All pins are not created equal: communicating skin cancer visually on Pinterest

All pins are not created equal: communicating skin cancer visually on Pinterest Abstract Skin cancer is the second most common cancer affecting women younger than 39 years in the USA. As a female-oriented social media, Pinterest could be effectively used in reaching this particular demographic group for the purpose of skin cancer education. We analyze the visual characteristics of skin cancer pins, including use of human image, use of fear-invoking image, pin composition, color, and text legibility. We also explore how these visual characteristics as well as information richness predict Pinterest users’ participative engagement. A combination of descriptive and predictive content analysis of 708 pins is conducted. The demographic characteristics of human models are consistent with epidemiology data. Text legibility in bodycopy is low. Information richness is a significant predictor of number of repins in all pins except pins on melanoma with human models. In the case of latter, pin composition, gender of human models, and fear-invoking images were associated with the number of repins. A number of visual characteristics as well as information richness significantly predict Pinterest users’ participant engagement with pins on skin cancer. Public health professionals should consider these factors in creating effective prevention messages to be circulated on Pinterest. Implications Research: This work proposes a comprehensive model for the analysis of visual characteristics of Pinterest contents. It demonstrates how different visual characteristics predict Pinterest users’ participative engagement. Practice: Public health professionals could match visual and textual characteristics of pins with types of skin cancer to create maximum participative engagement. Policy: More support is needed to fund future research on visual communication in the context of health promotion on social media. INTRODUCTION Pictures are essential to health education. Adding highly relevant pictures to health promotion texts generates more attention to and recall of the message [1]. Pictures showing the relationship among ideas increase the comprehension of health messages, especially among patients with low literacy [2]. People also rely on visual cues to assess the credibility of the health information online [3]. Today’s Internet users can easily share images online. Pinterest is one of the fastest growing social media dedicated to photo sharing with approximately 31 per cent of Internet users having used Pinterest in 2015 [4]. It is also the first social media platform that frames information with a female orientation [5]. Pinterest has a saturated population of white females under age 40 with high education [6]. Health is a prominent topic on Pinterest. Researchers have started to examine the presentation of a number of health-related topics on Pinterest, including vaccination [7], depression [8], and skin regiment [9]. Presented here is a study of the visual characteristics of pins about skin cancer on Pinterest.com. Skin cancer is the most common cancer affecting the U.S. population and is the second most common cancer for Caucasian women younger than 39 years [10]. Although melanoma skin cancer received the most attention due to its malignancy, nonmelanoma skin cancers (NMSCs), such as basal-cell carcinoma and squamous-cell carcinoma, are less known [11]. Guided by the theory of visual framing, the current study offers a comprehensive framework to analyze the visual characteristics of skin cancer messages on Pinterest. It explores how these visual characteristics as well as information richness predict users’ participative engagement. Practically, this study provides tangible guidelines for public health professionals in creating health promotion messages to be disseminated on Pinterest and possibly other visual-heavy social media. THE VISUAL FRAMING OF SKIN CANCER ON PINTEREST Media systematically influence the ways in which audiences perceive an event through framing [12]. Framing can be achieved through both verbal messages and visual images. Rodriguez and Dimitrova [13] identified four levels of visual framing analysis. On the most basic level, scholars examine the denotative meanings or the themes of visual images (e.g., what are the objects and elements included in the picture). The second level involves a stylistic analysis of the image in terms of camera angle, lighting, distance between the camera and the object, color, composition of the picture, etc. A reading of the connotative meaning of the image represents the third level of visual framing analysis. On this level, images are interpreted as symbols communicating socially constructed meanings. The fourth level focuses the ideological messages of the visual image. Visual framing has been used to analyze health-related visual images, including cancer [14], cancer treatment [15], genetically modified foods [16], and cigarette advertising [17]. Majority of these studies concentrate on the first and second levels of analysis, which is probably because the first and second levels of analysis allow relatively reliable coding [16]. The current study will focus on these two levels too. Use of human models The first level of visual framing analysis focuses on different types of objects included in in the visual image and their denotative meanings. The use of human images has been frequently examined in the study of health-related visual images. According to the social cognitive theory, individuals learn through the process of modeling and they are more likely to adopt the suggested behavior if they identify with the model [18]. Phillips et al. [15] found that the models used in cancer consumer magazines were more likely to be female, white, and younger, different from the demographics of actual cancer patients. They argued that such discrepancy would discourage patients from processing the information and adopting the recommendations provided in such magazines. Hence, it is important to examine whether pins on skin cancer include human models and whether these models reflect the population at high risk of skin cancer. This leads to the first two research questions (RQ1a and RQ1b): RQ1a: To what extent do pins on skin cancer use human models? RQ2b: What are the demographic characters of human models (in terms of age, gender, and skin color)? Use of fear-invoking images The use of fear-invoking images is another component of first-level visual framing. One of the most common strategies utilized in health promotion is fear appeal, especially through the use of scary visual images [19]. According to the Extended Parallel Processing Model, individuals would go through several processes when faced with the fear appeal. First, they will evaluate their susceptibility to the threat. The more they believe they are susceptible, the more likely they are going to assess the efficacy of the recommended behavior. High level of perceived threat and high level of perceived efficacy lead to more behavioral change. On the contrary, individuals are likely to deny the threat when they doubt the effectiveness of the recommended behavior [20]. Furthermore, women are more likely to respond to disgust- and fear-inducing images than men [21]. Since majority of Pinterest’s users are woman, it is especially important to examine the use of fear-invoking images in pins on skin cancer: RQ2: To what extent do pins on skin cancer include fear-invoking images? The analysis of stylistic features of images constitutes the second level of visual framing analysis. Past research identified several stylistic characters important to health-related visual images, including image composition, color, and text legibility. Pin composition Researchers have studied the composition of pins in terms of whether they are picture only, text only, video only, picture combined with texts, and infographics [7]. Pictures are important in health education as they have the abilities to generate more attention to the textual information and create more comprehension when pictures are related to texts [1]. Videos represent another effective channel of health communication in terms of increasing knowledge and influencing psychological variables associated with behavior change, such as awareness, self-efficacy, and personal vulnerability [22]. Inforgraphics visualize complicated information through a combination of texts, numbers, charts, and graphs [23]. The use of infographics increases the comprehension of message among people with low health literacy and numeracy and is effective in promoting health behaviors in general [24]. Since the composition of a pin is likely to affect the how much attention is paid to the pin and how information is processed, RQ3 is proposed: RQ3: What are the compositions of pins on skin cancer (picture only, text only, videoonly, picture and text, and inforgraphics)? Colors Color generates cognitive and emotional arousals affecting one’s assessment of an image [25]. Researchers often define colors in terms of warm and cool colors [26]. Across many cultures, cool colors such as blue and green are often considered peaceful and calming, whereas warm colors such as red and orange are perceived to be hot, vibrant, and emotional [27]. In the visual communication of health information, warm colors are often associated with heightened risks [28]. Lighter colors such as white, silver, and light blue were used in most of the cigarette advertisements across decades to convey the impression of “relaxation, freshness, and calmness” (p. 782) [17]. Thus, the next two RQs are proposed: RQ4a: What are the primary color schemes of pins on skin cancer? RQ4b: What are the color schemes of the fonts used in pins on skin cancer? Legibility Legibility of text is another visual element contributing to the quality of health messages [29]. Legibility increases individuals’ comfort level in their information use and is especially important on Pinterest as 80 per cent of its users reportedly use mobile phones as their primary device to access Pinterest [30]. A study on font size on handheld device revealed that the range of font size preference was greater among older generation than among young people; however, as long as the font size was greater than 8, both young and older people show satisfactory rating on legibility [31]. Having health information that is illegible obviously will prevent the proper processing of such information. This leads to the next RQ: RQ5: To what extent are the texts in pins legible? Participative engagement Social media enables audience’s participative engagement in terms of likes, comments, retweets, and shares. In the case of Pinterest, if a user likes a particular pin, she could repin it to one of her own boards. Social media content generating more participative engagement is usually more effective in influencing audience’s health beliefs and behaviors. Han et al. [32] found that the characteristics of a pin instead of characteristics of the user (such as number of followers) predict the number of repins. Since Pinterest relies heavily on the attractiveness of images and the platform is based on redistributing images among users [33], the visual characteristics of pins on skin cancer are likely to affect the likelihood these pins will be repined. In addition, since one of the major gratifications of Pinterest use is to store, categorize, and share information [34], information richness, that is, the amount of information about skin cancer in a pin, might also influence the perceived usefulness of the pin and its popularity in terms of number of repins. Studies [35, 36] suggested that the re-pin function was the strongest indication of user interaction on Pinterest and a positive correlation was found between the combined number of likes and comments and the number of repins. It is possible that both information richness and the visual characteristics will affect the likelihood an image gets repinned. Hence, the last RQ is proposed: RQ6: To what extent do information richness and visual characteristics of pins explain Pinterest users’ participative engagement in terms of the number of repins? METHOD A combination of descriptive and predictive content analysis was conducted to answer the research questions proposed above. Different from regular content analysis devoted to the description of media contents, predictive content analysis examines how content characteristics are likely to influence audiences’ psychological or behavioral outcomes [37]. Sampling Four keywords (skin cancer, melanoma, basal-cell carcinoma, and squamous-cell carcinoma) were used to search Pinterest.com. Since the platform of Pinterest did not allow to collect the data of a specific time frame, the only feasible option to collect data was scrolling down the page until no more pins would appear. The authors monitored the pages every day for 3 weeks between September 16 and October 9, 2015. A total of 708 pins were collected after excluding repetitive and nonrelevant pins. Each pin was assigned a unique ID number. Unit of analysis and measurement An individual pin was the unit of analysis and was coded based on the following items. Types of skin cancer Each pin was coded based on the type of skin cancer mentioned in the textual information: (i) skin cancer in general without mentioning any specific type, (ii) melanoma, (iii) squamous-cell carcinoma, and (iv) basal-cell carcinoma. Display of human body First, the display of human body was coded as present or absent. Types of human pictures were further coded into one of the following four categories: profile photo (upper torso with face shot), affected areas of body without face shot, drawing of a person (illustration), and multiple (a combination of more than one element). Next, age, skin color, and gender of human models were coded. Age was coded into four categories: old (when the model appeared to be older than 50 years old), young (when the model appeared to be younger than 50 years old), mixed (when both young and old models were present), and unable to tell (when the face of model was not shown). Skin color was coded into four different categories: white, black, brown, and mixed and these four categories were later merged into two categories: white and nonwhite (including any pins containing a nonwhite human model). Gender was coded as male, female, mixed, and unknown (when only part of the body was shown and gender of the model could not be decided). Fear-invoking images Each pin was coded in terms of whether it had images that could invoke fear, that is, visual images of skin cancer such as discoloration, blisters, and deformation of skin parts (present or absent). Next, pins containing fear-invoking messages were further coded into two groups: high fear and low fear. The mere presence of benign image of skin cancer such as discoloration or early signs of skin cancer (such as a mole) was coded as low fear, whereas clear indications of late stages of skin cancer (tumors, ulceration, or scars) were coded as high fear [38]. Pin composition The composition of pins was coded as one of the following: picture only, text only, video only, picture and text, and inforgraphic [7]. Picture and text were coded when the pin had prominent texts coupled with pictures. A pin was coded as an infographic when it communicated complicated information through a combination of texts, numbers, charts, and graphs [23]. Colors Two types of color are coded: primary picture color and font color. Primary picture color could be the color of a dominant object in the pin (for instance, skin or a piece of fruit) or the background color when such a dominant object is absent. Six primary picture colors are coded: black, white, cool colors, warm colors, skin color, and multiple colors. Cool color includes blue and green, whereas warm colors include red, yellow, orange, and pink [39]. Font colors are coded separately for headline and bodycopy. Five types of font colors were coded: black, white, cool colors, warm colors, and multiple colors. Font legibility When a pin contains texts (either as headline or bodycopy), the legibility of the texts was coded. Headline and bodycopy were evaluated separately. Legibility was coded as “high” if the text was readable on an iPhone without enlarging the picture and coded as “low” if the text was not readable. Information richness Textual information was coded in terms of susceptibility, severity, benefit, barriers, cue to action, self-efficacy, causes, symptoms, treatment, and prevention. Details of these items were reported in a previous study [40]. Information richness was calculated in terms of the number of these items present in a particular pin. The information richness scores ranged from 1 to 10 (M = 3.57, SD = 1.46). Participant engagement It was coded as the number of repins. Coding and intercoder