Abstract
Introduction
‘Store and forward’ teledermoscopy is a technology with potential advantages for melanoma screening. Any large-scale implementation of this technology is dependent on consumer acceptance.
Aim
To investigate preferences for melanoma screening options compared with skin self-examination in adults considered to be at increased risk of developing skin cancer.
Methods
A discrete choice experiment was completed by 35 consumers, all of whom had prior experience with the use of teledermoscopy, in Queensland, Australia. Participants made 12 choices between screening alternatives described by seven attributes including monetary cost. A mixed logit model was used to estimate the relative weights that consumers place on different aspects of screening, along with the marginal willingness to pay for teledermoscopy as opposed to screening at a clinic.
Results
Overall, participants preferred screening/diagnosis by a health professional rather than skin self-examination. Key drivers of screening choice were for results to be reviewed by a dermatologist; a higher detection rate; fewer non-cancerous moles being removed in relation to every skin cancer detected; and less time spent away from usual activities. On average, participants were willing to pay AUD110 to have teledermoscopy with dermatologist review available to them as a screening option.
Discussion and conclusions
Consumers preferentially value aspects of care that are more feasible with a teledermoscopy screening model, as compared with other skin cancer screening and diagnosis options. This study adds to previous literature in the area which has relied on the use of consumer satisfaction scales to assess the acceptability of teledermoscopy.
Introduction
The state of Queensland in Australia has the highest rate of skin cancer in the world including the highest rate for melanoma mortality. 1 In 2011, there were 3249 new cases of melanoma diagnosed in Queensland, which was the second leading form of cancer for both men and women. 2 In addition, around 133,000 cases of non-melanoma skin cancer are diagnosed in the state each year. 3 This presents a major challenge to health services in terms of the high number of cases, their differential diagnosis and often late presentation of tumours. The three main types of skin cancer detected are basal cell carcinoma (BCC), squamous cell carcinoma (SCC) and melanoma. 4 It is important to find all melanomas very early as recent results have shown that even those under 1 mm thick may lead to death. 5 BCCs and SCCs also benefit from earlier detection, but due to their lower mortality rates, less crucially so. Despite this there is currently no population-based screening programme implemented in Queensland or Australia. People are advised to check their own skin (skin self-examination (SSE)) and present to a doctor urgently if any spots or moles change.6,7 Opportunistic screening is also performed by General Practitioners (GPs; family doctors) and increasingly by skin cancer clinics, which are generally staffed by GPs, some with additional training. 8 Dermatologists cannot be accessed directly by consumers in Australia, requiring a referral from a GP.
Teledermoscopy is a technology that captures images of potential skin cancers using a hand-held dermatoscope. ‘Store and forward’ technology is used to upload and send the image for diagnosis.9,10 This technology is now available as a hardware addition to a mobile phone and consumers can submit good quality images for triage with minimal training. 11 Studies have found that results from teledermoscopy images, reviewed by an experienced teledermatologist, have high concordance with face-to-face diagnoses, as well as approximating 100% sensitivity and around 90% specificity to differentially diagnose melanoma and non-melanoma skin cancers.11–14 Thus, teledermoscopy has the potential to assist with managing clinician workloads, increasing access to dermatologists and potentially avoiding unnecessary biopsies and other investigations which occur at a higher rate when performed by non-dermatologist and less experienced clinicians. 15
However, consumer acceptance of this approach has yet to be established and is essential if broad implementation of this strategy is to be considered. Consumer preferences have predominantly been measured by ‘consumer satisfaction’ rating scales, with somewhat mixed results.16–22 A review paper by Demiris et al. 18 identified 14 studies related to consumer satisfaction and acceptance of teledermatology applications more broadly (including both the ‘store and forward’ technology as well as video conferencing). The authors concluded that the concept of satisfaction is multidimensional and should cover many underlying factors including convenience of the service; confidence in the result; and ease of communication with the clinician. 18 However, one of the main limitations with using satisfaction rating scales is that they are uni-dimensional.
A discrete choice experiment (DCE) is a type of survey which elicits consumer preferences around service delivery in a number of different disciplines, including increasingly in health.23–25 DCEs are designed to simulate real-life consumer choice situations where more or less attractive characteristics of a product or service are traded relative to alternative options. In this way, the key average drivers of choice for a population can be determined and the potential uptake of a new service estimated. In this study, we use DCE methodology to measure consumer preferences around the key characteristics of skin cancer screening options including teledermoscopy, in order to provide insights into their acceptability. In this study, we investigate the preferences for skin cancer screening options compared with skin self-examination in a group of people considered to be at increased risk of developing skin cancer.
