Uptake of and Engagement With an Online Sexual Health Intervention (HOPE eIntervention) Among African American Young Adults: Mixed Methods Study

Background: Regarding health technologies, African American young adults have low rates of uptake, ongoing usage, and engagement, which may widen sexual health inequalities. Objective: We aimed to examine rates of uptake and ongoing usage, and factors influencing uptake, ongoing usage, and engagement for a consumer health informatics (CHI) intervention for HIV/sexually transmitted infection (STI) prevention among African American young adults, using the diffusion of innovation theory, trust-centered design framework, and O’Brien and Toms’ model of engagement. Methods: This community-based participatory mixed methods study included surveys at four time points (n=315; 280 African American participants) among young adults aged 18 to 24 years involved in a blended offline/online HIV/STI prevention intervention (HIV Outreach, Prevention, and Education [HOPE] eIntervention), which was described as a “HOPE


Introduction
Background HIV and other sexually transmitted infections (STIs) are health concerns for African American young adults. In 2018, African Americans constituted 13% of the population but 42% of new HIV diagnoses in the United States [1], with young adults making up about 21% of new HIV diagnoses [2]. African American young adults are vulnerable to other STIs, having the highest rates of gonorrhea of all racial groups [3]. Consumer health informatics (CHI) technologies can support HIV/STI prevention by targeting outcomes such as sexual health literacy and HIV/STI testing rates [4].
However, previous experience with patient portals shows that CHI interventions may not achieve uptake and engagement among African Americans [5,6]. In an HIV/STI context, this may widen sexual health inequalities [7,8]. Numerous CHI interventions exist to prevent HIV [9][10][11][12][13][14][15] and STIs [16,17] among African American youth. Additionally, digital CHI interventions targeting African American populations can be effective at promoting safer sex behaviors [18,19] and can increase access to sexual health information [20,21]. However, few identify factors driving uptake and engagement [22]. While interventions have been effective at targeting African American youth [9,13,15], limited uptake and engagement can impede the promotion of behaviors, such as condom use [23,24], or the development of sexual health knowledge [25,26]. We define uptake as initiating contact with a CHI intervention, which is differentiated from ongoing usage of that intervention [27]. Uptake is rarely investigated, inconsistently defined, and often underreported [28,29]. Engagement has been defined by some as interactions over time [30], although information scientists use the term to demarcate the period in which a user interacts with an intervention [31]. Given our interest in experiences of CHI interventions, we use the latter definition of engagement in this study.
Factors driving uptake may explain disparities between racial groups in CHI intervention use [32]. Studies that examine uptake among African Americans suggest that it may be impeded by recruitment barriers, power differentials between researcher/clinician and participants, or researchers' poor cultural competence [33]. Medical distrust stemming from discrimination and historical mistreatment also plays a role [34]. However, targeted recruitment and individual tailoring may improve uptake among African American young adults. Peer-driven recruitment [35,36] and partnering with community-based organizations [37] may increase uptake. Tailoring web-based interventions to users' needs, interests, personalities, and contexts [38], and tailored messages from a personalized source [39] can also improve uptake.
CHI interventions should also foster ongoing usage. It can be difficult to differentiate between factors motivating uptake and ongoing usage; few studies have examined this. Some evidence shows that African American young adults' engagement may be impeded by distrust, lack of time, limited technology access, and limited cultural relevance [32].
Some CHI interventions have successfully engaged African American users via social media, peer-led interventions, and designing to capture user attention. Social media-based HIV-prevention interventions have shown low attrition [14]. Peer-led CHI interventions have maintained engagement through existing social networks [14]. High levels of interest [40] and perceived usefulness may also drive ongoing usage [41].
Yet, there is limited understanding of what motivates uptake and engagement in African American young adults participating in CHI interventions to prevent HIV/STIs. We identify motivating factors for a tailored website and social media accounts promoting safer sex for African American young adults, framing our analyses with the theories described below.

Diffusion of Innovations
The diffusion of innovations (DOI) theory describes a process of communicating about innovations among members of social systems [42]. Communication can lead to adoption (deciding to use an intervention; this aligns with our definition of uptake). Personal characteristics, such as demographics, influence adoption. For example, people with more education and higher social participation tend to adopt innovations earlier. Furthermore, homophily (social similarity) with adopters increases the likelihood that nonadopters will make the subsequent decision to adopt an innovation. We used DOI to identify variables that may predict uptake.
Several HIV-prevention interventions have used the DOI theory [43]. Similarly, it informed the design of the HIV Outreach, Prevention, and Education (HOPE) eIntervention in this study. This intervention asked African American young adults to invite peers to HOPE parties (described below), which introduced the innovations (both the eIntervention and healthy sexual behaviors).

