly after the first time the user posted a relevant tweet in the period. A network was then formed to include all users who tweeted about HPV vaccines from the three countries, and the follower relationships defined the social connections in an unweighted, directed network (online supplementary material, Section 2).

    The Macquarie University Human Research Ethics Committee (number 5201401028) and the University of Melbourne’s Research Ethics Board (number 1647488.1) provided ethics approval for data collection and analysis.

    Analysis

    Supervised machine learning methods were used to classify the tweets into two stages: (1) to identify tweets that expressed any concern and (2) to classify specific types of concerns. In the first stage, 1000 tweets were sampled from the set of all tweets to manually label those that expressed concerns. In the second stage, 1000 tweets were sampled from the set of tweets that were estimated to be concerns to manually label them by type. The manually labelled tweets were used to train classifiers to label any tweet by the type of concern expressed (online supplementary material, Section 3).

    The categories for types of tweets expressing concerns were determined using an inductive and deductive procedure. The Health Belief Model (HBM), one of the most widely used theories in health psychology, was used as the basis for coding the types of HPV vaccine concerns expressed on Twitter.57 The HBM has been used previously to evaluate the determinants of HPV vaccination and non-compliance by identifying perceived susceptibility to HPV, perceived severity of HPV, perceived benefits of HPV vaccination, perceived barriers of HPV vaccination (including tangible barriers such as logistical challenges and psychological barriers such as perceived harms of receiving the HPV vaccine) and cues to action (eg, influences prompting HPV vaccine uptake such as information from healthcare providers, family or friends).47 57–59 To account for additional prominent concerns that were not captured by the model, the constructs were also informed by previous content analyses of media and social media related to the HPV vaccine, as well as literature on vaccine hesitancy.28 33 37 41 42 45 47 60–68 The coding scheme was therefore extended to include mistrust, undermining of religious principles, undermining of civil liberties, additional concerns (not otherwise specified) and ambiguous tweets (table 1). The coding scheme was used by two investigators (GKS and RP) to code 12 types of concerns expressed in a second sample of 1000 tweets. After examining the proportions of different types of concerns in the sample and accuracy of the multiclass classifier (online supplementary material, Sections 3 and 4), types of concerns were combined to improve the performance that could be achieved by the machine learning classifiers (table 1). Combining categories was done based on conceptual similarity and trying to remain as true to the HBM as possible while attaining accuracy of the classifier.

    View this table:
    Table 1

    Coding scheme for the types of concerns expressed on Twitter

    The network of social connections formed by users who posted tweets about HPV vaccines was used to compare the proportions of local (within a country) and international (across countries) followers. The group of users for whom at least half of their relevant tweets were expressing concerns was assigned to one group (concern), and all other users were assigned to another (non-concern). The users were then also split by country, and the proportions of local and international followers were compared across groups.

    Results

    There were 129 286 Twitter users who posted at least one tweet about HPV vaccines during the period. The location inference method identified 2792 (2.2%) of those users located in Australia, 7237 (5.6%) located in Canada and 6760 (5.2%) located in the UK (table 2).

    View this table:
    Table 2

    The total number of users and tweets from Australia, Canada and the UK

    From the 16 789 users in the three countries, a total of 43 852 tweets about HPV vaccines were posted, of which 7173 (16.4%) were from Australia, 18 927 (43.2%) were from Canada and 17 752 (40.5%) were from the UK. This corresponded to an average of 2.57 tweets per user in Australia (range, 1–198), 2.61 tweets per user in Canada (range, 1–433) and 2.62 tweets per user in the UK (range, 1–501).

    Expressions of concern

    When labelling tweets that expressed concerns, the binary classifier (stage one) achieved a recall of 0.97 and a precision of 0.90. This indicates that the binary classifier missed 3% of tweets that were manually labelled as having expressed a concern, and 10% of tweets labelled as having expressed a concern were manually labelled otherwise. Because the multiclass classifier (at stage two) identified a proportion of these mislabelled tweets in the second round, the overall rate of error in stage one was within 5% of the correct proportion (online supplementary material, Section 4).

    The proportion of tweets posted about HPV vaccines from users in the three countries expressing concerns was 18.7% (8215 of 43 852 tweets), but there were differences in these proportions across the three countries (table 2). Canada had the lowest proportion of tweets expressing concerns at 14.9% (2818 of 18 927 tweets), followed by Australia at 19.3% (1388 of 7173 tweets). The UK had the highest proportion of tweets expressing concerns at 22.6% (4009 of 17 752 tweets). Tweets expressing concerns also tended to have smaller audiences compared with tweets not expressing concern about HPV vaccines (online supplementary material, Section 3).

    Types of concerns expressed

    When identifying concerns related to cues to action, the classifier respectively produced a precision of 0.81 and a recall of 0.74. For perceived barriers, the precision was 0.91, and the recall was 0.92. The classifier was less reliable for the remainder of the concern groups because these types of concerns made up a much smaller proportion, resulting in imbalance in the data, which affects the performance that can be achieved by the classifiers (online supplementary material, Section 5).

    Tweets expressing concerns about perceived barriers comprised the largest type of concern by both the proportion of tweets expressing concerns (table 3). The proportions of each group of concerns across the three countries were generally consistent.

