Moderators of the Effects of mySmartSkin, a Web-Based Intervention to Promote Skin Self-examination and Sun Protection Among Individuals Diagnosed With Melanoma

Abstract

Background

Identifying the characteristics of persons who benefit more from behavioral interventions can help health care providers decide which individuals should be offered particular interventions because this is the subgroup of persons who are more likely to derive greater benefit from the intervention and refine the underlying constructs of the model guiding the intervention.

Purpose

This study evaluated possible demographic, medical, knowledge and attitudinal, and psychosocial variables that may moderate the impact of an online intervention, called mySmartSkin (MSS), on engagement in skin self-examination (SSE) and sun protection behaviors among melanoma survivors.

Methods

Participants completed a baseline survey and were then randomized to the MSS condition or usual care. Follow-up surveys were completed by participants at 8-, 24-, and 48-week postrandomization.

Results

A greater impact of MSS on SSE was illustrated among participants with more phenotypic skin cancer risk factors and participants reporting lower baseline self-efficacy in conducting SSE. A more favorable response of MSS on sun protection behaviors was shown when initial knowledge about abnormal lesions and sun protection barriers were high. Greater use of MSS and more favorable evaluations of it were also associated with higher intervention response.

Conclusions

Future studies seeking to improve SSE and sun protection among melanoma survivors might benefit from focusing on survivors who report more skin cancer risk factors, lower self-efficacy in conducting SSE, less knowledge about what abnormal skin lesions look like, more perceived barriers to sun protection behaviors, and less worry about recurrence and cancer-related distress.

Melanoma survivors who had more skin cancer risk factors and lower self-efficacy in conducting skin self-exams benefited more from an online intervention to improve skin self-exams.

Introduction

Melanoma survivors are at elevated risk for disease recurrence, second primary melanomas, and keratinocyte carcinomas [1–3]. Therefore, regular total cutaneous exams (TCEs) and regular skin self-examination (SSE) are recommended [4, 5]. The vast majority of melanoma survivors report conducting a TCE. [6–8]. However, rates of regular and thorough SSE are lower, with reported rates ranging from 7% to 17% [6–10]. More than half of all recurrences and new primary melanomas are detected by survivors themselves [11, 12]. This provides a strong rationale for performance of regular and thorough SSE to enhance the likelihood of early detection by identifying suspicious lesions during the time period between TCEs. In addition to TCE and SSE, professional agencies recommend engagement in regular sun protection behaviors, such as staying in the shade, applying sunscreen with a sun protection factor of at least 30, and protective clothing (e.g., hats, long sleeves). However, engagement in regular sun protection behaviors among melanoma survivors is lower than recommended levels [6–10].

In a previous publication, we evaluated an online behavioral intervention, called mySmartSkin (MSS), the goal of which was to improve SSE and sun protection behaviors among melanoma survivors [13, 14]. MSS is a behaviorally based program which is delivered via the Internet, tailored to the user, and fully automated with no human clinical support. We compared MSS with usual care (UC). Results suggested that MSS had a significant and beneficial impact on SSE at the primary time point, 24 weeks. Long-term impact was similarly strong at the 48-week follow-up, with nearly 31% of participants enrolled in MSS reporting conducting a thorough SSE in the last 2 months, compared with 9%–13% of participants in the UC condition. The impact on sun protection was less strong. At the 24-week follow-up, participants enrolled in MSS reported engaging in significantly higher levels of sun protection behavior relative to UC, but there were not significant differences at the long-term 48-week follow-up.

Identifying the characteristics of persons who benefit more from behavioral interventions can help health care providers decide which individuals should be offered particular interventions because this is the subgroup of persons who are more likely to derive greater benefit from the intervention. From a theoretical perspective, identifying characteristics of a treatment response can refine constructs of the model guiding the intervention. The goal of this second publication from the larger intervention trial was to evaluate moderators for MSS effects on SSE and sun protection behaviors.

Our selection of potential moderators was based on the conceptual framework guiding MSS, the Preventive Health Model [15], as well as prior work evaluating factors associated with skin cancer surveillance and sun protection behaviors among melanoma survivors [6, 8–10, 16], and prior intervention studies that have identified moderators for interventions that improve SSE or sun protection among melanoma survivors (e.g., skin cancer risk) [17, 18]. We selected moderators representing four types of putative moderators: demographic, medical, knowledge and attitudes, and psychosocial factors.

