Globally, there is a large and growing number of people treated for melanoma localized to the skin who require ongoing surveillance for subsequent new melanoma [1]. Patients are recommended to attend routinely scheduled clinics at intervals varying between 3 and 12 months (clinician-led surveillance) to facilitate early detection of subsequent new primary or recurrent melanoma [2]. However, the optimum frequency and duration of follow-up and the clinical effectiveness of clinician-led surveillance are uncertain [3,4]. Many subsequent melanomas are detected by patients themselves, partners, or family members between scheduled visits [5,6].

These observations have led to the proposal of a new model of follow-up care called patient-led surveillance. This model involves regular and thorough skin self-examination (SSE), teledermatology facilitated monitoring, access to fast-tracked unscheduled clinic visits should the patient identify a lesion confirmed as concerning by the teledermatologist, and potentially fewer routinely scheduled clinic visits [7]. Mobile teledermoscopy is a mobile health store and forward technology in which patients use a mobile dermatoscope that attaches to their smartphone camera during their SSE [8]. A smartphone app is then used to process, track, and send high-quality images to a teledermatologist for assessment [9]. The adoption of mobile health technology interventions and telehealth is dependent upon their acceptance by patients and their treating clinicians [10,11]. Patients at risk of subsequent melanoma have reported that mobile teledermoscopy is acceptable when asked about its hypothetical use [9,12,13] and, in one study, after trying it out themselves (used on a one-off basis) [14].

The MEL-SELF pilot randomized controlled trial (RCT) [15] (ACTRN12616001716459) compared 6 months of patient-led surveillance in addition to usual care (intervention) with clinician-led surveillance (usual care; control). The intervention was found to increase SSE frequency and thoroughness, clinic visit frequency, skin lesion excision, and diagnoses of subsequent new primary melanoma ahead of routinely scheduled visits, with no detectable effect on adverse psychological outcomes. Adherence to the intervention was suboptimal, with only half of the patients submitting any images for teledermatology because of withdrawals and nonresponse. In this nested qualitative study among a subset of intervention arm participants, we aimed to explore patients’ perceptions and experiences of patient-led surveillance using mobile teledermoscopy, to understand possible determinants of adherence, and to identify opportunities for improving implementation of the intervention during the larger RCT.

The MEL-SELF pilot RCT was conducted from November 2018 to January 2020 and recruited 100 patients attending routine melanoma follow-up at 4 skin cancer clinics in Sydney and Newcastle, New South Wales, Australia (3 specialist-led clinics and 1 general practitioner–led clinic). Intervention arm participants were supported to undertake regular SSE and patient-performed teledermoscopy every 2 months. Teledermoscopy tools were provided by MetaOptima Technology Inc. [16], including a mobile dermatoscope (MoleScope I) that integrates with MoleScope (smartphone-based skin imaging app) [17] and DermEngine (a digital software system that facilitates the capture, storage, communication, and analysis of skin images by dermatologists) [18]. Each intervention arm participant also received a booklet of instructions and instructional videos. At the end of the 6-month study period, all 49 intervention arm participants were invited to participate in the qualitative study via postal mail and email, with follow-up invitations as needed.

This study was approved by the University of Sydney Human Research Ethics Committee (X15-0445) and the Royal Prince Alfred Hospital (HREC/15/RPAH/593). All participants provided informed consent. The reporting of this study followed the Standards for Reporting Qualitative Research [19].

Semistructured telephonic interviews were conducted between February and March 2020. An interview guide (Multimedia Appendix 1) was developed by the authors, including the 2 consumer investigators (CL and DL). The interviews were conducted by 3 members of the research team, all trained in qualitative interviewing (EH, AM, and ES). The interviews were audio recorded and transcribed verbatim using a transcription service. Quantitative data on demographic and clinical characteristics and on adherence were collected as part of the pilot RCT using web-based surveys (REDCap [Research Electronic Data Capture; Vanderbilt University]) and the data analytics on image submission from the trial’s teledermatology platform (DermEngine). RCT data regarding occupation were clarified and expanded upon in the qualitative interviews and then used to categorize the participants into occupation groups. Adherence data from the pilot RCT were used to group participants into categories of adherence, and then, we compared patient accounts between and within these categories. Preliminary codes were developed inductively from a subset of 6 transcripts [20] independently by 2 researchers, both experienced in thematic analysis (EH and DD). Preliminary codes and analytic memos were reviewed by the research team, which included researchers from a range of backgrounds, including clinical epidemiology, health psychology, behavioral science, and health economics. The emerging themes were identified as analogous to the constructs of the technology acceptance model (TAM). The general TAM framework posits that a person’s intent to use and actual use are predicated on their perception of the technology’s ease of use (usability) and usefulness (benefit) [21] and has been used previously to assess the acceptability of apps to support health care delivery [22-29]. Thereafter, we used an inductive and deductive coding approach based on TAM [30]. Agreement between coders (DD and EH) was high, and discrepancies were resolved through consensus. The framework analytic method was used to organize codes, identify themes, and explain how they relate to each other [31]. Data saturation and interpretation were determined through ongoing coding of the remaining transcripts and discussions with the research team. Coding was performed in Microsoft Word, and Microsoft Excel was used for the thematic analysis using a data matrix.