    View this table:
    Table 3

    Number of tweets by country and concern type

    Social connections among users

    Among users from the three countries who posted about HPV vaccines, 18.2% (3062 of 16 789) were labelled as having expressed concerns (at least half of the tweets about HPV vaccines they posted were labelled as having expressed a concern). The total number of follower connections among the set of 16 789 users was 502 629. Users from the three countries were disproportionately more likely to be followed by users from the same country, creating clusters of users by country (figure 1). Furthermore, users who expressed concerns about the HPV vaccines appear to be more tightly connected within the UK, compared with either Australia or Canada. Figure 1 also highlights that users discussing HPV vaccines in the UK are more often connected to users in Australia and Canada than users in Australia and Canada are connected to each other.

    Figure 1
    Figure 1

    The follower network for Twitter users posting about HPV vaccines is coloured by country (Australia, green; Canada, red; UK, blue). Each node represents a user, and the node sizes are proportional to the number of followers within the user’s network. Nodes are positioned by heuristic to be closer to nodes with which they are better connected, as a way of illustrating the community structure. Darker coloured nodes indicate users for whom at least 50% of their relevant tweets expressed concerns.

    To examine the proportion of followers of HPV vaccine tweets, figure 2 examines ‘concern’ and ‘non-concern’ tweets for each of the three countries (to produce six groups represented as circles). Relative to users who did not express concern about HPV, users that did express concerns had a higher proportion of international followers who also expressed concerns (figure 2). Among UK users expressing concerns, 26.1% of followers also expressed concerns (compared with 9.1% of followers among UK users not expressing concerns). Also among UK users expressing concerns, 28.0% of their followers also expressed concerns and were from Australia or Canada, and 9.9% of their followers did not express concerns and were from Australia or Canada. In comparison, among UK users not expressing concerns, only 5.8% of their followers were users not expressing concerns and from Australia or Canada, and 8.3% of their followers were users expressing concerns and from Australia or Canada (online supplementary material, Section 6). This pattern was consistent across each of the three countries. The results indicate that users who mostly expressed concerns were disproportionately well-connected to international users discussing HPV vaccines.

    Figure 2
    Figure 2

    The percentages of followers for all users by expression of concern from Australia, Canada and the UK. The circle represents a concern group of Twitter users, where the circle size is proportional to the number of users. The arrow represents the user following direction. The number represents the percentage of followers, where the number in a circle represents the percentage of followers from the same concern group. Only values above 1.5% are shown (online supplementary material, Section 5) for all values.

    Discussion

    This study found that in Australia, Canada and the UK, nearly 1 in 5 of the tweets about HPV vaccines was an expression of concern. Canadian Twitter users less often expressed concerns about HPV vaccines (14.9%) compared with Australia (19.3%) and the UK (22.6%) (table 2). There was a general consistency in the proportions of specific concerns across the three countries, and the most common concerns (46%) were related to ‘perceived barriers’ (ie, logistical challenges and psychological barriers such as perceived harms of receiving the HPV vaccine) (table 3). The results demonstrated that users expressing concerns about HPV vaccines tended to be relatively well-connected to users discussing HPV vaccine concerns in other countries, especially between Canada and the UK.

    Previous studies examining the representation of HPV vaccines on Twitter identified slightly higher proportions of negative tweets or tweets expressing concerns, but these studies captured different time periods and did not compare specific countries.52 56 For example, a study of 6 months of Twitter data in the United States (between October 2013 and April 2014) found that 25.1% of tweets were negative.51 Though greater research is required, the balance of positive and negative content appears to vary by source whereby the majority of news content,37 online comments (in response to news articles)42 and tweets have been found to be positive; the majority of YouTube content has been found to be negative.47 In examining the type of concern expressed about the HPV vaccine on other social media sites, researchers have also observed the predominance of perceived barrier (ie, logistical challenges and psychological barriers such as perceived harms of receiving the HPV vaccine).44 45 69 However, while the present research study found that concerns about safety were most common on Twitter, other research found safety to be surpassed or similar in salience to other prevalent themes including conspiracies/search for truth, mistrust for health system and promoting promiscuity.44 45 69 Surian et al analysed topics regarding HPV vaccines on Twitter and found that individuals who posted about ‘harms and conspiracies’ posted more often than other users, suggesting that some users are actively seeking to introduce concerns about HPV vaccines into the public domain.54 The predominance of HPV vaccine concerns about perceived barrier on Twitter indicates the importance of these concerns. It would be valuable to extend this work to examine differences in general vaccine concerns as well as compare concerns towards specific vaccines on Twitter.

