Following the International Patient Decision Aid Standards recommendations [29] for systematic user-centered development, we used UCD methodologies throughout the development process. This approach aligns with current best practices; a recent systematic review of 607 development studies found a widespread adoption of UCD principles, typically involving multiple rounds of patient feedback and prototyping [30]. UCD maximizes intervention usability for the target population by ensuring alignment with users’ needs, values, and preferred contexts [31,32], while improving satisfaction, engagement, and adoption rates [33].

MyLungHealth is a web application distributed through previsit questionnaires in the patient portal in preparation for SDM in upcoming primary care appointments. It provides educational content on LDCT, personalized risk assessments, and comprehensive information on screening benefits and risks in both text and video formats. The initial design was informed by previously identified information needs [11], with subsequent iterative refinement based on feedback from focus groups and interviews. The clinical impact of the resulting intervention was evaluated through a multisite randomized controlled trial at UHealth and New York University Langone Health (NYULH) [34].

We conducted focus groups and individual interviews to gather comprehensive user feedback through an iterative development process. Focus groups and interviews followed a semistructured format, meaning sessions used a standardized discussion guide with prepared questions while allowing flexibility for participants to raise additional topics and for the moderator to probe emerging issues.

The collected data were analyzed using Braun and Clarke’s principles for thematic analysis [35,36]. A multidisciplinary team of 4 researchers experienced in qualitative analyses engaged iteratively with the data to generate codes and develop themes through interpretive discussion and analytic reflection. Consistent with the reflexive thematic analysis, themes were conceptualized as analytic interpretations developed through researcher engagement with participant accounts rather than as preexisting categories awaiting discovery.

To enhance analytic rigor and transparency, we attended to multiple dimensions of trustworthiness throughout the analysis. Credibility was supported through prolonged engagement with the data, regular analytic debriefings, and iterative prototype revisions that functioned as a pragmatic form of member engagement, whereby updated prototypes were presented to subsequent participant groups. Dependability was established through the systematic documentation of coding decisions and the evolution of the analytic framework. Confirmability was supported by a hybrid deductive-inductive approach that anchored analysis in an existing conceptual framework while remaining open to emergent insights, alongside team-based analytic discussions that refined interpretations and thematic structures. Transferability was facilitated through a detailed description of the study context, participant characteristics, and methodological procedures across 2 geographically and demographically distinct sites.

Team members engaged in ongoing reflexivity, recognizing how their disciplinary backgrounds spanning clinical informatics, health communication, and SDM shaped analytic attention, interpretation, and theme development. Consistent with contemporary guidance for thematic analysis [36], these practices were intended to enhance analytic depth, transparency, and reflexive rigor.

The development began with an initial low-fidelity wireframe created using Balsamiq [37], a wireframing software. These low-fidelity prototypes presented broad design concepts with minimal detail [38] and were informed by insights from our prior study [11]. The foundational design was established through collaborative sessions with key stakeholders, including cancer communication experts.

Once the initial design and content were determined, we developed a high-fidelity interactive prototype using Adobe XD [39], which closely replicated the visual design and interactive features of the intended final tool [38]. The final tool was organized into 4 main sections: a brief introduction, educational content on LCS, personalized risk information, and a concluding segment outlining the next steps.

Based on these prototypes, we developed a fully functional web application called MyLungHealth, available in both Spanish and English. The Spanish version was developed using the University of Utah’s foreign language translation service, which provided professional translations of all tool content and associated materials. Throughout the focus groups and interviews, participants were shown the tool via live demonstrations or screenshots, allowing for interaction and feedback.

Recruitment was conducted from primary care patient populations at 2 sites: UHealth and NYULH. This dual-site approach was intentionally designed to capture diverse patient perspectives. UHealth serves a broad patient population across the Intermountain West region, including rural and underserved communities, while NYULH serves a large urban population in the New York City metropolitan area with diverse socioeconomic and cultural backgrounds.