Overall, we propose that melanoma survivors with fewer resources and less favorable attitudes about SSE and sun protection would benefit from the intervention. We also selected known correlates of lower engagement in SSE and/or less sun protection. In terms of demographic factors, we evaluated sex, education, marital status, and age. We predicted that MSS would have stronger effects among male (prior literature) [9, 19, 20], less educated (low resource; prior literature) [6, 21], unmarried (low resource), and older survivors (prior literature) [19]. With regard to medical factors, we predicted that MSS would have stronger impact among participants with a longer time since the melanoma surgery, a lower disease stage, and more phenotypic skin cancer risk factors (e.g., blonde hair, fair skin) (prior research) [10]. With regard to knowledge and attitudes, we proposed that MSS would be more effective for participants with lower baseline knowledge about melanoma and suspicious lesions (low resource, prior literature) [6, 19], more barriers to SSE/sun protection behaviors (low resource, negative attitude, prior literature) [6, 8], lower SSE/sun protection self-efficacy (low resource, prior literature) [6, 10, 19, 22], and lower perceived risk (negative attitude, prior literature) [19]. Finally, we proposed that survivors who experience more distress and worry about melanoma (low resource) will benefit more from the intervention. Our moderators were assessed at baseline. Our rationale was that survivors who had less resources when they start engaging with the intervention should be the individuals most strongly impacted by the intervention. For example, if an individual reports low levels of knowledge about what an abnormal lesion looks like prior to the intervention, the information provided by this intervention should be especially impactful on the individual’s subsequent engagement in SSE.

Methods

Intervention development and core content, study eligibility criteria, recruitment methods, randomization procedures, measures, and sample size considerations are described in detail elsewhere [13]. Briefly, MSS consisted of an orientation section, three cores, and a body mole map. Core 1 outlined the goals of the intervention, provided information about melanoma and risks of recurrence, skin cancer risk factors, how to do a SSE, and an overview of sun protection behaviors. Core 2 assessed prior experience with SSE, benefits and barriers to SSE, confidence in conducting skin self-check, and strategies for doing a skin self-check. An online body mole map was provided to record and track moles and other skin growths over time. Core 3 assessed participants’ current sun safe behaviors, guided them to set sun safety goals, and provided a sun safety action plan. There was also a section of the program where users can access printable documents from each core section and a summary of the most recent SSE and sun safe action plan. Tailored activities included selecting reasons why conducting SSE (and engaging in sun protection) is important to the participant, assessing barriers to engaging in SSE (and engaging in sun protection), and completing action plans for SSE and sun protection. Participants could also log in to use the program to help them complete their monthly SSE. Participants assigned to the UC condition received no additional intervention aside from their usual nonstudy clinical care, which consisted of scheduled total cutaneous skin exams by their physician.

Individuals diagnosed with stage 0–III melanoma who were 3–24 months postsurgery, had not completed a thorough SSE in the past 2 months and/or were not adherent to sun protection recommendations, (i.e., mean score <4 [which corresponds to “often”] on a 5-point scale [from “never” to “always”] that assesses the frequency of engaging in four sun protection behaviors [23]); were >18 years of age, able to speak and read English, and had access to a computer connected to the internet, participated in the study. Participants were recruited from several medical centers in New Jersey as well as via the New Jersey State Cancer Registry. Randomization was stratified by disease stage and the number of months since surgery and was implemented by the study’s biostatistician. After randomization, participants either received access to the MSS program or received no additional intervention. Follow-up surveys were emailed to participants at 8-, 24-, and 48-week postrandomization.

Measures

Primary Outcome (Baseline, 8-, 24-, and 48-Week Follow-ups)

Performance of thorough SSE

Participants reported if they had checked any part of their body for signs of skin cancer in the last 12 months (baseline survey), 2 months (for the 8-week survey), 4 months (for the 24-week survey), and 6 months (for the 48-week survey) [13, 20, 24]. Participants who reported checking their skin indicated how many times they had done so in the time frame for that assessment time point. Performance of thorough SSE was defined as examining each area of the body during the most recent SSE (yes/no).

Sun protection behaviors

Participants rated how often they engaged in four behaviors when outside on a sunny day: wearing sunscreen with an sun protection factor (SPF) >30, wearing a long-sleeved shirt, wearing a wide-brimmed hat, and staying in the shade [23].