    International networking on Twitter suggests that vaccine related controversies in one country could reverberate around the world and impact vaccine coverage. Public health professionals and policy-makers must therefore be able to monitor, rapidly identify and react to such concerns (eg, by providing evidence-based responses in real-time and strengthening their own international networks).34 50 70 This research provides public health practitioners and policy-makers with evidence that concerns about ‘perceived barriers’ on Twitter are widespread; effective communication campaigns could be designed and implemented to target this concern in locations where it is likely to have the greatest impact. However, it is important for further research to analyse results by type of sender. It would also be critical for future research to design and evaluate appropriate messaging of such a campaign so that this intervention does not ‘backfire’ and increase hesitancy.71

    This study also found that Twitter users expressing HPV vaccine concerns tended to have higher proportions of international connections compared with those not expressing concerns. Given the international connection between Twitter users who express concerns, public health organisations seeking to improve the uptake of HPV vaccines may benefit from tools that help them monitor the impact of vaccine scares on social media locally as well as in other countries in order to pre-empt and respond to misinformation. Greater research would be helpful to further investigate how public health organisations can monitor and intervene to address vaccine concerns. Such support could have been beneficial for Japan and Colombia when the media had a detrimental impact on HPV vaccine coverage.11 30 31

    Similar to other studies in this area, our research did not measure whether the expression of concerns on Twitter led to changes in decision-making and coverage. Further research would be beneficial to assess the pathway of HPV vaccine concerns, and whether such concerns have a real-world impact (eg, on vaccine coverage). In particular, future work should consider the relationship between the information about HPV vaccines that enters into the public discourse and the decision-making of individuals and populations. While Canada had the lowest proportion of tweets in which concerns were expressed of the three countries, it also has the lowest rate of HPV vaccine uptake.14 15 18 19 72 Accordingly, we echo Gollust et al’s recent call for greater experimental research designs to make causal assertions about the impact of the media on vaccine coverage.73 Although some studies have begun to do so,36 74 it would be helpful for future research to specifically evaluate the impact of Twitter messages and for moderating variables to also be evaluated.

    Together with other studies on the representation of HPV vaccines in the media, our results suggest that it would be useful to monitor early indications of negative influence on attitudes and beliefs on social media. Two studies have independently examined responses on Twitter to specific controversial events including US Representative Michele Bachmann’s claim that HPV vaccines could cause ‘mental retardation’ and Katie Couric’s television segment ‘HPV Vaccine Controversy’ that aired on 4 December 2013.53 75 Mahoney et al evaluated 200 social media posts before and after Bachmann’s comments on the Today Show and found that though most media was positive in tone, compared with Google News, Twitter disseminated more positive HPV vaccine articles and also used more personal accounts as a reference source.53 In contrast, using a random sample of 3595 tweets, Bahk et al found that most sentiment on Twitter towards HPV vaccines before Katie Couric’s episode was negative, and while there was a decrease of negative sentiment immediately after the show aired, negative sentiment returned to baseline after 2 weeks.75 Future research should also investigate how public health organisations should effectively intervene to curb misinformation or ‘fake news’ regarding HPV vaccination.

    There were several limitations to this study. First, the findings are specific to Twitter, and while Twitter represents one of the largest populations of social media users, the results are not necessarily representative of the broader public discourse about HPV vaccines in news and online social media.76–79 Twitter is an inherently biased representation of the broader population and is skewed both in terms of age and socioeconomics.80–83 Second, while the location inference method is a standard in the area,84 85 the methods are imperfect.86–89 Third, the study was limited to English-language tweets in three countries, and evaluations of other countries and other languages may have yielded different results. It would be beneficial for future research to expand the focus of analysis to examine diverse countries, as well as conduct more nuanced regional explorations of a single country. Finally, using networks based on which users follow each other does not necessarily capture all of the interactions that occur online. While some argue that interaction with content (liking or retweeting) is a better measure of impact than followers,90 others have argued that many users on Twitter are passive and do not interact with the content,91 and as such, followers may be a better indicator of impact. As this study examined follower networks, it would be helpful for future research to compare followers to different ways of interacting with content in order to better understand the impact of HPV vaccine tweets.

    Conclusions

    This study characterised the concerns about HPV vaccines expressed by Twitter users in three countries. The UK had the greatest proportion of tweets expressing concerns about HPV vaccines, and Canada had the least, and the types of concerns expressed were relatively consistent across the three countries. Users who expressed concerns about HPV vaccines were generally more closely connected to users in other countries who also expressed concerns, suggesting that controversies and misinformation may be rapidly shared across international boundaries. This research could be used to design public health interventions that address concerns about the HPV vaccine on Twitter. In particular, this study suggests that methods for addressing vaccine concerns may benefit from targeting concerns about perceived barriers to vaccination (including logistical challenges and psychological barriers such as vaccine pain, safety and side effects as a consequence of receiving the HPV vaccine). In addition, further coordination of public health agencies internationally may mitigate vaccine scares.