Participants were identified using EHR data from both institutions. Inclusion criteria followed the US Preventive Services Task Force guidelines for LCS [3], targeting adults aged 50 to 80 years with a 20+ pack-year smoking history who either currently smoke or have quit within the last 15 years. Spanish-speaking individuals were intentionally recruited for interviews to review the Spanish translation of the educational tool.

Recruitment was conducted using PHR messages at NYULH and email messages at UHealth, which were linked to a screening survey hosted in REDCap (Research Electronic Data Capture) [40,41] to assess interest and eligibility. The REDCap survey included a consent cover letter. Eligible participants were initially invited to focus groups. As the intervention matured to a stage where significant changes were no longer being made in response to focus group feedback, subsequent invitations were extended to both prior focus group participants and new participants for individual interviews.

We employed multiple qualitative assessment methods, including think-aloud protocols with contextual inquiry [42,43] and critical incident technique interviews, [44,45] in both focus groups and individual interviews.

This study protocol was approved by the institutional review boards (IRBs) of UHealth and NYULH under a single IRB (IRB_00153806), with the UHealth IRB serving as the IRB of record. All participants provided informed consent prior to participation, and participant privacy and confidentiality were maintained. Participants received a US $50 gift card as compensation for their participation.

We applied Braun and Clarke’s principles for thematic analysis to examine the data, as this method is well-suited for drawing out views, opinions, experiences, and values [35,36]. All focus group and interview discussions were audio-recorded, transcribed verbatim, and deidentified for analysis. Field notes were also taken during the focus group and interview sessions and used to inform iterative design modifications to the tool. Anonymized transcripts were securely stored and uploaded to Dedoose (SocioCultural Research Consultants, LLC), a qualitative data analysis platform [48].

Our coding approach combined a hybrid inductive-deductive method. Initially, we applied deductive codes based on the existing codebook from our prior study [11]. We then used an inductive process to develop additional codes for capturing emerging themes. Four team members with expertise in qualitative research, health care, and SDM collaborated in the coding process to ensure reflexive interpretation and minimize bias.

Regular team meetings were held to ensure consistent application of codes, refine definitions, and resolve discrepancies through discussion and consensus [36]. As the analysis progressed, the codebook was refined to identify new patterns and insights directly from the data. Once alignment was achieved among coders, the remaining transcripts were analyzed independently, with ongoing meetings to address emerging issues. At the conclusion of coding, all applied codes were reviewed collectively to ensure consensus, and similar codes were then synthesized into broader themes.

Between each focus group and before subsequent sessions, the research team discussed participant feedback and assessed it for feasibility, work effort, and appropriateness. These team discussions, rather than the formal thematic analysis, guided the immediate refinements to the MyLungHealth prototype. When feedback was clear and consistent with an obvious course of action, modifications were implemented before the next session. When feedback was more nuanced or conflicting, the research team reached a group consensus on whether and how to address the feedback. This real-time iterative approach allowed for rapid refinement based on user input while ensuring that changes aligned with the tool’s objectives and technical constraints. The formal thematic analysis described above was conducted after data collection was complete and served to identify broader patterns and insights across the entire development process, rather than driving the interim design modifications.

A total of 51 unique participants were recruited between May and December 2023: 24 from UHealth and 27 from NYULH. We conducted 7 focus groups with 34 participants and 18 individual interviews. One participant contributed to both a focus group and an interview. Participant demographic characteristics are summarized in Table 1. In the final stages of data collection, no novel themes emerged from either focus groups or interviews, confirming that thematic saturation had been achieved.

Our analysis of focus group and interview data revealed 6 major themes related to participants’ experiences and preferences regarding the MyLungHealth decision aid. Table 2 presents these themes, their associated subthemes, and representative participant quotes.

