In the United States, colorectal cancer (CRC) is the second-leading cause of cancer-related death among both men and women. Despite improvements in incidence and mortality rates nationally, Black men bear a disproportionate burden of disease compared to the national average [1-3]. CRC incidence rates for non-Hispanic Black men in the United States are 58.3 per 100,000 compared to 46.9 per 100,000 among all men, with mortality rates of 25.9 per 100,000 compared to 17.7 per 100,000 among all men with CRC [2]. Multiple factors contribute to these inequities among Black men, including poorer access to preventive screening, aversion to colonoscopy, and limited knowledge of alternative screening modalities [4,5]. These inequities result in Black men reporting one of the lowest screening completion rates. To reduce CRC disparities among Black men, there remains a critical need to develop scalable CRC-screening promotion interventions that are tailored to meet the informational, cultural, and decision-making needs specific to Black men [6-10].

Home stool screening, through the fecal immunochemical test (FIT) kit, has been demonstrated as an effective alternative to colonoscopy [11-14]. FIT is a non-invasive stool test that requires no preparation, can be done in the comfort of the patient’s home, and is inexpensive. Previous patient outreach studies have shown that when given a choice to screen [2] for CRC, screening rates are greater for FIT compared to colonoscopy among Black patients (43% vs 26.6%) [12].

Digital health interventions, including telehealth, are provisions of health and prevention resources to patients irrespective of their geographic location. The long-term benefits of telehealth indicate reductions in avoidable health care service utilization and related costs [14-17]. Short-term benefits include the ability to disseminate and personalize health information to patients in real time [14-16]. Telehealth interventions can effectively promote CRC screening but have been underutilized in the promotion of FIT screening. This study aimed to combine telehealth with FIT to overcome historical barriers associated with accessing CRC-screening resources among Black men.

The evolution of telehealth interventions now encompasses highly customized interactions with virtual clinicians (VCs). The VC featured in this study is entitled “Meet ALEX.” Agent Leveraging Empathy for eXams (ALEX) is an online virtual app using a visual representation of a health care provider that educates patients about the importance of CRC screening, identifies CRC-screening barriers, and provides a detailed description of the FIT kit. ALEX is interactive and replicates an in-person interaction using a combination of verbal and nonverbal behaviors. ALEX incorporates theory-driven conversational segments that include personalized risk information and barrier reduction strategies. Critically, ALEX can be tailored to reflect the demographic background (eg, race and gender) of the patient [18]. A growing body of literature has revealed demographic concordance between a health care provider and a racial/ethnic minority patient can increase adherence to recommended preventive testing and improve health outcomes [13,14]. Therefore, coupling a demographically concordant ALEX with FIT navigation offers promise to reduce CRC disparities among Black men. This method is only strengthened by incorporating Black men in the process of developing ALEX.

A rigorous user-centered design (UCD) process was conducted to incorporate feedback from Black men to ensure ALEX was perceived as credible [19]. To provide a framework to interpret this data, we adopted the modality, agency, interactivity, and navigability (MAIN) model. The MAIN model offers an organizational framework to understand how a technology’s affordances (ie, its capabilities) affect a user’s perceptions of an interface [20]. The four broad categories of technological affordances are modality, agency, interactivity, and navigability. These affordances cue cognitive heuristics that lead to positive or negative credibility judgments after interacting with the technology [20-22].

In the MAIN model, each affordance is external to the content but affects the users’ perception of the media’s quality and credibility. The MAIN model has been applied to internet search engines, social media, video games, and virtual reality interfaces [23-25]. What is not always highlighted when the model is applied are the cues. Cues give way to some type of interaction between the designers and receivers in computer-mediated communications [26]. Cues are also known as effectors on users’ interactions and perceptions [20,24,26-28]. In this study, we apply the MAIN model to ALEX, an online app that can be assessed by any device that can connect to the internet.

The purpose of this study was to (1) conduct a three-phase UCD process with Black men to develop and iteratively improve ALEX and (2) utilize the MAIN model to better understand heuristic cues identified by the Black male participants that lead to positive credibility judgments of ALEX before clinical dissemination.

