Validation of the “knowledge about melanoma early detection scale” in a sample of melanoma survivors

The purpose of this study was to describe the development and preliminary testing of a new scale assessing knowledge about melanoma risk factors and early detection/secondary prevention. Data was drawn from a longitudinal study assessing barriers and facilitators of skin self- examination among patients diagnosed with melanoma. For the current analysis, 191 patients who completed the new 9-item Knowledge About Melanoma Early Detection Scale and other study measures were included. Exploratory factor analysis with were conducted, which identified a robust scale comprised of 6 items with factor loadings ranging from .56 to .81. Higher scores on the Knowledge About Melanoma Early Detection Scale were associated with younger age and more positive attitudes about melanoma prevention, but not with biological sex, education, melanoma stage, or past self-administered and physician-provided skin checks. Further studies are needed to replicate these findings and to further establish the predictive validity of this scale and its usefulness for health research.

. It is therefore important to identify effective modalities to facilitate the performance of skin selfexaminations (SSE) among individuals at risk and to boost self-efficacy for seeking further medical advice after the selfidentification of problematic moles.
Several theoretical models of health behavior change stress the importance of delivering accurate information about risk factors and risk susceptibility, as well as clear and specific information about preventative behaviours that can be undertaken to prevent chronic diseases. For example, the health belief model (HMB; Rosenstock, 1974;Rosenstock, Strecher, & Becker, 1988) uses constructs such as "perceived susceptibility" to refer to one's subjective perception of the risk of acquiring the illness in the future and "perceived benefits" to refer to individual perceptions about the beneficial effects of undertaking certain actions or adopting behaviors that could reduce the risk of developing an illness. Further, the information-motivation-behavioral skills model (IMB; Fisher, Fisher, & Harman, 2003) uses the simple construct of "information" to incorporate both risk perceptions and preventative behaviors that are relevant in the context of preventing or managing a chronic illness.
Currently, there is no single melanoma-specific scale that addresses subjective perceptions of one's personal risk, effectiveness of prevention/early detection behaviors, as well as overall knowledge about melanoma early development and appearance on the skin. Measures of skin cancer knowledge used in previous research with at-risk individuals include assessment of risk factors (usually done in a multiple choice format), familiarity with the ABCDE criteria (which refer to melanoma characteristics in terms of Asymmetry, Border, Colour, Diameter, and Evolution), knowledge of disease progression and treatment options, common locations of melanoma on the skin, and attitudes toward melanoma detection behaviors (Geller et al., 2006;Gillespie, Watson, Emery, Lee, & Murchie, 2011;Manne & Lessin, 2006;Orringer et al., 2005;Swetter et al., 2009). There are several weaknesses associated with the existing knowledge scales and items such as the length of the measures (> 20 items); the inclusion of nonmelanoma-specific questions; and the mixing of items assessing knowledge, attitudes/perceptions, and self-report of skin cancer preventative behaviors (e.g., SSE, sun screen use) in one scale. It is important to separate the assessment of knowledge about melanoma (i.e., risk perception, awareness of prevention strategies, knowledge of melanoma development and appearance) from the actual practice of health behaviours (e.g., engagement in SSE, asking for a clinical skin exam) and from attitudes towards such behaviours, as they constitute distinct constructs that are typically studied separately in health intervention research and can be targeted differentially in interventions. For example, it was hypothesized that knowledge about melanoma (i.e., risk awareness, knowledge of prevention strategies) is a prerequisite for the adoption and maintenance of preventive health behaviours, such as SSE (Körner et al., 2013), it is possible that negative attitudes about SSE, among other constructs, might act as barriers to adopting and maintaining SSE over time. In sum, there are no standalone, standardized scales to assess melanoma knowledge among populations at high risk for melanoma. Such a scale could be used as a progress-monitoring tool in intervention studies targeting the delivery of information about melanoma risk reduction behaviours among high-risk individuals and as a clinical tool to identify knowledge gaps.