Our thematic analysis identified 6 key themes that characterized participants’ experiences with the MyLungHealth decision aid, as shown in Table 2. These 6 themes were organized into three overarching domains: (1) interpretation and impact of personalized risk information; (2) autonomy, privacy, and user interface preferences; and (3) integration with clinical workflows and patient portal systems.

Based on participant feedback, several significant design modifications were made to the MyLungHealth decision aid during the iterative development process. Table 3 presents the key interface elements that were modified and the specific changes implemented in response to user feedback.

The design changes focused on addressing key user preferences and concerns identified during focus groups and interviews. To enhance information accessibility, the risk presentation screen was substantially simplified, with detailed risk information moved to optional expandable sections that users could access if desired (Figure 1). This change was made in response to participant feedback about confusion and anxiety related to numerical risk data.

To accommodate different learning preferences, educational content was made available in both video and text formats, allowing users to engage with the material in their preferred way (Figure 2). To address participant concerns about provider awareness of their tool usage, we implemented an EHR alert system to notify clinicians when patients had accessed MyLungHealth (Figure 3).

These design modifications reflect the iterative, user-centered development process and demonstrate how participant feedback directly informed the evolution of the MyLungHealth decision aid from the initial prototype to final implementation.

In this study, the iterative design of MyLungHealth, a patient-centered LCS education tool, was informed by extensive user feedback gathered through 7 focus groups with 34 participants and 18 individual interviews. In a follow-up randomized clinical trial, the MyLungHealth tool was shown to significantly improve LCS ordering and collection of smoking information [49].

The thematic analysis of this feedback provided critical insights that can inform the development of future health care decision aids. Our findings are organized into three domains: (1) interpretation and impact of personalized risk information; (2) autonomy, privacy, and user interface preferences; and (3) integration with clinical workflows and patient portal systems.

A central theme across our findings concerned how participants interpreted and responded to personalized risk information. We found that presenting numerical risk information requires careful consideration, as some users found such data informative and motivational, while others found it difficult to understand, anxiety-provoking, or uncompelling. This observation aligns with our prior focus group study [11], a usability analysis conducted on similar web-based LCS decision aids for veteran populations [22], and a survey and focus group study on patients’ attitudes toward LCS and decision aids, which found that the magnitude of patients’ lung cancer risk and benefits was lower than they expected [10]. To address this concern, we relocated detailed numerical risk information to an optional “More Information” section, simplifying the presentation while still making comprehensive data available to those who desired it. These design decisions reflect a broader tension in risk communication: personalized risk estimates have the potential to motivate screening behavior, but their impact depends on how well they are understood and received by diverse patient populations.

Participants also consistently expressed a preference for simplicity in information presentation, which aligns with recommendations for simplified graphic presentation to enable effective communication [50], as well as recommendations for the design of decision aids by Fagerlin et al [51]. With respect to autonomy, participants valued the ability to update their smoking history within the tool, which has been associated with increased patient engagement and satisfaction [52]. However, this feature raised concerns about data privacy. Emphasizing that personal information would remain confidential and accessible only to their care team may help alleviate these concerns, suggesting that transparent messaging about data usage and privacy represents an area for potential improvement in building user trust.

These preferences for autonomy and privacy extended to how participants wanted information presented within the interface itself. Prior studies also found that patients had issues with navigating decision aids [22], so we ensured that MyLungHealth had a linear flow with a short introduction (education in either video or text form), presentation of risk information, and next steps. Following these best practices, we prioritized presenting less information rather than overwhelming patients with excessive details and offered risk information in both a numerical risk format and through visual aids.

Finally, participants highlighted the importance of integrating patient-facing tools into existing clinical workflows and patient portal systems. Participants expressed a strong desire for actionable follow-up options integrated within MyLungHealth, particularly direct scheduling and reminders. Although direct scheduling without clinician involvement is not feasible, given the need for SDM, this feedback highlights patients’ desire for a more seamless screening process. The implementation of an EHR alert addresses this need by notifying providers when patients use MyLungHealth, helping to integrate the tool into a standard clinical workflow rather than creating an additional step that might be bypassed. Similar to how integration into the patient portal may increase utilization [53], the alert system may increase provider involvement and engagement in SDM, creating a more comprehensive approach to screening discussions.