Black men participated across a three-phase UCD process, which included (1) needs investigation, (2) prototype development, and (3) evaluation [29,30]. A detailed overview of the UCD process has been published previously, which incorporated transdisciplinary expertise across communication science, computer science, a community advisory board, and a clinical team to develop the ALEX app [19].

Black men from the ages of 50 to 73 years were recruited between January 2017 and November 2018 using a purposive sampling strategy [20]. To be eligible for the study, participants had to be active patients in the university health care system; this inclusion criterion was established, given that the ALEX character would be used in a subsequent clinical trial involving a similar population to promote use of FIT. Eligible participants were identified via university-affiliated community engagement organizations and local senior centers from rural counties of North Florida. Participants were recruited by study staff through phone, email, and in-person methods [19].

Focus groups were conducted in person across various locations throughout the community. Each moderated focus group lasted between 60 and 90 minutes. To ensure rapport with the participants, the moderator was also a Black male. Institutional review board approval was granted for this study, and the participants provided informed consent prior to being audio-recorded. At the beginning of each interview, participants were provided information about the overall aim of the study and how the study team would incorporate participant feedback to help develop the Black male VC in the ALEX app. Each participant received a $40 gift card as compensation for their time and effort.

Feedback was collected across 8 focus groups to iteratively and systematically develop the Black male VC. Instead of recording the script and developing the interactive VC animation for the first set of focus groups, we obtained initial feedback on the still images of the VC and then applied a script and engaged in character development. The UCD ensured the involvement and perceptions of the Black men throughout the entire process, systematically tailoring the VC based on their feedback. Each focus group discussion began with identifying the participants’ familiarity, understanding, and knowledge of cancer, cancer screening, and virtual human technology. Participants in the first and second interviews viewed still images of the VC. Participants in the third, fourth, and fifth focus groups viewed iteration 1 of the VC prototype. Participants in the sixth and seventh focus groups viewed iteration 2 of the prototype, and participants in the eighth focus group provided feedback on the third and final iterations of the VC used in the ALEX app.

Audio recordings were transcribed verbatim and reviewed for accuracy. Transcripts were uploaded into NVivo (version 12.0; QSR International) to allow for electronic coding of the data. The data were analyzed using thematic analysis. Two coders (authors MV and DW) established an initial codebook with guidance from a senior author and expert in qualitative research (author JK). The codebook included a priori codes derived from the credibility literature and domains of inquiry from the interview guide. The full codebook is published in a previous publication [31].

Codes reflected theoretically informed components of credibility and guided a deductive coding approach for the presence of affordances, cues, and heuristics from the MAIN model, as seen in Figure 1. These cues prompted multiple heuristics such as social presence, authority, interaction, responsiveness, choice, control, and scaffolding, leading to an overall quality experience triggering enhanced credibility. The qualities associated with enhanced or positive credibility include uniqueness, trustworthiness, expertise, appearance, understandability, believability, clarity, importance, relevance, and representativeness. As shown in Figure 1, cues and heuristics were organized according to the affordance where modality examined how the content was conveyed through media (visual, text, audio, etc). Agency identified the source, that is, who or what was providing the content through the media interface (VC, text). Interactivity affordances identified cues that were associated with being in the presence of the digital media interface (the actions of clicking). Navigability examined the digital media interface’s ability to guide the user through the content (dialog boxes, links, etc).

During a line-by-line reading of the transcripts, emergent codes were added to the codebook through an inductive coding process. All codes were operationalized and assigned provisional definitions with exemplary quotes [31]. A formal coding framework was then established and used to determine interrater reliability (IRR). Two coders independently coded the transcript. An acceptable IRR was established by a κ coefficient above 0.8. The IRR was calculated using NVivo’s coding comparison query. Initially, 16 coding discrepancies were documented and sent to a third reviewer (JK) for feedback. After 3 rounds of coding, low κ scores (below 0.5) were determined to be due to unitization issues rather than disagreement on coding [27]. To address this, transcripts were segmented by both coders by isolating text in the transcript using box frames to create a unit. Each unit was coded in its entirety to ensure the coders would capture the same information from beginning to end. In the fourth and final round of coding, κ was above 0.8 for all codes, and the rest of the transcripts were coded.