Study Objective
The purpose of the current study was to report on the development and preliminary validation of a scale assessing knowledge about melanoma prevention among individuals with a prior diagnosis of melanoma. Knowledge about melanoma prevention was defined as 1) perceptions of risk factors for the development and/or recurrence of melanoma and 2) perceptions of benefits for melanoma early detection (or secondary prevention) behaviors, such as SSE; and 3) general information of melanoma early development and appearance (e.g., awareness of physical characteristics of suspicious moles). This study assessed the factorial structure and internal consistency reliability of the new scale and reported associations between the new scale and demographic and medical indicators previously identified as relevant to melanoma early detection. No a priori hypotheses were tested with respect to the factorial structure of the knowledge about melanoma early detection scale. It was anticipated that small positive associations with socio-demographic variables (i.e., younger age, female gender, higher education, higher income) and past skin checking behaviors, and moderate positive associations with disease severity (i.e., melanoma stage).

Participants and Procedures
This study uses baseline data collected for a longitudinal study investigating barriers and facilitators of SSE among individuals previously diagnosed with melanoma (Coroiu et al., 2020;Körner et al., 2013). English-and French-speaking participants were recruited from two hospitals in Montréal, Canada (active recruitment conducted from September 2012 to March 2014, longitudinal data collection completed in October 2016; see https://osf.io/FTW6V/). Participants completed baseline questionnaires assessing sociodemographic characteristics and a variety of other measures, including 9 items assessing knowledge about melanoma prevention. Medical information, such as date of diagnosis, melanoma stage and thickness, was extracted from hospital medical charts and tumor registry databases.

