Short and Long-term Barriers and Facilitators of SSE Among Individuals with Melanoma

Background Melanoma can be lethal if not detected early and treated. Early detection can be facilitated via skin self-examination (SSE) and as such, SSE is part of melanoma follow-up care for individuals with a prior history, who face a life-long risk of reoccurrence. The objective of the current study was to identify short- and long-term predictors of SSE among melanoma survivors to inform future prevention interventions in high-risk groups. Method This is an observational study with longitudinal assessments conducted with adult melanoma patients in active follow-up care. Primary Outcome Measures: Behavioral outcomes, comprehensive SSE (checking up to 5 body areas in the last 3 months) and optimal SSE (checking the entire body at least monthly in the last 3 months) and were assessed at 3, 12, and 24 months post a dermatological educational session on skin cancer prevention. T tests and chi square analyses were used to examine changes in outcomes from 3 to 12 and 24 months. Linear and logistic regression models were used to examine the association between predictors and the primary outcomes. Results Comprehensive SSE did not decrease significantly from 3 (M=2.7, SD=1.1) to 12 (M=2.6, SD=1.2) and 24 months (M=2.4, SD=1.2) post the education session, with the stronger predictor at all timepoints being intentions to perform SSE. Optimal SSE was higher at 3 months (59%) compared to 12 (46%) and 24 months (34%) post, with key predictors including self-efficacy and intentions to perform SSE and male sex at 3 months post; self-efficacy and reliance on medical advice at 12 months post; and (lower) education and self-efficacy at 24 months post. Conclusions The key findings of this study are that 1) survivors maintain SSE behaviour over time, but rates of SSE performed in agreement with medical recommendations are higher immediately post standard dermatological education (i.e. usual care) and decrease somewhat over a 24-month period; and 2) the strongest psycho-social predictors of SSE

3 are intentions and self-efficacy to perform the behavior, which are highly modifiable, for example via motivational interviewing and goal setting health interventions.
Melanoma survivors have a 9-fold increased risk to develop subsequent melanomas compared to the general population (range= 12.6 to 26.4-fold) and the risk remains elevated 20 years past the initial diagnosis (Bradford, Freedman, Goldstein, & Tucker, 2010;van der Leest et al., 2012). A study from Alberta, Canada, found that melanoma survivors have a 60% increased risk to develop a second primary melanoma compared to 4 people without a prior melanoma (Jung, Dover, & Salopek, 2014).
There is consensus within the clinical and scientific communities that 1) intervention strategies designed to reduce melanoma-related mortality must focus on early diagnosis of pre-metastatic tumours (Geller, Swetter, & Weinstock, 2015;Rhodes, 2006); and that 2) intervention strategies will have the highest impact if targeting high-risk individuals (Garbe et al., 2003;Geller et al., 2007;Markovic et al., 2007;Rhodes, 2006;Weinstock, 2006). Because melanoma develops with a visible, pre-clinical phase, it is amenable to early detection via visual inspection of the skin, by physicians and lay persons (Friedman, Rigel, & Kopf, 1985). Evidence-based clinical guidelines for melanoma follow-up and care developed in the United Kingdom recommend that melanoma patients should be a) given a full body skin exam (with palpation of the lymph nodes) at every follow-up appointment, b) given information (written and verbal) about types of skin cancer and instructions on skin self-examination (National Institute for Health and Care Excellence [NICE], 2015).
While clinical exams are undoubtedly useful for the early identification of cancerous skin lesions, most dermatological and cancer associations (American Academy of Dermatology, 2018; American Cancer Society, 2015; National Comprehensive Cancer Network, 2018) as well as most clinical guidelines for the prevention of melanoma (Marciano, Merlin, Bessen, & Street, 2014;Watts et al., 2015) recommend that individuals at increased risk perform regular skin self-exams (SSE) in between medical follow-ups and present for medical skin exam if suspicious lesions are identified during skin self-exams.

