PaperlayerView exact revision on arXiv

Source-linked AI summary

Tweet for Behavior Change: Using Social Media for the Dissemination of Public Health Messages

Aisling Gough, Ruth F Hunter, Oluwaseun Ajao, Anna Jurek, Gary McKeown, Jun Hong, Eimear Barrett, Marbeth Ferguson, Gerry McElwee, Miriam McCarthy, Frank Kee

arXiv:1703.08813v1cs.SI

TL;DR

Evidence was limited on how to develop and evaluate social-media health-promotion campaigns and define their success. This study tested a skin-cancer-prevention intervention and found trends toward improved knowledge and attitudes, with message framing affecting engagement.

  • Problem

    Evidence is limited on how to develop, implement, and evaluate social-media health-promotion campaigns, including what constitutes success.

  • Method

    The study tested a bespoke social-media skin-cancer-prevention intervention using a quasi-experimental interrupted time-series crossover design with Northern Ireland and Wales comparison areas.

  • Results

    Web-based surveys showed a trend toward improved skin-cancer knowledge and attitudes, while shocking messages produced the greatest impressions and engagements.

  • Takeaways & Limitations

    Social media was a feasible platform for delivering dynamic, tailored public health messages in real time.

  • Takeaways & Limitations

    Contamination across intervention phases and the unrepresentativeness of the Twitter population limit interpretation of the findings.

Abstract

from arXiv · show

Background: Social media public health campaigns have the advantage of tailored messaging at low cost and large reach, but little is known about what would determine their feasibility as tools for inducing attitude and behavior change. Objective: The aim of this study was to test the feasibility of designing, implementing, and evaluating a social media-enabled intervention for skin cancer prevention. Conclusions: Social media-disseminated public health messages reached more than 23% of the Northern Ireland population. A Web-based survey suggested that the campaign might have contributed to improved knowledge and attitudes toward skin cancer among the target population. Findings suggested that shocking and humorous messages generated greatest impressions and engagement, but information-based messages were likely to be shared most. The extent of behavioral change as a result of the campaign remains to be explored, however, the change of attitudes and knowledge is promising. Social media is an inexpensive, effective method for delivering public health messages. However, existing and traditional process evaluation methods may not be suitable for social media.

Introduction · Background

Social media offers public health organizations broad reach, audience data, targeted engagement, and dynamic communication, but its potential for raising awareness and changing behavior remains insufficiently understood. This study addresses these feasibility questions through a mass communication Twitter campaign for skin cancer prevention.

  • Background: Social media applications enable user-generated content, large-audience reach, access to user data, and monitoring of message recipients’ activities.It is defined as Internet-based applications built on Web 2.0 foundations that allow content creation and exchange.
  • Background: Because social media is ubiquitous and widely used to access Web content, it presents an ideal platform for health promotion and organizational communication.The passage notes that more people follow links on social media than access Web-based content through direct searches.
  • Background: Smartphone use has created near-instantaneous access to specialist information, prompting health care organizations to adopt social media and “always on” communication.Public Health England is cited as an example of switching to an “always on” approach using digital technologies.
  • Background: Social networks may influence health behaviors and outcomes, yet public health agencies have not fully harnessed their public health potential.The literature calls for interventions to harness social media’s participatory nature.
  • Background: Health promotion opportunities include targeted messaging, public interaction, reaching hard-to-reach groups, dynamic campaigns, and advice delivered directly to participants.Social connection and discussion opportunities are described as 14 times more effective with social media than with the written word.
  • Background: Although some evidence suggests social network interventions can improve health behavior-related outcomes, research on social media health promotion remains limited.It remains unclear how best to use social media to raise awareness and ultimately trigger behavioral change.
  • Background: Calls for further research reflect unanswered questions about the feasibility of using social media and communication technologies for public health promotion.The cited motivation centers on developing and implementing effective social media interventions.
  • Background: The study addresses these questions by reporting findings from a mass communication Twitter campaign for skin cancer prevention.The campaign provides the study’s empirical context for examining social media use in public health.