Trust-Centered Design Framework
The trust-centered design framework ( Figure 1) [12] outlines trust-based needs in relation to CHI interventions at the personal, group, technological, and institutional levels for African American young adults. It centers trust in CHI interventions, calls for addressing trust-related requirements at multiple user-experience levels, and highlights forms of trust that may influence uptake and engagement. We used this framework to inform HOPE technology design and implementation strategies, and to analyze qualitative data in this study.

User Engagement Model
O'Brien and Toms' model ( Figure 2) focuses specifically on the period, or session, of interacting with an intervention. Focusing on websites as an intervention type, it differentiates the following four stages of engagement: point of engagement, period of engagement, disengagement, and re-engagement [31]. It posits that there are different drivers at each stage, grouped into (1) website features and (2) user experience features. Points of engagement and re-engagement are influenced by web features of esthetics and novelty, and user features including motivation. The period of engagement is influenced by sensory appeal, interactivity, novelty, challenge, feedback, attention, awareness, control, and positive affect. Disengagement is driven by website features of usability and challenge, and user features of positive/negative affect, perceived time, and interruptions. These factors vary in influence over time. We used this model to analyze qualitative interview data.  [31]. Reprinted with permission.

Study Objectives
The aim of this study was to analyze uptake, ongoing usage, and engagement with a CHI intervention, the HOPE eIntervention, intended for African American young adults, which included a website and social media components. The research questions were as follows: RQ1: How much did HOPE participants use the eIntervention (uptake and ongoing usage)? RQ2: What factors were related to eIntervention uptake (initiating contact with the intervention)? RQ3: What factors were associated with ongoing usage and engagement while a participant interacted with the eIntervention?

Intervention Context
This mixed-methods, community-based participatory research (CBPR) [44] study included surveys and interviews with African American young adults who attended a community-based HIV/STI prevention intervention called a HOPE party.
Regarding the CBPR approach, YOUR Center, a local faith-based nonprofit organization, was involved in study design, recruitment, implementation, data collection, and result dissemination, and the Genesee County Health Department assisted in recruitment; this is consistent with use of the CBPR approach in health informatics [44]. The team had monthly meetings during the intervention timeframe, and community members were involved in the design of the HOPE eIntervention, from its inception to its implementation. Furthermore, YOUR Center staff led the in-person interventions, and community members assisted by leveraging their social networks for recruitment. A community advisory board helped to integrate community needs throughout the project; this also aligns with CBPR practice [44]. Further information about the partnership can be found in a previous report [44].

Intervention Description
HOPE parties were one-time, interactive, face-to-face HIV/STI prevention sessions hosted by African American young adults in their homes for their social networks. Attendees listened to didactic material, engaged in educational activities, and discussed safe sex practices with demonstrations of safe sex tools [41]. Local resources were also discussed, including information about the eIntervention. YOUR Center staff were trained to conduct the parties and used a facilitator's manual to ensure that core content was covered in each party; fidelity assessments were also used at each party. However, as expected, there was some variation across parties. Recruitment was conducted at Genesee County STI clinics by trained UM School of Public Health, Health Professions and Studies students and community residents. YOUR Center staff also recruited participants through naturally occurring face-to-face professional and personal social networks, and snowball recruitment in which hosts recruited others.
The eIntervention, available to HOPE party participants, aimed to reinforce information presented at the parties and foster commitment to HIV/STI prevention. Together, the face-to-face parties and eIntervention comprised a blended offline/online intervention. The eIntervention was designed in collaboration with YOUR Center's staff and community advisory board, and informed by design-oriented focus groups [12]. Based on the trust concerns when communicating about HIV/STIs that were raised in the trust-centered design framework developed as part of this study [12], the website allowed participants to remain anonymous to foster interactions that preserved privacy. Indeed, community partners expressed concerns about the acceptability of user accounts based on their interaction with the target audience for the intervention; thus, profiles were not required in order to access website content. Due to issues raised around inconsistent access to trusted information about sexual health [12,45], partnering with YOUR Center was intended to create an environment where African American young adult users would trust the information based on its institutional affiliation. Additionally, the blog features of the website allowed for users to access credible information via these short and casual posts. The website had a theme for mobile devices such as cellphones. A website plug-in detected what device was used to access the intervention and loaded either a desktop or mobile theme based on this detection. Users could toggle between the mobile and desktop versions of the website if desired.
At each HOPE party, YOUR Center staff and/or University of Michigan research staff introduced the eIntervention and distributed handouts that contained the website URL and directions on how to access social media accounts. The eIntervention consisted of a website with a blog, a Twitter account, and a Facebook page (examples of posts are included in Multimedia Appendix 1, Multimedia Appendix 2, and Multimedia Appendix 3). The website was updated at least once monthly. Facebook content included status updates, external links, photos, sexual health information, and/or questions. Tweets offered safer sex advice and memes reinforcing safer sex norms. Three social media contests incentivized eIntervention information sharing via social networks.