    References

    1. 1.↵
      2. Trottier H ,
      3. Franco EL
      . The epidemiology of genital human papillomavirus infection. Vaccine 2006;24:S1–15.doi:10.1016/j.vaccine.2005.09.054
      OpenUrlCrossRefPubMedWeb of Science
    2. 2.↵
      2. Tota JE ,
      3. Chevarie-Davis M ,
      4. Richardson LA , et al
      . Epidemiology and burden of HPV infection and related diseases: implications for prevention strategies. Prev Med 2011;53:S12–21.doi:10.1016/j.ypmed.2011.08.017
      OpenUrlCrossRefPubMed
    3. 3.↵
      2. Plummer M ,
      3. de Martel C ,
      4. Vignat J , et al
      . Global burden of cancers attributable to infections in 2012: a synthetic analysis. Lancet Glob Health 2016;4:e609–16.doi:10.1016/S2214-109X(16)30143-7
      OpenUrl
    4. 4.↵
      2. Forman D ,
      3. de Martel C ,
      4. Lacey CJ , et al
      . Global burden of human papillomavirus and related diseases. Vaccine 2012;30:F12–23.doi:10.1016/j.vaccine.2012.07.055
      OpenUrlCrossRefPubMedWeb of Science
    5. 5.↵
      National Advisory Committee on Immunization. 2012. Update On Human Papillomavirus Vaccines. Canada Communicable Disease. Report. 38.
    6. 6.↵
      Canadian Cancer Society. Canadian Cancer Statistics, 2016.
    7. 7.↵
      2. Markowitz LE ,
      3. Dunne EF ,
      4. Saraiya M , et al
      . Human papillomavirus vaccination: recommendations of the Advisory Committee on Immunization Practices. MMWR. Recommendations and reports: Morbidity and mortality weekly report. Recommendations and reports / Centers for Disease Control 2014;63:1–30.
      OpenUrl
    8. 8.↵
      2. Ferris D ,
      3. Samakoses R ,
      4. Block SL , et al
      . Long-term study of a quadrivalent human papillomavirus vaccine. Pediatrics 2014;134:e657–65.doi:10.1542/peds.2013-4144
      OpenUrlAbstract/FREE Full Text
    9. 9.↵
      2. Garland SM
      . The Australian experience with the human papillomavirus vaccine. Clin Ther 2014;36:17–23.doi:10.1016/j.clinthera.2013.12.005
      OpenUrlCrossRefPubMed
    10. 10.↵
      2. Brotherton JM ,
      3. Saville AM ,
      4. May CL , et al
      . Human papillomavirus vaccination is changing the epidemiology of high-grade cervical lesions in Australia. Cancer Causes Control 2015;26:953–4.doi:10.1007/s10552-015-0568-6
      OpenUrlCrossRefPubMed
    11. 11.↵
      2. Brotherton JML ,
      3. Bloem PJN
      . HPV vaccination: current global status. Curr Obstet Gynecol Rep 2015;4:220–33.doi:10.1007/s13669-015-0136-9
      OpenUrl
    12. 12.↵
      2. Drolet M ,
      3. Bénard É ,
      4. Boily MC , et al
      . Population-level impact and herd effects following human papillomavirus vaccination programmes: a systematic review and meta-analysis. Lancet Infect Dis 2015;15:565–80.doi:10.1016/S1473-3099(14)71073-4
      OpenUrlCrossRefPubMed
    13. 13.↵
      National Health Performance Authority. Healthy Communities: HPV immunisation rates for girls in 2013 (In Focus). 2015.
    14. 14.↵
      National HPV Vaccination Program Register. National (Australia) HPV 3 dose vaccination coverage for females turning 15 years of age in 2015. 2016.
    15. 15.↵
      National HPV Vaccination Program Register. National (Australia) HPV 3 dose vaccination coverage for males turning 15 years of age in 2015. 2016.
    16. 16.↵
      2. Shapiro GK ,
      3. Perez S ,
      4. Rosberger Z
      . Including males in Canadian human papillomavirus vaccination programs: a policy analysis. CMAJ 2016;188:881–6.doi:10.1503/cmaj.150451
      OpenUrlFREE Full Text
    17. 17.↵
      2. Perez S ,
      3. Shapiro GK ,
      4. Brown CA , et al
      . ‘I didn’t even know boys could get the vaccine’: parents’ reasons for human papillomavirus (HPV) vaccination decision making for their sons. Psychooncology 2015:1316–23.doi:10.1002/pon.3894
    18. 18.↵
      2. Perez S ,
      3. Tatar O ,
      4. Shapiro GK , et al
      . Psychosocial determinants of parental human papillomavirus (HPV) vaccine decision-making for sons: methodological challenges and initial results of a pan-Canadian longitudinal study. BMC Public Health 2016;16:1223.doi:10.1186/s12889-016-3828-9
      OpenUrl
    19. 19.↵
      2. Gilbert NL ,
      3. Gilmour H ,
      4. Dubé È , et al
      . Estimates and determinants of HPV non-vaccination and vaccine refusal in girls 12 to 14 y of age in Canada: results from the Childhood National Immunization Coverage Survey, 2013. Hum Vaccin Immunother 2016;12:1484–90.doi:10.1080/21645515.2016.1153207
      OpenUrl
    20. 20.↵
      2. Vishram B ,
      3. Byrne L ,
      4. White J , et al
      . Human Papillomavirus (HPV) vaccination coverage in adolescent females in England: 2014/15. London, England: Public Health England, 2015.