Aim … Why Twitter?

The study assessed whether a bespoke social media intervention could disseminate skin cancer prevention messages and influence knowledge and attitudes. It also examined message framing, diffusion strategies, evaluation measures, campaign motivation, and Twitter’s suitability for real-time public-domain data access.

  • Aim: The study tested the feasibility of designing, implementing, and evaluating a bespoke social media-enabled intervention to disseminate skin cancer prevention messages.
  • Research Objectives: The mixed-methods campaign focused on increasing knowledge and attitudes toward care in the sun.
  • Research Objectives: The objectives included assessing message frames, promoted messages, influencers, Thunderclap diffusion, demographic data access, and process evaluation measures.
  • Why a Skin Cancer Campaign?: More than 4000 skin cancer cases are diagnosed annually in Northern Ireland, motivating attention to prevention alongside established campaigns elsewhere.Australia’s “Slip, Slap, Slop” and Cancer Research UK’s “SunSmart” campaigns provided relevant precedents.
  • Why a Skin Cancer Campaign?: 750 respondents established baseline campaign parameters through a postcode-stratified household survey representative of Northern Ireland.Awareness that sun exposure could cause skin cancer was 80.7% (605/750), while awareness that skin cancer could lead to death was 88.9% (667/750).
  • Why a Skin Cancer Campaign?: 49.2% (369/750) considered a suntan healthy, and only 6.4% (48/750) reported frequent skin checks, identifying gaps in sun-safe attitudes.These gaps motivated the Public Health Agency to identify skin cancer as a priority for its regional social media campaign.
  • Why Twitter?: Twitter was chosen because posts are voluntary and public, while its APIs provide real-time access to large volumes of content for studying social media processes.Adoreboard enabled access to Twitter streaming data and preprocessing to minimize processing demands.

Design … Measures

The study used a quasi-experimental interrupted time series with comparison design, implemented through two campaign phases separated by a 2-week washout. Campaign activity was compared between Northern Ireland and Wales using Twitter analytics and predefined search terms, alongside measures of impressions and engagement.

  • Design: The quasi-experimental study used an interrupted time series with comparison design and a cross-over campaign across two Twitter phases.Phase 1 ran May 1–July 14, 2015, followed by a 2-week washout and Phase 2 from August 1–September 30, 2015.
  • Control Group: Northern Ireland received the campaign, while Wales served as a control area for comparing geographically located Twitter activity.Predefined skin cancer and sun-care keyword volumes were compared before and after the campaign in both areas.
  • Intervention Development and Implementation: Phase 1 used the regional cancer charity’s Twitter account to deliver informative, story, shock, humor, and opportunistic or responsive message frames.Messages addressed skin surveillance, care in the sun, and skin cancer prevention.
  • Intervention Development and Implementation: Phase 2 used a bespoke Twitter account to disseminate similar content on skin surveillance, sun care, and cancer prevention.The supplied passage identifies the bespoke account and similar thematic content but does not provide further implementation details.
  • Measures: Twitter analytics were collected before and after the campaign, during implementation, and through streaming data to assess campaign-related search-term use and exposure.The comparison time points were April 2015 and October 2015; dashboards tracked campaign analytics.
  • Measures: Measures included impressions, engagements, engagement rate, likes, and shares, with engagement rate defined as engagements divided by impressions.Impressions counted views, while engagements included clicks, likes, comments, shares, and retweets.
  • Intervention Development and Implementation: A Thunderclap enabled participating users’ accounts to automatically post a common campaign message at midday on September 1, 2015.The intervention used this Web-based “flash-mob” mechanism to coordinate message dissemination.
  • Measures: £10 paid-for promoted posts in both phases targeted Northern Ireland residents aged 18 years or older on a cost-per-click basis.Each promoted tweet remained active for the specified audience until the allocated budget ran out.