Of the 49 intervention arm participants invited, 43 (88%) responded and 20 (41%) agreed to participate in a telephone interview. Interviews ranged in duration from 13 to 36 minutes. Those who participated in an interview were more likely to have submitted at least one image (16/20, 80%) compared with 53% (26/49) of intervention arm participants. Participants’ demographics and frequency of image submission are summarized in Table 1. All interviewees thought that the intervention was a useful concept and a great idea for people treated for melanoma. They said that it could potentially provide quick access to an expert’s opinion between scheduled visits and save time on physician’s appointments, particularly for those who live at a distance from specialist services and would otherwise delay accessing care. We interpreted these hypothetical benefits (motivation to use the intervention) and reasons participants gave for not using the intervention at all as intention to use. Patterns of use throughout the trial follow-up (including no use) were interpreted as actual use. Figure 1 shows the adapted and extended TAM. Multimedia Appendix 2 includes additional illustrative quotes.

Of the 20 intervention arm participants interviewed, 16 (80%) used the intervention tools to image lesions and submit them to a teledermatologist for review. Of these 16 participants, 1 (6%) submitted images at all 3 time points and 9 (56%) submitted images at 2 time points. These 10 participants (10/20, 50% of interviewees) were categorized as more adherent. Of the 20 participants, 6 (30%) submitted images at 1 time point; they are referred to as less adherent. The remaining (4/20, 20%) participants did not submit any images and were referred to as nonadherent. These 4 participants provided important data to help understand the high rate of nonadherence in the trial. One participant did not have a compatible smartphone, and 2 participants did not use the tools because of competing time commitments and the unavailability of their skin check partner. The fourth participant reported a sense of tech overload or app fatigue and did not want to use yet another app. All 4 participants believed that their existing routinely scheduled clinic follow-up visits were sufficient for melanoma surveillance.

In this qualitative study of patients randomized to patient-led melanoma surveillance using teledermoscopy, we found that, in practice, among participants who were more adherent (submitted images at 2 or 3 time points), the intervention prompted increased awareness of their skin and SSE practice, reassurance, and early detection of subsequent melanoma. These more adherent participants were those who found the intervention easier to use because of working in a digital technology–related occupation and by having an effective skin check partner. Those who submitted images at only 1 time point found the tools too difficult to use. This outweighed the perceived potential benefits and impacted their intention to use the intervention tools. Although a few participants found the intervention tools easy to use from the start, most participants experienced varying degrees of difficulty in taking clear images and encountered several developmental glitches and navigational issues in the app. These participants needed repeated practice before they found the tools easier to use. Perceived usefulness was lower in people with many moles, especially when it prompted unnecessary clinic visits. Those who did not submit any images explained that their nonadherence was because of competing time commitments, not having an available skin check partner, not having a compatible mobile phone, or app fatigue.

In our study, anxiety was not a clear delineating factor between those who were more or less adherent; however, it was present in the experiences of most participants and their skin check partners. High perceived health threat associated with a personal history of melanoma, increased patient and skin check partner anxiety, particularly among those with many moles. Accounts of stress and anxiety among these patients in our study and in other studies suggest that patients with many moles may require more ongoing support to conduct patient-led melanoma surveillance because of difficulty in selecting moles and increased risk [14,33].

Aligning with the core hypotheses of the TAM framework [21], we found that ease of use and usefulness influenced intent to use and actual use. However, we also found that despite the initial intention to use, perceptions of ease of use and usefulness after the follow-up period were more influential in explaining actual use. Our findings also indicate that actual use impacted ease of use and usefulness, in that those who used the intervention over a longer period found it easier to use (after getting used to it) and experienced more of its benefits. TAM is commonly used to assess factors that influence the intention to use health technologies. Intention to use is interpreted as a measure of acceptability [23,25,27], even when participants are only asked about hypothetical use [26] or it is not clear if all participants have used the technology [29]. In studies that include actual use in their final TAM model, intention to use is not always a strong or statistically significant predictor of actual use [24,25]. Intention to use, measured hypothetically or after a short period of use, may not always be a good predictor of actual use behavior [28], particularly when digital technology requires a period of learning and is being used in the management of high-risk conditions. Qualitative assessment after implementation, and over time, may produce a more accurate and comprehensive understanding of context and patient and intervention characteristics that influence actual use behavior [34] to better inform implementation strategies.

Overwhelmingly, participants suggested that training and demonstration were necessary for themselves and their skin check partner. One-to-one training in SSE has been found to result in greater SSE skill acquisition compared with a paper workbook or electronic interactive training [35]. Previous research has found that a partner’s attendance at SSE skills training increased the frequency of SSE [36,37]. A patient and partner working together as a dyad has also been found to improve SSE practice [38]. In their assessment of mobile teledermoscopy, Horsham et al [14] found that most participants had the help of a family member to take photos and submit images. However, the necessity of having a skin check partner excludes people who do not have someone to help them take images regularly, and further consideration is needed on how to best support these people to undertake regular self-surveillance and act on their results.