Skin Self-Examination (SSE)
There is evidence that the practice of SSE is beneficial for high-risk individuals. Empirical cross-sectional studies have found that patients and family members detected up to 50-80% of all melanomas (Carli, De Giorgi, Betti, et al., 2003;De Giorgi et al., 2012;Swetter et al., 2012). Also, increased thoroughness (or extent of skin covered) of the skin selfexam was associated with thinner lesions (Pollitt et al., 2009), and patients who examined at least some parts of their body had thinner lesions at diagnosis compared to those who did not examine their skin (Carli, De Giorgi, Betti, et al., 2003;Pollitt et al., 2009). A case control study (423 melanoma patients and 678 matched controls) found that individuals who conducted SSE were twice as likely to self-detect melanoma and less likely to have thick (advanced) tumours at diagnosis compared to those who did not do SSE (Titus et al., 2013). Finally, in a study with 1,062 melanoma patients (stages I -II), among those who experienced a melanoma recurrence (19%), most recurrences were self-detected (55%) and led directly to seeking early medical advice (Dalal et al., 2008). Self-detection, not physician detection, independently predicted survival in this study.
Even though cutaneous melanomas are readily visible on the skin surface and it is wellknown that SSEs are related to better prognosis Geller et al., 2004;Koh, 1992;Robinson, Fisher, & Turrisi, 2002;Terushkin & Halpern, 2009), most melanoma survivors do not perform whole-body skin exams regularly (Berwick, Begg, Fine, Roush, & Barnhill, 1996;Carli, De Giorgi, Palli, et al., 2003;Manne & Lessin, 2006;Mujumdar et al., 2009). Reported SSE rates among melanoma survivors vary by the modality of asking about the exam and timeframe of assessment. For example, a large cohort study conducted in Australia (n=1433 confirmed melanoma cases) found that 57.4 % of participants had performed an SSE in the past 3 years (Olsen et al., 2015). A crosssectional study found that among 316 melanoma survivors, 28% reported having ever engaged in SSE, 16% reported doing monthly SSE, and 8% reported doing SSE every 2 6 months (Glenn, Chen, Chang, Lin, & Bastani, 2016). Another cross-sectional study (n=321 melanoma patients) found that 15% of individuals performed skin checks by examining their moles every 1-2 months, 18% checked their skin every 6 months, and 17 % checked their skin once a year (Pollitt et al., 2009). Coups and colleagues (Coups, Manne, Stapleton, Tatum, & Goydos, 2016) found that 72% of 176 melanoma patients had performed SSE during the last 2 months, but only 14% had examined their whole body.

Predictors of SSE
Personal characteristics that have been associated with SSE include a personal or family history of skin cancer, including melanoma Girgis, Campbell, Redman, & Sanson-Fisher, 1991;Glenn et al., 2016;Robinson et al., 2002), being female, and having a higher level of education (Carli, De Giorgi, Palli, et al., 2003;Glenn et al., 2016;Manne & Lessin, 2006;Olsen et al., 2015;Robinson et al., 2002). Unlike medical and demographic factors linked to SSE, which are not generally amenable to intervention, psychosocial and educational factors associated with SSE are potential targets of interventions that aim to improve adherence to SSE instructions. Psychosocial factors associated with SSE behaviours in melanoma survivors and other high-risk individuals include greater knowledge about melanoma and SSE (DiFronzo, Wanek, & Morton, 2001;Friedman, Rigel, Silverman, Kopf, & Vossaert, 1991;Robinson et al., 2002), higher perceived susceptibility to melanoma (Azzarello, Dessureault, & Jacobsen, 2006;Glenn et al., 2016;Olsen et al., 2015;Robinson et al., 2002), positive attitude towards SSE (Robinson et al., 2002;Robinson, Turrisi, & Stapleton, 2007), confidence in being able to perform an efficacious skin self-exam (Azzarello et al., 2006;Friedman et al., 1991;Robinson et al., 2002;Robinson et al., 2007), and having a physician recommend SSE (Chiu, Won, Malik, & Weinstock, 2006;Manne & Lessin, 2006;Robinson, Rigel, & Amonette, 1998;Robinson et al., 2007). There is also preliminary evidence that the level of anxiety and the psychosocial strain resulting from a melanoma diagnosis affect selfexam practice (Körner, Augustin, & Zschocke, 2011). Furthermore, being informed about SSE by a health care professional has been shown to be associated with SSE performance (Manne et al., 2004;Coups et al., 2016). Some of the limitations of the literature exploring predictors of SSE include the lack of a standardized operationalization of SSE (Coroiu et al., submitted), which directly affects the reported rates of this behaviour; the limited inclusion of psychosocial variables, such as distress, coping strategies and physician support, as only a few studies have addressed these constructs in relation to SSE; and limited duration of follow-up assessments, as research has shown that the performance of health behaviours decreases over longer time periods (DiMatteo, Giordani, Lepper, & Croghan, 2002;Glanz, Lewis, & Rimer, 2008).
In sum, there's a strong argument from the empirical literature that the early detection of melanoma is associated with less advanced disease and as such with lower melanomarelated mortality. While most melanomas are detected by patients, spouses, and other family members, physician-detected cancerous lesions tend to be thinner, representing an earlier disease stage, than self-detected lesions. Self-examination conducted in tandem with the clinical exam appears to be a desirable and more feasible approach to melanoma early detection than clinical exams alone. Despite SSE being an integral part of clinical guidelines for the prevention of melanoma among at-risk groups, many individuals at risk do not practice SSE regularly or thoroughly. Studies, including randomized controlled trials, have shown that rates of SSE can be improved through patient education, but little is known about those who adhere to SSE clinical recommendations versus those who do not. Acquiring knowledge on the strongest predictors of SSE practice will enable researchers and clinicians to design intervention protocols targeting core issues in melanoma prevention, and, thus, contribute to improved quality of life for patients, 8 decreased need for invasive (but rarely curative) treatments such as chemotherapy, and ideally improved survival.