reliability After training and test coding, the first author coded all the pins in the sample and the second author coded 10 per cent of the pins (n = 71) selected through systematic sampling. Percentage agreement was calculated and all of items reached over 85%–90% of agreement except for two items: primary color (84%) and bodycopy color (77%). Percentage agreement is more appropriate than Cohen’s κ or Krippendorff’s α when the distribution of the measure is highly skewed, because latter measurements tend to make the intercorder reliability artificially low even when coding is reasonably reliable [41]. Data analysis Descriptive statistics such as frequencies and percentages were used to answer RQs1–5. Chi-square tests were run to compare the differences among pins about skin cancer in general, melanoma, and NMSCs. Sequential bonferroni correction was used to account for the effects of multiple testing. Pair-wise comparison among these three types of pins showed that pins about skin cancer in general and pins about melanoma were very similar in that none of the chi-square tests yielded significant results after adjusting for multiple testing. Consequently, these two categories were collapsed and chi-square tests were run to compare two types of pins: melanoma (including skin cancer in general) and NMSC. To answer RQ6, negative binomial regression models were used to explore how information richness, stylistic variables (e.g., the compositions of pins, primary color schemes), human model-related factors (e.g., display of human body, demographic characters of human models), and fear-invoking image predicted the popularity of a pin (i.e., number of repins). Negative binomial regression analyses were chosen because the dependent variable (number of repins) is a count variable whose distribution was highly skewed (skewness = 7.344) and it contained a large number of zeros [42]. Data were analyzed using R Studio version 3.2.4. RESULTS Among the 708 pins in the sample, 27.4 per cent (n = 194) were about skin cancer in general, 22.5 per cent (n = 159) focused on melanoma, and the rest were about NMSCs (50.2%, n = 355) (see Table 1 for the descriptive statistics of all variables categorized by the types of skin cancer). Table 1 Results for descriptive analysis of pins on different types of skin cancer     Skin cancer (n = 194)  Melanoma (n = 159)  Nonmelanoma (n = 355)      n  %  n  %  n  %  RQ1a  Types of human images  Profile picture  27  13.9  23  14.5  46  13.0  Affected area  60  30.9  40  25.2  157  36.3  Drawing  2  1.0  6  3.8  4  1.1  Multiple  2  1.0  1  .6  1  .3  RQ1b  Gender  Male  13  6.7  7  4.4  57  16.1    Female  36  18.6  22  13.8  25  7.0    Unable to tell  39  20.1  34  21.4  119  27.1  Age  Young  13  6.7  7  4.4  57  16.1    Old  36  18.6  22  13.8  7  2.0  Skin color  White  89  45.9  62  39.0  195  54.9    Nonwhite  2  1.0  5  3.1  4  1.4  RQ2  Fear  No fear  137  70.6  116  73.0  198  55.8    High fear  11  5.7  5  3.1  31  8.7    Low fear  46  23.7  38  23.9  126  35.5  RQ3  Pin Composition  Picture only  83  42.8  44  27.7  216  60.8  Text only  10  5.2  10  6.3  7  2.0  Picture and text  71  36.6  81  50.9  106  29.9  Video only  5  2.6  2  1.3  24  6.8  Infographics  25  12.9  22  13.8  2  .6  RQ4a  Primary color  Skin  90  46.4  63  39.6  210  59.2    Black  14  7.2  35  22.0  21  5.9    White  10  5.2  10  6.3  51  14.4    Cool colors  30  15.5  24  15.1  36  10.1    Warm colors  37  19.1  18  11.3  30  8.5    Multiple colors  13  6.7  9  5.7  7  2.0  RQ4b  Headline color  Black  26  13.4  48  30.2  33  9.3  White  29  14.9  25  15.7  32  9.0  Cool colors  7  3.6  4  2.5  4  1.1  Warm colors  17  8.8  9  5.7  26  7.3  Multiple colors  14  7.2  16  10.1  14  3.9  Bodycopy color  Black  22  11.3  25  15.7  33  9.3  White  15  7.7  15  9.4  11  3.1  Cool colors  2  1.0  2  1.3  1  .3  Warm colors  4  2.1  4  2.5  7  2.0  Multiple colors  19  9.8  30  18.9  11  3.1  RQ5  Legibility  Headline-high  92  47.9  100  62.9  105  29.6  Headline-low  0  0  3  1.9  4  1.1  Bodycopy-high  33  17.0  40  25.2  40  11.3  Bodycopy-low  29  14.9  36  22.6  23  6.5      Skin cancer (n = 194)  Melanoma (n = 159)  Nonmelanoma (n = 355)      n  %  n  %  n  %  RQ1a  Types of human images  Profile picture  27  13.9  23  14.5  46  13.0  Affected area  60  30.9  40  25.2  157  36.3  Drawing  2  1.0  6  3.8  4  1.1  Multiple  2  1.0  1  .6  1  .3  RQ1b  Gender  Male  13  6.7  7  4.4  57  16.1    Female  36  18.6  22  13.8  25  7.0    Unable to tell  39  20.1  34  21.4  119  27.1  Age  Young  13  6.7  7  4.4  57  16.1    Old  36  18.6  22  13.8  7  2.0  Skin color  White  89  45.9  62  39.0  195  54.9    Nonwhite  2  1.0  5  3.1  4  1.4  RQ2  Fear  No fear  137  70.6  116  73.0  198  55.8    High fear  11  5.7  5  3.1  31  8.7    Low fear  46  23.7  38  23.9  126  35.5  RQ3  Pin Composition  Picture only  83  42.8  44  27.7  216  60.8  Text only  10  5.2  10  6.3  7  2.0  Picture and text  71  36.6  81  50.9  106  29.9  Video only  5  2.6  2  1.3  24  6.8  Infographics  25  12.9  22  13.8  2  .6  RQ4a  Primary color  Skin  90  46.4  63  39.6  210  59.2    Black  14  7.2  35  22.0  21  5.9    White  10  5.2  10  6.3  51  14.4    Cool colors  30  15.5  24  15.1  36  10.1    Warm colors  37  19.1  18  11.3  30  8.5    Multiple colors  13  6.7  9  5.7  7  2.0  RQ4b  Headline color  Black  26  13.4  48  30.2  33  9.3  White  29  14.9  25  15.7  32  9.0  Cool colors  7  3.6  4  2.5  4  1.1  Warm colors  17  8.8  9  5.7  26  7.3  Multiple colors  14  7.2  16  10.1  14  3.9  Bodycopy color  Black  22  11.3  25  15.7  33  9.3  White  15  7.7  15  9.4  11  3.1  Cool colors  2  1.0  2  1.3  1  .3  Warm colors  4  2.1  4  2.5  7  2.0  Multiple colors  19  9.8  30  18.9  11  3.1  RQ5  Legibility  Headline-high  92  47.9  100  62.9  105  29.6  Headline-low  0  0  3  1.9  4  1.1  Bodycopy-high  33  17.0  40  25.2  40  11.3  Bodycopy-low  29  14.9  36  22.6  23  6.5  View Large Table 1 Results for descriptive analysis of pins on different types of skin cancer     Skin cancer (n = 194)  Melanoma (n = 159)  Nonmelanoma (n = 355)      n  %  n  %  n  %  RQ1a  Types of human images  Profile picture  27  13.9  23  14.5  46  13.0  Affected area  60  30.9  40  25.2  157  36.3  Drawing  2  1.0  6  3.8  4  1.1  Multiple  2  1.0  1  .6  1  .3  RQ1b  Gender  Male  13  6.7  7  4.4  57  16.1    Female  36  18.6  22  13.8  25  7.0    Unable to tell  39  20.1  34  21.4  119  27.1  Age  Young  13  6.7  7  4.4  57  16.1    Old  36  18.6  22  13.8  7  2.0  Skin color  White  89  45.9  62  39.0  195  54.9    Nonwhite  2  1.0  5  3.1  4  1.4  RQ2  Fear  No fear  137  70.6  116  73.0  198  55.8    High fear  11  5.7  5  3.1  31  8.7    Low fear  46  23.7  38  23.9  126  35.5  RQ3  Pin Composition  Picture only  83  42.8  44  27.7  216  60.8  Text only  10  5.2  10  6.3  7  2.0  Picture and text  71  36.6  81  50.9  106  29.9  Video only  5  2.6  2  1.3  24  6.8  Infographics  25  12.9  22  13.8  2  .6  RQ4a  Primary color  Skin  90  46.4  63  39.6  210  59.2    Black  14  7.2  35  22.0  21  5.9    White  10  5.2  10  6.3  51  14.4    Cool colors  30  15.5  24  15.1  36  10.1    Warm colors  37  19.1  18  11.3  30  8.5    Multiple colors  13  6.7  9  5.7  7  2.0  RQ4b  Headline color  Black  26  13.4  48  30.2  33  9.3  White  29  14.9  25  15.7  32  9.0  Cool colors  7  3.6  4  2.5  4  1.1  Warm colors  17  8.8  9  5.7  26  7.3  Multiple colors  14  7.2  16  10.1  14  3.9  Bodycopy color  Black  22  11.3  25  15.7  33  9.3  White  15  7.7  15  9.4  11  3.1  Cool colors  2  1.0  2  1.3  1  .3  Warm colors  4  2.1  4  2.5  7  2.0  Multiple colors  19  9.8  30  18.9  11  3.1  RQ5  Legibility  Headline-high  92  47.9  100  62.9  105  29.6  Headline-low  0  0  3  1.9  4  1.1  Bodycopy-high  33  17.0  40  25.2  40  11.3  Bodycopy-low  29  14.9  36  22.6  23  6.5      Skin cancer (n = 194)  Melanoma (n = 159)  Nonmelanoma (n = 355)      n  %  n  %  n  %  RQ1a  Types of human images  Profile picture  27  13.9  23  14.5  46  13.0  Affected area  60  30.9  40  25.2  157  36.3  Drawing  2  1.0  6  3.8  4  1.1  Multiple  2  1.0  1  .6  1  .3  RQ1b  Gender  Male  13  6.7  7  4.4  57  16.1    Female  36  18.6  22  13.8  25  7.0    Unable to tell  39  20.1  34  21.4  119  27.1  Age  Young  13  6.7  7  4.4  57  16.1    Old  36  18.6  22  13.8  7  2.0  Skin color  White  89  45.9  62  39.0  195  54.9    Nonwhite  2  1.0  5  3.1  4  1.4  RQ2  Fear  No fear  137  70.6  116  73.0  198  55.8    High fear  11  5.7  5  3.1  31  8.7    Low fear  46  23.7  38  23.9  126  35.5  RQ3  Pin Composition  Picture only  83  42.8  44  27.7  216  60.8  Text only  10  5.2  10  6.3  7  2.0  Picture and text  71  36.6  81  50.9  106  29.9  Video only  5  2.6  2  1.3  24  6.8  Infographics  25  12.9  22  13.8  2  .6  RQ4a  Primary color  Skin  90  46.4  63  39.6  210  59.2    Black  14  7.2  35  22.0  21  5.9    White  10  5.2  10  6.3  51  14.4    Cool colors  30  15.5  24  15.1  36  10.1    Warm colors  37  19.1  18  11.3  30  8.5    Multiple colors  13  6.7  9  5.7  7  2.0  RQ4b  Headline color  Black  26  13.4  48  30.2  33  9.3  White  29  14.9  25  15.7  32  9.0  Cool colors  7  3.6  4  2.5  4  1.1  Warm colors  17  8.8  9  5.7  26  7.3  Multiple colors  14  7.2  16  10.1  14  3.9  Bodycopy color  Black  22  11.3  25  15.7  33  9.3  White  15  7.7  15  9.4  11  3.1  Cool colors  2  1.0  2  1.3  1  .3  Warm colors  4  2.1  4  2.5  7  2.0  Multiple colors  19  9.8  30  18.9  11  3.1  RQ5  Legibility  Headline-high  92  47.9  100  62.9  105  29.6  Headline-low  0  0  3  1.9  4  1.1  Bodycopy-high  33  17.0  40  25.2  40  11.3  Bodycopy-low  29  14.9  36  22.6  23  6.5  View Large Among the 708 pins, 52.1 per cent (n = 369) used human models. Out of them, 26 per cent (n = 96) included profile pictures with faces, 69.6 per cent (n = 257) included picture of body parts or affected area without showing the face, and 3.3 per cent (n = 12) used drawings of humans. Four pins included two or more of the categories mentioned above. Chi-square test showed that pins on NMSC were significantly more likely to show affected areas than other pins (see Table 2 for results of chi-square tests). Table 2 Results of chi-square tests comparing pins on melanoma (including skin cancer in general, n = 353) and nonmelanoma skin cancers (NMSC; n = 355)     X2  p  Adjusted p  RQ1a          Types of human image  Profile  .220  .639    Affected area  19.344*  .000  .000**  Drawing  1.379  .240    RQ1b  Gender  Male  19.715  .000  .000**    Female  15.075  0.00  .002**  Age  Young  3.807  .079      Old  5.653  .017    Skin color  White  10.463  .001  .026*    Non-white  4.840  .028    RQ2  Fear  Low fear  11.607  .001  .015*    High fear  5.037  .025    RQ3  Pin Composition  Picture only  43.825*  .000  .000**    Text only  6.584  .010      Picture and text  13.316*  .000  .006**    Video only  9.649*  .002  .039*    Inforgraphic  44.673*  .000  .000**  RQ4a  Primary color  Skin  17.713*  .000  .000**    Black  12.605*  .000  .009**    White  14.805*  .000  .003**    Cool colors  4.242  .039      Warm colors  8.518  .004      Multiple  8.179  .004    RQ4b  Headline color  Black  18.781*  .000  .000**    White  6.548  .010      Cool colors  3.378  .066      Warm colors  .000  .983      Multiple  6.301  .012    Bodycopy color  Black  2.852  .091      White  9.460*  .002  .042*    Cool colors  1.830  .176      Warm colors  .074  .786      Multiple  26.530*  .000  .000**  RQ5  Font legibility  Headline  1.412  .235      Bodycopy  1.972  .160        X2  p  Adjusted p  RQ1a          Types of human image  Profile  .220  .639    Affected area  19.344*  .000  .000**  Drawing  1.379  .240    RQ1b  Gender  Male  19.715  .000  .000**    Female  15.075  0.00  .002**  Age  Young  3.807  .079      Old  5.653  .017    Skin color  White  10.463  .001  .026*    Non-white  4.840  .028    RQ2  Fear  Low fear  11.607  .001  .015*    High fear  5.037  .025    RQ3  Pin Composition  Picture only  43.825*  .000  .000**    Text only  6.584  .010      Picture and text  13.316*  .000  .006**    Video only  9.649*  .002  .039*    Inforgraphic  44.673*  .000  .000**  RQ4a  Primary color  Skin  17.713*  .000  .000**    Black  12.605*  .000  .009**    White  14.805*  .000  .003**    Cool colors  4.242  .039      Warm colors  8.518  .004      Multiple  8.179  .004    RQ4b  Headline color  Black  18.781*  .000  .000**    White  6.548  .010      Cool colors  3.378  .066      Warm colors  .000  .983      Multiple  6.301  .012    Bodycopy color  Black  2.852  .091      White  9.460*  .002  .042*    Cool colors  1.830  .176      Warm colors  .074  .786      Multiple  26.530*  .000  .000**  RQ5  Font legibility  Headline  1.412  .235      Bodycopy  1.972  .160    df = 1 for all tests. *p <. 05; **p <. 01. View Large Table 2 Results of chi-square tests comparing pins on melanoma (including skin cancer in general, n = 353) and nonmelanoma skin cancers (NMSC; n = 355)     X2  p  Adjusted p  RQ1a          Types of human image  Profile  .220  .639    Affected area  19.344*  .000  .000**  Drawing  1.379  .240    RQ1b  Gender  Male  19.715  .000  .000**    Female  15.075  0.00  .002**  Age  Young  3.807  .079      Old  5.653  .017    Skin color  White  10.463  .001  .026*    Non-white  4.840  .028    RQ2  Fear  Low fear  11.607  .001  .015*    High fear  5.037  .025    RQ3  Pin Composition  Picture only  43.825*  .000  .000**    Text only  6.584  .010      Picture and text  13.316*  .000  .006**    Video only  9.649*  .002  .039*    Inforgraphic  44.673*  .000  .000**  RQ4a  Primary color  Skin  17.713*  .000  .000**    Black  12.605*  .000  .009**    White  14.805*  .000  .003**    Cool colors  4.242  .039      Warm colors  8.518  .004      Multiple  8.179  .004    RQ4b  Headline color  Black  18.781*  .000  .000**    White  6.548  .010      Cool colors  3.378  .066      Warm colors  .000  .983      Multiple  6.301  .012    Bodycopy color  Black  2.852  .091      White  9.460*  .002  .042*    Cool colors  1.830  .176      Warm colors  .074  .786      Multiple  26.530*  .000  .000**  RQ5  Font legibility  Headline  1.412  .235      Bodycopy  1.972  .160        X2  p  Adjusted p  RQ1a          Types of human image  Profile  .220  .639    Affected area  19.344*  .000  .000**  Drawing  1.379  .240    RQ1b  Gender  Male  19.715  .000  .000**    Female  15.075  0.00  .002**  Age  Young  3.807  .079      Old  5.653  .017    Skin color  White  10.463  .001  .026*    Non-white  4.840  .028    RQ2  Fear  Low fear  11.607  .001  .015*    High fear  5.037  .025    RQ3  Pin Composition  Picture only  43.825*  .000  .000**    Text only  6.584  .010      Picture and text  13.316*  .000  .006**    Video only  9.649*  .002  .039*    Inforgraphic  44.673*  .000  .000**  RQ4a  Primary color  Skin  17.713*  .000  .000**    Black  12.605*  .000  .009**    White  14.805*  .000  .003**    Cool colors  4.242  .039      Warm colors  8.518  .004      Multiple  8.179  .004    RQ4b  Headline color  Black  18.781*  .000  .000**    White  6.548  .010      Cool colors  3.378  .066      Warm colors  .000  .983      Multiple  6.301  .012    Bodycopy color  Black  2.852  .091      White  9.460*  .002  .042*    Cool colors  1.830  .176      Warm colors  .074  .786      Multiple  26.530*  .000  .000**  RQ5  Font legibility  Headline  1.412  .235      Bodycopy  1.972  .160    df = 1 for all tests. *p <. 05; **p <. 