Methods
Attributes and levels used in the discrete choice experiment.
The levels for the SSE were described in the survey text.
Constrained to be a GP who reviews the result if appearing with screening being performed at a skin cancer clinic or GP clinic.
Not including additional costs for biopsy or follow-up tests or treatment.
SSE: skin self-examination; GP: general practitioner
These alternatives, and the different attribute levels over which they vary, were chosen based on the results of three previous surveys undertaken in the target group, 11 as well as a literature review,14,16–22,27,28 and included important screening outcome, financial and convenience considerations. In each choice set participants were also asked to compare these service options with two ‘opt out’ options – either SSE (that is, performing a skin check on your own without the help of a health professional), or not to undertake any screening at all.
An example choice question.
The survey also collected information on socio-demographic characteristics; current and intended skin-screening practices; and the level of concern about developing skin cancer in the future. These questions had been validated before collection in previous questionnaires; however, the choice questions had not been tested previously.
Participant recruitment
As prior experience of a good or service can be a strong influence on choice, 31 it was decided to control for this by recruiting a sample who all had previous experience with teledermoscopy in a pilot teledermoscopy trial. 11 Participants from this trial who agreed to further contact were invited via email to complete the survey either online or on a paper–pencil copy we mailed to them. The inclusion criteria for the original study included: age 50–64 years; living in Queensland (residing in Brisbane or willing to travel); and considered to be at moderate or high risk for melanoma (they had to meet one of the following: fair eye, hair or skin type, previous skin excisions, or a personal or family history of melanoma). There were no exclusion criteria. 11
Analysis of the choice data
The DCE data were analysed using non-linear regression models, in which the attribute levels (independent variables) were used to explain participant screening choice (dependent variable). Here, we use a mixed-logit multinomial model (MMNL), which offers advantages to the more often used MNL.32,33 Details of the theoretical framework are provided in Appendix 1. A number of assumptions were made. First, as no participants chose the ‘I would prefer not to undertake any screening’ opt out option, this alternative was not included in the model. Instead, SSE was assumed to be the base alternative, the utility of which was assumed to be constant. All attributes except cost were effects coded to allow for non-linear relationships, especially for qualitative (unordered) variables. Costs were coded continuously. All attributes were initially modelled as random parameters. All distributions around random parameters were assumed to be normal and estimated using 1000 halton draws (this is a quasi-random or ‘intelligent’ method for simulating distributions often used in choice modelling 33 ). If the standard deviations of the random parameters were found to be non-significant at the 10% level, the attribute was modelled as a non-random parameter. Similar to linear regression, a positive coefficient means that the attribute level of interest is preferred over those with a negative coefficient.
Willingness to pay estimates
Using the outputs from the model derived above, an indicative marginal willingness to pay (MWTP) was estimated for the hypothetical policy change from a situation where only SSE, GP screening and skin cancer clinic screening is available, to a situation where the option of teledermoscopy with review by a dermatologist is additionally available. This provides an estimate of the monetary value for the welfare gain (or loss) for consumers by having teledermoscopy available to them as an alternative. Following Lancsar and Savage 34 and Ryan, 35 we use the method for compensating variation described by Small and Rosen 36 to estimate the welfare gain associated with different screening models. Compensating variation is a measure of how much money needs to be given or taken from a consumer after a price or quality change to leave them with their initial level of satisfaction (known as utility). 34 This method accounts for both the relative importance of a given attribute level(s) as well as the probability of choosing an alternative which contains that particular attribute. The Small and Rosen formula used is detailed in Appendix 2.
Results
Summary of participant characteristics.
All five skin cancers detected in the previous teledermoscopy trial were basal cell or squamous cell carcinomas. No melanomas were detected in the trial. 11
As can be seen from Table 3, participants were between 50 and 64 years, representing the age group with the highest incidence of melanoma (this was an inclusion criteria of the original trial). On average the participants had high levels of education and income and were predominantly residing in major cities.
There were no missing choice data. Approximately half of the sample completed each survey version (version 1, 16/35 or 46%; version 2, 19/35 or 54%). No respondents opted for the alternative ‘no screening’, indicating that early detection of melanoma was of high relevance to all participants. The ‘no screening’ option was therefore removed from the models. The SSE alternative was chosen on 36 separate occasions (choice sets) by six different participants. This is a small proportion (36/420; 8.6%) of the total possible 420 choice sets across all 35 participants, suggesting a strong preference in this cohort to take up a form of screening that involves a medical doctor.
Result of mixed multinomial logit model.
,**,* indicate significance at the 1%, 5% and 10% level respectively.