Data Collection
HOPE party participants (n=315; 280 African American; some participants invited non-African American friends to parties) completed four surveys during the study. These included (1) at the party, in-person (baseline); (2) 3 months postparty, online/via phone (T1); (3) 6 months postparty, online/via phone (T2); and (4) 12 months postparty, online/via phone (T3). There were 57 HOPE parties over 3 years, with 1 to 12 participants at each who responded to surveys. Follow-up survey requests were first attempted over the phone, and those participants whom we could not reach were contacted via US mail and by email. We attempted to contact all survey participants regardless of the status of completion for the other follow-up waves. Participants received US $10 for completing the baseline, T1, and T2 questionnaires, and US $15 for completing the T3 questionnaire. Participants who completed all four surveys were entered into a draw for a US $100 gift card. The survey asked about demographic characteristics, social network information, and questions regarding use of the HOPE eIntervention. Surveys were developed to measure network assessments such as network composition (gender homophily) [46], social influence of networks based on density [47], and type of social tie (ie, weak or strong) [48] as predictors of HOPE website use. Survey questions used in the data analysis are detailed further in Multimedia Appendix 4.
After survey and website server logs indicated low uptake and engagement, individual in-depth semistructured interviews [49] (n=19) were conducted with African American young adult party participants. Interviews were conducted at least 1 year after each participant's HOPE party. One graduate student (AB) from the University of Michigan who was trained in qualitative interviewing conducted the interviews using a semistructured interview guide that accommodated personalized follow-up questions. Using theoretical sampling [50], participants reporting low, medium, and high levels of eIntervention use on the survey were recruited via telephone or email. We contacted approximately 131 participants in total (56 participants with low levels of eIntervention use, 62 participants with medium levels of eIntervention use, and 12 participants with high levels of eIntervention use). Of the 131 participants, 19 participated in interviews (15% response rate).
We concluded interviews when we reached data saturation such that new data were no longer contributing new empirical insights [51]. Others have shown that data saturation can be achieved with as few as six to eight people in a relatively homogenous group such as that in this study (African American young adults in one geographic area) [52]. A timeline of the partnership with YOUR Center, development of the intervention, HOPE parties, follow-up surveys, and interviews is included in Multimedia Appendix 5.
Google Analytics measures included numbers of sessions, users, and page views; session duration; and trends. HOPE Twitter and Facebook data included numbers of followers/likes.

Dependent Variable
At each survey wave, the main outcome in the individual-level statistical model for predicting uptake was a binary variable for having used the HOPE website or social media since the party. For the party-level model, this outcome was an average of the number of party attendees who used the eIntervention at each wave.

Education
The DOI theory states that more education increases the speed of using innovations [42]; thus, educational level was a predictor. In the model, the variable had the following two levels: (1) less than high school diploma or General Education Diploma (GED) and (2) high school diploma or GED, some college, associate's degree, or bachelor's degree.

Gender
Gender was included as a binary predictor (male/female). This was included since men are more likely to have favorable technology attitudes [53], though these attitudes about technologies vary by age [54]. Further, women use health websites more [55] and are more likely to search for health information online [56].

Time Point
Participants' use by time point was included because we observed increases in overall usage of the eIntervention over time.

Gender Homophily
The DOI theory predicts that homophily increases the odds of adoption, and our previous work has shown that gender homophily is associated with HIV testing behavior [46,57]. Thus, gender homophily was used as an independent variable, which was calculated using the Krackhardt E-I index network measure [58] as the number of females minus the number of males at the party divided by total participants [58].