Pre- and Postintervention Web-Based Survey · Data Analysis

The study used pre- and postintervention Web-based surveys alongside social media analytics to assess campaign-related differences in usage, demographics, knowledge, attitudes, impressions, engagement, and sharing. Survey reporting followed CHERRIES, while descriptive and cross-tabulated analyses were used without significance testing.

  • Pre- and Postintervention Web-Based Survey: The Web-based survey followed the Checklist for Reporting Results of Internet E-Surveys (CHERRIES).The checklist was explicitly taken into account for reporting the surveys.
  • Pre- and Postintervention Web-Based Survey: Adults aged >18 years in Northern Ireland were invited through social-media advertisements in April and October 2015 to complete a Qualtrics survey.Participants could enter a draw to win an iPad Mini, and the survey took approximately 15-20 minutes.
  • Pre- and Postintervention Web-Based Survey: Paid-for “promoted” tweets were used to reach a wider audience for the survey recruitment.Those who clicked the advertisement were redirected to the Qualtrics survey website.
  • Data Analysis: Pre- and postintervention data were compared for social-media usage, demographics, and knowledge and attitudes toward UV exposure and skin cancer prevention.Responses were summarized using descriptive statistics and cross-tabulations by gender, age, and other sociodemographic characteristics.
  • Data Analysis: Tests of significance were omitted because of the study’s nature and concerns about the appropriateness of applying such tests to feasibility data.The supplied passage states that descriptive frequencies were tabulated and cross-tabulations were used.
  • Data Analysis: Social-media analytics were gathered for each post, message frame, and the overall campaign.Frequencies were used to compare impressions, engagement, and shares across message frames.

Measuring Twitter Analytics · Can a Bespoke Social Media Campaign on Skin Cancer Impact on Attitudes and Knowledge? · Demographic Characteristics

Twitter analytics captured tweet reach, engagement, hashtag use, retweets, and pre- versus postcampaign search-term frequencies. Survey respondents were more numerous postintervention but had similar age, marital-status, and educational-attainment distributions, with women overrepresented and respondents more educated than Northern Ireland’s population.

  • Measuring Twitter Analytics: Twitter Firehose data were used to measure tweet impressions, engagements, hashtag frequency, and message spread through retweets.Metadata were also searched for relevant terms to compare their frequencies before and after the campaign.
  • Measuring Twitter Analytics: Metadata searches compared relevant search-term frequencies before and after the campaign.The retrieved tweets and metadata supported these tabulations.
  • Demographic Characteristics: 429 participants completed the postintervention survey versus 337 who completed the precampaign survey.The survey participant counts were reported for the two campaign time points.
  • Demographic Characteristics: 41% of participants in both surveys were aged 18-29 years, indicating similar age distributions before and after the campaign.Age distributions were described as similar across the two surveys.
  • Demographic Characteristics: Women comprised 84.6% [281/337] of precampaign respondents and 80.4% [345/429] of postintervention respondents.Respondents were more likely to be female in both surveys.
  • Demographic Characteristics: More than half of respondents had similar marital-status and educational-attainment distributions before and after the campaign.The supplied passage states that both distributions were similar pre- and postcampaign.
  • Demographic Characteristics: 51% of Northern Ireland’s population were female and 49% were male according to the 2011 Census, unlike the survey samples’ female predominance.Campaign respondents were also more educated than the Northern Ireland population, where 29% aged 16+ years had no qualifications.
  • Demographic Characteristics: Respondents were of a similar age to Northern Ireland’s population, whose 2011 Census median age was 37 years.The campaign respondents were described as more educated than the population.

Attitudes to UV Exposure and Skin Cancer Prevention … Influence of Message Frames on Social Media

The campaign was associated with improved attitudes toward UV exposure, skin cancer prevention, and prevention knowledge. Message framing influenced social-media performance: shocking and humorous messages drove high impressions or engagement, while informative messages were shared most.