The lack of image submission by almost half of all intervention participants in the pilot RCT could be explained by the additional effort and time needed to learn to use the intervention and then use it routinely with a skin check partner, in addition to usual care. When combined with a high perceived health threat from their increased melanoma risk, this may mean that some patients prefer to rely solely on their physician for follow-up care [7,34]. However, early one-on-one training and demonstration may make the learning process less daunting, encourage participation of skin check partners, and create a supportive connection between trial staff and participants. This may encourage more participants to try the intervention and to continue to engage with it over a longer term. Among our study sample, those who used the intervention over a longer period reported more positively on ease of use and usefulness, highlighting the importance of supporting skill acquisition to increase self-efficacy.

Our findings have assisted in refining the design and implementation of a larger ongoing RCT on patient-led melanoma surveillance [39], and several changes have been made. To help overcome barriers to perceived usefulness, particularly for those with many moles, a target lesion will be selected by the treating clinician, an approach suggested in previous studies [9,40]. To reduce the potential for medical overuse [41], the need for a minimum number of lesions has been removed. If the patient does not have other lesions of concern, they will not need to submit any images other than those of the target lesion. Clinical practice guidelines consistently recommend that patients should be taught SSE, but the optimum frequency of SSE and teledermatology remains ill-defined [42]. The frequency of image submission requested in the pilot trial was assessed to be too high, as only a small proportion of participants in the pilot RCT, including only 1 qualitative study participant, were able to submit images at all 3 time points [15]. The frequency of image submission will be reduced from every 2 months to every 3 months. All intervention arm participants and their skin check partners will be encouraged to participate in one-to-one demonstration sessions with the study staff, in addition to receiving instructional videos and written instructions. The study staff will also be available for the duration of the follow-up period to assist patients with troubleshooting. In addition, the technology provider has made several improvements to the app and teledermatology feedback system, which addresses the technical difficulties experienced by the participants, including nonreceipt of the teledermatologist’s report.

The patient-led surveillance approach has the potential to partially (or completely) replace routinely scheduled visits. However, during this initial stage in which we are evaluating the safety and effectiveness of the intervention, we have implemented it in addition to usual care, both in the pilot trial that this study relates to and the larger trial that is ongoing. In the process of co-designing these studies, it was clear that both clinicians [43] and patients [12] preferred implementation in addition to usual care as a first step before considering circumstances in which it might replace (in part or in whole) routinely scheduled visits. Although we implemented the intervention in addition to usual care, we are surveying patients regarding their acceptance of a hypothetical reduction to their routinely scheduled clinic visits. We envision that this may inform situations in which the intervention might replace some routinely scheduled visits. Data from interviews with clinicians involved in the pilot trial were also informative. The clinicians indicated that after experiencing the actual use of the intervention in the pilot trial, they anticipated that in some clinical scenarios, it may replace routinely scheduled visits—in particular, where a specific lesion is being monitored for change (these findings have recently been corroborated; Drabarek, D, unpublished data, May 2022). In other scenarios, it may be used in addition to usual care to triage which patient concerns warrant review in the clinic—in particular, where a new lesion needs evaluation. Further exploration of these different uses of patient-led surveillance, the patients most likely to benefit from this approach, and integration with other approaches to surveillance [44] could help to define how it may be used in the most clinically effective and cost-effective way.

As our findings draw on patient experiences over the 6-month trial period, they provide novel insights into the implementation of patient-performed teledermoscopy interventions. As the study period allowed for repeated use of the intervention tools, we were able to interrogate a variety of adherence patterns and identify facilitators and barriers to these and their determinants. These findings have assisted in anticipating and mitigating the risk factors for low adherence in the larger MEL-SELF trial and may also be useful for future studies. Our findings also draw on the experiences of a study sample with variations in relevant demographic characteristics such as age, residing in metropolitan or regional areas, and digital technology self-efficacy. However, as an opt-in recruitment method was used, the qualitative substudy sample was much more adherent than those who did not agree to participate in an interview; thus, additional explanations for nonadherence may have been missed. Further research is necessary to understand the low uptake of mobile teledermoscopy interventions among patients. In addition, some aspects of the results reflect the use of software that was in its development phase, and the findings may not be transferrable to patients using teledermoscopy technologies created by different developers. A time frame longer than 6 months may have revealed further determinants of actual use. Finally, because interviews were conducted at the end of the pilot trial, experiences of learning to use the intervention tools at the beginning of the trial may not always have been reported accurately.

Patient-led surveillance is a complex behavioral change intervention. It aims to improve patients’ knowledge, skills, and confidence in performing SSE using digital technologies so that they are better able to detect and act on concerning changes to moles and other skin lesions. Understanding how and why patients do or do not use this intervention is fundamental to increasing adherence within a clinical trial setting and increasing uptake in clinical practice, if it is found to be a clinically beneficial and cost-effective method of melanoma surveillance. Ultimately, it may allow access to melanoma follow-up care regardless of geographical location [45] and could become a new normal method of surveillance after the COVID-19 pandemic [46].