Research Objectives
The primary aim of this study was to identify short-and long-term predictors of SSE in a sample of patients with melanoma, who had been advised to perform SSE during a standardized dermatological education session on skin cancer prevention. In order to better understand the challenges and opportunities for secondary prevention of melanoma in this high-risk group, we had two specific objectives: (a) to assess rates of SSE behavior over time in a naturalistic setting, and (b) to identify individual-level predictors of SSE in the short (3 months) and long-term (12 and 24 months) following standardized recommendations for SSE. We anticipated that the self-reported rates of SSE behavior would be higher immediately post the dermatological education session compared to the rates reported at 12-and 24-month follow-ups. Hypothesized predictors of SSE in the short-and long-term included non-modifiable (i.e., biological sex, age, education, melanoma stage) and modifiable individual characteristics (i.e., knowledge about melanoma early detection, intentions to perform SSE, self-efficacy about SSE, physician support, psychological distress, and coping). We did not hypothesize a direction or magnitude for the associations between the non-modifiable predictors and SSE given that previous literature was inconclusive. We hypothesized that increased knowledge, intentions, and self-efficacy about performing SSE would be associated with higher rates of SSE in the short and long term, based on theoretical models of health behavior change that link these constructs to behavioral changes.

Study Design
This is an observational study with longitudinal follow-up (5 time points). We followed the 9 STROBE guidelines for reporting of observational studies (Von Elm et al., 2007). A detailed study protocol is available (Korner et al., 2013). Personal characteristics (e.g., age, sex, education, ethnicity, years lived in Canada, marital status, mother tongue) were collected at enrolment (time 1). Disease-specific information (e.g., date of diagnosis, melanoma stage and depth) were collected from the patient medical charts and the pathology reports. A 20-minute standardized educational session on skin cancer preventive behaviours modelled after best-practice guidelines of care for individuals at high-risk for As there is no standardized method to assess SSE behaviour (Coroiu et al., submitted), we developed a scale based on items previously used in melanoma prevention research. We assessed SSE behaviour using 7 items, which inquired about the frequency of examining the skin on the entire body for problematic moles during the 10 previous 3 months (e.g., "In the last 3 months, how often did you examine …"). The first 5 items inquired about the examination of separate body areas: 1) head and neck (face, neck and scalp), 2) front upper body (stomach, check, arms and shoulders), 3) front lower body (legs, genital/hip areas, top and bottom of feet, between toes), 4) back upper body (upper and lower back), 5) back lower body (buttocks and back of legs). Two additional items assessed whether participants had correctly examined the back areas, by using the help of another person ("In the last three months, how often did you have someone else help you with the skin self-exam") or mirrors ("In the last three months, how often did you use a mirror for skin self-exams). Responses were scored on a 6-point scale: 0 ('never'), 1 ('once every 3 months'), 2 ('once every 2 months'), 3 ('once a month'), 4 ('once a week'), 5 ('more often'). To score the SSE behaviour scale, first we collapsed answer choices 4 ('once a week') and 5 ('more often') into one answer due to no prior evidence to suggest a benefit of "weekly" or "more frequent SSE" compared to "monthly SSE" (which is the recommended frequency for SSE, as per most clinical guidelines) and inconsistent distribution of scores for the "weekly" or "more often" categories in the current sample.