    21. 21.↵
      Australian Government Department of Health. Australian Immunisation Register: Current Data. In: Program IA, ed: Commonwealth of Australia; 2017.
    22. 22.↵
      Screening & Immunisations Team-NHS Digital. NHS Immunisation Statistics England 2015-16 , 2016.
    23. 23.↵
      Government of Canada. Vaccine coverage in Canadian children: highlights from the 2013 childhood National Immunization Coverage Survey (cNICS). 2016.
    24. 24.↵
      2. Holman DM ,
      3. Benard V ,
      4. Roland KB , et al
      . Barriers to human papillomavirus vaccination among US adolescents: a systematic review of the literature. JAMA Pediatr 2014;168:76–82.doi:10.1001/jamapediatrics.2013.2752
      OpenUrl
    25. 25.↵
      2. Kumar D ,
      3. Chandra R ,
      4. Mathur M , et al
      . Vaccine hesitancy: understanding better to address better. Isr J Health Policy Res 2016;5.doi:10.1186/s13584-016-0062-y
    26. 26.↵
      2. Mason BW ,
      3. Donnelly PD
      . Impact of a local newspaper campaign on the uptake of the measles mumps and rubella vaccine. J Epidemiol Community Health 2000;54:473–4.doi:10.1136/jech.54.6.473
      OpenUrlFREE Full Text
    27. 27.↵
      2. Gangarosa EJ ,
      3. Galazka AM ,
      4. Wolfe CR , et al
      . Impact of anti-vaccine movements on pertussis control: the untold story. Lancet 1998;351:356–61.doi:10.1016/S0140-6736(97)04334-1
      OpenUrlCrossRefPubMedWeb of Science
    28. 28.↵
      2. Larson HJ ,
      3. Jarrett C ,
      4. Schulz WS , et al
      . Measuring vaccine hesitancy: the development of a survey tool. Vaccine 2015;33:4165–75.doi:10.1016/j.vaccine.2015.04.037
      OpenUrl
    29. 29.↵
      2. Gollust SE ,
      3. Attanasio L ,
      4. Dempsey A , et al
      . Political and news media factors shaping public awareness of the HPV vaccine. Womens Health Issues 2013;23:e143–51.doi:10.1016/j.whi.2013.02.001
      OpenUrlCrossRefPubMedWeb of Science
    30. 30.↵
      2. Hanley SJ ,
      3. Yoshioka E ,
      4. Ito Y , et al
      . HPV vaccination crisis in Japan. Lancet 2015;385:2571.doi:10.1016/S0140-6736(15)61152-7
      OpenUrlCrossRefPubMed
    31. 31.↵
      2. Tiempo E
      . Caso de niñas del Carmen de Bolívar desplomó vacunación contra el VPH. 2015 http://www.eltiempo.com/estilo-de-vida/salud/vacuna-contra-el-papiloma-humano-cayo-58-puntos-porcentuales/15246061
    32. 32.↵
      2. Gabarron E ,
      3. Serrano JA ,
      4. Wynn R , et al
      . Tweet content related to sexually transmitted diseases: no joking matter. J Med Internet Res 2014;16:e228.doi:10.2196/jmir.3259
      OpenUrl
    33. 33.↵
      2. Britt RK ,
      3. Hatten KN ,
      4. Chappuis SO
      . Perceived behavioral control, intention to get vaccinated, and usage of online information about the human papillomavirus vaccine. Health Psychol Behav Med 2014;2:52–65.doi:10.1080/21642850.2013.869175
      OpenUrl
    34. 34.↵
      2. Amicizia D ,
      3. Domnich A ,
      4. Gasparini R , et al
      . An overview of current and potential use of information and communication technologies for immunization promotion among adolescents. Hum Vaccin Immunother 2013;9:2634–42.doi:10.4161/hv.26010
      OpenUrl
    35. 35.↵
      2. Betsch C ,
      3. Renkewitz F ,
      4. Betsch T , et al
      . The influence of vaccine-critical websites on perceiving vaccination risks. J Health Psychol 2010;15:446–55.doi:10.1177/1359105309353647
      OpenUrlCrossRefPubMedWeb of Science
    36. 36.↵
      2. Nan X ,
      3. Madden K
      . HPV vaccine information in the blogosphere: how positive and negative blogs influence vaccine-related risk perceptions, attitudes, and behavioral intentions. Health Commun 2012;27:829–36.doi:10.1080/10410236.2012.661348
      OpenUrlCrossRefPubMedWeb of Science
    37. 37.↵
      2. Perez S ,
      3. Fedoruk C ,
      4. Shapiro GK , et al
      . Giving boys a shot: the HPV vaccine’s portrayal in Canadian newspapers. Health Commun 2016;31:1527–38.doi:10.1080/10410236.2015.1089466
      OpenUrlCrossRefPubMed
    38. 38.↵
      2. Robbins SC ,
      3. Pang C ,
      4. Leask J
      . Australian newspaper coverage of human papillomavirus vaccination, October 2006–December 2009. J Health Commun 2012;17:149–59.doi:10.1080/10810730.2011.585700
      OpenUrlCrossRefPubMedWeb of Science
    39. 39.↵
      2. Krieger JL ,
      3. Katz ML ,
      4. Eisenberg D , et al
      . Media coverage of cervical cancer and the HPV vaccine: implications for geographic health inequities. Health Expect 2013;16:e1–e12.doi:10.1111/j.1369-7625.2011.00721.x
      OpenUrlCrossRefPubMedWeb of Science
    40. 40.↵
      2. Quintero Johnson J ,
      3. Sionean C ,