Measures
Development of the Knowledge about Melanoma Early Detection Scale The research team identified several key "knowledge" domains relevant to melanoma prevention and early detection in high-risk groups. This was based on a comprehensive literature search including epidemiological and review studies of melanoma risk factors and predictors of melanoma early detection behaviors (e.g., Hamidi, Peng, & Cockburn, 2010;Psaty, Scope, Halpern, & Marghoob, 2010;Rhodes, 1995Rhodes, , 2006Rhodes, Weinstock, Fitzpatrick, Mihm Jr, & Sober, 1987) and original empirical research studies assessing knowledge about melanoma and/or skin cancer (e.g., Carli, De Giorgi, Palli, & et al., 2003;Geller et al., 2006;Manne & Lessin, 2006;Orringer et al., 2005). The knowledge domains included knowledge of unmodifiable melanoma risk factors, e.g., a personal or family history of melanoma (Rhodes, 2006), knowledge of melanoma detection practices, including who is most likely to detect melanoma, i.e., patients and their family members/ friends/social networks (Carli, De Giorgi, & Palli, 2003) and how most melanoma are detected, i.e., by checking the skin (Geller, Swetter, Oliveria, Dusza, & Halpern, 2011) and using the ACBDE criteria to identify problematic moles (Hamidi et al., 2010). They also included knowledge of melanoma development, i.e., as a new skin lesion and from an existing mole (Rivers, 2004) and melanoma location, i.e., anywhere on the body as opposed to just the areas exposed to the sun (Caini et al., 2009;Garbe & Leiter, 2009), individual appraisal of melanoma being a serious disease, e.g. with a highly increased metastatic potential when compared to most other human cancers (Yang et al., 2009), and of the efficacy of treatments if the disease were to be detected early (Markovic et al., 2007).
The first (AC) and senior (AK) authors created a pool of 9 different items assessing several of these knowledge domains, with a particular focus on 1) perceptions of risk factors for the development and/or recurrence of melanoma and 2) perceptions of benefits for melanoma early detection behaviors, including who is most likely to detect melanoma (i.e., patients) and by which means (i.e., SSE); and 3) general information of melanoma early development and appearance (e.g., awareness of physical characteristics of suspicious moles). For each item, participants were instructed to check off the statement that they believe is correct (sample item, "Melanoma can develop a) only as a new growth on the skin or b) as a new growth on the skin or in an existing mole or c) I don't know"). A neutral answer choice ("I don't know") was added in order to minimize the tendency to guess the correct answer when unsure or to skip the item altogether. Responses to the 9 items were scored by the research team as "correct" or "incorrect". For data analyses, a Correct/Incorrect dichotomy was used whereby the "I don't know option" was collapsed with the "Incorrect" option, as they both capture the incorrect answer choice. This decision was consistent with the conceptualization of this scale as measuring objective knowledge, as opposed to attitudes, beliefs or perceptions about melanoma. This scoring method is also in line with the scoring of previously used melanoma knowledge questionnaires (Gillespie et al., 2011).
The final 9 items were originally created in English and later translated into Canadian French to reflect the language needs of the Québécois patients. A strict protocol for forward and backward translations was followed (Dominique, Pauline, & Yvette, 2000), using the services of an authorized French translator with prior experience in health research to translate the English items into French (forward translation). Further, one bilingual (English-French) research assistant who identified as a native English speaker translated the French items back into English (backward translation). Subsequently, two bilingual members from the research team evaluated the forward and backward translations against each other and suggested edits, which were discussed with the French authorized translator who then adapted the final French version. Finally, two independent bilingual reviewers familiar with psychosocial oncology research and one community member ensured the clarity and relevance of the items and approved the final English and French versions.
Skin Self-Exams (SSE; Körner et al., 2013) Skin selfexamination (SSE) behaviour was assessed using 5 items inquiring about the frequency of checking the skin of 5 body areas prior to receiving a diagnosis of melanoma (i.e., "Prior to your melanoma diagnosis, how often did you carefully and purposefully examine … your head / front upper body/ front lower body/ back upper body, back lower body" ). Items were rated on the 4-point scale, including 0 (never), 1 (a few times a year), 2 (once a month), and 3 (more often than monthly). Two additional items were used to inquire about use of mirrors and help from someone else during the self-exam. Answers on the items asking about the back areas were adjusted to be at least equal to the scores reporting on the help items (using mirrors or someone's help) under the assumptions that participants could not, in fact, adequately check the skin on their lower and upper backs without using mirrors and/or being assisted by someone else. A sum score across the 5 body items was used in analyses, with possible scores ranging from 0 to 20 and higher scores reflecting more comprehensive SSE. Cronbach's alpha for the SSE behavior scale was .86 in the current study.
Medical Skin Exams and Recommandations for Skin Self-Exams (CSE; Körner et al., 2013) Clinical skin exams (CSE) prior to diagnosis, as reported by patients, was assessed using a single item ("Prior to your melanoma diagnosis, how often did a physician/nurse performed a skin examination of your body") rated on a 0 to 3 range (never, once, once a year, a few times a year). Healthcare professional recommendations ("How often did your physician/nurse suggest that you should self-examine your skin") and demonstration of skin exams ("How often did your physician/nurse show you how to identify suspicious changes in your skin") were assessed via single items rated on the same Likert-type scale. Cronbach's alpha for the 3-item CSE scale was .81 in the current study. The items were used separately in analyses.
Attitudes about SSE (Manne & Lessin, 2006) Four items, two positive ("By doing SSE, I can find moles/growths on my skin that are problematic") and two negative ("I do not need to do SSE because I protect my skin from the sun") were used to assess attitudes towards SSE. The items were rated on a scale ranging from 0 (strongly disagree) to 3 (strongly agree) and were used separately in analyses. Cronbach's alpha for this 4item scale was .63 in the current study.