Demographic Moderator Measures

Participants reported sex, age, income, and education level (4-year college degree or more versus some college education or less), and marital status (married or living with partner, single, divorced, or widowed), which was recoded to married/cohabitating versus others for analyses.

Medical Moderator Measures

Time since surgery was calculated from date of surgery in the medical chart and stage of disease was collected from medical chart. The number of phenotypic skin cancer risk factors was assessed with eight items (eye color, hair color, skin tone, skin reactivity to sun, freckles, moles, use of tanning bed, and melanoma family history). Scores ranged from 0 to 8.

Knowledge and Attitudes Measures

Melanoma knowledge was assessed using 13 true–false items [8] (e.g., “Melanoma is the most common form of skin cancer”). The number of correct answers was calculated.

Knowledge about characteristics of abnormal lesions (ABCDEF). Six multiple choice items assessed knowledge of the ABCDEFs of SSE [6, 25]. The ABCDEFs refer to characteristics of abnormal lesions: asymmetrical, irregular border, inconsistent color, large diameter, evolving or changing, and funny looking. The score reflected the number of correct answers.

Barriers to SSE [8, 26] were measured using an 11-item Likert scale from 1 = strongly disagree to 5 = strongly agree). (e.g., “Doing skin self-examination would be very embarrassing”), a = .81.

SSE self-efficacy was assessed using a 12-item Likert scale from 1 = strongly disagree to 5 = strongly agree, alpha = .94. Items [6] assessed confidence in examining different parts of the body for signs of skin cancer and confidence in telling the difference between a normal mole or skin growth and a melanoma.

Barriers for sun protection behaviors were measured by the mean of 23 Likert items rated from 1 = strongly disagree to 5 = strongly agree (e.g., “For me, using sunscreen with a SPF of 30 or more when I am outside on a sunny day is inconvenient”), α = 0.903 [27].

Self-efficacy for sun protection [28] was assessed through the mean of 12 Likert items from 1 = Not at all confident to 5 = Very confident (e.g., “How confident are you that when you are outside on a sunny day, you can wear sunscreen with a SPF of 30 or more?), a = 0.938 [28].

Perceived risk for melanoma recurrence was assessed using a four-item Likert scale from 1= strongly agree to 5 = strongly agree (e.g., “I feel I will experience melanoma again”), a = 0.81 [8].

Psychological Factors

Distress about melanoma was measured using a single Likert scale from 1 = not at all distressed to 10 = extremely distressed (“Select the number that describes how distressed you are currently about your melanoma”) [8].

Worry about melanoma recurrence was assessed using a four-item Likert scale from 1= never to 6 = all the time) adapted from Vickberg et al. [29] (e.g., “How often do you worry about the possibility you could have melanoma again?”), a = 0.92.

Measures Completed by MSS Participants Only

Internet experience

Six items assessed comfort using the internet (1 = very uncomfortable, 5= very comfortable), emails sent and received daily (1 = none, 7 = more than 200), frequency of internet use (1 = less than every few weeks, 7 = at least 4 times a day), whether participant has a cell or mobile phone (yes/no), whether participant has a “smart” phone (1 = yes, 2 = no, 3 = not sure), and average number of text messages sent per day (1= none, 7 = more than 200), a = 0.67.

Evaluation of MSS

At the 8-week follow-up, participants enrolled in the MSS intervention completed three evaluation surveys: (a) A 20-item Impact and Effectiveness measure [30, 31] assessing the degree to which MSS helped the participant learn how and be prepared to conduct SSE and engage in sun protection behaviors as well as feel in control of his/her health and feel less worried about melanoma (1 = not at all, 5 = very); (b) A 22-item intervention barriers measure [30], which evaluated technical, personal, general, and intervention specific barriers (0 = Not a problem, 1= a little problem, and 2 = a major problem); and (c) A 15-item Evaluation and Utility survey [30, 31], which assessed usefulness, ease of use, worry about privacy, ease of navigation, and other features of MSS (1 = not at all, 5 = very).

Usage of MSS

We calculated the number of cores completed, the number of SSEs done using the skin self-check program (used to predict only SSE), the number of logins to sun safe action plan (used to predict sun protection behavior only), and the number of logins not associated with any of these activities or survey completion, which indicates reviewing or returning to material without completion of additional activities.