      4. Scott AM
      . Exploring the presentation of news information about the HPV vaccine: a content analysis of a representative sample of U.S. newspaper articles. Health Commun 2011;26:491–501.doi:10.1080/10410236.2011.556080
      OpenUrlCrossRefPubMedWeb of Science
    41. 41.↵
      2. Habel MA ,
      3. Liddon N ,
      4. Stryker JE
      . The HPV vaccine: a content analysis of online news stories. J Womens Health 2009;18:401–7.doi:10.1089/jwh.2008.0920
      OpenUrlCrossRefWeb of Science
    42. 42.↵
      2. Feinberg Y ,
      3. Pereira JA ,
      4. Quach S , et al
      . Understanding public perceptions of the HPV vaccination based on online comments to Canadian news articles. PLoS One 2015;10:e0129587.doi:10.1371/journal.pone.0129587
    43. 43.↵
      2. Coloma PM ,
      3. Becker B ,
      4. Sturkenboom MC , et al
      . Evaluating social media networks in medicines safety surveillance: two case studies. Drug Saf 2015;38:921–30.doi:10.1007/s40264-015-0333-5
      OpenUrl
    44. 44.↵
      2. Penţa MA ,
      3. Băban A
      . Dangerous agent or saviour? HPV vaccine representations on online discussion forums in Romania. Int J Behav Med 2014;21:20–8.doi:10.1007/s12529-013-9340-z
      OpenUrlCrossRefPubMed
    45. 45.↵
      2. Keelan J ,
      3. Pavri V ,
      4. Balakrishnan R , et al
      . An analysis of the human papilloma virus vaccine debate on MySpace blogs. Vaccine 2010;28:1535–40.doi:10.1016/j.vaccine.2009.11.060
      OpenUrlCrossRefPubMedWeb of Science
    46. 46.↵
      2. Ache KA ,
      3. Wallace LS
      . Human papillomavirus vaccination coverage on YouTube. Am J Prev Med 2008;35:389–92.doi:10.1016/j.amepre.2008.06.029
      OpenUrlCrossRefPubMed
    47. 47.↵
      2. Briones R ,
      3. Nan X ,
      4. Madden K , et al
      . When vaccines go viral: an analysis of HPV vaccine coverage on YouTube. Health Commun 2012;27:478–85.doi:10.1080/10410236.2011.610258
      OpenUrlCrossRefPubMedWeb of Science
    48. 48.↵
      2. Gunasekaran B ,
      3. Jayasinghe Y ,
      4. Fenner Y , et al
      . Knowledge of human papillomavirus and cervical cancer among young women recruited using a social networking site. Sex Transm Infect 2013;89:327–9.doi:10.1136/sextrans-2012-050612
      OpenUrlAbstract/FREE Full Text
    49. 49.↵
      2. McNeill A ,
      3. Harris PR ,
      4. Briggs P
      . Twitter influence on UK vaccination and antiviral uptake during the 2009 H1N1 pandemic. Front Public Health 2016;4:26.doi:10.3389/fpubh.2016.00026
      OpenUrl
    50. 50.↵
      2. Bahk CY ,
      3. Cumming M ,
      4. Paushter L , et al
      . Publicly available online tool facilitates real-time monitoring of vaccine conversations and sentiments. Health Aff 2016;35:341–7.doi:10.1377/hlthaff.2015.1092
      OpenUrlAbstract/FREE Full Text
    51. 51.↵
      2. Dunn AG ,
      3. Surian D ,
      4. Leask J , et al
      . Mapping information exposure on social media to explain differences in HPV vaccine coverage in the United States. Vaccine 2017;35:3033–40.doi:10.1016/j.vaccine.2017.04.060
      OpenUrl
    52. 52.↵
      2. Zhou X ,
      3. Coiera E ,
      4. Tsafnat G , et al
      . Using social connection information to improve opinion mining: identifying negative sentiment about HPV vaccines on Twitter. Stud Health Technol Inform 2015;216:761–5.
      OpenUrl
    53. 53.↵
      2. Mahoney LM ,
      3. Tang T ,
      4. Ji K , et al
      . The digital distribution of public health news surrounding the human papillomavirus vaccination: a longitudinal infodemiology study. JMIR Public Health Surveill 2015;1:e2.doi:10.2196/publichealth.3310
      OpenUrl
    54. 54.↵
      2. Surian D ,
      3. Nguyen DQ ,
      4. Kennedy G , et al
      . Characterizing Twitter discussions about HPV vaccines using topic modeling and community detection. J Med Internet Res 2016;18:e232.doi:10.2196/jmir.6045
      OpenUrl
    55. 55.↵
      2. Salathé M ,
      3. Vu DQ ,
      4. Khandelwal S , et al
      . The dynamics of health behavior sentiments on a large online social network. EPJ Data Sci 2013;2.doi:10.1140/epjds16
    56. 56.↵
      2. Dunn AG ,
      3. Leask J ,
      4. Zhou X , et al
      . Associations between exposure to and expression of negative opinions about human papillomavirus vaccines on social media: an observational study. J Med Internet Res 2015;17:e144.doi:10.2196/jmir.4343
      OpenUrlCrossRefPubMed
    57. 57.↵
      2. Rosenstock IM
      . Historical origins of the Health Belief Model. Health Educ Monogr 1974;2:328–35.doi:10.1177/109019817400200403
      OpenUrlCrossRefWeb of Science
    58. 58.↵
      2. Donadiki EM ,
      3. Jiménez-García R ,
      4. Hernández-Barrera V , et al