Qualitative focus groups (included 7 focus groups and 1 individual interview) were conducted with a total of 25 Black men. The 1 individual interview occurred because 2 of the invited participants rescheduled on the day of the focus group. Therefore, he was the only participant during his focus group [19,29,32]. Demographic information of the men recruited can be found in Table 1. In summary, Black men were between the ages of 50 and 73 years (median age 61 [6.12]), with income ranging from less than $10,000 to over $100,000 annually. Their education levels ranged from eighth grade level to postsecondary and professional degrees.

Four of the focus groups were exclusively Black men, with three focus groups including a combination of Black and White men (Table 2). Feedback from the Black men was extracted and analyzed from the combination groups, as also shown in Table 2. For the developmental iterations of ALEX, the results are presented according to the affordances associated with the MAIN model: modality, agency, interaction, and navigability.

In iteration 2, 3 focus groups engaged with ALEX on a Samsung smartphone. In this iteration, participants viewed a computer software program that created an initial rendering of the VC, as shown in Figure 3. The VC could speak to participants, and the app contained subtitles corresponding to the speech.

Feedback from the previous focus groups that examined iteration 2 of ALEX led to a re-evaluation and development of the overall agent. This redevelopment of ALEX transpired over 9 months. The VC was further modified by a 3D artist to improve the appearance and match the health care providers' clothing. Nonverbal animations were recorded by recruiting gender-matched actors using the Vicon motion capture system. The virtual character’s voice was recorded by professional voice talents who were race- and gender-matched with the virtual character. To make the environment like a real health care provider’s office, we used a background image of a real clinical room. Changes were made to the Black male VC’s eye shape and color, mouth, figure, hair, and posture, and more realistic movements were added to improve anthropomorphism. He was given props, such as a lab coat, name tag, and stethoscope, to convey an authoritative level of medical expertise, as shown in Figure 4. Facial expressions and hand gestures were added to the VC so that he could emotionally connect with the user.

The final focus group viewed the fourth iteration, which was the third and final version of the ALEX app, as shown in Figure 5, and the alterations were well received. Recommendations and comments from the previous focus groups concerning ALEX’s movement and facial expressions were incorporated to increase the humanistic behavior of the VC. As the script was read by the hired voice actors, their facial expressions were recorded and the computer science team of the project applied the movements to ALEX, focusing on eye contact and mouth movements. The spatial usage of the patient room was also enhanced with a picture on the wall, as recommended by a participant, and an upgraded computational setup, and a patient bed was added for a more realistic experience, as shown in Figure 5. Near and far recordings were aligned with different parts of the script to assist with an enhanced interaction. In addition, a video of African American male physician introducing ALEX, virtual human technology, and the purpose of the intervention was added to the app to address previous concerns and comments addressing not using a person to deliver the intervention. This enabled changes in the script that allowed enabled ALEX to focus on CRC instead of having to explain the concept of a virtual human.

One of the interesting outcomes from all focus groups was the attitude change toward screening after engaging with ALEX and learning more about CRC and the FIT kit. The content itself was novel to the men, and participants were interested in learning more, talking to their doctors, and sharing the ALEX app with others. One participant who viewed iteration 3 commented,

It was important for the men to get the message about CRC screening out to those who could benefit:

Specifically, the VC app was seen as supplemental information that may not be obtained otherwise.

The novelty of the app was engaging and helped men see the benefits of CRC screening:

The demonstration of how to use the FIT kit was well received and informed one of the participants who had the kit at home but did not know how to use it”

For other men, this information was viewed as novel:

In addition, it provided a new way of looking at CRC screening:

In this study, Black men informed the development of a Black male VC, ALEX, that provided education on CRC, at-home screening recommendations, and risk factors. Participants influenced the visual, audio, and textual content of this app, ensuring it would be attractive and informative for future men who would interact with this app as part of a clinical trial.

The MAIN model was applied to this study to assess the credibility of ALEX. Feedback from the participants was categorized according to the four affordances.