Second, we compared the answers to the items inquiring about the back areas (items 3, 4, 5) against the answers to the items inquiring about getting help during the exam from another person (item 6) or using mirrors (item 7) and adjusted the responses to the respective body areas to match the highest answer on either item 6 or 7. Our rationale for scoring the SSE behaviour variable in this manner was that in reality, participants could only adequately examine their backs if they received help from someone else or used two mirrors. Third, we used two separate conceptualizations of SSE in analyses: a) comprehensive SSE, i.e., the frequency of and the extent of skin covered during the skin self-exam, and b) optimal SSE, i.e., whole-body SSE performed at least monthly, as per our standardized patient education based on clinical care guidelines for individuals at 11 increased risk.
Comprehensive SSE. Comprehensive SSE was operationalized as frequency of examining up to 5 body areas in the last three months. A sum score was computed across the 5 body part items (items 1-5), using corrected scores for the items asking about the back areas (3, 4, 5) against items 6 or 7, as per scoring methodology described above. Possible scores ranged from 0 to 20, where higher scores indicated more comprehensive SSE.
Optimal SSE. Optimal SSE was operationalized as whole-body self-exam performed monthly or more often in the last 3 months. To compute this variable, we used the 5 body part items with corrected scores for the items asking about the back areas, as per our scoring methodology described above, which we dichotomized into 1= if all of the 5 body parts were checked at least monthly ("monthly" or "more often") or 0 = if any of the 5 body parts were checked less often than monthly ("once every 3 months" or "once every 3 months").

Other Measures: Independent Variables
Knowledge about Melanoma Early Detection (Coroiu, Moran, Kwakkenbos, Thombs, & Körner, Submitted). Melanoma knowledge was assessed using a 6-item self-report measure covering melanoma risk factors and melanoma preventive behaviours (sample item, "Melanoma can develop a) on any skin surface; b) only on parts of the skin exposed to the sun"). The items were scored as "True" or "False", with total sum scores computed across the 6 items ranging from 0 to 6. Validation analyses conducted in the current sample found that higher knowledge scores were associated with younger age and more positive attitudes about SSE, but was not associated with educational attainment, melanoma stage, or past SSE (Coroiu et al.). . Self-confidence in performing effective SSE was measured using a 5-item self-report questionnaire. Response options ranged from 0 ("strongly disagree") to 3 ("strongly agree"), with possible total scores ranging from 0 to 15 and higher total scores indicating higher levels of self-efficacy for SSE. Item 3 ("There are so many moles and freckles on my body that performing skin self-exams would be difficult") was reverse coded. A previous investigation of the psychometric properties of the scale in data collected at enrolment found that it was reliable (a=0.74) and positively associated with physician support and intentions to perform SSE . (Manne & Lessin, 2006). Intentions to perform SSE were assessed using 1 item: "How likely are you to self-examine your skin on a regular basis in the coming year?". The item was scored on a 5-point Likert scale ranging from 1 ("very unlikely") to 5 ("very likely"), with higher scores indicating stronger intentions to perform SSE. In a study assessing intentions and adoption of SSE practice in patients with melanoma, higher intention to perform SSE was associated with female gender, physician recommendation of SSE, and patient perception of barriers and benefits of SSE (Manne & Lessin, 2006). (Kroenke, Spitzer, Williams, & Löwe, 2009).