01. View Large Skin colors, gender, and age were coded to assess human models’ demographic characteristics. Out of 357 pins with identifiable skin colors, majority of pins had models with fair skin color (n = 346, 97%). Pins on NMSC were significantly more likely to use white models than other pins. Out of 160 pins with available gender information, approximately 51.8 per cent (n = 83) of pins had female models, whereas 48.2 per cent (n = 77) of pins featured male models. Pins on NMSC were significantly more likely to use male models, whereas pins on melanoma (including skin cancer in general) were more likely to use female models. Finally, out of 142 pins with age information, 54.2 per cent (n = 77) of pins featured younger models and 45.8 per cent of pins (n = 65) had older models. No statistical difference was found between pins on melanoma (including skin cancer in general) and pins on NMSC in terms of the age of the models used. RQ2 focused on the use of fear-invoking images. Only 36.3 per cent (n = 257) included fear-invoking images. Among them, the majority had images of low level of fear (n = 210, 81.7%) and only 18.3 per cent (n = 47) pins had pictures invoking high level of fear. Pins on NMSC were significantly more likely to show low fear appeal than pins on melanoma (including skin cancer in general) and more likely to use high fear appeal as well, even though the difference was not statistically significant. The composition of pins was examined in RQ3. Almost half of the pins were picture only (48.4%, n = 343) and around a third of the pins were picture with texts (36.4%, n = 258). Infographics only made up 6.9 per cent (n = 49) of all pins, followed by video only (4.4%, n = 31) and text only (3.8%, n = 27). Chi-square tests indicated that pins on melanoma skin (including skin cancer in general) were more likely to use infographics as well as picture with texts, whereas pins on NMSC were more likely to be picture only and video only. RQ4a and RQ4b examined the colors of images and texts. In terms of the primary colors of images, skin color (n = 363, 51.3%) was most dominant, followed by cool colors (n = 90, 12.7%), warm colors (n = 85, 12.0%), white (n = 71, 10.0%), black (n = 70, 9.9%), and multiple (n = 29, 4.1%). Pins on NMSC were more likely to use skin color and white as their primary colors, whereas pins on melanoma (including skin cancer in general) were more likely to use black as their primary color. The latter is not surprising since black is the signature color of melanoma. When font colors were concerned, 304 out of 708 pins had a headline and only 201 pins included bodycopy. For headline, black was the most prevalent color (n = 107, 35.2%) followed by white (n = 86, 28.2%), warm color (n = 52, 17.1%), multiple (n = 44, 14.5%), and cool color (n = 15, 4.93%). For bodycopy, black was the most dominant color (n = 80, 39.8%) followed by multiple (n = 58, 29.4%), white (n = 41, 20.4%), warm (n = 17, 8.46%), and cool (n = 5, 2.5%). Although black was the most frequently used font color in headline and bodycopy, cool colors (e.g., blue and green) were the least used. Pins on melanoma (including skin cancer in general) were significantly more likely to use black in its headlines and use white and multiple colors in its bodycopy than pins on NMSC. In terms of the legibility of texts (RQ5), headlines were usually in font-sizes that were easy to read. Among the 306 pins with headlines, approximately 97 per cent (n = 297) had headlines that were highly legible. However, legibility became a serious problem when bodycopy was concerned. Among the 201 pins with bodycopies, 43.8 per cent (n = 88) had bodycopy of low legibility (i.e., impossible to read on a handheld device without enlarging it). There was no significant difference between pins on melanoma (including skin cancer in general) and pins on nonmelanoma in terms of the legibility of either headlines or bodycopy. RQ6 sought to explore how information richness and visual characteristics of pins predicted participative engagement. To obtain a more nuanced and practically useful understanding of the predictors of popularity, we divided data into four different groups: (a) melanoma pins without human models; (b) melanoma pins with human models; (c) nonmelanoma pins without human models; and (d) nonmelanoma pins with human models (Because a large percentage of pins did not contain any human image, indiscriminately combining these pins with pins with human images would introduce too many missing data and render many significant predictions insignificant in the negative binomial regression models.). Then we developed four distinctive multivariate negative binomial regression models using each group of data. For pins without human models, we modeled the number of repins with information richness, stylistic variables (e.g., the compositions of pins, primary color schemes) and the presence/absence of fear-invoking images. For pins with human models, we modeled the number of repins with information richness, stylistic variables, human model-related factors (e.g., display of human body, demographic characters of human models), and the presence/absence of fear-invoking images. As shown in Table 3, among pins about melanoma without human models, only information richness (coefficient = 0.12, p = .04) significantly predicted the number of repins. In contrast, among pins about melanoma with human models, the number of repins was significantly associated with pin composition (p = 2.17e-13 < .001), gender of the human model (p = .02 < .05), and presence of fear-invoking images (p = 0.01 < .05). Post hoc Tukey test showed that picture-only pins and video-only pins received fewer repins than inforgraphics and picture combined with texts, and pins with female models received more repins than those using male models (coefficient = −0.87, p = .0497 < .05), and pins with fear-invoking images had more repins than those without fear-invoking images (coefficient = 0.76, p = .014 < .05). In other words, among melanoma-related pins featuring human models, the most popular ones were inforgraphics or picture combined with texts, pins with female models, and pins with fear-invoking images. Table 3 Results of negative binomial regression models Dependent variable: number of pins  Pins about melanoma coefficient (SE)  Pins about NMSC coefficient (SE)  No human models  With human models  No human models  With human models  INTERCEPT  4.15 (0.47)  5.39 (0.95)  1.14 (1.16)  −0.61 (0.55)  p-value  p < 2e-16***  p = 1.26e-08***  p = .33  p = .26  Richness  0.12 (0.06)  0.02 (0.08)  0.23 (0.09)  0.61 (0.08)  p-value  p = .04*  p = .78  p = .016*  p = 3.04e-14***  Pin composition  p-value  p = .12  p = 2.17e-13***  p = .45  p = .15  Infographic  0a  0a  0a  NAb   Picture + Text  −0.64 (0.27)  −1.49 (0.79)  −0.62 (1.04)  0a   Picture only  −0.55 (0.30)  −3.21 (0.80)***  −0.15 (1.02)  0.62 (0.37)   Text only  −0.30 (0.35)  NAb  −0.95 (1.18)  NAb   Video only  −2.01 (1.23)  −3.83 (0.95)***  −0.99 (1.22)  0.16 (0.46)  Primary color  p-value  p = .55  p = .83  p = .12  p = .26   Skin  0a  0a  0a  0a   Warm color  0.36 (0.34)  −0.60 (0.53)  0.30 (0.45)  1.03 (1.48)   Cool color  0.55 (0.33)  −0.35 (0.45)  −0.11 (0.45)  −0.66 (0.83)   Black  0.66 (0.34)  −0.21 (0.63)  0.66 (0.50)  −1.64 (1.37)   White  0.28 (0.39)  NAb  0.03 (0.45)  −2.00 (1.13)   Multiple  0.32 (0.41)  −0.05 (0.86)  −1.59 (0.82)*  −1.42 (1.16)  Human model  p-value    p = .24    p = .17   Profile    0a    0a   Affected area    −0.57 (0.36)    0.09 (0.42)   Drawing    −0.17 (0.63)    2.01 (0.99)   Multiple    −1.69 (0.83)    1.75 (1.86)  Gender  p-value    p = .02*    p = .08   Male    0a    0a   Female    1.02 (0.34)**    0.28 (0.33)   Mixed    −0.01 (0.60)    −0.19 (0.80)   Unknown    −0.91 (0.48)    0.88 (0.32)  Age  p-value    p = .74    p = .04*   Young    0a    0a   Old    0.45 (0.41)    0.09 (0.32)   Mixed    −0.02 (0.78)    −0.97 (1.00)   Unknown    −0.06 (0.44)    −1.13 (0.40)**  Skin color  p-value    p = .59    p = .06   White    0a    0a   Nonwhite    0.33 (0.60)    −1.05 (0.52)  Fear  p-value  p = .19  p = .0099**  p = .10  p = 0.67   Present  −0.66 (0.45)  0.76 (0.30)*  −1.82 (1.04)  −0.16 (0.34)   Absent  0a  0a  0a  0a  Model p -value  p = .02*  p = 1.02e-11***  p = .0003***  p = 7.70e-12***  Likelihood ratio χ2  22.45  92.89  34.31  91.48  Degree of freedom  11  19  11  19  Number of cases  192  161  147  208  Dependent variable: number of pins  Pins about melanoma coefficient (SE)  Pins about NMSC coefficient (SE)  No human models  With human models  No human models  With human models  INTERCEPT  4.15 (0.47)  5.39 (0.95)  1.14 (1.16)  −0.61 (0.55)  p-value  p < 2e-16***  p = 1.26e-08***  p = .33  p = .26  Richness  0.12 (0.06)  0.02 (0.08)  0.23 (0.09)  0.61 (0.08)  p-value  p = .04*  p = .78  p = .016*  p = 3.04e-14***  Pin composition  p-value  p = .12  p = 2.17e-13***  p = .45  p = .15  Infographic  0a  0a  0a  NAb   Picture + Text  −0.64 (0.27)  −1.49 (0.79)  −0.62 (1.04)  0a   Picture only  −0.55 (0.30)  −3.21 (0.80)***  −0.15 (1.02)  0.62 (0.37)   Text only  −0.30 (0.35)  NAb  −0.95 (1.18)  NAb   Video only  −2.01 (1.23)  −3.83 (0.95)***  −0.99 (1.22)  0.16 (0.46)  Primary color  p-value  p = .55  p = .83  p = .12  p = .26   Skin  0a  0a  0a  0a   Warm color  0.36 (0.34)  −0.60 (0.53)  0.30 (0.45)  1.03 (1.48)   Cool color  0.55 (0.33)  −0.35 (0.45)  −0.11 (0.45)  −0.66 (0.83)   Black  0.66 (0.34)  −0.21 (0.63)  0.66 (0.50)  −1.64 (1.37)   White  0.28 (0.39)  NAb  0.03 (0.45)  −2.00 (1.13)   Multiple  0.32 (0.41)  −0.05 (0.86)  −1.59 (0.82)*  −1.42 (1.16)  Human model  p-value    p = .24    p = .17   Profile    0a    0a   Affected area    −0.57 (0.36)    0.09 (0.42)   Drawing    −0.17 (0.63)    2.01 (0.99)   Multiple    −1.69 (0.83)    1.75 (1.86)  Gender  p-value    p = .02*    p = .08   Male    0a    0a   Female    1.02 (0.34)**    0.28 (0.33)   Mixed    −0.01 (0.60)    −0.19 (0.80)   Unknown    −0.91 (0.48)    0.88 (0.32)  Age  p-value    p = .74    p = .04*   Young    0a    0a   Old    0.45 (0.41)    0.09 (0.32)   Mixed    −0.02 (0.78)    −0.97 (1.00)   Unknown    −0.06 (0.44)    −1.13 (0.40)**  Skin color  p-value    p = .59    p = .06   White    0a    0a   Nonwhite    0.33 (0.60)    −1.05 (0.52)  Fear  p-value  p = .19  p = .0099**  p = .10  p = 0.67   Present  −0.66 (0.45)  0.76 (0.30)*  −1.82 (1.04)  −0.16 (0.34)   Absent  0a  0a  0a  0a  Model p -value  p = .02*  p = 1.02e-11***  p = .0003***  p = 7.70e-12***  Likelihood ratio χ2  22.45  92.89  34.31  91.48  Degree of freedom  11  19  11  19  Number of cases  192  161  147  208  a = Set to zero because this parameter is redundant (baseline category). b = Empty cell: sample size = 0. *p < .05; **p < .01; ***p < .001. View Large Table 3 Results of negative binomial regression models Dependent variable: number of pins  Pins about melanoma coefficient (SE)  Pins about NMSC coefficient (SE)  No human models  With human models  No human models  With human models  INTERCEPT  4.15 (0.47)  5.39 (0.95)  1.14 (1.16)  −0.61 (0.55)  p-value  p < 2e-16***  p = 1.26e-08***  p = .33  p = .26  Richness  0.12 (0.06)  0.02 (0.08)  0.23 (0.09)  0.61 (0.08)  p-value  p = .04*  p = .78  p = .016*  p = 3.04e-14***  Pin composition  p-value  p = .12  p = 2.17e-13***  p = .45  p = .15  Infographic  0a  0a  0a  NAb   Picture + Text  −0.64 (0.27)  −1.49 (0.79)  −0.62 (1.04)  0a   Picture only  −0.55 (0.30)  −3.21 (0.80)***  −0.15 (1.02)  0.62 (0.37)   Text only  −0.30 (0.35)  NAb  −0.95 (1.18)  NAb   Video only  −2.01 (1.23)  −3.83 (0.95)***  −0.99 (1.22)  0.16 (0.46)  Primary color  p-value  p = .55  p = .83  p = .12  p = .26   Skin  0a  0a  0a  0a   Warm color  0.36 (0.34)  −0.60 (0.53)  0.30 (0.45)  1.03 (1.48)   Cool color  0.55 (0.33)  −0.35 (0.45)  −0.11 (0.45)  −0.66 (0.83)   Black  0.66 (0.34)  −0.21 (0.63)  0.66 (0.50)  −1.64 (1.37)   White  0.28 (0.39)  NAb  0.03 (0.45)  −2.00 (1.13)   Multiple  0.32 (0.41)  −0.05 (0.86)  −1.59 (0.82)*  −1.42 (1.16)  Human model  p-value    p = .24    p = .17   Profile    0a    0a   Affected area    −0.57 (0.36)    0.09 (0.42)   Drawing    −0.17 (0.63)    2.01 (0.99)   Multiple    −1.69 (0.83)    1.75 (1.86)  Gender  p-value    p = .02*    p = .08   Male    0a    0a   Female    1.02 (0.34)**    0.28 (0.33)   Mixed    −0.01 (0.60)    −0.19 (0.80)   Unknown    −0.91 (0.48)    0.88 (0.32)  Age  p-value    p = .74    p = .04*   Young    0a    0a   Old    0.45 (0.41)    0.09 (0.32)   Mixed    −0.02 (0.78)    −0.97 (1.00)   Unknown    −0.06 (0.44)    −1.13 (0.40)**  Skin color  p-value    p = .59    p = .06   White    0a    0a   Nonwhite    0.33 (0.60)    −1.05 (0.52)  Fear  p-value  p = .19  p = .0099**  p = .10  p = 0.67   Present  −0.66 (0.45)  0.76 (0.30)*  −1.82 (1.04)  −0.16 (0.34)   Absent  0a  0a  0a  0a  Model p -value  p = .02*  p = 1.02e-11***  p = .0003***  p = 7.70e-12***  Likelihood ratio χ2  22.45  92.89  34.31  91.48  Degree of freedom  11  19  11  19  Number of cases  192  161  147  208  Dependent variable: number of pins  Pins about melanoma coefficient (SE)  Pins about NMSC coefficient (SE)  No human models  With human models  No human models  With human models  INTERCEPT  4.15 (0.47)  5.39 (0.95)  1.14 (1.16)  −0.61 (0.55)  p-value  p < 2e-16***  p = 1.26e-08***  p = .33  p = .26  Richness  0.12 (0.06)  0.02 (0.08)  0.23 (0.09)  0.61 (0.08)  p-value  p = .04*  p = .78  p = .016*  p = 3.04e-14***  Pin composition  p-value  p = .12  p = 2.17e-13***  p = .45  p = .15  Infographic  0a  0a  0a  NAb   Picture + Text  −0.64 (0.27)  −1.49 (0.79)  −0.62 (1.04)  0a   Picture only  −0.55 (0.30)  −3.21 (0.80)***  −0.15 (1.02)  0.62 (0.37)   Text only  −0.30 (0.35)  NAb  −0.95 (1.18)  NAb   Video only  −2.01 (1.23)  −3.83 (0.95)***  −0.99 (1.22)  0.16 (0.46)  Primary color  p-value  p = .55  p = .83  p = .12  p = .26   Skin  0a  0a  0a  0a   Warm color  0.36 (0.34)  −0.60 (0.53)  0.30 (0.45)  1.03 (1.48)   Cool color  0.55 (0.33)  −0.35 (0.45)  −0.11 (0.45)  −0.66 (0.83)   Black  0.66 (0.34)  −0.21 (0.63)  0.66 (0.50)  −1.64 (1.37)   White  0.28 (0.39)  NAb  0.03 (0.45)  −2.00 (1.13)   Multiple  0.32 (0.41)  −0.05 (0.86)  −1.59 (0.82)*  −1.42 (1.16)  Human model  p-value    p = .24    p = .17   Profile    0a    0a   Affected area    −0.57 (0.36)    0.09 (0.42)   Drawing    −0.17 (0.63)    2.01 (0.99)   Multiple    −1.69 (0.83)    1.75 (1.86)  Gender  p-value    p = .02*    p = .08   Male    0a    0a   Female    1.02 (0.34)**    0.28 (0.33)   Mixed    −0.01 (0.60)    −0.19 (0.80)   Unknown    −0.91 (0.48)    0.88 (0.32)  Age  p-value    p = .74    p = .04*   Young    0a    0a   Old    0.45 (0.41)    0.09 (0.32)   Mixed    −0.02 (0.78)    −0.97 (1.00)   Unknown    −0.06 (0.44)    −1.13 (0.40)**  Skin color  p-value    p = .59    p = .06   White    0a    0a   Nonwhite    0.33 (0.60)    −1.05 (0.52)  Fear  p-value  p = .19  p = .0099**  p = .10  p = 0.67   Present  −0.66 (0.45)  0.76 (0.30)*  −1.82 (1.04)  −0.16 (0.34)   Absent  0a  0a  0a  0a  Model p -value  p = .02*  p = 1.02e-11***  p = .0003***  p = 7.70e-12***  Likelihood ratio χ2  22.45  92.89  34.31  91.48  Degree of freedom  11  19  11  19  Number of cases  192  161  147  208  a = Set to zero because this parameter is redundant (baseline category). b = Empty cell: sample size = 0. *p < .05; **p < .01; ***p < .001. View Large Among pins about nonmelanoma without human models, information richness was the only significant predictor of number of repins: images with higher levels of information richness tended to have more repins compared with those with lower levels of richness (coefficient = 0.23, p = .016 < .05). Among pins about nonmelanoma that used human models, the number of repins was significantly associated with information richness (coefficient = 0.61, p = 3.04e-14 < .001) and age of human models (p = .04 < .05). Post hoc Tukey test showed that among nonmelanoma pins with human models, pins with human images for which age cannot be identified received less pins than both pins featured young models (coefficient = −1.13, p = .02) and pins featured older models (coefficient = −1.06, p = .01). DISCUSSION This study examines the portrayal of skin cancer on Pinterest with a focus on the visual characteristics of pins. Around half of the pins on skin cancer used human images. Among them, nearly 70 per cent only show a body part without the face and only 26 per cent used profile picture with a person’s face and upper torso. Using headless pictures of human models might focus audiences’ attention on areas affected by skin cancer and provide anonymity to cancer patients. However, it might also stigmatize patients with cancer in characterizing them by their illness [43]. Pins on NMSC were more likely to show affected areas without human face than other pins. This is actually an accurate depiction of NMSC since it often affected areas other than neck and head [44]. Contrary to the assumption that human models used on Pinterest will be mainly young women based on the demographics of Pinterest users, this study finds an almost 50-50 split between men and women and between young and old (50 years or older). Pins on NMSC are more likely to use male models. This also reflects the demographics of NMSC patients [44]. In terms of the skin color of models, 97 per cent of the human models have fair skin color associated with Caucasians and sometimes Hispanics. Epidemiological data indicate that Caucasians are the primary victims of skin cancer. However, people of color are also at risk of skin cancer [45]. Hu et al. [46] found that the diagnosis of melanoma is often delayed among African Americans and Hispanics. The predominance of fair-skinned human models used in pins on skin cancer might contribute the misconception that colored people are immune to skin cancer and create further gaps in early diagnosis and effective treatment of skin cancer among different racial groups. Visual images invoking fear are found in one third of the pins; however, only 5 per cent of pins contain highly disturbing or fearful images. This is consistent with the use of Pinterest for visual curation. Highly disturbing or fearful images are probably not favored by users who care about the aesthetics of their pin boards. Around half of all pins on skin cancer are picture only and another third are pictures with texts. This confirms the centrality of visual images on Pinterest. Another 7 per cent of the pins are infographics. Future research should further examine how Pinterest users respond to these different formats and which format holds the greatest potential in getting attention and aiding comprehension. Color is