One unique aspect of this study was the overlapping heuristic cues identified. Traditionally specific heuristic cues were associated with specific affordances [20,21,27,28]. Some heuristics such as social presence and identity were found to be associated or cued by both agency and interaction affordances in different mediums [20,21,27,28]. In this study, multiple cues triggered overlapping heuristics that led to overall enhanced user perception and credibility judgement with the VC and usage of the FIT kit. The converging cues triggered multiple heuristics associated with multiple affordances. The heuristics then led to positive qualities, including expertise, understandability, believability, clarity, importance, relevance, appearance, and representative, which were all referred to by the focus group participants at one time or another throughout this study. For instance, the multiple modalities associated with the modality affordance triggered the being there, realism, and novelty heuristics and also triggered social presence and identity when the men called ALEX their “brother-man” or would rather interact with ALEX instead of their doctor. They perceived him as a member of their inner circle because multiple modalities were provided. They could visually see ALEX and watch his movements, eye contact, and mannerisms. They could also hear the audio and his words, diction, and tone. This increased the credibility of the ALEX app.

Through their interaction with the app, the anthropomorphism of ALEX improved the most throughout the focus groups. The realism and authority heuristics were cued from the final corrections needed to establish anthropomorphism. ALEX’s movement was one of the largest distractions for the focus group members. Once the movement of the VC perfectly aligned with the audio, the men were able to focus more on the content of the app, as shown in the results.

Since the goal of this app is to encourage the user to make a positive decision regarding their health and screening for CRC, a novel affordance emerged, the content itself. The content examined the characteristics associated with conveying the right message. The men spoke of accuracy of the content related to having gone through FIT previously. The importance and understandability of the content motivated the men, and they wanted to share the app or obtain more information from their doctors. All possible cues (understandability, clarity, accuracy) for the content are examples of converging cues, where they lead to authority, helper, own-ness, and novelty heuristics associated with other affordances.

ALEX was seen both as an authority or a helper by the men, where it could serve as a resource at the doctor’s office. Others saw it as an intelligent entity and would rather engage with the app instead of a doctor. The own-ness heuristic, defined as the ability to adapt the content to the perceived interest of the users, enabled the men to make recommendations for the content and identify things that were confusing to them. We were then able to address those concerns and personalize and tailor the information for the intended users as part of the clinical trial.

The FIT kit as an option for CRC screening was highly novel to most of the focus group participants. Multiple studies have explored the methods of promoting screening via at-home fecal testing within underserved populations [6,7,33-36]. A systematic review published in 2018 identified 27 unique CRC-screening interventions using FIT or the fecal occult blood test with low-income and rural populations in different settings [36]. The authors identified the most effective studies focused on increasing community access to fecal testing, increasing or providing delivery of testing kits, developing tailored decision aids promoting screening, or interacting with patient navigators, clinicians, or community members to increase understanding of fecal testing [36]. The ALEX app captures all of those efforts in one setting.

ALEX serves as a virtual patient navigator that was designed by the targeted population, promotes screening, and provides delivery of the testing kit. Overall, the novelty and coolness of the ALEX app promoting FIT kits was well received and led to increased interest in screening, which is the goal of the app. The Black men in our study were not only interested in the FIT kit after interacting with ALEX but also wanted to share the app. This interface was liked and found trustworthy by the Black male focus group participants.

In this study, the feedback from Black men helped enhance the design of the ALEX app. Future implications of the ALEX app include involvement as part of a clinical trial to deliver tailored messages promoting FIT uptake. This study entailed a great deal of reflexivity on the part of the research team. The current social climate has made direct conversations around race challenging, which is one reason we think this work is so important. The research team is racially and ethnically diverse, and we engaged in weekly conversations about both the conduct of the research study as well as the larger context. We also drew on the experience and wisdom of a racially and ethnically diverse community advisory board. We have built this relationship over the past 4 years, which enabled us to have direct and meaningful conversations about the methods as well as the results.

As digital health care interventions continue to advance, involving specific populations in the design and development of these technologies is critical. Targeted interventions, such as the one in this study, can lay the groundwork for engaging Black men in science. Long term, it can increase participation in health care interventions by promoting trust, inclusion, and responsiveness.