Patient Health Questionnaire-4 (PHQ-4)
The PHQ-4 is a 4-item scale assessing symptoms of depression (sample item, "Little interest or pleasure in doing things") and anxiety (sample item, "Not being able to stop or control worrying") over the past 2 weeks. The items are scored on a three-point scale, ranging from 0 ('not at all') to 3 ('nearly every day'). Total scores range from 0 to 12, with higher scores indicating higher distress levels. In a sample of patients seeking treatment in primary-care settings, PHQ-4 scores were strongly associated with functional impairment and higher healthcare usage (Kroenke et al., 2009). Higher PHQ-4 scores were also associated with longer duration of hospital stay, higher likelihood of re-hospitalization within 90 days and of death in patients with advanced cancer (Nipp et al., 2017). (Rhee et al., 2006). Disease-specific emotional, social and appearance-related distress were assessed using a 15-item measure. This selfadministered questionnaire asks about skin cancer worries in the past month (sample item, 'During the past month, how much have you… felt anxious about your skin cancer').

Skin Cancer Index (SCI)
Response options range from 1 ('very much') to 5 ('not at all'), for a total possible sum score of 15 -75. Items were reverse coded to improve comparability to other measures in the current study, where higher scores indicate higher levels of the measured construct. This scale demonstrated a high level of internal consistency (0.82>a>0.92) and good convergent and divergent validity among skin cancer patients (Rhee et al., 2006), and was sensitive to detect changes in distress post-surgical treatment (Rhee et al., 2007).

Reliance on Medical Advice (FKV-2). The two-item subscale "compliance/ trust in doctor" of the Freiburg Questionnaire of Coping with Illness (FKV) [Freiburger Fragebogen
zur Krankheitsverarbeitung] (Muthny, 1989) was used to assess coping with melanoma by relying on medical advice ("I follow the medical advice exactly" and "I trust my doctors").
Response options ranged from 1 ('not at all') to 4 ('very much') with possible total scores between 2 and 8. Higher scores indicate more use of coping by adhering to medical advice. Items from this German scale were translated to English using forward-backward translation procedure (Acquadro et al., 2008;Cha, Kim, & Erlen, 2007). The total FKV scale (35 items) has shown good psychometric properties in a variety of samples with chronic illnesses (Muthny, 1989) and the here used subscale had an acceptable internal consistency of 0.69 in a cancer sample (Hardt et al., 2003).
Constructive Attitudes about Health. The 5-items constructive attitudes subscale of the Health Education Impact Questionnaire (heiQ) (Osborne, Elsworth, & Whitfield, 2007) was used to assess constructive attitudes and approaches to managing challenges of the cancer experience (sample item, "I try not to let my health problems stop me from enjoying life"). Response options range from 0 ('strongly disagree') to 3 ('strongly agree'), with possible total scores between 0 and 15 and higher scores indicating increased attempts to minimize detrimental effects of illness upon one's life. The heiQ questionnaire was originally developed to evaluate patient self-management and education programs (Osborne et al., 2007). Its adaptation to the cancer context was found to be reliable and valid in a large sample of Canadian cancer survivors, where sum scores of the constructive attitudes and approaches subscale were associated with self-efficacy and productive communication, and improved emotional and mental health (Maunsell et al., 2014).

Data Analysis Plan
Descriptive statistics (means, standard deviations, percentages, percentiles) were computed for all study variables. Chi square and t tests were used to test changes in the two behavioral outcomes over time (3-12, 12-24, 3-24 months). Stepwise regression models, with all variables entered in one step were conducted with the two behavioural outcomes separately. Optimal SSE was modeled as a binary outcome and comprehensive SSE was modelled as a continuous outcome. The predictors were age, gender, education, melanoma stage, melanoma knowledge, SSE intentions, SSE self-efficacy, the skin cancer index (SCI), the PHQ-4, constructive attitudes and approaches subscale of the heiQ subscale, and over-reliance on medical advice (FKV-2). Given that we added a data collection (assessment) timepoint after the publication of the protocol and we experienced higher than anticipated attrition rates, we were not able to analyze the data, as per the plan in the protocol. We opted for descriptive, exploratory analyses as opposed to hypothesis-driven analysis of change in the outcome over time due to low sample size per   Table 2. SSE Optimal. The percentage of individuals who performed optimal SSE (defined as checking all 5 body parts at least monthly vs. less often than monthly in the last 3 months) decreased from 57% (3 months) to 44% (12 months) to 37% (24 months) post the dermatological session, as shown in Table 1. Changes in optimal SSE behavior from 3-12 months (χ 2 (1)=4.21, p=0.04) and from 3-24 months (χ 2 (1)=14.72, p<0.001) were statistically significant, but those from 12-24 months were not (χ 2 (1)=3.66, p=0.055).
In analyses controlling for all variables, significant predictors of optimal SSE ranked by strength of association with the outcome at 3 months post the dermatological education session included SSE self-efficacy (OR = 15.31), SSE intentions (OR = 4.48, and female sex (OR = 0.32); at 12 months, significant predictors included reliance on medical advice (OR=4.76) and SSE self-efficacy (OR = 2.95); and at 24 months, predictors included SSE self-efficacy (OR=5.04) and (lower) education (OR=0.69). Detailed statistics per timepoint are included in Table 3. Age, melanoma stage, knowledge about detection, distress, and constructive attitudes about health were not related to optimal SSE.