Data Analysis Plan
Exploratory Factor Analysis (EFA) with weighted least squares parameter estimation (MLSMV) and oblique rotation was conducted with initial items from the knowledge about melanoma early detection scale in MPlus v.7 (Muthén & Muthén, 2007). Prior to the EFA, items in which one of the response options were endorsed by over 90% of the sample were removed from the analysis (items #6, # 7). Complete responses included in EFA analyses were defined as completion of at least 7 out of the 9 items.
To assess model fit, we used the chi-square test, which is highly sensitive to sample size and potentially leading to erroneous rejection of the model fit (Reise, Widaman, & Pugh, 1993) in tandem with a combination of other indices, i.e., the Tucker-Lewis Index (TLI; Tucker & Lewis, 1973), the Comparative Fit Index (CFI; Bentler, 1990) and the Root Mean Square Error of Approximation (RMSEA; Steiger, 1990). Good fitting models are indicated by a TLI and CFI ≥ 0.95 and RMSEA ≤0.06 (Hu & Bentler, 1999), although a CFI and TLI of .90 or above (Kline, 2005) and a RMSEA of .08 or less (Browne & Cudeck, 1993) are regarded as indicators of an adequate model fit. Descriptive statistics, Pearson's correlations, and Cronbach's alpha were computed using SPSS, version 20. The magnitude of the correlations was interpreted following Cohen's effect size descriptors, i.e., r ≤ 0.10 indicating small, r = 0.30 indicating moderate, and r = 0.50 indicating large differences (Cohen, 1988).
A preliminary EFA with 7 items scored as administered, on a three-point scale (correct, incorrect, I don't know) revealed a less than adequate fit for a 1-factor solution χ 2 (N = 191, df = 14) = 40.21, p < .001, RMSEA = .10, 90% CI [.06, .14], CFI = .94, TLI = .91 (Appendix , Table S1). In line with our a priori decisions about scoring this measure, the final EFA analyses included items scored dichotomously, with the neutral answer choice ("I don't know") collapsed with the "incorrect" answer option. This decision reflects most closely the purpose of the scale, which was to assess presence versus absence of accurate knowledge on the early prevention of melanoma.

Sample Characteristics
In total, 242 participants diagnosed with melanoma consented and were eligible for the longitudinal study. The current analysis included 191 participants who provided complete data on the knowledge about melanoma early detection scale (79% of the total sample recruited for a longitudinal project) and were therefore included in factor analyses. Participants with incomplete responses on other study measures were removed from validity analysis using listwise deletion, leaving a subsample of 154 patients who provided complete answers to all of the study measures. Socio-demographic characteristics for the full sample and descriptive statistics for the Knowledge about Melanoma Early Detection Scale are included in Tables 1 and 2 (n = 191), respectively. Descriptive statistics for study measures used in validity analyses are included in Table 3 (n = 154).

Validity and Reliability
Cronbach's alpha for the 6-item scale was .69 (M = 4.31; SD = 1.65; Sum score Range = 0-6, n = 191). The sum score on the Knowledge about Melanoma Early Detection scale was positively associated with beliefs that doing SSE can facilitate early detection of melanoma (attitude 1) and that SSE is very important for melanoma survivors (attitude 4), and negatively associated with age and the belief that those who use sun protection do not need to perform SSE (attitude 2). The correlations (included in Table 4) were in the small-to-moderate range. The Knowledge about Melanoma Early Detection scale was not associated with sex, education level, SSE behavior and health care professionals' recommendation for SSE, performance of a clinical exam in the past, or demonstration of a self-exam. Test-retest reliability, which was measured at 3-6 months post initial administration, was .62.