another important aspect of visual communication. Black is the most used font color for both headlines and bodycopy. This is encouraging given the evidence that dark shade of fonts on a light background was more visible and readable especially when black and dark blue fonts were used [47]. Text legibility is a potential problem. Nearly half of pins with bodycopy have texts that are illegible on a handheld device such as an iPhone. This poses barriers to the comprehension of skin cancer information on Pinterest as 80 per cent of users report using mobile phones as the primary device to access information on Pinterest [30]. The reason for such low legibility could be that a lot of the pins are created as advertisements or public service announcements and not intended for devices with small screens. If public health professionals want to utilize Pinterest or other similar visual based social media platforms to disseminate information, they need to make sure that their messages include legible texts. This study contributes to the research on health-related visual images because it explores the relationship between visual characteristics of pins on skin cancer and audiences’ participative engagement measured by the number of repins. It finds different models for pins on melanoma and NMSC, with and without the use of human models. For pins on melanoma without a human model, only information richness is significant in predicting number of repins. For pins on melanoma with a human model, information richness is no longer significant, whereas visual characteristics including pin composition, gender of the human model, and presence of fear-invoking images predict the popularity of the pins. Among them, pins with picture only and video only are less likely to be repined than inforgraphics as well as picture with texts. One possible explanation is that pins without any texts might appear confusing and thus are less likely to be repined. Future efforts on melanoma prevention and skin cancer prevention in general on Pinterest should use the format of infographic or picture blended with texts because these two compositions generate more participation. Female models increase the popularity of pins on melanoma skin cancer with human models. It could be that since Pinterest users are predominantly female, they are more likely to repin those pins with human models they identify with. Interestingly, fear-invoking images increase repins. Notice that a large majority of these images are categorized as of low fear, typically pictures of benign images of early signs of skin cancer. These images make the pins more instead of less popular probably because pins with such images are considered informative. For pins on NMSC with or without of human models, informational richness is positively associated with the number of repins. This finding supports the belief that more information is better. However, for pins on NMSC with a human model, age of human models predicts the number of repins. Pins with human models that age are unidentifiable are less likely to be repined than pins with either young or old models. It shows that in creating popular pins about NMSC, public health professionals should use images with vivid human faces and avoid using images with drawings of human models or images of affected areas only. Overall, the primary color of pins does not predict their popularity across all four models. This finding is surprising because past research has suggested that cool colors are preferred over warm colors [48]. One possible explanation is that one’s evaluation of color varies by a number of factors such as age, gender, and culture [49]. Future research should investigate how Pinterest users with different demographic characteristics respond to colors differently. Limitations Most obviously, the dependent variable examined is the number of repins, or the popularity of pins. Repining a pin on her own board does not necessarily mean that a user is going to actually process the information included in the pin or to adopt the behaviors suggested. Future experiment studies should be conducted to further examine how information richness as well as different visual characteristics influence not only the likelihood users are going to repin a pin about skin cancer or other health-related topics, but also the ways in which they are going to process the information, their enjoyment of the pins, their recall of information, and their intentions for behavior change. Second, the models about the popularity of pins proposed in this paper rely on an imperfect dependent variable: number of repins. One important factor that is likely to influence the number of repins is the amount of time since the pictures used in these pins have been created and made available online. A picture that has been online for a relatively short period of time is likely to have less repins than a picture that has been there for years. However, the length of a picture’s availability online is unfortunately unavailable because Pinterest users can essentially pin any picture online outside of Pinterest. CONCLUSIONS Visual-based social media sites such as Pinterest hold great potential in disseminating health information and promoting behavioral changes. This paper provides an overview of the visual characteristics of pins on skin cancer in terms of pin composition, text legibility, color, use of human models, and use of fear-invoking messages. Furthermore, it proposes the first model of the relationship between visual characteristics and information richness and the popularity of pins. The popularity of pins about different types of skin cancer (melanoma with human models, melanoma without human models, NMSC with human models, and NMSC without human models) is predicted by different variables. Overall, information richness is the only predictor of number of repins among pins that do not use human models (including pins on both melanoma and NMSC). Information-rich pins are more likely to be repined. For pins about melanoma with human models, pin composition, gender of human model, and the use of fear-invoking images are associated with the number of repins. Pins in the form of infographic or picture combined with texts use female models and include fear-invoking messages that are more popular than other pins. Among pins about NMSC with human models, those pins that show the age of the model (either old or young) are more likely to be repined than pins that do not show the face of the model. These findings provide useful tools for public health professionals who wish to create and disseminate skin cancer–related information via social media. Compliance with Ethical Standards Author contributions: Sung-Eun Park and Lu Tang designed the study and coded the pins. Bijie Bie conducted data analysis. Degui Zhi provided advice on the statistical model and analysis and the interpretation of results. Sung-Eun Park and Lu Tang drafted the manuscript, and all authors read, edited, and approved the final manuscript. Conflict of Interest: Sung-Eun Park, Lu Tang, Bijie Bie, and Degui Zhi declare that they have no conflicts of interest. Ethical Approval: This article does not contain any studies with human participants or animals performed by any of the authors. Informed Consent: This article has no human participants and therefore informed consent was not required. References 1. Houts PS, Doak CC, Doak LG, Loscalzo MJ. The role of pictures in improving health communication: a review of research on attention, comprehension, recall, and adherence. Patient Educ Couns . 2006; 61( 2): 173– 190. Google Scholar CrossRef Search ADS PubMed  2. Dowse R, Ehlers MS. The evaluation of pharmaceutical pictograms in a low-literate South African population. Patient Educ Couns . 2001; 45( 2): 87– 99. Google Scholar CrossRef Search ADS PubMed  3. Robins D, Holmes J, Stansbury M. Consumer health information on the Web: the relationship of visual design and perceptions of credibility. J Am Soc Inf Sci Technol . 2010; 61 (1): 13– 29. Google Scholar CrossRef Search ADS   4. Pew Research Center. Mobile Messaging and Social Media 2015; 2015. Available at http://www.pewinternet.org/2015/08/19/mobile-messaging-and-social-media-2015/. Accessibility verified October 1, 2016. 5. Opallo D. Pin down Pinterest’s value. DMN . 2012; 34( 8): 19– 20. 6. Duggan M, Brenner J. The Demographics of Social Media Users . Washington, DC: Pew Research Center’s Internet & American Life Project; 2013. 7. Guidry JP, Carlyle K, Messner M, Jin Y. On pins and needles: how vaccines are portrayed on Pinterest. Vaccine . 2015; 33( 39): 5051– 5056. Google Scholar CrossRef Search ADS PubMed  8. Guidry J, Zhang Y, Jin Y, Parrish C. Portrayals of depression on Pinterest and why public relations practitioners should care. Public Relat Rev . 2016; 42( 1): 232– 236. Google Scholar CrossRef Search ADS   9. Whitsitt J, Mattis D, Hernandez M, Kollipara R, Dellavalle RP. Dermatology on pinterest. Dermatol Online J . 2015; 21( 1). 10. American Cancer Society. Cancer facts & figures; 2015. Available at http://www.cancer.org/research/cancerfactsstatistics/cancerfactsfigures2015/ . Accessibility verified October 1, 2016 11. Lomas A, Leonardi-Bee J, Bath-Hextall F. A systematic review of worldwide incidence of nonmelanoma skin cancer. Br J Dermatol . 2012; 166( 5): 1069– 1080. Google Scholar CrossRef Search ADS PubMed  12. Entman RM. Framing: towards clarification of a fractured paradigm. J Commun . 1993; 43( 4): 51– 58. Google Scholar CrossRef Search ADS   13. Rodriguez L, Asoro RL. Visual representation of genetic engineering and genetically modified organisms in the online media. Visual Commun Quart . 2012; 19(4): 232– 245. Google Scholar CrossRef Search ADS   14. King AJ. A content analysis of visual cancer information: prevalence and use of photographs and illustrations in printed health materials. Health Commun . 2015; 30( 7): 722– 731. Google Scholar CrossRef Search ADS PubMed  15. Phillips SG, Della LJ, Sohn SH. What does cancer treatment look like in consumer cancer magazines? An exploratory analysis of photographic content in consumer cancer magazines. J Health Commun . 2011; 16( 4): 416– 430. Google Scholar CrossRef Search ADS PubMed  16. Rodriguez L, Dimitrova DV. The levels of visual framing. J Visual Literacy . 2011; 30: 48– 65. Google Scholar CrossRef Search ADS   17. Paek HJ, Reid LN, Choi H, Jeong HJ. Promoting health (implicitly)? A longitudinal content analysis of implicit health information in cigarette advertising, 1954-2003. J Health Commun . 2010; 15( 7): 769– 787. Google Scholar CrossRef Search ADS PubMed  18. Bandura A. Organizational applications of social cognitive theory. Aus J Management . 1988; 13(2): 275– 302. Google Scholar CrossRef Search ADS   19. Leshner G, Bolls P, Thomas E. Scare’ em or disgust ‘em: the effects of graphic health promotion messages. Health Commun . 2009; 24( 5): 447– 458. Google Scholar CrossRef Search ADS PubMed  20. Witte K, Allen M. A meta-analysis of fear appeals: implications for effective public health campaigns. Health Educ Behav . 2000; 27( 5): 591– 615. Google Scholar CrossRef Search ADS PubMed  21. Schienle A, Schäfer A, Stark R, Walter B, Vaitl D. Gender differences in the processing of disgust- and fear-inducing pictures: an fMRI study. Neuroreport . 2005; 16( 3): 277– 280. Google Scholar CrossRef Search ADS PubMed  22. Occa A, Suggs LS. Communicating breast cancer screening with young women: an experimental test of didactic and narrative messages using video and infographics. J Health Commun . 2016; 21( 1): 1– 11. Google Scholar CrossRef Search ADS PubMed  23. Featherstone RM. Visual research data: an infographics Primer. J Can Health Libr Assoc . 2014; 35(3): 147– 150. Google Scholar CrossRef Search ADS   24. Couper MP, Conrad FG, Tourangeau R. Visual context effects in web surveys. Public Opin Q . 2007; 71(4): 623– 634. Google Scholar CrossRef Search ADS   25. Moshagen M, Thielsch MT. Facets of visual aesthetics. Int J Hum Comput Stud . 2010; 68(10): 689– 709. Google Scholar CrossRef Search ADS   26. Cyr D, Head M, Larios H. Colour appeal in website design within and across cultures: a multi-method evaluation. Int J Hum Comput Stud . 2010; 68: 1– 21. Google Scholar CrossRef Search ADS   27. Madden TJ, Hewett K, Roth MS. Managing images in different cultures: a cross- national study of color meanings and preferences. J Int Market . 2000; 8( 4): 90– 107. Google Scholar CrossRef Search ADS   28. Welhausen CA. Visualizing a non-pandemic: considerations for communicating public health risks in intercultural contexts. Tech Commun . 2015; 62: 244– 257. 29. Cline RJ, Haynes KM. Consumer health information seeking on the Internet: the state of the art. Health Educ Res . 2001; 16( 6): 671– 692. Google Scholar CrossRef Search ADS PubMed  30. Novet J. 80 percent of Pinterest’s traffic comes from mobile devices. VentureBeat . Feb 18, 2016. Available at http://venturebeat.com/2015/03/31/80-percent-of-pinterests-traffic-comes-from-mobile-devices/. Accessibility verified October 1, 2016. 31. Darroch I, Goodman J, Brewster S, Gray P. The effect of age and font size on reading text on handheld computers. In: Costabile MF, Paternò F, eds. Human-Computer Interaction-INTERACT 2005 . Berlin, Germany: Springer; 2005: 253– 266. Google Scholar CrossRef Search ADS   32. Han J, Choi D, Chun BG, Kwon TT, Kim HC, Choi Y. Collecting, organizing, and sharing pins in Pinterest: interest-driven or social-driven? Paper presented at: The 2014 ACM international conference on Measurement and modeling of computer systems; June 2014 , Austin, Texas. 33. Bernardini C, Silverston T, Festor O. A Pin is worth a thousand words: characterization of publications in Pinterest. In: Wireless Communications and Mobile Computing Conference (IWCMC), 2014 International IEEE; Nicosia, Cyprus: August, 2014. 322– 327. 34. Mull IR, Lee SE. “PIN” pointing the motivational dimensions behind Pinterest. Comput Human Behav . 2014; 33: 192– 200. Google Scholar CrossRef Search ADS   35. Guidry JP, Carlyle K, Messner M, Jin Y. On pins and needles: how vaccines are portrayed on Pinterest. Vaccine . 2015; 33( 39): 5051– 5056. Google Scholar CrossRef Search ADS PubMed  36. Gilbert E, Bakhshi S, Chang S, Terveen L. I need to try this?: a statistical overview of pinterest. In: Proceedings of the SIGCHI conference on human factors in computing systems; Apr 27-May 2, 2013; Paris, France. 2427– 2436. 37. Neuendorf KA. The Content Analysis Guidebook . Thousand Oaks, CA: Sage; 2002. 38. Stephenson MT, Witte K. Fear, threat, and perceptions of efficacy from frightening skin cancer messages. Public Health Rev . 1998; 26( 2): 147– 174. Google Scholar PubMed  39. Bellizzi JA, Crowley AE, Hasty RW. The effects of color in store design. J Retail . 1983; 59(1): 21– 45. 40. Tang L, Park SE. Sun exposure, tanning beds, and herbs that cure: an examination of skin cancer on pinterest. Health Commun . 2017; 32( 10): 1192– 1200. Google Scholar CrossRef Search ADS PubMed  41. Potter J, Levine-Donnerstein D. Rethinking validity and reliability in content analysis. J Applied Commun Res . 1999; 27(3): 258– 284. Google Scholar CrossRef Search ADS   42. Allison PD. Fixed Effects Regression Models . Thousand Oaks, CA: Sage; 2009. Google Scholar CrossRef Search ADS   43. Puhl RM, Peterson JL, DePierre JA, Luedicke J. Headless, hungry, and unhealthy: a video content analysis of obese persons portrayed in online news. J Health Commun . 2013; 18( 6): 686– 702. Google Scholar CrossRef Search ADS PubMed  44. Diepgen TL, Mahler V. The epidemiology of skin cancer. Br J Dermatol . 2002; 146( Suppl 61): 1– 6. Google Scholar CrossRef Search ADS PubMed  45. Gloster HMJr, Neal K. Skin cancer in skin of color. J Am Acad Dermatol . 2006; 55( 5): 741– 760. Google Scholar CrossRef Search ADS PubMed  46. Hu S, Parmet Y, Allen G, et al.   Disparity in melanoma: a trend analysis of melanoma incidence and stage at diagnosis among whites, Hispanics, and blacks in Florida. Arch Dermatol . 2009; 145( 12): 1369– 1374. Google Scholar CrossRef Search ADS PubMed  47. Greco M, Stucchi N, Zavagno D, Marino B. On the portability of computer-generated presentations: the effect of text-background color combinations on text legibility. Hum Factors . 2008; 50( 5): 821– 833. Google Scholar CrossRef Search ADS PubMed  48. Marcus A, Gould EW. Crosscurrents: cultural dimensions and global Web user-interface design. Interactions . 2000; 7( 4): 32– 46. Google Scholar CrossRef Search ADS   49. Ou LC, Luo MR, Woodcock A, Wright A. A study of colour emotion and colour preference. Part I: colour emotions for single colours. Color Res App . 2004; 29 (3): 232– 240. Google Scholar CrossRef Search ADS   © Society of Behavioral Medicine 2018. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Translational Behavioral Medicine Oxford University Press

All pins are not created equal: communicating skin cancer visually on Pinterest

Loading next page...