      . Health Belief Model applied to non-compliance with HPV vaccine among female university students. Public Health 2014;128:268–73.doi:10.1016/j.puhe.2013.12.004
      OpenUrl
    59. 59.↵
      2. Krawczyk A ,
      3. Knäuper B ,
      4. Gilca V , et al
      . Parents’ decision-making about the human papillomavirus vaccine for their daughters: I. Quantitative results. Hum Vaccin Immunother 2015;11:322–9.doi:10.1080/21645515.2014.1004030
      OpenUrl
    60. 60.↵
      2. Larson H ,
      3. Leask J ,
      4. Aggett S , et al
      . A multidisciplinary research agenda for understanding vaccine-related decisions. Vaccines 2013;1:293–304.doi:10.3390/vaccines1030293
      OpenUrl
    61. 61.↵
      2. Casciotti DM ,
      3. Smith KC ,
      4. Tsui A , et al
      . Discussions of adolescent sexuality in news media coverage of the HPV vaccine. J Adolesc 2014;37:133–43.doi:10.1016/j.adolescence.2013.11.004
      OpenUrlCrossRefPubMedWeb of Science
    62. 62.↵
      2. Casciotti DM ,
      3. Smith KC ,
      4. Andon L , et al
      . Print news coverage of school-based human papillomavirus vaccine mandates. J Sch Health 2014;84:71–81.doi:10.1111/josh.12126
      OpenUrl
    63. 63.↵
      2. Madden K ,
      3. Nan X ,
      4. Briones R , et al
      . Sorting through search results: a content analysis of HPV vaccine information online. Vaccine 2012;30:3741–6.doi:10.1016/j.vaccine.2011.10.025
      OpenUrlCrossRefPubMedWeb of Science
    64. 64.↵
      2. Nan X ,
      3. Daily K
      . Biased assimilation and need for closure: examining the effects of mixed blogs on vaccine-related beliefs. J Health Commun 2015;20:462–71.doi:10.1080/10810730.2014.989343
      OpenUrlPubMed
    65. 65.↵
      2. Thomson A ,
      3. Robinson K ,
      4. Vallée-Tourangeau G
      . The 5As: a practical taxonomy for the determinants of vaccine uptake. Vaccine 2016;34:1018–24.doi:10.1016/j.vaccine.2015.11.065
      OpenUrl
    66. 66.↵
      2. Wolfe RM ,
      3. Sharp LK ,
      4. Lipsky MS
      . Content and design attributes of antivaccination web sites. JAMA 2002;287:3245–8.doi:10.1001/jama.287.24.3245
      OpenUrlCrossRefPubMedWeb of Science
    67. 67.↵
      2. Keelan J ,
      3. Pavri-Garcia V ,
      4. Tomlinson G , et al
      . YouTube as a source of information on immunization: a content analysis. JAMA 2007;298:2481–4.doi:10.1001/jama.298.21.2482
      OpenUrlCrossRefPubMed
    68. 68.↵
      2. Abdelmutti N ,
      3. Hoffman-Goetz L
      . Risk messages about HPV, cervical cancer, and the HPV vaccine Gardasil: a content analysis of Canadian and U.S. national newspaper articles. Women Health 2009;49:422–40.doi:10.1080/03630240903238776
      OpenUrlCrossRef
    69. 69.↵
      2. Kata A
      . A postmodern Pandora’s box: anti-vaccination misinformation on the Internet. Vaccine 2010;28:1709–16.doi:10.1016/j.vaccine.2009.12.022
      OpenUrlCrossRefPubMedWeb of Science
    70. 70.↵
      2. Larson HJ ,
      3. Wilson R ,
      4. Hanley S , et al
      . Tracking the global spread of vaccine sentiments: the global response to Japan’s suspension of its HPV vaccine recommendation. Hum Vaccin Immunother 2014;10:2543–50.doi:10.4161/21645515.2014.969618
      OpenUrlCrossRefPubMed
    71. 71.↵
      2. Nyhan B ,
      3. Reifler J ,
      4. Richey S , et al
      . Effective messages in vaccine promotion: a randomized trial. Pediatrics 2014;133:e835–e842.doi:10.1542/peds.2013-2365
      OpenUrlAbstract/FREE Full Text
    72. 72.↵
      2. Vishram B ,
      3. Byrne L ,
      4. White J , et al
      . Human Papillomavirus (HPV) vaccination coverage in adolescent females in England: 2014/15. London, England: Public Health England, 2015.
    73. 73.↵
      2. Gollust SE ,
      3. LoRusso SM ,
      4. Nagler RH , et al
      . Understanding the role of the news media in HPV vaccine uptake in the United States: synthesis and commentary. Hum Vaccin Immunother 2016;12:1430–4.doi:10.1080/21645515.2015.1109169
      OpenUrlPubMed
    74. 74.↵
      2. Betsch C
      . Social media targeting of health messages. A promising approach for research and practice. Hum Vaccin Immunother 2014;10:2636–7.doi:10.4161/hv.32234
      OpenUrl
    75. 75.↵
      2. Bahk CY ,
      3. Cumming M ,
      4. Paushter L , et al
      . Publicly available online tool facilitates real-time monitoring of vaccine conversations and sentiments. Health Aff 2016;35:341–7.doi:10.1377/hlthaff.2015.1092
      OpenUrlAbstract/FREE Full Text
    76. 76.↵
      2. Weeg C ,
      3. Schwartz HA ,
      4. Hill S , et al
      . Using Twitter to measure public discussion of diseases: a case study. JMIR Public Health Surveill 2015;1:e6.doi:10.2196/publichealth.3953
