The Health Belief Model (HBM) guided the study.14 HBM theorizes that knowledge and perceptions about disease susceptibility and severity, perceived barriers and benefits, and self-efficacy drive ultimately health behaviors, and that intention to enact a behavior predicts behavioral action. According to the model, those with higher levels of perceived susceptibility and perceived benefits, as well as self-efficacy, are more likely to engage in health behaviors, as are those who perceive low barriers to the behavior.
Sample and Setting: We recruited a convenience sample of English-speaking women ages 18-26 years who were residents of public housing in two Massachusetts cities. We elected to recruit women from public housing in order to reach women with low levels of income; in the state of Massachusetts, those eligible to live in public housing must meet specific income requirements (i.e.,< 80% of the area median family income).15 We did not require that women have an existing Twitter account, but they had to be willing to do so for the study. We also did not restrict eligibility to those who had not previously received the vaccine, as we were interested in assessing the feasibility of offering the Twitter campaign and in addition, wanted to assess cervical cancer screening practices in this high-risk population. Our goal was to recruit 50 women over a two-month period. To recruit women, we mailed information about the study, posted fliers in common use areas in housing developments, and with the housing authorities, transmitted email or phone announcements about the study. Women were informed that this was a a research study on “women’s health.”
Intervention: Women received a daily tweet over a period of one-month that contained messages from educational materials produced by the Centers for Disease Control and Prevention, National Cancer Institute and the Massachusetts Department of Public Health. Messages were selected to align with the Health Belief Model and primarily addressed HPV vaccination, although there were also messages promoting cervical cancer screening.
Data Collection: Pre/post surveys were administered by phone or online and assessed cervical cancer screening, HPV and HPV vaccine knowledge, perceived risk for cervical cancer, perceived barriers to HPV vaccination, and HPV vaccination intentions. Upon completion of the campaign, we contacted women through Twitter to let them know that we would be contacting them to complete the post-survey. Women received a $50 gift card for participation. All procedures and protocols were approved by the Institutional Review Board at Tufts University.
Assessment of Feasibility: Feasibility was assessed with standard metrics, for example, by determining the numbers of women who: were recruited relative to recruitment goals, had Twitter accounts at the time of recruitment, continued to receive our messages over the one-month period (i.e., did not block messages), and who completed the post-test survey. Twitter analytics were used to track re-tweets and “likes.” To assess receipt and recall of messages, participants were asked at the end of the post-test to indicate whether they received daily tweets from the campaign. The first question was: “Did you receive Twitter messages about HPV from the Tufts Women's Health Study?” (yes/no). Next, we asked: “Which of the following messages did you receive from the Tufts Women's Health Study?” Response options included true statements that were emphases of the campaign (e.g., “HPV vaccine can prevent cervical cancer”) and also false statements not included in the campaign (e.g., “Most women will feel symptoms if they have the HPV infection, so it is often detected early”). In addition, women were asked about the acceptability of receiving Twitter messages at the end of the study.
Measures of Health Belief Model Constructs
When standardized questions were not available, we adapted items from our prior studies, which had high reliability.16 Items to assess receipt of the HPV vaccine, Pap testing, and sociodemographic characteristics were taken from the BRFSS.17 Intention to be vaccinated among those who had not completed the 3-dose series was assessed by asking “How likely is it that you will get vaccinated against HPV in the future? In the next 6 months? In the next 12 months?” with responses on a 4-point Likert scale from very likely to very unlikely. Additional survey items assessed usual source of care, health insurance, and selected sociodemographic based on standard items from national surveys.18
HPV and cervical cancer knowledge was assessed with 13 items that addressed knowledge about HPV, the HPV vaccine, and cervical cancer; HPV and cervical cancer risk factors and HPV transmission.16 For each correct answer, respondents received one point; for each “don’t know” or incorrect answer, respondents received no point. Total points were then divided by the maximum number of points and multiplied by 100 to arrive at a scale with a range of 0-100%.
Perceived susceptibility to HPV and perceived susceptibility to cervical cancer were each assessed with 2 items that addressed overall perceived risk (e.g., “Overall, how would you rate your chance of developing HPV?”) and risk compared to similar-aged peers (e.g., “How would you rate your chances of developing cervical cancer compared to average women your age?”).19 Response options were on a 5-point Likert scale (ranging from 1 = “very low” to 5 = “very high”). “Don’t know” responses were coded as 0. Points were summed up for the 2 items in each score such that higher values reflect higher perceived risk (range 0-10).
Perceived benefits of HPV vaccination were assessed with two composite scores: vaccine efficacy and vaccine safety.16 Vaccine efficacy was assessed with 3 items that examined the potential for vaccination to prevent HPV infection, genital warts, and cervical cancer (e.g., “In your opinion, how effective is the HPV vaccine in preventing HPV infection?”). Response options were on a 4-point Likert scale (ranging from 1 = “not at all effective” to 4 = “very effective”). “Don’t know” responses were coded as 0. Points were summed up for the 3 items such that higher values reflect greater perceived efficacy (range 0-12). Vaccine safety was assessed with two items that addressed safety and likelihood of the vaccine causing other health problems. Safety response options were on a 4-point Likert scale (ranging from 1 = “not at all safe” to 4 = “very safe”), while likelihood of other health problems was on a 5-point Likert scale (ranging from 1 = “very likely” to 5 = “never”). “Don’t know” responses were coded as 0. Points were summed up for the 2 items such that higher values reflect greater perceived safety (range 0-9).
Perceived barriers to vaccination included assessment of potential pain and cost associated with vaccination with 2 items: “In your opinion, how painful [expensive] would it be to receive the HPV vaccine?” with response options on a 4-point Likert scale (ranging from 1 = “not at all painful [expensive]” to 4 = “very painful [expensive]”).20 “Don’t know” responses were coded as 0. Responses were combined such that higher scores reflect greater perceived barriers (range 0-8).
Decision self-efficacy with regard to obtaining and comprehending information about the vaccine was assessed with 11 items adapted from the Decision Self-Efficacy Scale21 (e.g., “How confident do you feel to get the facts about the risks of the HPV vaccine?). Responses were on a 2-point Likert scale (ranging from 0 = “not at all confident” to 2 = “very confident”). Total points were then divided by the maximum number of points and multiplied by 100 to arrive at a scale with a range of 0-100%, with higher scores indicating higher confidence.
Analysis
Data from pre- and post-test surveys were analyzed using R and RStudio.22,23 Descriptive statistics, including means and standard deviations (SD) were used to describe the sample in terms of sociodemographic characteristics and HPV vaccination knowledge, attitudes, and behaviors. Assessment of feasibility involved descriptive data (e.g., number of women who recalled receiving messages) as well as examination of Twitter analytics (e.g., likes, retweets). Changes between pre- and post-test were analyzed with paired t-tests for continuous variables (i.e., knowledge, perceived susceptibility, perceived benefits of and barriers to HPV vaccination, and decision self-efficacy) and McNemar’s test for categorical variables (i.e., intention to get the HPV vaccine in the future among those who had not completed vaccination).