 
/lp/ou_press/all-pins-are-not-created-equal-communicating-skin-cancer-visually-on-80Z0mQxHWe
Copyright
© Society of Behavioral Medicine 2018. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
ISSN
1869-6716
eISSN
1613-9860
D.O.I.
10.1093/tbm/iby044
Publisher site
See Article on Publisher Site

Abstract

Abstract Skin cancer is the second most common cancer affecting women younger than 39 years in the USA. As a female-oriented social media, Pinterest could be effectively used in reaching this particular demographic group for the purpose of skin cancer education. We analyze the visual characteristics of skin cancer pins, including use of human image, use of fear-invoking image, pin composition, color, and text legibility. We also explore how these visual characteristics as well as information richness predict Pinterest users’ participative engagement. A combination of descriptive and predictive content analysis of 708 pins is conducted. The demographic characteristics of human models are consistent with epidemiology data. Text legibility in bodycopy is low. Information richness is a significant predictor of number of repins in all pins except pins on melanoma with human models. In the case of latter, pin composition, gender of human models, and fear-invoking images were associated with the number of repins. A number of visual characteristics as well as information richness significantly predict Pinterest users’ participant engagement with pins on skin cancer. Public health professionals should consider these factors in creating effective prevention messages to be circulated on Pinterest. Implications Research: This work proposes a comprehensive model for the analysis of visual characteristics of Pinterest contents. It demonstrates how different visual characteristics predict Pinterest users’ participative engagement. Practice: Public health professionals could match visual and textual characteristics of pins with types of skin cancer to create maximum participative engagement. Policy: More support is needed to fund future research on visual communication in the context of health promotion on social media. INTRODUCTION Pictures are essential to health education. Adding highly relevant pictures to health promotion texts generates more attention to and recall of the message [1]. Pictures showing the relationship among ideas increase the comprehension of health messages, especially among patients with low literacy [2]. People also rely on visual cues to assess the credibility of the health information online [3]. Today’s Internet users can easily share images online. Pinterest is one of the fastest growing social media dedicated to photo sharing with approximately 31 per cent of Internet users having used Pinterest in 2015 [4]. It is also the first social media platform that frames information with a female orientation [5]. Pinterest has a saturated population of white females under age 40 with high education [6]. Health is a prominent topic on Pinterest. Researchers have started to examine the presentation of a number of health-related topics on Pinterest, including vaccination [7], depression [8], and skin regiment [9]. Presented here is a study of the visual characteristics of pins about skin cancer on Pinterest.com. Skin cancer is the most common cancer affecting the U.S. population and is the second most common cancer for Caucasian women younger than 39 years [10]. Although melanoma skin cancer received the most attention due to its malignancy, nonmelanoma skin cancers (NMSCs), such as basal-cell carcinoma and squamous-cell carcinoma, are less known [11]. Guided by the theory of visual framing, the current study offers a comprehensive framework to analyze the visual characteristics of skin cancer messages on Pinterest. It explores how these visual characteristics as well as information richness predict users’ participative engagement. Practically, this study provides tangible guidelines for public health professionals in creating health promotion messages to be disseminated on Pinterest and possibly other visual-heavy social media. THE VISUAL FRAMING OF SKIN CANCER ON PINTEREST Media systematically influence the ways in which audiences perceive an event through framing [12]. Framing can be achieved through both verbal messages and visual images. Rodriguez and Dimitrova [13] identified four levels of visual framing analysis. On the most basic level, scholars examine the denotative meanings or the themes of visual images (e.g., what are the objects and elements included in the picture). The second level involves a stylistic analysis of the image in terms of camera angle, lighting, distance between the camera and the object, color, composition of the picture, etc. A reading of the connotative meaning of the image represents the third level of visual framing analysis. On this level, images are interpreted as symbols communicating socially constructed meanings. The fourth level focuses the ideological messages of the visual image. Visual framing has been used to analyze health-related visual images, including cancer [14], cancer treatment [15], genetically modified foods [16], and cigarette advertising [17]. Majority of these studies concentrate on the first and second levels of analysis, which is probably because the first and second levels of analysis allow relatively reliable coding [16]. The current study will focus on these two levels too. Use of human models The first level of visual framing analysis focuses on different types of objects included in in the visual image and their denotative meanings. The use of human images has been frequently examined in the study of health-related visual images. According to the social cognitive theory, individuals learn through the process of modeling and they are more likely to adopt the suggested behavior if they identify with the model [18]. Phillips et al. [15] found that the models used in cancer consumer magazines were more likely to be female, white, and younger, different from the demographics of actual cancer patients. They argued that such discrepancy would discourage patients from processing the information and adopting the recommendations provided in such magazines. Hence, it is important to examine whether pins on skin cancer include human models and whether these models reflect the population at high risk of skin cancer. This leads to the first two research questions (RQ1a and RQ1b): RQ1a: To what extent do pins on skin cancer use human models? RQ2b: What are the demographic characters of human models (in terms of age, gender, and skin color)? Use of fear-invoking images The use of fear-invoking images is another component of first-level visual framing. One of the most common strategies utilized in health promotion is fear appeal, especially through the use of scary visual images [19]. According to the Extended Parallel Processing Model, individuals would go through several processes when faced with the fear appeal. First, they will evaluate their susceptibility to the threat. The more they believe they are susceptible, the more likely they are going to assess the efficacy of the recommended behavior. High level of perceived threat and high level of perceived efficacy lead to more behavioral change. On the contrary, individuals are likely to deny the threat when they doubt the effectiveness of the recommended behavior [20]. Furthermore, women are more likely to respond to disgust- and fear-inducing images than men [21]. Since majority of Pinterest’s users are woman, it is especially important to examine the use of fear-invoking images in pins on skin cancer: RQ2: To what extent do pins on skin cancer include fear-invoking images? The analysis of stylistic features of images constitutes the second level of visual framing analysis. Past research identified several stylistic characters important to health-related visual images, including image composition, color, and text legibility. Pin composition Researchers have studied the composition of pins in terms of whether they are picture only, text only, video only, picture combined with texts, and infographics [7]. Pictures are important in health education as they have the abilities to generate more attention to the textual information and create more comprehension when pictures are related to texts [1]. Videos represent another effective channel of health communication in terms of increasing knowledge and influencing psychological variables associated with behavior change, such as awareness, self-efficacy, and personal vulnerability [22]. Inforgraphics visualize complicated information through a combination of texts, numbers, charts, and graphs [23]. The use of infographics increases the comprehension of message among people with low health literacy and numeracy and is effective in promoting health behaviors in general [24]. Since the composition of a pin is likely to affect the how much attention is paid to the pin and how information is processed, RQ3 is proposed: RQ3: What are the compositions of pins on skin cancer (picture only, text only, videoonly, picture and text, and inforgraphics)? Colors Color generates cognitive and emotional arousals affecting one’s assessment of an image [25]. Researchers often define colors in terms of warm and cool colors [26]. Across many cultures, cool colors such as blue and green are often considered peaceful and calming, whereas warm colors such as red and orange are perceived to be hot, vibrant, and emotional [27]. In the visual communication of health information, warm colors are often associated with heightened risks [28]. Lighter colors such as white, silver, and light blue were used in most of the cigarette advertisements across decades to convey the impression of “relaxation, freshness, and calmness” (p. 782) [17]. Thus, the next two RQs are proposed: RQ4a: What are the primary color schemes of pins on skin cancer? RQ4b: What are the color schemes of the fonts used in pins on skin cancer? Legibility Legibility of text is another visual element contributing to the quality of health messages [29]. Legibility increases individuals’ comfort level in their information use and is especially important on Pinterest as 80 per cent of its users reportedly use mobile phones as their primary device to access Pinterest [30]. A study on font size on handheld device revealed that the range of font size preference was greater among older generation than among young people; however, as long as the font size was greater than 8, both young and older people show satisfactory rating on legibility [31]. Having health information that is illegible obviously will prevent the proper processing of such information. This leads to the next RQ: RQ5: To what extent are the texts in pins legible? Participative engagement Social media enables audience’s participative engagement in terms of likes, comments, retweets, and shares. In the case of Pinterest, if a user likes a particular pin, she could repin it to one of her own boards. Social media content generating more participative engagement is usually more effective in influencing audience’s health beliefs and behaviors. Han et al. [32] found that the characteristics of a pin instead of characteristics of the user (such as number of followers) predict the number of repins. Since Pinterest relies heavily on the attractiveness of images and the platform is based on redistributing images among users [33], the visual characteristics of pins on skin cancer are likely to affect the likelihood these pins will be repined. In addition, since one of the major gratifications of Pinterest use is to store, categorize, and share information [34], information richness, that is, the amount of information about skin cancer in a pin, might also influence the perceived usefulness of the pin and its popularity in terms of number of repins. Studies [35, 36] suggested that the re-pin function was the strongest indication of user interaction on Pinterest and a positive correlation was found between the combined number of likes and comments and the number of repins. It is possible that both information richness and the visual characteristics will affect the likelihood an image gets repinned. Hence, the last RQ is proposed: RQ6: To what extent do information richness and visual characteristics of pins explain Pinterest users’ participative engagement in terms of the number of repins? METHOD A combination of descriptive and predictive content analysis was conducted to answer the research questions proposed above. Different from regular content analysis devoted to the description of media contents, predictive content analysis examines how content characteristics are likely to influence audiences’ psychological or behavioral outcomes [37]. Sampling Four keywords (skin cancer, melanoma, basal-cell carcinoma, and squamous-cell carcinoma) were used to search Pinterest.com. Since the platform of Pinterest did not allow to collect the data of a specific time frame, the only feasible option to collect data was scrolling down the page until no more pins would appear. The authors monitored the pages every day for 3 weeks between September 16 and October 9, 2015. A total of 708 pins were collected after excluding repetitive and nonrelevant pins. Each pin was assigned a unique ID number. Unit of analysis and measurement An individual pin was the unit of analysis and was coded based on the following items. Types of skin cancer Each pin was coded based on the type of skin cancer mentioned in the textual information: (i) skin cancer in general without mentioning any specific type, (ii) melanoma, (iii) squamous-cell carcinoma, and (iv) basal-cell carcinoma. Display of human body First, the display of human body was coded as present or absent. Types of human pictures were further coded into one of the following four categories: profile photo (upper torso with face shot), affected areas of body without face shot, drawing of a person (illustration), and multiple (a combination of more than one element). Next, age, skin color, and gender of human models were coded. Age was coded into four categories: old (when the model appeared to be older than 50 years old), young (when the model appeared to be younger than 50 years old), mixed (when both young and old models were present), and unable to tell (when the face of model was not shown). Skin color was coded into four different categories: white, black, brown, and mixed and these four categories were later merged into two categories: white and nonwhite (including any pins containing a nonwhite human model). Gender was coded as male, female, mixed, and unknown (when only part of the body was shown and gender of the model could not be decided). Fear-invoking images Each pin was coded in terms of whether it had images that could invoke fear, that is, visual images of skin cancer such as discoloration, blisters, and deformation of skin parts (present or absent). Next, pins containing fear-invoking messages were further coded into two groups: high fear and low fear. The mere presence of benign image of skin cancer such as discoloration or early signs of skin cancer (such as a mole) was coded as low fear, whereas clear indications of late stages of skin cancer (tumors, ulceration, or scars) were coded as high fear [38]. Pin composition The composition of pins was coded as one of the following: picture only, text only, video only, picture and text, and inforgraphic [7]. Picture and text were coded when the pin had prominent texts coupled with pictures. A pin was coded as an infographic when it communicated complicated information through a combination of texts, numbers, charts, and graphs [23]. Colors Two types of color are coded: primary picture color and font color. Primary picture color could be the color of a dominant object in the pin (for instance, skin or a piece of fruit) or the background color when such a dominant object is absent. Six primary picture colors are coded: black, white, cool colors, warm colors, skin color, and multiple colors. Cool color includes blue and green, whereas warm colors include red, yellow, orange, and pink [39]. Font colors are coded separately for headline and bodycopy. Five types of font colors were coded: black, white, cool colors, warm colors, and multiple colors. Font legibility When a pin contains texts (either as headline or bodycopy), the legibility of the texts was coded. Headline and bodycopy were evaluated separately. Legibility was coded as “high” if the text was readable on an iPhone without enlarging the picture and coded as “low” if the text was not readable. Information richness Textual information was coded in terms of susceptibility, severity, benefit, barriers, cue to action, self-efficacy, causes, symptoms, treatment, and prevention. Details of these items were reported in a previous study [40]. Information richness was calculated in terms of the number of these items present in a particular pin. The information richness scores ranged from 