Discussion
Preliminary exploratory analyses with the Knowledge about Melanoma Early Detection scale, which included risk perception, perceived benefits of skin exams, general knowledge about melanoma appearance and development, and importance of early detection revealed that a one factor fit the data reasonably well and that internal consistency reliability for the final 6-item scale was slightly lower than what is typically considered acceptable for a newly developed scale (.70;Nunnally, 1978). However, researchers have suggested that values as low as .60 are acceptable in the early stages of scale development (Nunnally, 1978) and for exploratory research (Hair, Black, Babin, & Anderson, 2010). The alpha coefficient value is sensitive to the number of items included in a scale, which might explain why the alpha value is below the accepted cut-off. More importantly, the construct of knowledge is qualitatively different from other psychological constructs, such as attitudes or personality traits, which usually show much higher alpha's, in that variability in answers across the items might be greater compared to that shown in other scales (i.e., less consistent pattern in answering the questions, as a function of individual knowledge level). Not surprisingly, younger individuals reported increased levels of knowledge in this study, which may be related to easy access to various medical and prevention information on the web, including through use of social media. Further, more positive attitudes about skin checking were associated with increased knowledge about prevention and early detection. This is reassuring from the viewpoint that people who hold positive attitudes about preventive behaviours are also the ones with better, more accurate knowledge about other aspects of disease prevention and early detection. Conversely, knowledge was not associated with biological sex or disease severity (assessed using melanoma stage), past skin self-exam, or past clinical exams, recommendations to perform skin exams or demonstration on how to perform a skin exam. It is not surprising that knowledge about melanoma prevention and early detection was not associated with behavioral outcomes, skin self-checking or medical skin exams: in skin cancer prevention and lifestyle change literature (e.g., physical activity, dieting), information alone and/or various knowledge constructs have shown weak or non-significant associations with health behaviors (Carpenter, 2010;Manne & Lessin, 2006;Schwarzer et al., 2007).

Limitations and Future Directions
Some of the limitations of the current study include the absence of a qualitative component during the development

Implications
The Knowledge About Melanoma Early Detection Scale is a brief, but robust tool designed to help various healthcare professionals involved in melanoma prevention and care to assess and quantify patient knowledge, by specifically demystifying information about melanoma prevention (e.g., melanoma can develop as new mole, not just in an existing mole; melanoma can appear anywhere on the body, not just in areas exposed to the sun). This tool can also be used to identify knowledge gaps among individuals at risk and those who would benefit from further, more specific education about melanoma in clinical practice. For instance, the results of this study suggest that certain patient groups (i.e., older individuals) have lower levels of knowledge about melanoma-related risk factors and early detection strategies. These individuals may benefit from targeted interventions from nurses to increase their risk awareness and illness susceptibility, knowledge of beneficial  Attitude 3 = I go to all of my melanoma follow up appointments, therefore I don't have to do SSE; Attitude 4 = SSE is very important because I have had melanoma a Correlation strength was interpreted using Cohen's effect size descriptors, i.e., r ≤ 0.10 indicating small, r = 0.30 indicating moderate, and r = 0.50 indicating large differences (Cohen, 1988), and CI rather than p value alone. s preventative behaviors. Lastly, this scale was designed primarily for research purposes as a tool to measure gaps in patient knowledge of melanoma secondary prevention, and as such, it could be used as a progress-monitoring tool in intervention research with individuals at increased risk for melanoma.
Author's Contribution AC, AK, BT Designed the study. AC, CM, LK Conducted data analysis. AC completed the first draft of the manuscript and incorporated feedback from co-authors.
CM, LK, BT, AK Contributed critical feedback to earlier drafts of the manuscript, reviewed, and approved the final draft of the manuscript.
Funding The current study was funded by operating grants from the Fonds de recherche du Québec -Santé (FRQS) and the Canadian Institutes of Health Research (CIHR) awarded to Annett Körner. Adina Coroiu's research training was supported by graduate and post graduate awards from FRQS and CIHR. Dr. Thombs was supported by a FRQS researcher salary award. The funding agencies had no bearing in the design and conduct of the study or the write-up and publication of the study results.
Data Availability The datasets generated during and/or analysed during the current study are available in the Open Science Framework repository at https://osf.io/au5tp/.

Compliance with Ethical Standards
Conflict of Interest The authors declare that there is no conflict of interest.
Ethical Statement Institutional Research Boards of McGill University and participating hospitals approved the study.
Informed Consent Was obtained in writing from all participants before study enrollment.