Data Analytic Approach

Because this analysis focuses on moderation by baseline attributes, there was little missing data for these analyses. Nonetheless, as recommended by Lang and Little [32] we used multiple imputation with 50 imputed samples in our analyses to handle missing data for the outcomes at T2, T3, and T4. Moderated regression models were used to predict sun protection behavior as a function of treatment condition, the moderating variable (grand-mean centered if continuous or effect coded if categorical), and the interaction between condition and the moderating variable. All models also included sex, education level, cancer stage at diagnosis, months since surgery, and baseline sun protection as covariates. Models predicting thorough SSE performed were estimated using moderated binary logistic regression models. As with sun protection, predictors included treatment condition, the moderating variable, and the interaction between condition and the moderating variable. Covariates included sex, education, stage, months since surgery, and whether the individual had conducted a thorough SSE prior to baseline. Significant interactions were followed-up using simple slopes analyses at ±1 SD from the mean of the moderating factor.

For analyses examining the extent to which MSS efficacy varied as a function of usage levels and evaluations, a separate data set that included only participants in MSS was created and multiple imputation was used to handle missing data. Models predicting sun protection behavior used standard multiple regression in which sun protection was predicted to be a function of the evaluation/usage variable, baseline sun protection, stage, months since surgery, sex, and education. Similar models predicted thorough SSE using binary logistic regression. These models also included an evaluation/usage variable and controlled for sex, education, stage, months since surgery, and baseline thorough SSE.

Results

Participants

A more detailed description of the sample is available in the study’s outcomes publication [14]. The sample consisted of 441 participants, with 224 participants randomized to MSS condition and 217 participants randomized to the UC condition. Ninety-eight percent of participants were non-Hispanic White, 68.1% completed a college-level education or higher, 34.7% reported an annual income of $150,000 or higher, and almost 80% were married. Approximately half of the sample had melanoma surgery between 3 and 14 months prior to completion of the survey, and 86.4% were diagnosed with early-stage melanoma (0 or 1). Following established guidelines for estimating the proportion of individuals of unknown eligibility status who were in fact eligible for the study, the participant response rate was 40.9% [33]. Follow-up survey completion rates for MSS were: 88.4%, 84.3%, and 80.3% for the 8-, 24-, and 48-week follow-ups, respectively. Follow-up survey completion rates for the UC condition were 99.1%, 95.4%, and 93.1% for the 8-, 24-, and 48-week follow-ups, respectively.

In terms of sample characteristics, participants were primarily non-Hispanic white, 51% male, relatively well educated, and primarily diagnosed with early-stage melanoma. Comparisons between participants and study refusers revealed only one difference: study refusers were more likely to be male (58%) compared with study participants (51%) (chi-square = 5.48, p = .019).

Moderation of the Intervention Effect on Thorough SSE

There was no evidence that the effects of the intervention on performance of a thorough SSE were moderated by demographic variables. For the medical variables, although there was no moderation by stage or months since surgery, number of risk factors did moderate the effect of condition on SSE at 8 weeks, b = 0.257, SE = 0.110, p = .019, exp(b) = 1.293. This interaction is shown in the top panel of Fig. 1. When the number of phenotypic risk factors was high, there was a significant effect of condition on likelihood of conducting a thorough SSE, b = 1.563, SE = 0.312, p < .001, exp(b) = 4.773, such that individuals in MSS were more likely than those in UC to conduct an SSE at 8 weeks, but when the number of risk factors was low, the condition effect (though statistically significant) was considerably smaller, b = 0.722, SE = 0.222, p = .001, exp(b) = 2.058. The condition by risk factors interaction was not statistically significant for 24- or 48-week SSE, but the pattern of modestly stronger condition effects for individuals with higher risk was consistent with the 8-week findings.

Fig. 1.

Likelihood of having conducted a thorough skin self-examination at 8 weeks as a function of treatment condition and baseline risk factors (top panel), baseline worry (middle panel), and baseline distress (bottom panel).