MMNL: mixed multinomial logit; CI: confidence interval; RP: random parameter; dist.: distribution; SE: standard error; SD: standard deviation
Participants strongly preferred their screening results to be reviewed by a dermatologist rather than a GP; considered there to be a high chance (>95%) of detection during screening; considered that fewer non-cancerous moles (three) are removed to detect one skin cancer, rather than more (10); and that there would be lower cost (all p-values < 0.05). Not spending greater than four hours away from usual activities was preferred, but this was only significant at the 10% level. No significant difference was found between the different screening methods in terms of their effect on screening choice (teledermoscopy compared with visiting a skin cancer clinic or a GP). However, teledermoscopy results were the only results reviewed by a dermatologist in the available choices and participants had been made aware of this in the survey. There was a non-linear preference observed for the levels of the attribute ‘length of time to receive results’, with a wait time of up to one day preferred to less than four hours; however, this relationship was only significant at the 10% level. Participating in a screening service involving a doctor’s opinion either during a face-to-face visit or by telediagnosis was strongly preferred over SSE as shown by the negative and statistically significant constant for this alternative. We tested a range of additional variables to investigate whether they explained the variation in participant choices for screening. These included the participant socio-demographics shown in Table 3, whether participants had previously been diagnosed with skin cancer, if they had a skin cancer detected during the trial, if they were more worried about developing skin cancer in the future and if they currently performed SSE. None of these variables were shown to significantly explain heterogeneity around choices and they were therefore excluded from the final model.
Willingness-to-pay for teledermoscopy to be available as an additional alternative.
The ‘initial state of the world’ includes three available alternatives: SSE, skin cancer clinic screening and GP screening. The ‘state of world’ after policy change includes four available alternatives: SSE, skin cancer clinic screening, GP screening and teledermoscopy. The estimates for probability of uptake assume differences in three attributes: (i) the type attribute; (i) who reviews the results; and (ii) the benign attribute. The level of the benign attribute for the teledermoscopy alternative is set to the ‘three non-cancerous moles removed for one skin cancer’ level, whereas this is ‘five non-cancerous moles’ for GP and skin cancer clinic alternatives. Dermatologist review is assumed for the teledermoscopy alternative, GP review is assumed for the skin cancer and GP clinic alternatives. The best levels for all other attributes are chosen and held constant across the GP, skin cancer clinic and teledermoscopy alternatives. The Small and Rosen formula used to calculate compensating variation is shown in Appendix 2.
SSE: skin self-examination; GP: general practitioner
Here, we consider an initial ‘state of the world’ as comprising three different screening options, which were chosen as being most indicative of the current options available: SSE, skin cancer clinic or GP screening. We assume a one to two hour distraction from usual activities, greater than 95% chance of detecting a skin cancer if one is present, a wait time of less than four hours for results, that the results are reviewed by a GP, and that five non-cancerous moles are removed for every skin cancer detected for services provided in a skin cancer or GP clinic. The introduction of a teledermoscopy alternative as an additional (fourth) screening model assumes review by a dermatologist and that only three non-cancerous moles need to be removed to detect one skin cancer. The use of a lower rate of non-cancerous mole removal by dermatologists has been justified by previous findings. 15 This change, from three screening options to four screening options, is associated with an average welfare improvement of AUD110 (estimated using the Small and Rosen compensating variation equation (Appendix 2). This is interpreted as consumers being willing to pay an average of AUD110 to move the current situation where they can choose between SSE, skin cancer clinic and GP screening alternatives only, to a situation where all of these options plus teledermoscopy with dermatologist review being available. The likelihood of uptake of the different alternatives also changes (Table 5). Whilst the skin cancer clinic model is the most demanded initially (0.548 probability of uptake), the new teledermoscopy model is the most likely to be taken up after its introduction (probability of uptake 0.668). This likelihood of uptake and welfare gain is driven primarily by having review of the results by a dermatologist rather than a GP.
Given the strong preference for skin examination by a dermatologist, we conducted further hypothetical modelling assuming that dermatologists would staff skin cancer clinics and achieve a lower rate of mole removal (three for every skin cancers detected). In this scenario, the welfare improvement associated with the teledermoscopy (fourth) option is estimated to be AUD52 (results not presented). Thus, teledermoscopy still provides additional welfare gains to consumers beyond access to a dermatologist and a lower rate of mole removal.