Discussion
To our knowledge, this is the first study in the literature to follow participants in melanoma follow-up care for as long as 24 month and periodically assess their self-surveillance (skin checking or SSE) behaviour in order to identify the key predictors of this behaviour under the condition of best-practice care in the context of secondary prevention of melanoma. Specifically, the first objective of this observational study with longitudinal follow-up was to assess the prevalence of SSE behaviours over time, i.e., SSE comprehensive and SSE optimal performance, following a standardized educational session on melanoma prevention and early detection via SSE. Comprehensive SSE was conceptualized as frequency and extent of skin covered during the skin exam while accounting for help while checking the back areas. SSE optimal was conceptualized as at least monthly whole-body exam and accounting for help while checking the back areas.
The standardized educational session was designed to match the prevention strategies (information on early signs of melanoma and demonstration of how to perform SSE) recommended by dermatology associations and clinical care guidelines for patients in melanoma follow-up care. We found that comprehensive SSE remained stable from 3 to 12 and 24 months post the educational session, while optimal SSE decreased over time.
Taken together, these results suggest that while individuals continued to examine their bodies for problematic moles to some extent throughout the study period, the percentage of those who adhered fully to the recommendations received during the educational session (and consequently to the recommendations included in the clinical guidelines of care for individuals with a prior history of melanoma) decreased from 3 to 12 months, but decreased less from 12 to 24 months. Of note, the reported rates of SSE behaviour found in this study are higher than those previously reported in cross-sectional studies (Olsen et al., 2015;Glenn et al., 2016;Pollitt et al., 2009). This is potentially due to the fact that we offered all of our participants a standardized educational session on how to adequately examine one's skin to identify problematic lesions, which may boosted everyone's performance of SSE, some of which was maintained over time (see reported results for SSE comprehensive). Moreover, our repeated assessments may have acted as reminder or motivator for performing skin self-exams.
The second objective of this study was to assess the predictors of comprehensive and optimal SSE at 3, 12, and 24 months following a standardized educational session on SSE.
Intention to perform SSE was the strongest predictor of comprehensive SSE in the short (3 months) and long-term (12 and 24 months). The strongest predictors of optimal SSE were self-efficacy and intentions for SSE in the short term (3 months) and self-efficacy for SSE in the long term (12 and 24 months). A possible explanation for these findings is that we used a high-risk sample in active follow-up by a dermatologist. Thus, our participants may have experienced higher levels of motivation (conceptualized as self-efficacy or confidence to perform a preventative behaviour and planning or intending to perform the behavior) to perform preventive behaviours, such as SSE, to begin with. It is also possible that our educational session boosted people's motivation through the reinforcement of the benefits of such behaviour (e.g., early detection of melanoma). Further, at our recruiting hospitals, dermatologists routinely recommended lifelong SSE to all of their melanoma patients and occasionally demonstrated how to perform a skin exam, which might have contributed to elevated levels of self-efficacy for SSE throughout the study compared to patients followed by a general practitioner.
Other variables that showed a small association with increased comprehensiveness of the skin exam included male sex, increased self-efficacy, more advanced cancer stage, and increased reliance on medical advice at 3 months follow-up; and advanced stage, increased self-efficacy, and lower education level at 12 and 24 months follow-ups. Age, knowledge about detection, distress (general and melanoma-specific), and constructive attitudes about health were not related to either comprehensive nor optimal SSE. Of note, previous studies have found females, rather than males, to be more likely to perform SSE and higher levels of education to be associated with SSE (Carli, De Giorgi, Palli, et al., 2003;Glenn et al., 2016;Manne & Lessin, 2006;Olsen et al., 2015;Robinson et al., 2002).
Given the low strength of the association between these variables and the SSE outcome in the current study, we recommend caution when interpreting these effects and their clinical relevance at this point.