Melanoma knowledge, knowledge of abnormal lesions, perceived risk, and SSE barriers did not moderate the effect of the intervention. There was evidence that SSE self-efficacy moderated the treatment effect at 24 weeks, b = −0.414, SE = 0.180, p = .021, exp(b) = 0.661 (see Fig. 2). Simple slopes analyses indicated that, although the condition effect was significant for both high (b = 0.462, SE = 0.175, p = .008, exp(b) = 1.588) and low (b = 1.222, SE = 0.301, p < .001, exp(b) = 3.395) baseline SSE self-efficacy, the condition effect was greater for those who began the study with lower SSE self-efficacy. The interaction effect was similar in nature at 8 weeks, but it was not statistically significant (p = .051).

Fig. 2.

Likelihood of having conducted a thorough skin self-examination at 24 weeks as a function of treatment condition and baseline self-efficacy for conducting skin self-examinations.

Worry and distress also moderated the effects of the intervention on thorough SSE, but these interactions emerged only at the 8-week follow-up. These interactions are shown in the middle and bottom panels of Fig. 1. For worry, the interaction effect was b = −0.311, SE = 0.141, p = .028, exp(b) = 0.733. Simple slopes indicated that although the condition effect was statistically significant for both high and low worry, individuals who were higher in worry at baseline showed a smaller effect of the treatment, b = 0.767, SE = 0.216, p < .001, exp(b) = 2.153, than did individuals who were lower in worry at baseline, b = 1.559, SE = 0.323, p < .001, exp(b) = 4.755. The interaction with initial cancer distress was similar, b = −0.154, SE = 0.076, p = .042, exp(b) = 0.858, with individuals higher in baseline distress showing a weaker effect of the treatment, b = 0.771, SE = 0.223, p = .001, exp(b) = 2.163, than those low in baseline distress, b = 1.463, SE = 0.293, p < .001, exp(b) = 4.319.

Moderation of the Intervention Effect on Sun Protection Behaviors

Of the four demographic variables (sex, education, married/partnered status, and age), only education moderated the effect of condition on sun protection behavior, b = −0.072, SE = 0.032, pooled t = 2.23, p = .026, and that interaction occurred only for sun protection at 24 weeks. The top panel of Fig. 3 presents the estimated marginal means for this interaction. Simple slopes analyses indicated that for those without a 4-year college degree, individuals in MSS engaged in significantly more sun protection behaviors 24 weeks after the intervention than those in UC, b = 0.185, SE = 0.053, pooled t = 3.496, p < .001, but those with a 4-year degree or more education there was no significant difference between MSS and UC (b = 0.041, SE = 0.037, pooled t = 1.121, p = .262). There was no evidence that the medical variables (stage at diagnosis, months since surgery, or risk factors) moderated the effects of the intervention on sun protection behaviors.

Fig. 3.

Top panel: sun protection behavior at 24 weeks as a function of treatment condition and education. Middle panel: sun protection behavior at 8 weeks as a function of treatment condition and baseline knowledge about abnormal lesions (low = 1 SD below average; high = 1 SD above average). Bottom panel: sun protection behavior at 48 weeks as a function of treatment condition and baseline sun protection barriers (low = 1 SD below average; high = 1 SD above average) [7].

For the knowledge and attitudes variables, although melanoma knowledge did not moderate the effects of the intervention on sun protection, baseline knowledge about abnormal lesions showed significant moderation of the treatment effect at 8 weeks after the intervention with b = −0.039, SE = 0.017, pooled t =2.313, p = .021 (see the middle panel of Fig. 3). When initial knowledge was high, the effect of condition at 8 weeks was not statistically significant (b = −0.024, SE = 0.044, pooled t = −0.543, p = .587), but when knowledge at baseline was low, the effect of condition was significant (b = 0.120, SE = 0.044, pooled t = 2.731, p = .006). Thus, when initial knowledge about abnormal lesions was low, the MSS intervention resulted in significantly higher sun protection behavior than UC at 8 weeks, but when initial knowledge was high, no such condition effect emerged. Although the interaction was not significant for sun protection at 24 or 48 weeks (ps = .052 and.107, respectively), the pattern of results was the same (i.e., significant differences for condition at low baseline knowledge but no differences at high baseline knowledge).