Network Density
Network density [47] was included since it is linked to social influence [57]. This standard network measure [59] quantifies the number of participants at a party who have relationships divided by the number of total possible relationships between party members. Higher levels of network density are associated with health behavior transmission between youth [60,61]. This was calculated as the total number of attendee ties (any self-reported relationship before the party) divided by total possible ties [62].

Strong Tie Proportion
Strong ties [48] are also linked to social influence. They tell us the proportion of relationships between people at a party that are based on close interpersonal relationships [57]. Using a standard measure of tie strength, this survey question asked participants how close they felt to other party attendees, with the options being very close (strong ties), somewhat close, and not close [63]. There is ample evidence that people with whom one is emotionally intimate, such as close friends, influence one's sexual risk behaviors [64][65][66]. This was measured by the number to whom each attendee felt very close divided by the number of existing ties between party members.

Qualitative Interview Protocol
Interviews included (1) a demographic and technology use questionnaire, and (2) interviews about party experience, eIntervention uptake and engagement, internet and social media use, and HIV/STI-related communication practices.

Survey Analysis
Survey data were explored using descriptive statistics. Data analysis was conducted after the conclusion of T3. To identify uptake predictors, Stata was used to fit a generalized linear mixed-effects model [67] with binomial distribution to account for (1) the binary outcome variable (eIntervention use) and (2) repeated measures across the three time points. This method accounts for both within-participant and across-participant variability. The coefficients in this model can be interpreted as the log odds of predicted HOPE eIntervention use by variable. Due to our focus on African Americans, only the sample of African Americans (n=280) was entered into this model.
A second generalized linear mixed-effects model with a beta distribution using party-level indicators was fit. Uptake trends were also analyzed with respect to independent party-level variables such as gender homophily, network density, and strong tie proportion. There was a random effect for the person-level variable and were fixed effects for party ID (identifier) and within-party variables (gender homophily, strong tie proportion, and cohesion density) to predict average uptake at the party level while controlling for party size. For this model, the entire sample (n=315) was entered into the model as parties had both African American and non-African American attendees, although African Americans were the focus of the intervention design.

Interview Data Analysis
Interviews, averaging 43 minutes, were audio-recorded and transcribed. Among the 19 interview participants, 8 (42%) had low self-reported usage levels, 6 (32%) had medium self-reported usage levels, and 5 (26%) had high self-reported usage levels, with representation across usage levels. Transcripts were analyzed using deductive and inductive approaches in NVivo. Deductively, O'Brien and Toms' [31] model of engagement and the trust-centered design framework [12] provided codes; line-by-line coding resulted in inductive codes. One coder initially coded the interviews, and a second coder reviewed the codes and reached consensus with the first. This resulted in emergent categories regarding information sharing, party experience, eIntervention experience, and social media [68]. Analytical memos were used to develop themes after reviewing and combining or collapsing codes [68]. Categories were also developed deductively using O'Brien and Toms' model of engagement in the period, or session, in which users interacted with an intervention. Their model includes initial engagement, period of engagement, re-engagement, and disengagement. In this context, initial engagement is conceptualized as the beginning of the interaction with the eIntervention; period of engagement is the remaining time through which a participant uses the eIntervention; re-engagement is when a participant returns to engage with the eIntervention after days, weeks, or months; and disengagement is when a participant discontinues use of the eIntervention in a given session. Re-engagement and disengagement can be thought of as components of ongoing usage.

Website/Social Media Usage Data Analysis
Frequency counts were compiled for Google Analytics, Twitter, and Facebook data.

Characteristics of the Participants
As Table 1 shows, 280 African American participants responded to the survey, and 19 African American young adult interviewees were drawn from this pool of respondents. Of the 280 survey respondents, 164 (58.6%) were female and 110 (39.3%) completed high school or equivalent, and the mean age was 21.13 years. Of the 19 interview participants, 14 (74%) were female, with similar education levels and a mean age of 24 years. As Table 2 shows, there were 57 parties in total, with an average of 5.63 (SD 2.56) participants per party who responded to the survey. Of the total 315 attendees who responded to the survey, 280 (88.9%) were African American.   b There were 57 total parties, but one party did not have any respondents fill out the survey at baseline, so 56 were included in the analyses. c Includes only party participants who completed a survey.