Discussion
This study uses discrete choice methods to elicit consumer preferences for skin cancer screening services, including novel teledermoscopy screening. This new technology offers consumers the chance to ‘store and send’ an image of any potential skin cancer for review by a health professional. The findings suggest that people aged 50–64 years at high risk of skin cancer strongly preferred their results to be reviewed by a dermatologist – an option that was only available using teledermoscopy in the survey, and would likely only be available via teledermoscopy for the majority of consumers in the Australian health system, where family doctors must be consulted first before any specialist appointment. Results also show that a higher skin cancer detection rate and lower rate of removal of non-cancerous lesions are also strongly preferred, as is a shorter time away from usual activities. There was an unexpected non-linear preference observed for levels of the attribute length of time to receive results. However, this was not significant at the conventional 5% level. Furthermore, any such trend could indicate that people prefer a fast, but also thorough, assessment of their skin lesions.
Many of the results found supporting the importance of particular attributes in this study are in concordance with the previous literature focusing on consumer satisfaction with teledermoscopy or teledermatology more broadly.14,18,19,27,28 For example, Whited et al. found that the majority of consumers agreed that they had confidence that dermatologists can diagnose teledermatology pictures and that a teledermatology consult is more convenient than going to a dermatologist clinic. 22 Qualitative and quantitative results from Collins et al. show no difference in the overall satisfaction depending on the type of consultation (face to face compared with teledermatology) 17 and that both groups were happy with their care. This is despite concerns by some consumers that they would like more personal or face-to-face communication. In relation to waiting time, Azfar et al. 16 asked specifically ‘how many days would you be willing to wait to get a response from mobile teledermatology in exchange for the convenience of not having to travel so far or wait for a face-to-face consultation?’ Interestingly, the majority (40%) of participants answered one to three days compared with 19% who wanted the answer on the same day. This is similar in some ways to the wait attribute results found in this study where the level ‘up to one day’ was preferred over both ‘less than four hours’ and ‘up to three days’ (although, this only reached significance at the 10% level). It is uncertain why participants may prefer not to receive results straight away, but this may reflect people’s assumptions about how long it will take for results to be processed. Future research should explore this question in more detail.
The use of a DCE methodology here is an advance on the existing literature which relied on the use of satisfaction rating scales, which do not allow for the relative importance of different aspects of the service to be directly compared. 37 This is important when designing how teledermoscopy might be incorporated into the current health system as it helps to inform decisions about who reviews the results, in what time frame and the acceptance of any out-of-pocket costs. The indicative welfare gain of AUD110 to have teledermoscopy available as an additional screening option strongly suggests this screening mode could provide a societal welfare gain, even after considering the comparative costs of providing this service. Nevertheless, given the limitations associated with the sample (discussed below), this estimate requires confirmation in larger representative samples before being implemented. The results however indicate that participants found the option of teledermoscopy valuable and would be likely to take it up if it was available, all other things being equal. This estimate is also sensitive to the assumption that skin cancer clinics are staffed by GPs rather than dermatologists and the estimate drops to AUD52 if this assumption is not met in practice. Given the relative shortage of dermatologists, we think the implementation of teledermoscopy is more feasible than skin cancer clinics being predominantly staffed by dermatologists, at least in the short to medium term.
In terms of limitations, this study used a small homogenous sample of participants at moderate to high risk of melanoma, all with experience of using teledermoscopy in a previous component of this study. 11 As such, the generalizability of these results is limited. Nevertheless, the preferences elicited in this DCE are indicative of the likely preferences of a group of consumers who are at moderate to high risk of skin cancer and who are therefore a population of particular interest for targeted screening strategies. Further, due to the particular nature of the Australian health service, where GP visits are publicly funded and access to a dermatologist requires a GP referral, results cannot inform policy makers in other countries without further confirmation. However, this study outlines an approach to undertaking a similar study in a larger international sample, which would be advisable to inform implementation of teledermoscopy on a global scale.
Overall, we found that many of the aspects of skin cancer screening and skin cancer triage offered by teledermoscopy are valued by consumers at moderate to high risk of skin cancer, which is an important consideration for health service design. Teledermoscopy is an innovative technology which potentially offers at least equivalent health outcomes for consumers, delivered in a more acceptable way, and may offer improved health service efficiency.
Footnotes
Acknowledgements
The authors would like to acknowledge the contribution of Professor David Whiteman to this project as well as Mr Riyaz Shakya for his early assistance with survey development.
Declaration of Conflicting Interests
The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: HPS is a shareholder of e-derm consult GmbH and MoleMap by Dermatologists Ltd Pty. He provides teledermatological reports regularly for both companies. MJ was funded by a National Health and Medical Research Council (NHMRC) career development award (no. 1045247). JS and JW have nothing to declare.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the Centre of Research Excellence in Telehealth funded by NHMRC (grant ID: APP1061183) and an internal grant from Menzies Health Institute Queensland, Griffith University.