1 to 10 (M = 3.57, SD = 1.46). Participant engagement It was coded as the number of repins. Coding and intercoder reliability After training and test coding, the first author coded all the pins in the sample and the second author coded 10 per cent of the pins (n = 71) selected through systematic sampling. Percentage agreement was calculated and all of items reached over 85%–90% of agreement except for two items: primary color (84%) and bodycopy color (77%). Percentage agreement is more appropriate than Cohen’s κ or Krippendorff’s α when the distribution of the measure is highly skewed, because latter measurements tend to make the intercorder reliability artificially low even when coding is reasonably reliable [41]. Data analysis Descriptive statistics such as frequencies and percentages were used to answer RQs1–5. Chi-square tests were run to compare the differences among pins about skin cancer in general, melanoma, and NMSCs. Sequential bonferroni correction was used to account for the effects of multiple testing. Pair-wise comparison among these three types of pins showed that pins about skin cancer in general and pins about melanoma were very similar in that none of the chi-square tests yielded significant results after adjusting for multiple testing. Consequently, these two categories were collapsed and chi-square tests were run to compare two types of pins: melanoma (including skin cancer in general) and NMSC. To answer RQ6, negative binomial regression models were used to explore how information richness, stylistic variables (e.g., the compositions of pins, primary color schemes), human model-related factors (e.g., display of human body, demographic characters of human models), and fear-invoking image predicted the popularity of a pin (i.e., number of repins). Negative binomial regression analyses were chosen because the dependent variable (number of repins) is a count variable whose distribution was highly skewed (skewness = 7.344) and it contained a large number of zeros [42]. Data were analyzed using R Studio version 3.2.4. RESULTS Among the 708 pins in the sample, 27.4 per cent (n = 194) were about skin cancer in general, 22.5 per cent (n = 159) focused on melanoma, and the rest were about NMSCs (50.2%, n = 355) (see Table 1 for the descriptive statistics of all variables categorized by the types of skin cancer). Table 1 Results for descriptive analysis of pins on different types of skin cancer     Skin cancer (n = 194)  Melanoma (n = 159)  Nonmelanoma (n = 355)      n  %  n  %  n  %  RQ1a  Types of human images  Profile picture  27  13.9  23  14.5  46  13.0  Affected area  60  30.9  40  25.2  157  36.3  Drawing  2  1.0  6  3.8  4  1.1  Multiple  2  1.0  1  .6  1  .3  RQ1b  Gender  Male  13  6.7  7  4.4  57  16.1    Female  36  18.6  22  13.8  25  7.0    Unable to tell  39  20.1  34  21.4  119  27.1  Age  Young  13  6.7  7  4.4  57  16.1    Old  36  18.6  22  13.8  7  2.0  Skin color  White  89  45.9  62  39.0  195  54.9    Nonwhite  2  1.0  5  3.1  4  1.4  RQ2  Fear  No fear  137  70.6  116  73.0  198  55.8    High fear  11  5.7  5  3.1  31  8.7    Low fear  46  23.7  38  23.9  126  35.5  RQ3  Pin Composition  Picture only  83  42.8  44  27.7  216  60.8  Text only  10  5.2  10  6.3  7  2.0  Picture and text  71  36.6  81  50.9  106  29.9  Video only  5  2.6  2  1.3  24  6.8  Infographics  25  12.9  22  13.8  2  .6  RQ4a  Primary color  Skin  90  46.4  63  39.6  210  59.2    Black  14  7.2  35  22.0  21  5.9    White  10  5.2  10  6.3  51  14.4    Cool colors  30  15.5  24  15.1  36  10.1    Warm colors  37  19.1  18  11.3  30  8.5    Multiple colors  13  6.7  9  5.7  7  2.0  RQ4b  Headline color  Black  26  13.4  48  30.2  33  9.3  White  29  14.9  25  15.7  32  9.0  Cool colors  7  3.6  4  2.5  4  1.1  Warm colors  17  8.8  9  5.7  26  7.3  Multiple colors  14  7.2  16  10.1  14  3.9  Bodycopy color  Black  22  11.3  25  15.7  33  9.3  White  15  7.7  15  9.4  11  3.1  Cool colors  2  1.0  2  1.3  1  .3  Warm colors  4  2.1  4  2.5  7  2.0  Multiple colors  19  9.8  30  18.9  11  3.1  RQ5  Legibility  Headline-high  92  47.9  100  62.9  105  29.6  Headline-low  0  0  3  1.9  4  1.1  Bodycopy-high  33  17.0  40  25.2  40  11.3  Bodycopy-low  29  14.9  36  22.6  23  6.5      Skin cancer (n = 194)  Melanoma (n = 159)  Nonmelanoma (n = 355)      n  %  n  %  n  %  RQ1a  Types of human images  Profile picture  27  13.9  23  14.5  46  13.0  Affected area  60  30.9  40  25.2  157  36.3  Drawing  2  1.0  6  3.8  4  1.1  Multiple  2  1.0  1  .6  1  .3  RQ1b  Gender  Male  13  6.7  7  4.4  57  16.1    Female  36  18.6  22  13.8  25  7.0    Unable to tell  39  20.1  34  21.4  119  27.1  Age  Young  13  6.7  7  4.4  57  16.1    Old  36  18.6  22  13.8  7  2.0  Skin color  White  89  45.9  62  39.0  195  54.9    Nonwhite  2  1.0  5  3.1  4  1.4  RQ2  Fear  No fear  137  70.6  116  73.0  198  55.8    High fear  11  5.7  5  3.1  31  8.7    Low fear  46  23.7  38  23.9  126  35.5  RQ3  Pin Composition  Picture only  83  42.8  44  27.7  216  60.8  Text only  10  5.2  10  6.3  7  2.0  Picture and text  71  36.6  81  50.9  106  29.9  Video only  5  2.6  2  1.3  24  6.8  Infographics  25  12.9  22  13.8  2  .6  RQ4a  Primary color  Skin  90  46.4  63  39.6  210  59.2    Black  14  7.2  35  22.0  21  5.9    White  10  5.2  10  6.3  51  14.4    Cool colors  30  15.5  24  15.1  36  10.1    Warm colors  37  19.1  18  11.3  30  8.5    Multiple colors  13  6.7  9  5.7  7  2.0  RQ4b  Headline color  Black  26  13.4  48  30.2  33  9.3  White  29  14.9  25  15.7  32  9.0  Cool colors  7  3.6  4  2.5  4  1.1  Warm colors  17  8.8  9  5.7  26  7.3  Multiple colors  14  7.2  16  10.1  14  3.9  Bodycopy color  Black  22  11.3  25  15.7  33  9.3  White  15  7.7  15  9.4  11  3.1  Cool colors  2  1.0  2  1.3  1  .3  Warm colors  4  2.1  4  2.5  7  2.0  Multiple colors  19  9.8  30  18.9  11  3.1  RQ5  Legibility  Headline-high  92  47.9  100  62.9  105  29.6  Headline-low  0  0  3  1.9  4  1.1  Bodycopy-high  33  17.0  40  25.2  40  11.3  Bodycopy-low  29  14.9  36  22.6  23  6.5  View Large Table 1 Results for descriptive analysis of pins on different types of skin cancer     Skin cancer (n = 194)  Melanoma (n = 159)  Nonmelanoma (n = 355)      n  %  n  %  n  %  RQ1a  Types of human images  Profile picture  27  13.9  23  14.5  46  13.0  Affected area  60  30.9  40  25.2  157  36.3  Drawing  2  1.0  6  3.8  4  1.1  Multiple  2  1.0  1  .6  1  .3  RQ1b  Gender  Male  13  6.7  7  4.4  57  16.1    Female  36  18.6  22  13.8  25  7.0    Unable to tell  39  20.1  34  21.4  119  27.1  Age  Young  13  6.7  7  4.4  57  16.1    Old  36  18.6  22  13.8  7  2.0  Skin color  White  89  45.9  62  39.0  195  54.9    Nonwhite  2  1.0  5  3.1  4  1.4  RQ2  Fear  No fear  137  70.6  116  73.0  198  55.8    High fear  11  5.7  5  3.1  31  8.7    Low fear  46  23.7  38  23.9  126  35.5  RQ3  Pin Composition  Picture only  83  42.8  44  27.7  216  60.8  Text only  10  5.2  10  6.3  7  2.0  Picture and text  71  36.6  81  50.9  106  29.9  Video only  5  2.6  2  1.3  24  6.8  Infographics  25  12.9  22  13.8  2  .6  RQ4a  Primary color  Skin  90  46.4  63  39.6  210  59.2    Black  14  7.2  35  22.0  21  5.9    White  10  5.2  10  6.3  51  14.4    Cool colors  30  15.5  24  15.1  36  10.1    Warm colors  37  19.1  18  11.3  30  8.5    Multiple colors  13  6.7  9  5.7  7  2.0  RQ4b  Headline color  Black  26  13.4  48  30.2  33  9.3  White  29  14.9  25  15.7  32  9.0  Cool colors  7  3.6  4  2.5  4  1.1  Warm colors  17  8.8  9  5.7  26  7.3  Multiple colors  14  7.2  16  10.1  14  3.9  Bodycopy color  Black  22  11.3  25  15.7  33  9.3  White  15  7.7  15  9.4  11  3.1  Cool colors  2  1.0  2  1.3  1  .3  Warm colors  4  2.1  4  2.5  7  2.0  Multiple colors  19  9.8  30  18.9  11  3.1  RQ5  Legibility  Headline-high  92  47.9  100  62.9  105  29.6  Headline-low  0  0  3  1.9  4  1.1  Bodycopy-high  33  17.0  40  25.2  40  11.3  Bodycopy-low  29  14.9  36  22.6  23  6.5      Skin cancer (n = 194)  Melanoma (n = 159)  Nonmelanoma (n = 355)      n  %  n  %  n  %  RQ1a  Types of human images  Profile picture  27  13.9  23  14.5  46  13.0  Affected area  60  30.9  40  25.2  157  36.3  Drawing  2  1.0  6  3.8  4  1.1  Multiple  2  1.0  1  .6  1  .3  RQ1b  Gender  Male  13  6.7  7  4.4  57  16.1    Female  36  18.6  22  13.8  25  7.0    Unable to tell  39  20.1  34  21.4  119  27.1  Age  Young  13  6.7  7  4.4  57  16.1    Old  36  18.6  22  13.8  7  2.0  Skin color  White  89  45.9  62  39.0  195  54.9    Nonwhite  2  1.0  5  3.1  4  1.4  RQ2  Fear  No fear  137  70.6  116  73.0  198  55.8    High fear  11  5.7  5  3.1  31  8.7    Low fear  46  23.7  38  23.9  126  35.5  RQ3  Pin Composition  Picture only  83  42.8  44  27.7  216  60.8  Text only  10  5.2  10  6.3  7  2.0  Picture and text  71  36.6  81  50.9  106  29.9  Video only  5  2.6  2  1.3  24  6.8  Infographics  25  12.9  22  13.8  2  .6  RQ4a  Primary color  Skin  90  46.4  63  39.6  210  59.2    Black  14  7.2  35  22.0  21  5.9    White  10  5.2  10  6.3  51  14.4    Cool colors  30  15.5  24  15.1  36  10.1    Warm colors  37  19.1  18  11.3  30  8.5    Multiple colors  13  6.7  9  5.7  7  2.0  RQ4b  Headline color  Black  26  13.4  48  30.2  33  9.3  White  29  14.9  25  15.7  32  9.0  Cool colors  7  3.6  4  2.5  4  1.1  Warm colors  17  8.8  9  5.7  26  7.3  Multiple colors  14  7.2  16  10.1  14  3.9  Bodycopy color  Black  22  11.3  25  15.7  33  9.3  White  15  7.7  15  9.4  11  3.1  Cool colors  2  1.0  2  1.3  1  .3  Warm colors  4  2.1  4  2.5  7  2.0  Multiple colors  19  9.8  30  18.9  11  3.1  RQ5  Legibility  Headline-high  92  47.9  100  62.9  105  29.6  Headline-low  0  0  3  1.9  4  1.1  Bodycopy-high  33  17.0  40  25.2  40  11.3  Bodycopy-low  29  14.9  36  22.6  23  6.5  View Large Among the 708 pins, 52.1 per cent (n = 369) used human models. Out of them, 26 per cent (n = 96) included profile pictures with faces, 69.6 per cent (n = 257) included picture of body parts or affected area without showing the face, and 3.3 per cent (n = 12) used drawings of humans. Four pins included two or more of the categories mentioned above. Chi-square test showed that pins on NMSC were significantly more likely to show affected areas than other pins (see Table 2 for results of chi-square tests). Table 2 Results of chi-square tests comparing pins on melanoma (including skin cancer in general, n = 353) and nonmelanoma skin cancers (NMSC; n = 355)     X2  p  Adjusted p  RQ1a          Types of human image  Profile  .220  .639    Affected area  19.344*  .000  .000**  Drawing  1.379  .240    RQ1b  Gender  Male  19.715  .000  .000**    Female  15.075  0.00  .002**  Age  Young  3.807  .079      Old  5.653  .017    Skin color  White  10.463  .001  .026*    Non-white  4.840  .028    RQ2  Fear  Low fear  11.607  .001  .015*    High fear  5.037  .025    RQ3  Pin Composition  Picture only  43.825*  .000  .000**    Text only  6.584  .010      Picture and text  13.316*  .000  .006**    Video only  9.649*  .002  .039*    Inforgraphic  44.673*  .000  .000**  RQ4a  Primary color  Skin  17.713*  .000  .000**    Black  12.605*  .000  .009**    White  14.805*  .000  .003**    Cool colors  4.242  .039      Warm colors  8.518  .004      Multiple  8.179  .004    RQ4b  Headline color  Black  18.781*  .000  .000**    White  6.548  .010      Cool colors  3.378  .066      Warm colors  .000  .983      Multiple  6.301  .012    Bodycopy color  Black  2.852  .091      White  9.460*  .002  .042*    Cool colors  1.830  .176      Warm colors  .074  .786      Multiple  26.530*  .000  .000**  RQ5  Font legibility  Headline  1.412  .235      Bodycopy  1.972  .160        X2  p  Adjusted p  RQ1a          Types of human image  Profile  .220  .639    Affected area  19.344*  .000  .000**  Drawing  1.379  .240    RQ1b  Gender  Male  19.715  .000  .000**    Female  15.075  0.00  .002**  Age  Young  3.807  .079      Old  5.653  .017    Skin color  White  10.463  .001  .026*    Non-white  4.840  .028    RQ2  Fear  Low fear  11.607  .001  .015*    High fear  5.037  .025    RQ3  Pin Composition  Picture only  43.825*  .000  .000**    Text only  6.584  .010      Picture and text  13.316*  .000  .006**    Video only  9.649*  .002  .039*    Inforgraphic  44.673*  .000  .000**  RQ4a  Primary color  Skin  17.713*  .000  .000**    Black  12.605*  .000  .009**    White  14.805*  .000  .003**    Cool colors  4.242  .039      Warm colors  8.518  .004      Multiple  8.179  .004    RQ4b  Headline color  Black  18.781*  .000  .000**    White  6.548  .010      Cool colors  3.378  .066      Warm colors  .000  .983      Multiple  6.301  .012    Bodycopy color  Black  2.852  .091      White  9.460*  .002  .042*    Cool colors  1.830  .176      Warm colors  .074  .786      Multiple  26.530*  .000  .000**  RQ5  Font legibility  Headline  1.412  .235      Bodycopy  1.972  .160    df = 1 for all tests. *p <. 05; **p <. 01. View Large Table 2 Results of chi-square tests comparing pins on melanoma (including skin cancer in general, n = 353) and nonmelanoma skin cancers (NMSC; n = 355)     X2  p  Adjusted p  RQ1a          Types of human image  Profile  .220  .639    Affected area  19.344*  .000  .000**  Drawing  1.379  .240    RQ1b  Gender  Male  19.715  .000  .000**    Female  15.075  0.00  .002**  Age  Young  3.807  .079      Old  5.653  .017    Skin color  White  10.463  .001  .026*    Non-white  4.840  .028    RQ2  Fear  Low fear  11.607  .001  .015*    High fear  5.037  .025    RQ3  Pin Composition  Picture only  43.825*  .000  .000**    Text only  6.584  .010      Picture and text  13.316*  .000  .006**    Video only  9.649*  .002  .039*    Inforgraphic  44.673*  .000  .000**  RQ4a  Primary color  Skin  17.713*  .000  .000**    Black  12.605*  .000  .009**    White  14.805*  .000  .003**    Cool colors  4.242  .039      Warm colors  8.518  .004      Multiple  8.179  .004    RQ4b  Headline color  Black  18.781*  .000  .000**    White  6.548  .010      Cool colors  3.378  .066      Warm colors  .000  .983      Multiple  6.301  .012    Bodycopy color  Black  2.852  .091      White  9.460*  .002  .042*    Cool colors  1.830  .176      Warm colors  .074  .786      Multiple  26.530*  .000  .000**  RQ5  Font legibility  Headline  1.412  .235      Bodycopy  1.972  .160        X2  p  Adjusted p  RQ1a          Types of human image  Profile  .220  .639    Affected area  19.344*  .000  .000**  Drawing  1.379  .240    RQ1b  Gender  Male  19.715  .000  .000**    Female  15.075  0.00  .002**  Age  Young  3.807  .079      Old  5.653  .017    Skin color  White  10.463  .001  .026*    Non-white  4.840  .028    RQ2  Fear  Low fear  11.607  .001  .015*    High fear  5.037  .025    RQ3  Pin Composition  Picture only  43.825*  .000  .000**    Text only  6.584  .010      Picture and text  13.316*  .000  .006**    Video only  9.649*  .002  .039*    Inforgraphic  44.673*  .000  .000**  RQ4a  Primary color  Skin  17.713*  .000  .000**    Black  12.605*  .000  .009**    White  14.805*  .000  .003**    Cool colors  4.242  .039      Warm colors  8.518  .004      Multiple  8.179  .004    RQ4b  Headline color  Black  18.781*  .000  .000**    White  6.548  .010      Cool colors  3.378  .066      Warm colors  .000  .983      Multiple  6.301  .012    Bodycopy color  Black  2.852  .091      White  9.460*  .002  .042*    Cool colors  1.830  .176      Warm colors  .074  .786      Multiple  26.530*  .000  .000**  RQ5  Font legibility  Headline  1.412  .235      Bodycopy  1.972  .160    df = 1 for all tests. *p <. 05; **p <. 