Survey Results
Initial uptake of the eIntervention and ongoing usage were low (Table 3). Self-reported use increased at subsequent time points. Initially, only 21 of 280 (8%) participants reported ever using the eIntervention (uptake), with an increase to 35 of 280 participants (13%) at T2 and 42 of 280 participants (15%) at T3. The most frequent engagements were visiting the website (T1: n=9, T2: n=24, T3: n=34), visiting the Facebook page (T1: n=1, T2: n=8, T3: n=11), and tweeting (T1: n=2, T2: n=4, T3: n=5). However, large amounts of survey data were missing either due to nonresponse at that time point or due to skipping that question (Table 3). Of the 359 total HOPE party participants, 315 participants took at least one survey. Regarding response at each time period, among 359 HOPE party participants, 178 responded at T1 (49.6% response rate), 180 responded at T2 (50.1% response rate), and 186 responded at T3 (51.8% response rate). Among the 280 African American party participants who ever answered the survey, 20 (7%) answered at only T1, 14 (5%) answered at only T2, and 16 (6%) answered at only T3. Table 3 outlines responses from the 280 African American respondents who ever responded to a survey. surveys, the "general use" question was eliminated due to overlap. e Open responses for "other" were categorized by "I did not know about it" and "I forgot about it."

Log File and Social Media Results
According to the Google Analytics report, the website had 2432 sessions and 5754 total page views from July 2011 to March 2014, with an average session duration of 2 minutes 52 seconds. The Facebook page was updated with five to seven posts weekly and had 81 followers. HOPE Twitter was active from July 2011 to April 2014, with 131 followers. Inconsistency in reported usage may be due to missing survey responses or to eIntervention use by non-HOPE party participants.

Survey Results
To identify uptake predictors across the three time points, we fitted a generalized linear mixed-effects model at the individual and party levels ( Table 4). Individual-level results suggested that having some college education was a marginally significant predictor of eIntervention use. No other individual-level factors predicted uptake in our analysis of HOPE eIntervention use after the conclusion of T3 surveys, although there was an increase in usage from T1 to T2, as well as an increase from T1 to T3. In the party-level model, gender homophily was the only marginally significant group-level variable, indicating a weak positive effect of a high proportion of females (gender homophily) on average uptake rates across participants in a given party. Table 3, most participants who reported not having used the HOPE eIntervention in the past 30 days (of whom 100% had never used the intervention) at T1-T3 said this was due to lack of time (54-58 participants, 41.9%-47.2%). Following this, in frequency, were not having access to a computer (31-38 participants, 24.4%-29.5%), not wanting to (13-23 participants, 10.1%-18.1%), and other (17-24 participants, 13.5%-18.6%). Other responses concerned forgetting about the eIntervention or not being aware of it. Given their frequency, we interpreted perceived time and technology access as "necessary conditions" for uptake. Necessary conditions are facts that must be true for intervention uptake; however, they are insufficient on their own to promote uptake.

Awareness
At least three interview participants were unaware of the eIntervention, although one mentioned visiting the website to see what was on it (Multimedia Appendix 6). Three participants suggested that uptake could be improved through more promotion.

Motivation
Some used the eIntervention with the goal of referring members of their community or family to it, believing they would benefit from learning about prevention (Multimedia Appendix 6). Furthermore, 12 interviewees shared information learned from the website with friends and family. Conversations with friends regarding HOPE usually involved sharing information about HIV/STI symptoms or testing locations (Multimedia Appendix 6).

Necessary Conditions
Necessary conditions were not typically enough to motivate uptake of the eIntervention, but their lack meant that people would not use it.

Technology Access
According to survey results, while some interview participants had internet access on varied devices, three did not have high-quality technology access, with two accessing it through others. Some participants also primarily accessed the internet on a mobile device; this limited their activities (Multimedia Appendix 6).

Perceived Time
Like survey respondents, some interviewees said they were too busy from work, school, and/or child care for the internet in general or for the HOPE eIntervention in particular (Multimedia Appendix 6).

Trust-Institutional
YOUR Center, which facilitated the parties, was considered a trustworthy information source (Multimedia Appendix 6). This perception was aided by personal connections. Three participants mentioned knowing someone linked to YOUR Center, and participants mentioned connections between HOPE parties and their churches. Attending parties organized and facilitated by individuals from trusted institutions made participants more likely to see the information from both the parties and eIntervention as credible.