      OpenUrl
    77. 77.↵
      Pew Research Center. Social networking fact sheet. http://www.pewinternet.org/fact-sheets/social-networking-fact-sheet/
    78. 78.↵
      2. Radzikowski J ,
      3. Stefanidis A ,
      4. Jacobsen KH , et al
      . The measles vaccination narrative in Twitter: a quantitative analysis. JMIR Public Health Surveill 2016;2:e1.doi:10.2196/publichealth.5059
      OpenUrl
    79. 79.↵
      2. Paul MJ ,
      3. Dredze M
      . You are what you tweet: analyzing Twitter for public health. Association for the Advancement of Artificial Intelligence, 2011.
    80. 80.↵
      2. Sadah SA ,
      3. Shahbazi M ,
      4. Wiley MT , et al
      . A study of the demographics of web-based health-related social media users. J Med Internet Res 2015;17:e194.doi:10.2196/jmir.4308
      OpenUrl
    81. 81.↵
      2. Mislove A ,
      3. Lehmann S ,
      4. Ahn Y-Y , et al.
      , 2011. Understanding the demographics of Twitter users. Proceedings of the 5th International AAAI Conference on Weblogs and Social Media
    82. 82.↵
      2. Sloan L ,
      3. Morgan J ,
      4. Burnap P , et al
      . Who tweets? Deriving the demographic characteristics of age, occupation and social class from Twitter user meta-data. PLoS One 2015;10:e0115545.doi:10.1371/journal.pone.0115545
    83. 83.↵
      2. Sloan L ,
      3. Morgan J
      . Who tweets with their location? Understanding the relationship between demographic characteristics and the use of geoservices and geotagging on Twitter. PLoS One 2015;10:e0142209.doi:10.1371/journal.pone.0142209
    84. 84.↵
      2. Kumar S ,
      3. Morstatter F ,
      4. Marshall G , et al.
      , 2012. Navigating Information Facets on Twitter (NIF-T). Proceedings of the 18th ACM International Conference on Knowledge Discovery and Data Mining (SIGKDD)
    85. 85.↵
      2. Kounadi O ,
      3. Lampoltshammer TJ ,
      4. Groff E , et al
      . Exploring Twitter to analyze the public’s reaction patterns to recently reported homicides in London. PLoS One 2015;10:e0121848.doi:10.1371/journal.pone.0121848
    86. 86.↵
      2. Elmongui HG ,
      3. Morsy H ,
      4. Mansour R
      , 2015. Inference models for Twitter user’s home location prediction. IEEE/ACS International Conference of Computer Systems and Applications
    87. 87.↵
      2. Bo H ,
      3. Paul C ,
      4. Timothy B
      . Text-based Twitter user geolocation prediction. J AI Research 2014;49:451–500.
      OpenUrl
    88. 88.↵
      2. Jurgens D
      , 2013. That’s what friends are for: inferring location in online social media platforms based on social relationships. Proceedings of the AAAI International Conference on Weblogs and Social Media (ICWSM)
    89. 89.↵
      2. Cheng Z ,
      3. Caverlee J ,
      4. Lee K
      , 2010. You are where you tweet: a content-based approach to geo-locating Twitter users. Proceedings of the 19th ACM International Conference on Information and Knowledge Management
    90. 90.↵
      2. Cha M ,
      3. Haddadi H ,
      4. Benevenuto F , et al
      . Measuring user influence in Twitter: the million follower fallacy Association for the Advancement of Artificial Intelligence. 2010.
    91. 91.↵
      2. Romero DM ,
      3. Galuba W ,
      4. Asur S
      . Influence and passivity in social media. In: Gunopulos D , Hofmann T , Malerba D , Vazirgiannis M , et al . eds. Machine Learning and Knowledge Discovery in Databases, 2011:18–33.

    Footnotes

    • Contributors GKS, AGD and MK conceptualised and designed the study. DS and AGD acquired the data. GKS drafted the manuscript. All authors analysed and interpreted the data, critically revised the manuscript for important intellectual content and approved the final version.

    • Funding GKS was supported by the CIHR Vanier Canada Graduate Scholarship and Queen Elizabeth II Diamond Jubilee Scholarship programmes. AD is an investigator on a related NHMRC Project Grant (APP1128968). The funding source had no role in the design of the study; collection, management, analysis and interpretation of the data; preparation, review or approval of the manuscript and the decision to submit the manuscript for publication.

    • Competing interests None declared.

    • Ethics approval Macquarie University Human Research Ethics Committee (number 5201401028) and the University of Melbourne’s Research Ethics Board (number 1647488.1).

    • Provenance and peer review Not commissioned; externally peer reviewed.

    • Data sharing statement No additional data are available. Due to ethical approvals and the Twitter Terms of Service, the individual identifiers for Twitter posts and users cannot be provided with their manual or automatic labels.