01. View Large Skin colors, gender, and age were coded to assess human models’ demographic characteristics. Out of 357 pins with identifiable skin colors, majority of pins had models with fair skin color (n = 346, 97%). Pins on NMSC were significantly more likely to use white models than other pins. Out of 160 pins with available gender information, approximately 51.8 per cent (n = 83) of pins had female models, whereas 48.2 per cent (n = 77) of pins featured male models. Pins on NMSC were significantly more likely to use male models, whereas pins on melanoma (including skin cancer in general) were more likely to use female models. Finally, out of 142 pins with age information, 54.2 per cent (n = 77) of pins featured younger models and 45.8 per cent of pins (n = 65) had older models. No statistical difference was found between pins on melanoma (including skin cancer in general) and pins on NMSC in terms of the age of the models used. RQ2 focused on the use of fear-invoking images. Only 36.3 per cent (n = 257) included fear-invoking images. Among them, the majority had images of low level of fear (n = 210, 81.7%) and only 18.3 per cent (n = 47) pins had pictures invoking high level of fear. Pins on NMSC were significantly more likely to show low fear appeal than pins on melanoma (including skin cancer in general) and more likely to use high fear appeal as well, even though the difference was not statistically significant. The composition of pins was examined in RQ3. Almost half of the pins were picture only (48.4%, n = 343) and around a third of the pins were picture with texts (36.4%, n = 258). Infographics only made up 6.9 per cent (n = 49) of all pins, followed by video only (4.4%, n = 31) and text only (3.8%, n = 27). Chi-square tests indicated that pins on melanoma skin (including skin cancer in general) were more likely to use infographics as well as picture with texts, whereas pins on NMSC were more likely to be picture only and video only. RQ4a and RQ4b examined the colors of images and texts. In terms of the primary colors of images, skin color (n = 363, 51.3%) was most dominant, followed by cool colors (n = 90, 12.7%), warm colors (n = 85, 12.0%), white (n = 71, 10.0%), black (n = 70, 9.9%), and multiple (n = 29, 4.1%). Pins on NMSC were more likely to use skin color and white as their primary colors, whereas pins on melanoma (including skin cancer in general) were more likely to use black as their primary color. The latter is not surprising since black is the signature color of melanoma. When font colors were concerned, 304 out of 708 pins had a headline and only 201 pins included bodycopy. For headline, black was the most prevalent color (n = 107, 35.2%) followed by white (n = 86, 28.2%), warm color (n = 52, 17.1%), multiple (n = 44, 14.5%), and cool color (n = 15, 4.93%). For bodycopy, black was the most dominant color (n = 80, 39.8%) followed by multiple (n = 58, 29.4%), white (n = 41, 20.4%), warm (n = 17, 8.46%), and cool (n = 5, 2.5%). Although black was the most frequently used font color in headline and bodycopy, cool colors (e.g., blue and green) were the least used. Pins on melanoma (including skin cancer in general) were significantly more likely to use black in its headlines and use white and multiple colors in its bodycopy than pins on NMSC. In terms of the legibility of texts (RQ5), headlines were usually in font-sizes that were easy to read. Among the 306 pins with headlines, approximately 97 per cent (n = 297) had headlines that were highly legible. However, legibility became a serious problem when bodycopy was concerned. Among the 201 pins with bodycopies, 43.8 per cent (n = 88) had bodycopy of low legibility (i.e., impossible to read on a handheld device without enlarging it). There was no significant difference between pins on melanoma (including skin cancer in general) and pins on nonmelanoma in terms of the legibility of either headlines or bodycopy. RQ6 sought to explore how information richness and visual characteristics of pins predicted participative engagement. To obtain a more nuanced and practically useful understanding of the predictors of popularity, we divided data into four different groups: (a) melanoma pins without human models; (b) melanoma pins with human models; (c) nonmelanoma pins without human models; and (d) nonmelanoma pins with human models (Because a large percentage of pins did not contain any human image, indiscriminately combining these pins with pins with human images would introduce too many missing data and render many significant predictions insignificant in the negative binomial regression models.). Then we developed four distinctive multivariate negative binomial regression models using each group of data. For pins without human models, we modeled the number of repins with information richness, stylistic variables (e.g., the compositions of pins, primary color schemes) and the presence/absence of fear-invoking images. For pins with human models, we modeled the number of repins with information richness, stylistic variables, human model-related factors (e.g., display of human body, demographic characters of human models), and the presence/absence of fear-invoking images. As shown in Table 3, among pins about melanoma without human models, only information richness (coefficient = 0.12, p = .04) significantly predicted the number of repins. In contrast, among pins about melanoma with human models, the number of repins was significantly associated with pin composition (p = 2.17e-13 < .001), gender of the human model (p = .02 < .05), and presence of fear-invoking images (p = 0.01 < .05). Post hoc Tukey test showed that picture-only pins and video-only pins received fewer repins than inforgraphics and picture combined with texts, and pins with female models received more repins than those using male models (coefficient = −0.87, p = .0497 < .05), and pins with fear-invoking images had more repins than those without fear-invoking images (coefficient = 0.76, p = .014 < .05). In other words, among melanoma-related pins featuring human models, the most popular ones were inforgraphics or picture combined with texts, pins with female models, and pins with fear-invoking images. Table 3 Results of negative binomial regression models Dependent variable: number of pins  Pins about melanoma coefficient (SE)  Pins about NMSC coefficient (SE)  No human models  With human models  No human models  With human models  INTERCEPT  4.15 (0.47)  5.39 (0.95)  1.14 (1.16)  −0.61 (0.55)  p-value  p < 2e-16***  p = 1.26e-08***  p = .33  p = .26  Richness  0.12 (0.06)  0.02 (0.08)  0.23 (0.09)  0.61 (0.08)  p-value  p = .04*  p = .78  p = .016*  p = 3.04e-14***  Pin composition  p-value  p = .12  p = 2.17e-13***  p = .45  p = .15  Infographic  0a  0a  0a  NAb   Picture + Text  −0.64 (0.27)  −1.49 (0.79)  −0.62 (1.04)  0a   Picture only  −0.55 (0.30)  −3.21 (0.80)***  −0.15 (1.02)  0.62 (0.37)   Text only  −0.30 (0.35)  NAb  −0.95 (1.18)  NAb   Video only  −2.01 (1.23)  −3.83 (0.95)***  −0.99 (1.22)  0.16 (0.46)  Primary color  p-value  p = .55  p = .83  p = .12  p = .26   Skin  0a  0a  0a  0a   Warm color  0.36 (0.34)  −0.60 (0.53)  0.30 (0.45)  1.03 (1.48)   Cool color  0.55 (0.33)  −0.35 (0.45)  −0.11 (0.45)  −0.66 (0.83)   Black  0.66 (0.34)  −0.21 (0.63)  0.66 (0.50)  −1.64 (1.37)   White  0.28 (0.39)  NAb  0.03 (0.45)  −2.00 (1.13)   Multiple  0.32 (0.41)  −0.05 (0.86)  −1.59 (0.82)*  −1.42 (1.16)  Human model  p-value    p = .24    p = .17   Profile    0a    0a   Affected area    −0.57 (0.36)    0.09 (0.42)   Drawing    −0.17 (0.63)    2.01 (0.99)   Multiple    −1.69 (0.83)    1.75 (1.86)  Gender  p-value    p = .02*    p = .08   Male    0a    0a   Female    1.02 (0.34)**    0.28 (0.33)   Mixed    −0.01 (0.60)    −0.19 (0.80)   Unknown    −0.91 (0.48)    0.88 (0.32)  Age  p-value    p = .74    p = .04*   Young    0a    0a   Old    0.45 (0.41)    0.09 (0.32)   Mixed    −0.02 (0.78)    −0.97 (1.00)   Unknown    −0.06 (0.44)    −1.13 (0.40)**  Skin color  p-value    p = .59    p = .06   White    0a    0a   Nonwhite    0.33 (0.60)    −1.05 (0.52)  Fear  p-value  p = .19  p = .0099**  p = .10  p = 0.67   Present  −0.66 (0.45)  0.76 (0.30)*  −1.82 (1.04)  −0.16 (0.34)   Absent  0a  0a  0a  0a  Model p -value  p = .02*  p = 1.02e-11***  p = .0003***  p = 7.70e-12***  Likelihood ratio χ2  22.45  92.89  34.31  91.48  Degree of freedom  11  19  11  19  Number of cases  192  161  147  208  Dependent variable: number of pins  Pins about melanoma coefficient (SE)  Pins about NMSC coefficient (SE)  No human models  With human models  No human models  With human models  INTERCEPT  4.15 (0.47)  5.39 (0.95)  1.14 (1.16)  −0.61 (0.55)  p-value  p < 2e-16***  p = 1.26e-08***  p = .33  p = .26  Richness  0.12 (0.06)  0.02 (0.08)  0.23 (0.09)  0.61 (0.08)  p-value  p = .04*  p = .78  p = .016*  p = 3.04e-14***  Pin composition  p-value  p = .12  p = 2.17e-13***  p = .45  p = .15  Infographic  0a  0a  0a  NAb   Picture + Text  −0.64 (0.27)  −1.49 (0.79)  −0.62 (1.04)  0a   Picture only  −0.55 (0.30)  −3.21 (0.80)***  −0.15 (1.02)  0.62 (0.37)   Text only  −0.30 (0.35)  NAb  −0.95 (1.18)  NAb   Video only  −2.01 (1.23)  −3.83 (0.95)***  −0.99 (1.22)  0.16 (0.46)  Primary color  p-value  p = .55  p = .83  p = .12  p = .26   Skin  0a  0a  0a  0a   Warm color  0.36 (0.34)  −0.60 (0.53)  0.30 (0.45)  1.03 (1.48)   Cool color  0.55 (0.33)  −0.35 (0.45)  −0.11 (0.45)  −0.66 (0.83)   Black  0.66 (0.34)  −0.21 (0.63)  0.66 (0.50)  −1.64 (1.37)   White  0.28 (0.39)  NAb  0.03 (0.45)  −2.00 (1.13)   Multiple  0.32 (0.41)  −0.05 (0.86)  −1.59 (0.82)*  −1.42 (1.16)  Human model  p-value    p = .24    p = .17   Profile    0a    0a   Affected area    −0.57 (0.36)    0.09 (0.42)   Drawing    −0.17 (0.63)    2.01 (0.99)   Multiple    −1.69 (0.83)    1.75 (1.86)  Gender  p-value    p = .02*    p = .08   Male    0a    0a   Female    1.02 (0.34)**    0.28 (0.33)   Mixed    −0.01 (0.60)    −0.19 (0.80)   Unknown    −0.91 (0.48)    0.88 (0.32)  Age  p-value    p = .74    p = .04*   Young    0a    0a   Old    0.45 (0.41)    0.09 (0.32)   Mixed    −0.02 (0.78)    −0.97 (1.00)   Unknown    −0.06 (0.44)    −1.13 (0.40)**  Skin color  p-value    p = .59    p = .06   White    0a    0a   Nonwhite    0.33 (0.60)    −1.05 (0.52)  Fear  p-value  p = .19  p = .0099**  p = .10  p = 0.67   Present  −0.66 (0.45)  0.76 (0.30)*  −1.82 (1.04)  −0.16 (0.34)   Absent  0a  0a  0a  0a  Model p -value  p = .02*  p = 1.02e-11***  p = .0003***  p = 7.70e-12***  Likelihood ratio χ2  22.45  92.89  34.31  91.48  Degree of freedom  11  19  11  19  Number of cases  192  161  147  208  a = Set to zero because this parameter is redundant (baseline category). b = Empty cell: sample size = 0. *p < .05; **p < .01; ***p < .001. View Large Table 3 Results of negative binomial regression models Dependent variable: number of pins  Pins about melanoma coefficient (SE)  Pins about NMSC coefficient (SE)  No human models  With human models  No human models  With human models  INTERCEPT  4.15 (0.47)  5.39 (0.95)  1.14 (1.16)  −0.61 (0.55)  p-value  p < 2e-16***  p = 1.26e-08***  p = .33  p = .26  Richness  0.12 (0.06)  0.02 (0.08)  0.23 (0.09)  0.61 (0.08)  p-value  p = .04*  p = .78  p = .016*  p = 3.04e-14***  Pin composition  p-value  p = .12  p = 2.17e-13***  p = .45  p = .15  Infographic  0a  0a  0a  NAb   Picture + Text  −0.64 (0.27)  −1.49 (0.79)  −0.62 (1.04)  0a   Picture only  −0.55 (0.30)  −3.21 (0.80)***  −0.15 (1.02)  0.62 (0.37)   Text only  −0.30 (0.35)  NAb  −0.95 (1.18)  NAb   Video only  −2.01 (1.23)  −3.83 (0.95)***  −0.99 (1.22)  0.16 (0.46)  Primary color  p-value  p = .55  p = .83  p = .12  p = .26   Skin  0a  0a  0a  0a   Warm color  0.36 (0.34)  −0.60 (0.53)  0.30 (0.45)  1.03 (1.48)   Cool color  0.55 (0.33)  −0.35 (0.45)  −0.11 (0.45)  −0.66 (0.83)   Black  0.66 (0.34)  −0.21 (0.63)  0.66 (0.50)  −1.64 (1.37)   White  0.28 (0.39)  NAb  0.03 (0.45)  −2.00 (1.13)   Multiple  0.32 (0.41)  −0.05 (0.86)  −1.59 (0.82)*  −1.42 (1.16)  Human model  p-value    p = .24    p = .17   Profile    0a    0a   Affected area    −0.57 (0.36)    0.09 (0.42)   Drawing    −0.17 (0.63)    2.01 (0.99)   Multiple    −1.69 (0.83)    1.75 (1.86)  Gender  p-value    p = .02*    p = .08   Male    0a    0a   Female    1.02 (0.34)**    0.28 (0.33)   Mixed    −0.01 (0.60)    −0.19 (0.80)   Unknown    −0.91 (0.48)    0.88 (0.32)  Age  p-value    p = .74    p = .04*   Young    0a    0a   Old    0.45 (0.41)    0.09 (0.32)   Mixed    −0.02 (0.78)    −0.97 (1.00)   Unknown    −0.06 (0.44)    −1.13 (0.40)**  Skin color  p-value    p = .59    p = .06   White    0a    0a   Nonwhite    0.33 (0.60)    −1.05 (0.52)  Fear  p-value  p = .19  p = .0099**  p = .10  p = 0.67   Present  −0.66 (0.45)  0.76 (0.30)*  −1.82 (1.04)  −0.16 (0.34)   Absent  0a  0a  0a  0a  Model p -value  p = .02*  p = 1.02e-11***  p = .0003***  p = 7.70e-12***  Likelihood ratio χ2  22.45  92.89  34.31  91.48  Degree of freedom  11  19  11  19  Number of cases  192  161  147  208  Dependent variable: number of pins  Pins about melanoma coefficient (SE)  Pins about NMSC coefficient (SE)  No human models  With human models  No human models  With human models  INTERCEPT  4.15 (0.47)  5.39 (0.95)  1.14 (1.16)  −0.61 (0.55)  p-value  p < 2e-16***  p = 1.26e-08***  p = .33  p = .26  Richness  0.12 (0.06)  0.02 (0.08)  0.23 (0.09)  0.61 (0.08)  p-value  p = .04*  p = .78  p = .016*  p = 3.04e-14***  Pin composition  p-value  p = .12  p = 2.17e-13***  p = .45  p = .15  Infographic  0a  0a  0a  NAb   Picture + Text  −0.64 (0.27)  −1.49 (0.79)  −0.62 (1.04)  0a   Picture only  −0.55 (0.30)  −3.21 (0.80)***  −0.15 (1.02)  0.62 (0.37)   Text only  −0.30 (0.35)  NAb  −0.95 (1.18)  NAb   Video only  −2.01 (1.23)  −3.83 (0.95)***  −0.99 (1.22)  0.16 (0.46)  Primary color  p-value  p = .55  p = .83  p = .12  p = .26   Skin  0a  0a  0a  0a   Warm color  0.36 (0.34)  −0.60 (0.53)  0.30 (0.45)  1.03 (1.48)   Cool color  0.55 (0.33)  −0.35 (0.45)  −0.11 (0.45)  −0.66 (0.83)   Black  0.66 (0.34)  −0.21 (0.63)  0.66 (0.50)  −1.64 (1.37)   White  0.28 (0.39)  NAb  0.03 (0.45)  −2.00 (1.13)   Multiple  0.32 (0.41)  −0.05 (0.86)  −1.59 (0.82)*  −1.42 (1.16)  Human model  p-value    p = .24    p = .17   Profile    0a    0a   Affected area    −0.57 (0.36)    0.09 (0.42)   Drawing    −0.17 (0.63)    2.01 (0.99)   Multiple    −1.69 (0.83)    1.75 (1.86)  Gender  p-value    p = .02*    p = .08   Male    0a    0a   Female    1.02 (0.34)**    0.28 (0.33)   Mixed    −0.01 (0.60)    −0.19 (0.80)   Unknown    −0.91 (0.48)    0.88 (0.32)  Age  p-value    p = .74    p = .04*   Young    0a    0a   Old    0.45 (0.41)    0.09 (0.32)   Mixed    −0.02 (0.78)    −0.97 (1.00)   Unknown    −0.06 (0.44)    −1.13 (0.40)**  Skin color  p-value    p = .59    p = .06   White    0a    0a   Nonwhite    0.33 (0.60)    −1.05 (0.52)  Fear  p-value  p = .19  p = .0099**  p = .10  p = 0.67   Present  −0.66 (0.45)  0.76 (0.30)*  −1.82 (1.04)  −0.16 (0.34)   Absent  0a  0a  0a  0a  Model p -value  p = .02*  p = 1.02e-11***  p = .0003***  p = 7.70e-12***  Likelihood ratio χ2  22.45  92.89  34.31  91.48  Degree of freedom  11  19  11  19  Number of cases  192  161  147  208  a = Set to zero because this parameter is redundant (baseline category). b = Empty cell: sample size = 0. *p < .05; **p < .01; ***p < .001. View Large Among pins about nonmelanoma without human models, information richness was the only significant predictor of number of repins: images with higher levels of information richness tended to have more repins compared with those with lower levels of richness (coefficient = 0.23, p = .016 < .05). Among pins about nonmelanoma that used human models, the number of repins was significantly associated with information richness (coefficient = 0.61, p = 3.04e-14 < .001) and age of human models (p = .04 < .05). Post hoc Tukey test showed that among nonmelanoma pins with human models, pins with human images for which age cannot be identified received less pins than both pins featured young models (coefficient = −1.13, p = .02) and pins featured older models (coefficient = −1.06, p = .01). DISCUSSION This study examines the portrayal of skin cancer on Pinterest with a focus on the visual characteristics of pins. Around half of the pins on skin cancer used human images. Among them, nearly 70 per cent only show a body part without the face and only 26 per cent used profile picture with a person’s face and upper torso. Using headless pictures of human models might focus audiences’ attention on areas affected by skin cancer and provide anonymity to cancer patients. However, it might also stigmatize patients with cancer in characterizing them by their illness [43]. Pins on NMSC were more likely to show affected areas without human face than other pins. This is actually an accurate depiction of NMSC since it often affected areas other than neck and head [44]. Contrary to the assumption that human models used on Pinterest will be mainly young women based on the demographics of Pinterest users, this study finds an almost 50-50 split between men and women and between young and old (50 years or older). Pins on NMSC are more likely to use male models. This also reflects the demographics of NMSC patients [44]. In terms of the skin color of models, 97 per cent of the human models have fair skin color associated with Caucasians and sometimes Hispanics. Epidemiological data indicate that Caucasians are the primary victims of skin cancer. However, people of color are also at risk of skin cancer [45]. Hu et al. [46] found that the diagnosis of melanoma is often delayed among African Americans and Hispanics. The predominance of fair-skinned human models used in pins on skin cancer might contribute the misconception that colored people are immune to skin cancer and create further gaps in early diagnosis and effective treatment of skin cancer among different racial groups. Visual images invoking fear are found in one third of the pins; however, only 5 per cent of pins contain highly disturbing or fearful images. This is consistent with the use of Pinterest for visual curation. Highly disturbing or fearful images are probably not favored by users who care about the aesthetics of their pin boards. Around half of all pins on skin cancer are picture only and another third are pictures with texts. This confirms the centrality of visual images on Pinterest. Another 7 per cent of the pins are infographics. Future research should further examine how Pinterest users respond to these different formats and which format holds the greatest potential in