Trust-Technological
Participants had trust in the website's information, a form of technological trust [12,15]. Seven participants mentioned understanding the information easily, perceiving it as clear and having an appropriate length (Multimedia Appendix 6). However, participants mentioned negative experiences on social media that may have shaped their perceptions of discussing sexual health information online, potentially resulting in lower uptake of eIntervention social media components. People expressed concern about potential technology-facilitated privacy breaches and online fighting. Furthermore, gossip was a major concern. Ten discussed negative experiences with "Facebook Exposed" pages, a user-generated network of pages featuring incendiary posts disclosing personal information about sexual promiscuity, sexual identity, and HIV/STI status.

Esthetics and Sensory Appeal
Participants responded positively to the website's esthetics. Four liked its visual layout, stating it was not too simple and had multiple colors (Multimedia Appendix 6).

Challenge/Ease of Use
Participants valued that information on the website was direct, short, and engaging. Its content was easy to use. Compared to other websites about sexual health, the HOPE website was easier to use and thus more accessible (Multimedia Appendix 6).

Interactivity
The interactivity present on the website, such as the blog, was appreciated. However, three participants suggested making the website more interactive by imitating social media platforms. One suggested user profiles, much like social networking sites (Multimedia Appendix 6). This approach could also offer novelty.

Novelty
As Multimedia Appendix 6 shows, two participants who were initially regular website users stopped visiting due to a lack of consistent updates.

Promotion of the Intervention Through Contests
Social media contests were intended to incentivize participants to visit the website to share information with peers; these fostered some limited re-engagement. Two interviewees participated in the contest, and one male participant said that he was interested in the financial reward and thought it was a good way to engage people with the content (Multimedia Appendix 6).

Principal Results
This study identified factors promoting uptake and engagement with a CHI intervention for HIV/STI prevention among African American young adults. Uptake and ongoing usage were low overall; website uptake increased as more participants entered the study (Table 3), but always remained low. The only marginally significant individual-level positive predictor of uptake was education. There was also a marginally significant positive relationship between party gender homophily and party-level uptake. Awareness and motivation to share information with others positively influenced uptake. Necessary conditions undermined uptake when absent; these included technology access, perceived lack of time, and technological trust, especially regarding social media-based discussions about HIV/STIs. Visual appeal of the website, information with the appropriate level of challenge, and interactivity positively affected ongoing usage, although the website was not interactive enough for some participants. A social media contest also increased re-engagement (a component of ongoing usage), with limited reach. Lack of novelty was linked to disengagement (another component of ongoing usage).