MyLungHealth is a Health Level Seven International Substitutable Medical Applications Reusable Technologies (SMART) [54] on the Fast Healthcare Interoperability Resources (FHIR) [55] application distributed via previsit questionnaires from the patient portal. MyLungHealth follows the SIMPLE (Standards-based Implementation Maximizing Portability Leveraging the EHR) architectural pattern for integrating patient-facing apps into clinical workflows [56]. By leveraging the Health Level Seven International SMART and FHIR standards, SIMPLE allows seamless application integration into the PHR, without requiring additional software installation, ensuring that the integration supports UCD through its streamlined, user-friendly nature that maximizes intervention usability by the target population.

A strength of this study is the substantial and diverse participant pool, which allowed us to reach saturation for insights. Notably, 24 (47%) participants came from racial or ethnic minority backgrounds, enhancing the generalizability of our findings to diverse populations.

This study has limitations. We recruited only participants who were comfortable engaging in focus groups or interviews via Zoom, potentially skewing our findings toward individuals with higher digital literacy and health literacy. However, this sample selection was appropriate for our intervention, as MyLungHealth requires PHR access for use. Additionally, the study was conducted at only 2 academic medical centers, which may limit generalizability. Nevertheless, the inclusion of sites from different regions (Northeast and Mountain West), including a highly diverse area (New York City), strengthens the applicability of our findings.

The insights gained from this study have supported the iterative enhancement of MyLungHealth in a multi-institutional patient-randomized controlled trial at UHealth and NYULH to assess its clinical impact on LCS rates and patient engagement [34].

Additionally, although participants were screened for comfort in reviewing Spanish-language materials to evaluate the translated version of the tool, only 1 participant ultimately completed an interview using the Spanish-language version. While 7 additional participants indicated initial comfort engaging with Spanish-language educational content during screening, they ultimately declined to do so, explaining that, although they could understand conversational Spanish, they lacked confidence with medical terminology or did not feel fluent enough to provide detailed feedback. As a result, feedback from the single Spanish-language interview was insufficient to identify themes specific to Spanish-speaking users or to independently guide substantial modifications to the translated version. Instead, design changes identified through feedback from English-speaking participants were applied to both language versions. To maintain accuracy, the University of Utah’s foreign language translation service updated the Spanish-language materials after each iterative modification to ensure revisions were appropriately reflected in the translated tool. Future studies with larger samples of Spanish-speaking participants are needed to assess language-specific usability considerations, comprehension of medical terminology, and cultural appropriateness of the tool.

This study reports on the user-centered development process of MyLungHealth, a patient-facing educational tool designed to enhance SDM for LCS. Through focus groups and interviews with diverse participants from 2 health care systems, we identified 6 major themes reflecting user experiences and preferences: challenges with risk information interpretation, variable impact of personalized risk data on motivation, desire for autonomy over health data, preference for straightforward interfaces with multiple format options, expectations for integration with clinical workflows, and mixed experiences with patient portal systems. These insights aligned with prior findings and recommendations by ourselves [11] and others [10,22,50,51] and directly informed iterative design modifications, including simplified risk presentation, multimodal educational content options, and provider notification features.

The findings highlight important considerations for developing patient-centered health care decision aids. Effective tools must balance providing comprehensive information with preventing cognitive overload, offer appropriate contextualization of risk data, ensure seamless integration with clinical workflows, and maintain user control over personal health information. By implementing these design principles within an interoperable SMART on FHIR architecture, MyLungHealth exemplifies how patient-facing applications can be effectively integrated into digital health infrastructure while supporting meaningful SDM. The recently completed randomized controlled trial provides evidence regarding the tool's impact, potentially offering a model for similar interventions addressing other preventive services [49].