Baseline sun protection barriers moderated the effects of the intervention at 48 weeks, b = 0.079, SE = 0.039, pooled t = 2.02, p = .044 (see the bottom panel of Fig. 3). When initial barriers were low there was no effect of the intervention (b = 0.021, SE = 0.041, pooled t = 505, p = .614), but when initial barriers were high, individuals in MSS engaged in more sun protection than did those in UC at 48 weeks (b = 0.138, SE = 0.041, pooled t = 3.325, p = .001). The same pattern emerged at 24 weeks (a significant effect of condition for those high in initial barriers and no effect of condition for those low in initial barriers), although the barriers by condition interaction was not statistically significant (p = .094). No significant treatment interactions with baseline sun protection self-efficacy, perceived risk, worry, or distress emerged.

Effects of Evaluation and Usage on SSE and Sun Protection Behavior

Table 1 presents the results predicting whether the participant completed a thorough SSE using logistic regression in which the usage or evaluation variable, along with stage, months since surgery, sex, education, and baseline SSE performance were included as covariates. In terms of usage, individuals who completed more of the MSS cores were more likely to conduct a thorough SSE, but this effect was only significant at the 24-week follow-up. However, those who viewed MSS more were more likely to do a thorough SSE at each follow-up. Individuals who reported using the MSS mole map were also more likely to conduct thorough SSEs at each follow-up.

Table 1.

Logistic regression coefficients predicting thorough SSE at 8, 24, and 28 weeks after the intervention for MSS participants only as a function of usage and evaluation

	Week of assessment	b	SE	Exp(b)	p
Usage
Highest core completed	8	0.290	0.193	1.336	.133
	24	0.591	0.234	1.805	.012
	48	0.398	0.240	1.489	.099
Number of views of MSSa	8	0.144	0.041	1.155	.000
	24	0.199	0.046	1.220	.000
	48	0.135	0.041	1.144	.001
Number SSEs using the mole map	8	0.229	0.072	1.257	.001
	24	0.344	0.084	1.411	.000
	48	0.227	0.072	1.255	.002
Evaluation
Barriers to use	8	−1.028	0.938	0.358	.273
	24	−4.220	1.456	0.015	.004
	48	−5.637	1.708	0.004	.001
Impact/effectiveness	8	1.144	0.336	3.139	.001
	24	1.879	0.416	6.546	.000
	48	1.248	0.355	3.485	.000
Evaluation/utility	8	0.968	0.331	2.632	.004
	24	1.724	0.401	5.609	.000
	48	1.185	0.356	3.270	.001

MSS mySmartSkin; SSE skin self-examination.

^aViews of MSS in which the participant did not complete a core, survey, SSE, or sun safe action plan. All models included sex, education, stage, months since surgery, and baseline SSE as covariates. Only results for evaluation and usage variables are shown.

As shown in Table 1, individuals who found MSS to be more impactful and perceived it as more useful, easy to navigate, easy to understand, and trustworthy were more likely to conduct an SSE at each follow-up. Individuals who reported more barriers to using MSS were less likely to conduct a thorough SSE at 24 and 48 weeks after the intervention.

Table 2 presents the regression results predicting sun protection behaviors as a function of usage and evaluation variables for MSS participants at each follow-up. Note that although only the regression coefficients for the evaluation and usage variables are shown, each regression model also included cancer stage at diagnosis, months since surgery, sex, education, and baseline sun protection as covariates. Individuals who completed more MSS cores and those who viewed MSS more times (i.e., views during which the participant did not complete a core, survey, SSE, or sun safe action plan) engaged in more sun protection behaviors. Individuals who made more sun safe action plans did not engage in more sun protection at any follow-ups. In terms of treatment evaluation, individuals who reported that MSS was more impactful and perceived it as more useful, easy to navigate, easy to understand, and trustworthy were more likely to engage in sun protection at each follow-up. However, an impact of perceived barriers to use on sun protection behaviors was only seen at the 8-week follow-up.

Table 2.

Regression coefficients predicting sun protection behavior at 8, 24, and 48 weeks after the intervention as a function of usage and evaluation for MSS participants only

	Week of assessment	b	SE	Pooled t	p
Usage
Highest core completed	8	0.133	0.050	2.658	.008
	24	0.099	0.044	2.262	.024
	48	0.112	0.046	2.422	.016
Number of views of MSSa	8	0.041	0.011	3.832	.000
	24	0.028	0.010	2.820	.005
	48	0.033	0.010	3.211	.001
Number of sun safe action plans done	8	0.081	0.049	1.666	.096
	24	0.063	0.044	1.432	.152
	48	0.083	0.044	1.880	.060
Intervention evaluation
Barriers to use	8	−0.510	0.254	−2.009	.045
	24	−0.313	0.254	−1.234	.218
	48	−0.430	0.244	−1.760	.079
Impact and effectiveness	8	0.192	0.067	2.867	.004
	24	0.172	0.063	2.732	.006
	48	0.180	0.063	2.846	.005
Evaluation and utility	8	0.179	0.074	2.425	.016
	24	0.196	0.069	2.852	.005
	48	0.195	0.069	2.830	.005

MSS mySmartSkin; SSE skin self-examination.

