Sample and study design
This study was submitted to and approved as exempt research by the Indiana University IRB. Participants gave verbal consent as signed consent is not required for exempt research according to Indiana University IRB. The participants were provided a verbal and written description of the study. The written description included the following information: that subjects were being asked to participate in research; a description of the study procedures; a statement regarding any potential risks or benefits of participation; a statement that participation is voluntary; and the name, affiliation, and contact information of the researchers.
The data were collected from women who attended the health fair portion of the Annual Indiana Black Expo, a large cultural event that draws an estimated 40,000 attendees. Eligibility inclusion criteria included women who were 1.) Black, 2.) between the ages of 21–65 (age range for cervical cancer screening according to USPSTF guidelines), and 3.) could read and write English. Exclusion criteria included women with a hysterectomy. A sample of 98 women were included in analysis. The sample size was limited by the inclusion criteria, the availability of research personnel, and the many health vendors with displays at the health fair (i.e., many potential participants may have not wanted to take the extra time required to complete the survey). Participants were given the choice between a computer survey administered on REDCap (n = 58) or paper surveys (n = 40) after determining eligibility and explanation of the study. Participants were compensated with a $20 gift card after completing the survey.
Basic sociodemographic characteristics were assessed by self-report, including income (divided at the $30,000 point, which represents approximately 200% of the poverty level for a 2-person household), education, and marital status of participants. Then, participants were given a description of self-sampling for HPV testing and asked to evaluate hypothetical self-sampling scenarios that varied along 3 dimensions: delivery of an HPV self-sampling kit (mail, pharmacy pick-up, clinic pick-up), return of the HPV self-sampling kit (mail, pharmacy drop-off, clinic drop-off), and HPV result delivery (mail, phone call, or text message). A full factorial design would have required women to rate 27 scenarios, which would have imposed undue burden on the respondents. Therefore, we generated a fractional factorial design using the conjoint analysis procedure available in SPSS version 24 (IBM Corp., Armonk, NY) to create nine representative, independent scenarios that allowed us to evaluate the main effects of the dimensions. An example of a scenario presented to participants was “An HPV self-sampling kit delivered by mail that could be returned by mail with results delivered by text message. If self-sampling were available today, and you had time, how likely would you be to complete self-sampling in the given scenario?”. Participants rated the scenarios on an 11-point scale in intervals of 10 points (1–100) where 0 represented that they would never complete self-sampling for HPV testing and 100 meant that they would definitely complete self-sampling for HPV testing. Acceptability for HPV self-sampling in general was evaluated by creating a scale score based on the mean value across the 9 items illustrating hypothetical self-sampling scenarios (Cronbach’s alpha = .94).
We used SPSS 24 to describe sociodemographic characteristics for all 98 participants and the 63 participants who did not assign the same ratings to all scenarios. Then, RBCA was used to examine how HPV self-sample kit characteristics influenced ratings for participants (n = 63) who did not assign the same ratings to all scenarios. Participants who assigned the same ratings to all scenarios were not included in analyses. Nine hypothetical self-sampling scenarios were each defined along 3 dimensions: delivery of HPV self-sampling kit (mail, pharmacy pick-up, or clinic pick-up), return of HPV self-sampling kit (mail, pharmacy drop-off, or clinic drop-off), and delivery of HPV results (mail, phone call, text message). Income, education, and marital status were regressed on the HPV testing acceptability measure (created from the 9 items illustrating different hypothetical self-sampling for HPV testing scenarios among all 98 women) in a linear regression model.
RBCA is a regression-based technique used to understand how product preferences are influenced by product attributes and has been validated for use in previous health-related research [18,19,20]. RBCA allows respondents to consider attributes simultaneously so that respondents can make trade-offs. Conjoint analysis of the 9 scenarios revealed the relative preferences, named part-worth utilities in conjoint analysis, participants placed on each dimension. For example, in the dimension of HPV self-sampling kit return, the preference placed on the attribute of mailed delivery of HPV self-sampling kit is reflected in a higher part-worth utility score compared to the part-worth utility scores of clinic or pharmacy pick-up. A negative part-worth utility score indicates a relative dislike for an attribute (such as clinic pick-up of HPV self-sampling kit) and a positive part-worth utility score shows a relative preference for an attribute. A wider range of part-worth utility attribute scores across a given dimension has a greater influence on importance scores than a dimension with a smaller range in values. The sum of the part-worth utilities needs to equal zero in each of the 3 dimensions. Importance scores were calculated by the relative ranges of part-worth utilities across the 3 dimensions, and in this approach, the sum of importance scores across dimensions must equal 100. The higher the importance score for a given dimension (such as HPV self-sampling kit return), the greater the influence on acceptability of a given scenario.