getting attention and aiding comprehension. Color is another important aspect of visual communication. Black is the most used font color for both headlines and bodycopy. This is encouraging given the evidence that dark shade of fonts on a light background was more visible and readable especially when black and dark blue fonts were used [47]. Text legibility is a potential problem. Nearly half of pins with bodycopy have texts that are illegible on a handheld device such as an iPhone. This poses barriers to the comprehension of skin cancer information on Pinterest as 80 per cent of users report using mobile phones as the primary device to access information on Pinterest [30]. The reason for such low legibility could be that a lot of the pins are created as advertisements or public service announcements and not intended for devices with small screens. If public health professionals want to utilize Pinterest or other similar visual based social media platforms to disseminate information, they need to make sure that their messages include legible texts. This study contributes to the research on health-related visual images because it explores the relationship between visual characteristics of pins on skin cancer and audiences’ participative engagement measured by the number of repins. It finds different models for pins on melanoma and NMSC, with and without the use of human models. For pins on melanoma without a human model, only information richness is significant in predicting number of repins. For pins on melanoma with a human model, information richness is no longer significant, whereas visual characteristics including pin composition, gender of the human model, and presence of fear-invoking images predict the popularity of the pins. Among them, pins with picture only and video only are less likely to be repined than inforgraphics as well as picture with texts. One possible explanation is that pins without any texts might appear confusing and thus are less likely to be repined. Future efforts on melanoma prevention and skin cancer prevention in general on Pinterest should use the format of infographic or picture blended with texts because these two compositions generate more participation. Female models increase the popularity of pins on melanoma skin cancer with human models. It could be that since Pinterest users are predominantly female, they are more likely to repin those pins with human models they identify with. Interestingly, fear-invoking images increase repins. Notice that a large majority of these images are categorized as of low fear, typically pictures of benign images of early signs of skin cancer. These images make the pins more instead of less popular probably because pins with such images are considered informative. For pins on NMSC with or without of human models, informational richness is positively associated with the number of repins. This finding supports the belief that more information is better. However, for pins on NMSC with a human model, age of human models predicts the number of repins. Pins with human models that age are unidentifiable are less likely to be repined than pins with either young or old models. It shows that in creating popular pins about NMSC, public health professionals should use images with vivid human faces and avoid using images with drawings of human models or images of affected areas only. Overall, the primary color of pins does not predict their popularity across all four models. This finding is surprising because past research has suggested that cool colors are preferred over warm colors [48]. One possible explanation is that one’s evaluation of color varies by a number of factors such as age, gender, and culture [49]. Future research should investigate how Pinterest users with different demographic characteristics respond to colors differently. Limitations Most obviously, the dependent variable examined is the number of repins, or the popularity of pins. Repining a pin on her own board does not necessarily mean that a user is going to actually process the information included in the pin or to adopt the behaviors suggested. Future experiment studies should be conducted to further examine how information richness as well as different visual characteristics influence not only the likelihood users are going to repin a pin about skin cancer or other health-related topics, but also the ways in which they are going to process the information, their enjoyment of the pins, their recall of information, and their intentions for behavior change. Second, the models about the popularity of pins proposed in this paper rely on an imperfect dependent variable: number of repins. One important factor that is likely to influence the number of repins is the amount of time since the pictures used in these pins have been created and made available online. A picture that has been online for a relatively short period of time is likely to have less repins than a picture that has been there for years. However, the length of a picture’s availability online is unfortunately unavailable because Pinterest users can essentially pin any picture online outside of Pinterest. CONCLUSIONS Visual-based social media sites such as Pinterest hold great potential in disseminating health information and promoting behavioral changes. This paper provides an overview of the visual characteristics of pins on skin cancer in terms of pin composition, text legibility, color, use of human models, and use of fear-invoking messages. Furthermore, it proposes the first model of the relationship between visual characteristics and information richness and the popularity of pins. The popularity of pins about different types of skin cancer (melanoma with human models, melanoma without human models, NMSC with human models, and NMSC without human models) is predicted by different variables. Overall, information richness is the only predictor of number of repins among pins that do not use human models (including pins on both melanoma and NMSC). Information-rich pins are more likely to be repined. For pins about melanoma with human models, pin composition, gender of human model, and the use of fear-invoking images are associated with the number of repins. Pins in the form of infographic or picture combined with texts use female models and include fear-invoking messages that are more popular than other pins. Among pins about NMSC with human models, those pins that show the age of the model (either old or young) are more likely to be repined than pins that do not show the face of the model. These findings provide useful tools for public health professionals who wish to create and disseminate skin cancer–related information via social media. Compliance with Ethical Standards Author contributions: Sung-Eun Park and Lu Tang designed the study and coded the pins. Bijie Bie conducted data analysis. Degui Zhi provided advice on the statistical model and analysis and the interpretation of results. Sung-Eun Park and Lu Tang drafted the manuscript, and all authors read, edited, and approved the final manuscript. Conflict of Interest: Sung-Eun Park, Lu Tang, Bijie Bie, and Degui Zhi declare that they have no conflicts of interest. Ethical Approval: This article does not contain any studies with human participants or animals performed by any of the authors. Informed Consent: This article has no human participants and therefore informed consent was not required. References 1. Houts PS, Doak CC, Doak LG, Loscalzo MJ. The role of pictures in improving health communication: a review of research on attention, comprehension, recall, and adherence. Patient Educ Couns . 2006; 61( 2): 173– 190. Google Scholar CrossRef Search ADS PubMed  2. Dowse R, Ehlers MS. The evaluation of pharmaceutical pictograms in a low-literate South African population. Patient Educ Couns . 2001; 45( 2): 87– 99. Google Scholar CrossRef Search ADS PubMed  3. Robins D, Holmes J, Stansbury M. Consumer health information on the Web: the relationship of visual design and perceptions of credibility. J Am Soc Inf Sci Technol . 2010; 61 (1): 13– 29. Google Scholar CrossRef Search ADS   4. Pew Research Center. Mobile Messaging and Social Media 2015; 2015. Available at http://www.pewinternet.org/2015/08/19/mobile-messaging-and-social-media-2015/. Accessibility verified October 1, 2016. 5. Opallo D. Pin down Pinterest’s value. DMN . 2012; 34( 8): 19– 20. 6. Duggan M, Brenner J. The Demographics of Social Media Users . Washington, DC: Pew Research Center’s Internet & American Life Project; 2013. 7. Guidry JP, Carlyle K, Messner M, Jin Y. On pins and needles: how vaccines are portrayed on Pinterest. Vaccine . 2015; 33( 39): 5051– 5056. Google Scholar CrossRef Search ADS PubMed  8. Guidry J, Zhang Y, Jin Y, Parrish C. Portrayals of depression on Pinterest and why public relations practitioners should care. Public Relat Rev . 2016; 42( 1): 232– 236. Google Scholar CrossRef Search ADS   9. Whitsitt J, Mattis D, Hernandez M, Kollipara R, Dellavalle RP. Dermatology on pinterest. Dermatol Online J . 2015; 21( 1). 10. American Cancer Society. Cancer facts & figures; 2015. Available at http://www.cancer.org/research/cancerfactsstatistics/cancerfactsfigures2015/ . Accessibility verified October 1, 2016 11. Lomas A, Leonardi-Bee J, Bath-Hextall F. A systematic review of worldwide incidence of nonmelanoma skin cancer. Br J Dermatol . 2012; 166( 5): 1069– 1080. Google Scholar CrossRef Search ADS PubMed  12. Entman RM. Framing: towards clarification of a fractured paradigm. J Commun . 1993; 43( 4): 51– 58. Google Scholar CrossRef Search ADS   13. Rodriguez L, Asoro RL. Visual representation of genetic engineering and genetically modified organisms in the online media. Visual Commun Quart . 2012; 19(4): 232– 245. Google Scholar CrossRef Search ADS   14. King AJ. A content analysis of visual cancer information: prevalence and use of photographs and illustrations in printed health materials. Health Commun . 2015; 30( 7): 722– 731. Google Scholar CrossRef Search ADS PubMed  15. Phillips SG, Della LJ, Sohn SH. What does cancer treatment look like in consumer cancer magazines? An exploratory analysis of photographic content in consumer cancer magazines. J Health Commun . 2011; 16( 4): 416– 430. Google Scholar CrossRef Search ADS PubMed  16. Rodriguez L, Dimitrova DV. The levels of visual framing. J Visual Literacy . 2011; 30: 48– 65. Google Scholar CrossRef Search ADS   17. Paek HJ, Reid LN, Choi H, Jeong HJ. Promoting health (implicitly)? A longitudinal content analysis of implicit health information in cigarette advertising, 1954-2003. J Health Commun . 2010; 15( 7): 769– 787. Google Scholar CrossRef Search ADS PubMed  18. Bandura A. Organizational applications of social cognitive theory. Aus J Management . 1988; 13(2): 275– 302. Google Scholar CrossRef Search ADS   19. Leshner G, Bolls P, Thomas E. Scare’ em or disgust ‘em: the effects of graphic health promotion messages. Health Commun . 2009; 24( 5): 447– 458. Google Scholar CrossRef Search ADS PubMed  20. Witte K, Allen M. A meta-analysis of fear appeals: implications for effective public health campaigns. Health Educ Behav . 2000; 27( 5): 591– 615. Google Scholar CrossRef Search ADS PubMed  21. Schienle A, Schäfer A, Stark R, Walter B, Vaitl D. Gender differences in the processing of disgust- and fear-inducing pictures: an fMRI study. Neuroreport . 2005; 16( 3): 277– 280. Google Scholar CrossRef Search ADS PubMed  22. Occa A, Suggs LS. Communicating breast cancer screening with young women: an experimental test of didactic and narrative messages using video and infographics. J Health Commun . 2016; 21( 1): 1– 11. Google Scholar CrossRef Search ADS PubMed  23. Featherstone RM. Visual research data: an infographics Primer. J Can Health Libr Assoc . 2014; 35(3): 147– 150. Google Scholar CrossRef Search ADS   24. Couper MP, Conrad FG, Tourangeau R. Visual context effects in web surveys. Public Opin Q . 2007; 71(4): 623– 634. Google Scholar CrossRef Search ADS   25. Moshagen M, Thielsch MT. Facets of visual aesthetics. Int J Hum Comput Stud . 2010; 68(10): 689– 709. Google Scholar CrossRef Search ADS   26. Cyr D, Head M, Larios H. Colour appeal in website design within and across cultures: a multi-method evaluation. Int J Hum Comput Stud . 2010; 68: 1– 21. Google Scholar CrossRef Search ADS   27. Madden TJ, Hewett K, Roth MS. Managing images in different cultures: a cross- national study of color meanings and preferences. J Int Market . 2000; 8( 4): 90– 107. Google Scholar CrossRef Search ADS   28. Welhausen CA. Visualizing a non-pandemic: considerations for communicating public health risks in intercultural contexts. Tech Commun . 2015; 62: 244– 257. 29. Cline RJ, Haynes KM. Consumer health information seeking on the Internet: the state of the art. Health Educ Res . 2001; 16( 6): 671– 692. Google Scholar CrossRef Search ADS PubMed  30. Novet J. 80 percent of Pinterest’s traffic comes from mobile devices. VentureBeat . Feb 18, 2016. Available at http://venturebeat.com/2015/03/31/80-percent-of-pinterests-traffic-comes-from-mobile-devices/. Accessibility verified October 1, 2016. 31. Darroch I, Goodman J, Brewster S, Gray P. The effect of age and font size on reading text on handheld computers. In: Costabile MF, Paternò F, eds. Human-Computer Interaction-INTERACT 2005 . Berlin, Germany: Springer; 2005: 253– 266. Google Scholar CrossRef Search ADS   32. Han J, Choi D, Chun BG, Kwon TT, Kim HC, Choi Y. Collecting, organizing, and sharing pins in Pinterest: interest-driven or social-driven? Paper presented at: The 2014 ACM international conference on Measurement and modeling of computer systems; June 2014 , Austin, Texas. 33. Bernardini C, Silverston T, Festor O. A Pin is worth a thousand words: characterization of publications in Pinterest. In: Wireless Communications and Mobile Computing Conference (IWCMC), 2014 International IEEE; Nicosia, Cyprus: August, 2014. 322– 327. 34. Mull IR, Lee SE. “PIN” pointing the motivational dimensions behind Pinterest. Comput Human Behav . 2014; 33: 192– 200. Google Scholar CrossRef Search ADS   35. Guidry JP, Carlyle K, Messner M, Jin Y. On pins and needles: how vaccines are portrayed on Pinterest. Vaccine . 2015; 33( 39): 5051– 5056. Google Scholar CrossRef Search ADS PubMed  36. Gilbert E, Bakhshi S, Chang S, Terveen L. I need to try this?: a statistical overview of pinterest. In: Proceedings of the SIGCHI conference on human factors in computing systems; Apr 27-May 2, 2013; Paris, France. 2427– 2436. 37. Neuendorf KA. The Content Analysis Guidebook . Thousand Oaks, CA: Sage; 2002. 38. Stephenson MT, Witte K. Fear, threat, and perceptions of efficacy from frightening skin cancer messages. Public Health Rev . 1998; 26( 2): 147– 174. Google Scholar PubMed  39. Bellizzi JA, Crowley AE, Hasty RW. The effects of color in store design. J Retail . 1983; 59(1): 21– 45. 40. Tang L, Park SE. Sun exposure, tanning beds, and herbs that cure: an examination of skin cancer on pinterest. Health Commun . 2017; 32( 10): 1192– 1200. Google Scholar CrossRef Search ADS PubMed  41. Potter J, Levine-Donnerstein D. Rethinking validity and reliability in content analysis. J Applied Commun Res . 1999; 27(3): 258– 284. Google Scholar CrossRef Search ADS   42. Allison PD. Fixed Effects Regression Models . Thousand Oaks, CA: Sage; 2009. Google Scholar CrossRef Search ADS   43. Puhl RM, Peterson JL, DePierre JA, Luedicke J. Headless, hungry, and unhealthy: a video content analysis of obese persons portrayed in online news. J Health Commun . 2013; 18( 6): 686– 702. Google Scholar CrossRef Search ADS PubMed  44. Diepgen TL, Mahler V. The epidemiology of skin cancer. Br J Dermatol . 2002; 146( Suppl 61): 1– 6. Google Scholar CrossRef Search ADS PubMed  45. Gloster HMJr, Neal K. Skin cancer in skin of color. J Am Acad Dermatol . 2006; 55( 5): 741– 760. Google Scholar CrossRef Search ADS PubMed  46. Hu S, Parmet Y, Allen G, et al.   Disparity in melanoma: a trend analysis of melanoma incidence and stage at diagnosis among whites, Hispanics, and blacks in Florida. Arch Dermatol . 2009; 145( 12): 1369– 1374. Google Scholar CrossRef Search ADS PubMed  47. Greco M, Stucchi N, Zavagno D, Marino B. On the portability of computer-generated presentations: the effect of text-background color combinations on text legibility. Hum Factors . 2008; 50( 5): 821– 833. Google Scholar CrossRef Search ADS PubMed  48. Marcus A, Gould EW. Crosscurrents: cultural dimensions and global Web user-interface design. Interactions . 2000; 7( 4): 32– 46. Google Scholar CrossRef Search ADS   49. Ou LC, Luo MR, Woodcock A, Wright A. A study of colour emotion and colour preference. Part I: colour emotions for single colours. Color Res App . 2004; 29 (3): 232– 240. Google Scholar CrossRef Search ADS   © Society of Behavioral Medicine 2018. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

Translational Behavioral MedicineOxford University Press

Published: Apr 17, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off