Comparison With Prior Work
Based on these findings, we extended O'Brien and Toms' model of engagement [31] in a "Model of CHI Intervention Uptake and Engagement for African American Young Adults" ( Figure  3). Our model reinforces O'Brien and Toms' inclusion of motivation, challenge/ease of use, esthetics/sensory appeal, and novelty as factors influencing engagement; we found this in our interviews. To make the model more appropriate for this group, we added uptake. Factors driving uptake were identified through interviews (awareness and motivation) and survey data (education level and party network homophily). The model also newly incorporated necessary conditions and intervention context, including promotion efforts and the intervention's platform on social media, which was linked to technological trust. Multimedia Appendix 7 presents which details support each element of the new model in Figure 3. With these additions, we offer interventionists a framework to identify culturally relevant factors for CHI design and implementation among African American young adults. Of the three social influence variables studied, gender homophily weakly predicted uptake, aligning with the DOI theory. This comports with studies that found that network homophily influenced technology use [69,70] and increased adoption of new health behaviors in an online intervention [46]. Strong tie proportion and network density did not influence uptake, contrary to previous findings [47,48]. In prior work, strong ties and network density often resulted in the formation of subjective norms around technology use [48]. With the HOPE eIntervention, it is possible that no strong subjective norms existed. This could explain the insignificant results of the other social influence measures [71]. Future CHI intervention studies should promote normative behaviors regarding use to drive uptake.
Sharing HIV/STI information within social networks motivated uptake and engagement. This is consistent with the design of the parties as educational events within pre-existing social networks and our use of social media contests, as well as with prior work showing motivation to share health information within African American populations [72,73]. African American young adults often have limited access to HIV/STI information [45,74]. Sharing within social networks may help fill this gap, and this could be leveraged more effectively in future work. These results suggest that CHI interventions that target prevention may need to provide easily sharable information [75] applicable to the lives of the audience.
Trust in information sources is fundamental for African American technology users [12,76,77]. HOPE users' trust in YOUR Center had an effect on uptake of the HOPE parties, which in turn affected eIntervention uptake. The website incorporated design elements aligned with recommendations posited by a meta-analysis [78] to influence trust, such as lack of login, in order to protect privacy [74]. CHI interventions for this population must incorporate trust-related considerations, including privacy and security [79].
While the website was designed to protect privacy [12], uptake may have been affected by participants' prior experiences with the discussion of HIV/STI information online, and other trust-related concerns observed in prior work. Observing negative interactions shapes how individuals engage with online interventions [80] and manage online identities [81]. Therefore, these experiences might have particularly deterred use of interactive intervention elements, such as commenting on the blog. Furthermore, interactions, such as liking a Facebook page or following a Twitter account, may have been identifying [82] enough to discourage doing so. While African American young adults may believe that social media can helpfully disseminate health information [83], HIV/STI information may be too sensitive to discuss publicly.
Participants expressed interest in increased interactivity and similarity to sites such as Facebook. Modeling interventions after social media sites may promote engagement. Young adults have the highest social media usage among age groups [84] and may expect interventions to mirror their interactivity. Notably, although a blog provided interactivity on the website, decreasing use of blogs among youth over time [85] indicates that blogs may not have had the requisite popularity with the audience to provide them with desired interactivity. Additionally, given the low uptake of social media-based HOPE accounts, websites targeting HIV prevention or other health issues for youth could implement other components, such as newsfeeds, thus allowing for passive engagement [86,87].
The website fostered initial engagement with esthetically pleasing visuals and audience-appropriate information. Webpage attractiveness may be as important as content for engaging African American young adults [17], who prefer attention-getting colors [88]. Perceived information quality and ease of use also drove engagement. CHI interventions targeting specific populations should present content that is culturally relevant, accessible, and perceived as trustworthy, such as through links to social networks and trusted institutions [75].
Others have shown that lack of time and technology access may be uptake barriers [29,37]. Many survey respondents said that they lacked time to use the eIntervention. More advantaged participants are more likely to have time to engage with health promotion efforts [89], so young African Americans may particularly need CHI designs with low time burden. Additionally, a significant minority of party participants said they did not use the eIntervention because they lacked technology access. African Americans are less likely than white individuals to own computers or have broadband and are more likely to be "smartphone only" internet users [90]. Interventions targeting African American young adults or other groups that may not have access to consistent home broadband should be optimized for cellphones.
Despite attempts to increase uptake and re-engagement through face-to-face introductions to the eIntervention at HOPE parties and social media contests with financial incentives, low uptake may have resulted from insufficient promotion or limited usefulness [91]. Research on African American health intervention use found that face-to-face interactions have a greater impact on health outcomes [92]. Including more in-person promotional campaigns before implementing an intervention may be helpful.

Limitations
This study has some limitations. African American young adults were sampled from one county in one state. While we cannot generalize the results to broader populations, they could be extended to groups at a high risk for acquiring HIV/STIs in other urban settings in the United States. The survey and interview relied on participant self-report, which may have desirability effects; however, this mixed-methods approach allowed for diverse data to inform our analysis. Large amounts of data were missing in the self-report questionnaires, largely due to difficulties with follow-up despite attempts using listed phone numbers and mailing addresses, which is a common obstacle in community-engaged research. Additionally, while Google Analytics, Facebook, and Twitter data provided descriptive information about website use, we could not ensure that individual visitors were party participants because we did not implement user accounts due to privacy concerns and the website was accessible via public search [12]. Additionally, many of the respondents reported being unemployed at the point of the baseline survey, but they may have begun employment or schooling during the follow-up period.

Conclusion
Our study identified factors driving uptake and engagement within an HIV-prevention CHI intervention with African American young adults. We affirm and extend O'Brien and Toms' model to include uptake, individual factors, necessary conditions, and context as displayed in Figure 3. Findings revealed that CHI interventions for prevention among African American young adults should facilitate peer information sharing, be implemented through trusted organizations, be interactive, and offer attention-grabbing esthetic designs, accessible information, and regular updates. Intervention design and implementation must also foster trust and address barriers such as lack of time.