  • Attitudes to UV Exposure and Skin Cancer Prevention: Postcampaign agreement that respondents “like to tan” fell from 60.5% [202/334] to 55.6% [238/428].
  • Attitudes to UV Exposure and Skin Cancer Prevention: Agreement that a tanned person looks more healthy or attractive declined from 55.9% [186/333] to 52.7% [225/427] and from 48.6% [162/333] to 43.7% [186/426], respectively.
  • Attitudes to UV Exposure and Skin Cancer Prevention: Agreement that sun protection can help avoid skin cancer increased from 62.6% [209/334] to 65.0% [278/428].
  • Knowledge of Skin Cancer Prevention: Awareness that skin cancer is the most common cancer increased from 28.4% [95/335] to 39.3% [168/428], while recognition that melanoma is most serious rose from 49.1% [165/336] to 55.5% [238/429].
  • Knowledge of Skin Cancer Prevention: Awareness that the sun’s rays are strongest at midday increased from 91.3% [306/335] to 93.5% [400/428].
  • Influence of Message Frames on Social Media: Among four message frames, shock or disgust generated the most impressions (n=2369), humor the most engagements (n=148), and informative content the most retweets (17 followers).Median results likewise favored shocking messages for impressions (565), engagements (15.5), and retweets (2.5), while humor produced the highest median engagement rate (2.5%).

Influencers … Twitter Analytics

The campaign’s Twitter analytics showed that influencer-inclusive messages produced the strongest impression counts, while opportunistic messages generated more limited engagement and retweet totals. Thunderclap participation achieved substantial social reach, and overall campaign activity produced 417,678 impressions, 11,213 engagements, and 1,211 retweets.

  • Influencers: Influencer-inclusive tweets generated greater impression counts, reaching 11,349 impressions for a #eek post and 9,612 for a #story post.Promoted posts did not notably increase impressions, engagements, or retweets.
  • Opportunistic Messages: Opportunistic messages reached a maximum of 2,993 impressions, 103 engagements, and 8 retweets for a single message.
  • Thunderclap: Thunderclap tweets including an influencer achieved 11,740 impressions, compared with 2,527 impressions for tweets without one.
  • Twitter Analytics: Campaign activity totaled 417,678 tweet impressions, 11,213 post engagements, and 1,211 retweets, including 92 Thunderclap retweets.
  • Twitter Analytics: A single tweet achieved the campaign’s highest impression count, 11,349, and also produced the most engagements, 811.The maximum number of retweets for one post was 17.
  • Twitter Analytics: Table 3 reports Twitter analytic attributes for message frames, including influencer, promoted, and Thunderclap categories.

Is There an Appropriate Control Group for a Social Media Campaign? · Principal Findings · Investigating the Impact of a Bespoke Social Media Campaign on Skin Cancer Attitudes and Knowledge

The study found that Twitter was a feasible platform for disseminating tailored public health messages, while evidence suggested campaign-related improvements in skin cancer knowledge and attitudes. Northern Ireland and Wales keyword data did not provide an uncomplicated control comparison, although campaign hashtags were absent from Welsh search results.

  • Is There an Appropriate Control Group for a Social Media Campaign?: 15,964 and 14,168 tweets relating to sun exposure and skin cancer were returned for Northern Ireland in April and October 2015, respectively.The corresponding Wales search returned 50,164 and 51,634 tweets for April and October, respectively.
  • Is There an Appropriate Control Group for a Social Media Campaign?: Postcampaign keyword counts increased for Wales but decreased for Northern Ireland, complicating their use as a control comparison.Northern Ireland had a population of 1.8 million and Wales had a population of 3.0 million.
  • Is There an Appropriate Control Group for a Social Media Campaign?: The designated campaign hashtags did not appear in the Welsh keyword-search results.This finding distinguished the Welsh search results from the campaign’s designated hashtags.
  • Principal Findings: The study aimed to develop, implement, and evaluate a social media public health campaign and assess message frames, promotion techniques, and evaluation measures.The feasibility assessment focused on using Twitter to disseminate public health messages.
  • Principal Findings: The findings suggested that social media was a feasible platform for delivering a public health campaign.The study specifically evaluated social media as a platform for public health message dissemination.
  • Investigating the Impact of a Bespoke Social Media Campaign on Skin Cancer Attitudes and Knowledge: Pre- and postcampaign Web-based surveys showed a trend toward improved skin cancer knowledge and attitudes.The survey results should be interpreted cautiously.
  • Investigating the Impact of a Bespoke Social Media Campaign on Skin Cancer Attitudes and Knowledge: Awareness improved that sun protection can reduce skin cancer risk and that skin cancer is severe.The campaign used dynamic and tailored messages delivered to an audience in real time.