^aViews of MSS in which the participant did not complete a core, survey, SSE, or sun safe action plan. All models included sex, education, stage, months since surgery, and baseline sun protection as covariates. Only results for evaluation and usage variables are shown.

Discussion

Understanding the characteristics of melanoma survivors who benefit from a behavioral intervention designed to enhance their skin cancer surveillance and sun protection prevention practices would help target that intervention to those who could benefit most and refine theoretical models that frame the intervention content. In this study, we examined the role of demographic, medical, knowledge and attitudes, and psychological factors in the effects of MSS on SSE and sun protection behaviors that were reported in the initial outcomes study [14]. We hypothesized that melanoma survivors with fewer resources and less favorable attitudes about surveillance and prevention would show stronger intervention effects. Overall, our findings were relatively consistent with this prediction. For SSE, a more favorable response to MSS was shown among survivors with more phenotypic skin cancer risk factors (only at the 8-week follow-up) and among those reporting lower baseline self-efficacy in conducting SSE (at the 8- and 24-week follow-ups). For sun protection, a more favorable response of MSS was shown among those with lower education (at 24-week follow-up), when initial knowledge about abnormal lesions was low (at the 8-week follow-up) and when initial sun protection barriers were high (at the 8- and 24-week follow-ups). Moderator effects were most consistent at the first follow-up, which might indicate that these variables play a less important role in the long-term effects of MSS.

It is interesting to note that psychological factors played a different moderating role than predicted. MSS participants who reported more worry about recurrence and higher distress about melanoma derived less benefit from MSS on their performance of thorough SSE at the first follow-up. Greater worry about recurrence has been associated with higher levels of thorough SSE in some studies focusing on melanoma survivors [8, 9], but not in others [8, 14, 34]. In order to evaluate potential explanations, in a post hoc examination, we compared participants reporting high versus low baseline distress and high versus low versus cancer worry using median splits. These comparisons suggest that participants reporting high distress were more likely to be female, significantly younger, possess fewer phenotypic risk factors, more knowledge about what a suspicious lesion looked like, higher perceived risk for recurrence, more SSE barriers, and fewer SSE benefits. There were no differences in their usage or evaluations of the MSS intervention. The pattern of associations was similar for baseline worry, with females, younger survivors, survivors reporting more phenotypic risk factors, higher perceived risk for recurrence, and more SSE self-efficacy reporting significantly more baseline worry. Among the evaluation and usage variables, MSS participants reporting more worry also reported more personal barriers to using MSS. Although these are only post hoc comparisons and firm conclusions cannot be drawn, it is possible that distressed survivors may require more attention to psychological factors such as distress and worry about their disease interfering with engagement in SSE.

Greater utilization of MSS was consistently associated with better outcomes for both SSE and sun protection behaviors, which is consistent with a large body of literature indicating engagement in online interventions predicts a stronger impact for a variety of behavioral and psychological outcomes [35–41]. The only exception to this pattern of results was the sun safe action plan. Participants who made a sun safe action plan did not engage in more sun protection behavior. However, this was likely due to the fact that only 12% of MSS participants opened the sun safety action plan module, suggesting low engagement. Taken together with the less consistent and weaker magnitude effects of MSS on sun protection outcomes reported in the parent intervention trial and the lower engagement in Core 3 [14], these findings suggest that explicating the strategies to effectively engage melanoma survivors in sun safety goal setting, planning, and making sun protection behavior changes will require more attention. Finally, evaluations of MSS’ ease of navigation, usefulness, ease of understanding, trust in the content, interesting content, and perceived impact on SSE and sun safety behaviors were associated with better SSE and sun protection outcomes. Our findings are consistent with prior research suggesting that intervention satisfaction and acceptability are associated with better outcomes for online interventions [36, 42–44].