Limitations
There are some limitations of this study, which primarily relate to our sampling procedures, the study design, and the measurement of the behavioural outcomes. First, we aimed to include as many eligible participants as were seen at our recruitment's centers during the active phase of the study. However, the skin cancer clinics were extremely busy, and some patients stayed on the premises strictly for the medical checkup, which made it difficult for us to approach them. It is possible that we missed some participants, who would have been eligible and could have provided valuable data to the study. Second, this is an observational study with longitudinal design, so we anticipated a 30% attrition rate. However, as the study extended to 24 months post the educational session, that is at least 27 months post enrolment, we experienced higher than expected loss to follow-up. For the current analyses we did not do imputations and chose to report results from study completers, which reduces our sample size considerably over time.
Smaller numbers at subsequent time points affect the precision of the estimates reported and might affect the generalizability of our findings to other populations. In addition, it is possible that SSE behaviours were lower among individuals, who did not complete the study. Last, this study assessed the behavioral outcomes, comprehensive and optimal SSE via unvalidated self-report measures given that no standardized method of assessment currently exists. While we created the items based on items used previously in research and after consultation with experts, this might nonetheless affect the validity of our 20 findings. Furthermore, recall bias might also play a role in how accurately SSE was reported.

Clinical Implications and Directions for Future Research
The key findings of this study are 1) individuals at increased risk for melanoma, such as melanoma survivors in active follow-up care, maintain SSE behaviour over time, but rates of SSE performed in agreement with medical recommendations are higher immediately after the delivery of a brief standardized dermatological education on skin checking (recommendations for SSE and live demo) and tend to decrease somewhat over a 24month period; and 2) the strongest psycho-social predictors of SSE are intentions and selfefficacy, which can be circumscribed to the larger construct of personal motivation to perform health behaviors. In order to boost adherence to recommended guidelines for SSE performance, future intervention studies with longer follow-up should include reminders and booster sessions. Long-term adherence to recommended guidelines (monthly wholebody SSE and timely seeking of medical opinion when problematic moles are identified) is crucial, as melanoma patients continue to experience an increased risk even 20 years past their diagnosis (Bradford, Freedman, Goldstein, & Tucker, 2010;van der Leest et al., 2012).
To improve self-efficacy for the skin exam specifically, future intervention studies should include specific recommendations about how SSE should be performed, ideally accompanied by in-person demonstration of how to inspect the skin for early signs of melanoma, as well as concrete instructions about when to seek a medical opinion following a self-exam. Future research is needed to optimize these recommendations and instructions for different at-risk populations. To facilitate the setting of intentions to perform skin self-exams and subsequently ask for clinical exams, future studies could employ Motivational Interviewing strategies (Rollnick, Miller, & Butler, 2008) to address 21 individual barriers to change and/or to the adoption of skin cancer screening behaviors, followed by interventions focused on effective goal setting (e.g., implementation intentions: when, how, where to perform the SSE; Gollwitzer, 1993).

Conclusions
To our knowledge, this is the first study in the literature addressing the short (3 months) and long-term (12 and 24 months) predictors of SSE behaviour in a sample of patients with melanoma in active follow-up. We found that the comprehensiveness of the skin exam did not decrease substantially over time and that its strongest predictor in the short and long term was the intention to perform the skin exam. We also found that optimal skin examination (or whole-body SSE performed at least monthly, as per recommendations) decreased over time. The strongest predictors of optimal skin examination were selfefficacy and intentions for the skin exam in the short term and self-efficacy and reliance on medical advice in the long term. These results have implications for the design of future melanoma prevention interventions, as intentions for SSE and self-efficacy for SSE are highly modifiable characteristics tapping personal motivation to perform preventative behaviours and can be targeted by psycho-social interventions (e.g., Motivational Interviewing and/or Implementation Intentions).