Investigating the Impact of Employing Different Message Frames on Social Media · Promoted Messages · Influencers

Message framing shaped social-media performance: shocking messages maximized impressions, humorous messages increased engagement, and informative messages were most retweeted. Promoted tweets underperformed, whereas influencer-seeded posts showed suggestive gains in impressions and engagement.

  • Investigating the Impact of Employing Different Message Frames on Social Media: Shocking (#eek) messages generated the most impressions, while humorous (#geg) messages produced greater public engagement than personal-story messages.These findings concern social-media message frames used in the skin cancer prevention campaign.
  • Investigating the Impact of Employing Different Message Frames on Social Media: Shocking messages attracted attention through a fear-based approach, although such approaches may also produce dissonance or desensitization.The study connected its greater impressions from shocking messages with prior work on fear-based public-health campaigns.
  • Investigating the Impact of Employing Different Message Frames on Social Media: The campaign’s most retweeted message was informative, despite prior research suggesting that information is relatively unimportant as a retweeting motivation.Twitter users may retweet selectively based on message content and other motivations.
  • Promoted Messages: Promoted tweets produced fewer impressions and retweets than organic posts and influencer-included posts, providing little overall value in this study.Promoted posts nevertheless offered potential targeting by location, age, gender, and interests.
  • Influencers: Influencer use showed suggestive evidence of increasing impressions and engagements compared with posts without influencers.Influencers or seeds were included following feedback from focus groups and co-design workshops, alongside a unique campaign hashtag.
  • Influencers: Identifying influential users commonly considers followers, friends, account age, prior tweets, and previous mentions, but more empirical research is needed.These factors were described as indicators of user influence in social networks.
  • Influencers: The study indicates that influencer seeds increase message impressions, while machine-learning methods could automate influence identification and propagation prediction.Such methods may enhance future campaign assessment and increase social-media campaign impact.

Thunderclap · Determining the Appropriate Process Evaluation Measures and Access to Data for a Social Media Campaign

The Thunderclap exceeded its supporter target and reached more than 450,000 people, while the campaign’s readily available Twitter metrics raised concerns about whether social media engagement adequately evaluates public health impact. The findings support careful promotion and explanation of Thunderclap campaigns and the development of adapted social media evaluation methods.

  • Thunderclap: 122 supporters exceeded the Thunderclap campaign’s target of 100 supporters.The target was described as somewhat arbitrary but exceeded the number achieved by previous Regional Public Health Agency campaigns.
  • Thunderclap: More than 450,000 people saw the campaign message after users pledged support for its posting from their chosen social media accounts.The campaign was considered useful for spreading awareness when correctly utilized, adequately promoted, and explained before launch.
  • Thunderclap: Campaign messages were posted on Twitter 3–4 days per week, with the same or slightly varied message posted up to 4 times daily.Scheduling reflected focus-group input and the host charity’s social media-account availability.
  • Determining the Appropriate Process Evaluation Measures and Access to Data for a Social Media Campaign: The study evaluated the campaign using Twitter analytics for impressions, engagements, likes, and shares.These measures were commonly used in the literature, but their appropriateness for public health campaign evaluation was questioned.
  • Determining the Appropriate Process Evaluation Measures and Access to Data for a Social Media Campaign: Liking or retweeting a message cannot reliably show that users support the advice, intend to share it, or will apply sunscreen.The passage cautions that these behaviors are not good barometers of impending behavioral change.
  • Determining the Appropriate Process Evaluation Measures and Access to Data for a Social Media Campaign: New social media research methods are needed, potentially by adapting traditional approaches for eHealth and melanoma interventions.The study delivered its Web-based survey via social media, while netnography was emerging to inform social media search terms.