Despite the support for a moderating role of some baseline variables, MSS was equally effective in increasing SSE across different age groups, income levels, sexes, marital statuses, disease stages, time since surgery, knowledge levels, levels of perceived risk, and levels of perceived barriers. Similarly, MSS was equally effective in increasing sun protection across different sexes, education levels, marital statuses, disease stages, time since surgery, levels of phenotypic risk, melanoma knowledge levels, and levels of self-efficacy, perceived risk, melanoma distress, and worry about recurrence. These findings suggest that MSS has broad applicability for melanoma survivors. It is not clear why some demographic, medical, and psychosocial resources made a difference, whereas others did not. Future research into why certain subgroups of participants benefit and others do not would be beneficial.

Before closing, it is important to discuss the strengths and limitations of this study. The sample size of melanoma survivors was large, and the assessments followed participants for almost a year after baseline, providing a test of the durability of our moderator effects on outcomes. The follow-up survey completion rate was high, with the intervention considered highly trusted, useful, well-liked, interesting, easy to understand, easy to navigate, and easy to use, and the online delivery platform was considered appropriate by participants. Limitations include the self-reported SSE and sun protection outcomes, which are standard for these constructs but not without potential bias. The cancer distress measure consisted of a single item, and future work could use a more comprehensive measure. The sample was primarily non-Hispanic white, which is similar to the demographics of melanoma survivors, but not representative of the U.S. population.

Despite these limitations, our findings provide preliminary evidence that survivors who may benefit most from MSS report more skin cancer risk factors, lower self-efficacy in conducting SSE, less knowledge about what abnormal skin lesions look like, more perceived barriers to sun protection behaviors, and less worry about recurrence and cancer-related distress. However, because many proposed variables did not moderate treatment effects, the majority of the moderator effects were at the first follow-up and not consistent across the postintervention period, these findings should be considered preliminary. Since the primary outcomes were the 24-week follow-up assessment, the short-term moderator effects at 8 and 48 weeks should be interpreted with caution.

To improve clinical care in real world settings, potential strategies could include brief screening questions about confidence in performing SSE and phenotypic risk factors completed at the time of treatment completion or at follow-up appointments, followed by the provision of intensive education about how to perform SSE, what to look for during self-checks, and reminders about the importance for survivors with these risk factors. Providers wishing to improve sun protection could assess known barriers to sun protection as well as knowledge about what skin cancer looks like and provide information about suspicious lesions and barriers counseling. Since usage of the intervention was a consistent predictor of engagement in thorough SSE and better sun protection behavior, further iterations of MSS could include identification of ways to improve engagement such as reinforcements and incentives for completing MSS modules and greater emphasis on sun safety planning earlier in the intervention content. Gathering iterative feedback from survivors when redesigning MSS may foster greater engagement and ultimately improve its impact.

Funding

This research was supported by National Cancer Institute (National Institutes of Health) grant R01CA171666 awarded to Elliot J. Coups, PhD and to Sharon L. Manne, PhD, and by the Biometrics Shared Resource (NCI-CCSG P30CA072720-5918) and the Population Science Research Support Shared Resource at Rutgers Cancer Institute of New Jersey. This research was facilitated by the New Jersey State Cancer Registry, Cancer Epidemiology Services, New Jersey Department of Health, which is funded by the Surveillance, Epidemiology and End Results (SEER) Program of the National Cancer Institute under contract HHSN261201300021I and control no. N01-PC-2013-00021, the National Program of Cancer Registries (NPCR), Centers for Disease Control and Prevention under grant NU5U58DP006279-02-00 as well as the State of New Jersey and the Rutgers Cancer Institute of New Jersey.

Compliance with Ethical Standards

Authors’ Statement of Conflict of Interest and Adherence to Ethical Standards Authors Sharon Manne, Carolyn J. Heckman, Deborah Kashy, Lee Ritterband, Frances Thorndike, Carolina Lozada, and Elliot J. Coups declare that they have no conflict of interest.

Authors’ Contributions Deborah Kashy: data analysis, data interpretation, study write up. Carolyn J. Heckman: data interpretation, study write up. Elliot J. Coups: methods, measures, intervention design, data collection. Sharon Manne: methods, measures, data interpretation, study write up. Lee Ritterband: Intervention design, data collection, study write up. Frances Thorndike: Intervention design, data collection. Carolina Lozada: data collection, intervention design.