Investigating Whether There Is an Appropriate “Control Group” for a Social Media Campaign · Limitations · Conflicts of Interest

The feasibility study could not establish a suitable control group and identified methodological challenges in evaluating social-media interventions. Key concerns included geographic comparability, dependent tweet data, phase contamination, message-frame agreement, and funding context.

  • Investigating Whether There Is an Appropriate “Control Group” for a Social Media Campaign: Wales showed virtually no campaign footprint, but local geography and weather may affect message reach and engagement, limiting its suitability as a control group.The campaign used specific hashtags and expected no impact in Wales.
  • Investigating Whether There Is an Appropriate “Control Group” for a Social Media Campaign: Traditional evaluation principles may not transfer directly to social media, although emerging methods assess weather effects and social-media sentiment.Examples included instrumental-variable analysis and emoji-based sentiment measurement.
  • Investigating Whether There Is an Appropriate “Control Group” for a Social Media Campaign: Clustered and dependent tweets may make traditional statistical analyses inappropriate, so the study sought statistical guidance for handling Twitter data.The authors advised interpreting the results within the context of other study limitations.
  • Investigating Whether There Is an Appropriate “Control Group” for a Social Media Campaign: Shocking tweets achieved the most median impressions, but the highest raw impression count was likely driven by an associated influencer.The authors noted that median impressions were substantially lower than the exceptional raw count.
  • Investigating Whether There Is an Appropriate “Control Group” for a Social Media Campaign: Future research should assess agreement on message frames because the research team determined tweet content and participant judgments of humor may differ.Content was verified through a codesign workshop, but the authors recommended further assessment.
  • Investigating Whether There Is an Appropriate “Control Group” for a Social Media Campaign: Updated sunlight-exposure guidelines essentially echoed the intervention’s aims, while the campaign was tailored, piloted, and integrated with local promotion programs.The guidelines changed after the intervention was completed.
  • Limitations: Contamination across intervention phases was a key validity concern, prompting recommendations for phase-based pre-post analysis with an adequate wash-out period.The authors identified contamination as a broader concern for social-media research.
  • Limitations: The intervention received support from public-health and cancer organizations, including Medical Research Council funding and collaboration with Northern Ireland partners.Additional support came from the Centre of Excellence for Public Health and a National Institute of Health Research fellowship.

Multimedia Appendix 1 … Abbreviations

The supplementary materials comprise four multimedia appendices covering intervention development, survey reporting, and Twitter advertising, followed by an abbreviations section defining UV and CHERRIES.

  • Multimedia Appendix 2: Multimedia Appendix 2 addresses intervention development and is provided as a 194KB Adobe PDF file.The appendix is listed as publichealth_v3i1e14_app2.pdf.
  • Multimedia Appendix 3: Multimedia Appendix 3 contains the Checklist for Reporting Results of Internet E-Surveys (CHERRIES) in a 56KB Adobe PDF file.The appendix is listed as publichealth_v3i1e14_app3.pdf.
  • Multimedia Appendix 4: Multimedia Appendix 4 presents the survey advertisement on Twitter in a 160KB Adobe PDF file.The appendix is listed as publichealth_v3i1e14_app4.pdf.
  • Abbreviations: The abbreviations section defines UV as Ultraviolet and CHERRIES as Checklist for Reporting Results of Internet E-Surveys.Both abbreviations are presented together in the source passage.
  • Abbreviations: The article was accepted on 21.01.17 and published on 23.03.17 after peer review and revision.The editorial record lists submission, peer review, comments, revision, acceptance, and publication dates.
  • Abbreviations: The article is open access under the Creative Commons Attribution License, permitting unrestricted use and distribution in any medium, provided conditions are met.The passage identifies the authors and links to the license terms.
Loading 1703.08813v1…