USF Libraries
USF Digital Collections

Investigating the role of appearance-based factors in predicting sunbathing and tanning salon use

MISSING IMAGE

Material Information

Title:
Investigating the role of appearance-based factors in predicting sunbathing and tanning salon use
Physical Description:
Book
Language:
English
Creator:
Cafri, Guy
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla
Publication Date:

Subjects

Subjects / Keywords:
Skin cancer
Sunbathing
Indoor tanning
Body image
Theory of planned behavior
Dissertations, Academic -- Psychology -- Doctoral -- USF   ( lcsh )
Genre:
non-fiction   ( marcgt )

Notes

Abstract:
ABSTRACT: Understanding the motives for sunbathing and indoor tanning is an extremely important public health issue. UV exposure via sunbathing and utilization of sun lamps and tanning beds are considered important risk factors for the development of skin cancer. Psychosocial models of UV exposure are often based on theories of health behavior, but theory from the body image field can be useful in understanding motives to UV expose as well. The current study examines models that prospectively predict sunbathing and indoor tanning behaviors using constructs and interrelationships derived from the tripartite theory of body image (Thompson et al., 1999), as well as those from the theory of reasoned action (Ajzen & Fishbein, 1980), health belief model (Rosenstock, 1974), revised protection motivation theory (Rogers, 1983), and a proposed integration of several health behavior models (Fishbein, 2000). The results generally support a model in which intentions mediate the relationship between appearance attitudes and tanning behaviors, appearance reasons to tan and intentions mediate the relationship between sociocultural influences and tanning behaviors, and appearance reasons not to tan and intentions mediate the role of perceived threat on behaviors. The implications of these findings yield important information relevant to the understanding of motives to UV expose, which can useful to the development of novel prevention and early intervention programs geared toward the reduction of skin cancer risk.
Thesis:
Dissertation (Ph.D.)--University of South Florida, 2008.
Bibliography:
Includes bibliographical references.
System Details:
Mode of access: World Wide Web.
System Details:
System requirements: World Wide Web browser and PDF reader.
Statement of Responsibility:
by Guy Cafri.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 41 pages.
General Note:
Includes vita.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001992183
oclc - 315883757
usfldc doi - E14-SFE0002356
usfldc handle - e14.2356
System ID:
SFS0026674:00001


This item is only available as the following downloads:


Full Text
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001992183
003 fts
005 20090316134023.0
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 090316s2008 flu s 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0002356
035
(OCoLC)315883757
040
FHM
c FHM
049
FHMM
090
BF121 (Online)
1 100
Cafri, Guy.
0 245
Investigating the role of appearance-based factors in predicting sunbathing and tanning salon use
h [electronic resource] /
by Guy Cafri.
260
[Tampa, Fla] :
b University of South Florida,
2008.
500
Title from PDF of title page.
Document formatted into pages; contains 41 pages.
Includes vita.
502
Dissertation (Ph.D.)--University of South Florida, 2008.
504
Includes bibliographical references.
516
Text (Electronic dissertation) in PDF format.
3 520
ABSTRACT: Understanding the motives for sunbathing and indoor tanning is an extremely important public health issue. UV exposure via sunbathing and utilization of sun lamps and tanning beds are considered important risk factors for the development of skin cancer. Psychosocial models of UV exposure are often based on theories of health behavior, but theory from the body image field can be useful in understanding motives to UV expose as well. The current study examines models that prospectively predict sunbathing and indoor tanning behaviors using constructs and interrelationships derived from the tripartite theory of body image (Thompson et al., 1999), as well as those from the theory of reasoned action (Ajzen & Fishbein, 1980), health belief model (Rosenstock, 1974), revised protection motivation theory (Rogers, 1983), and a proposed integration of several health behavior models (Fishbein, 2000). The results generally support a model in which intentions mediate the relationship between appearance attitudes and tanning behaviors, appearance reasons to tan and intentions mediate the relationship between sociocultural influences and tanning behaviors, and appearance reasons not to tan and intentions mediate the role of perceived threat on behaviors. The implications of these findings yield important information relevant to the understanding of motives to UV expose, which can useful to the development of novel prevention and early intervention programs geared toward the reduction of skin cancer risk.
538
Mode of access: World Wide Web.
System requirements: World Wide Web browser and PDF reader.
590
Advisor: Joel Kevin Thompson, Ph.D.
653
Skin cancer
Sunbathing
Indoor tanning
Body image
Theory of planned behavior
690
Dissertations, Academic
z USF
x Psychology
Doctoral.
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.2356



PAGE 1

Investigating the Role of Appearance-Ba sed Factors in Predicting Sunbathing and Tanning Salon Use by Guy Cafri A dissertation submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy Department of Psychology College of Arts and Sciences University of South Florida Major Professor: Joel Kevin Thompson, Ph.D. Paul Jacobsen, Ph.D. Thomas Brandon, Ph.D. William Sacco, Ph.D. Michael Brannick, Ph.D. Rita Debate, Ph.D. Date of Approval: March 24, 2008 Keywords: Skin cancer, sunbathing, i ndoor tanning, body image, theory of planned behavior Copyright 2008, Guy Cafri

PAGE 2

i Table of Contents List of Tables ii List of Figures iii Abstract iv Introduction 1 Method 4 Participants 4 Procedure 4 Measures 4 Appearance Factors 4 Perceived Threat-Skin Cancer 5 Skin Cancer Risk 5 UV Exposure Behaviors 5 UV Exposure Intentions 5 Missing Data 6 Results 7 Attrition 7 Outliers and Normality 7 Model Comparisons 8 Interpretation of the Final Model 9 Discussion 10 Implications on Theory 10 Implications for the De sign of Interventions 11 Footnotes 13 References 15 Appendices 20 Appendix A: Scale Items 21 Appendix B: Tables and Figures 24 About the Author End Page

PAGE 3

ii List of Tables Table 1 Means, Standard Deviations, and Reliability of Scales 24 Table 2 Correlations Among Scales 27 Table 3 Tests of Alternative Mo dels (without item deletion) 28 Table 4 Tests of Alternative Models (with item deletion) 29 Table 5 Tests of Alternative Models (outliers deleted) 30 Table 6 Tests of Indirect E ffects (without item deletion) 31 Table 7 Tests of Indirect Eff ects (with item deletion) 32 Table 8 Tests of Indirect Eff ects (with outliers deleted) 33

PAGE 4

iii List of Figures Figure 1 Hypothesized Model 34 Figure 2 Alternative Models for th e Role of Intentions/Behaviors 35 Figure 3 Alternative Model for the Role of Sociocultural Influence 36 Figure 4 Alternative Models for th e Role of Perceived Threat 37 Figure 5 Alternative Model for the Role of Skin Cancer Risk 38 Figure 6 Standardized Path Coeffi cients (without item deletion) 39 Figure 7. Standardized Path Coef ficients (with item deletion) 40 Figure 8. Standardized Path Coef ficients (with outlier deletion) 41

PAGE 5

iv Investigating the Role of Appearance-Ba sed Factors in Predicting Sunbathing and Tanning Salon Use Guy Cafri ABSTRACT Understanding the motives for sunbathing and indoor tanning is an extremely important public health issue. UV exposure vi a sunbathing and utiliz ation of sun lamps and tanning beds are considered important ri sk factors for the development of skin cancer. Psychosocial models of UV exposure are often based on theories of health behavior, but theory from the body image field can be useful in understanding motives to UV expose as well. The current study examin es models that prospectively predict sunbathing and indoor tanning be haviors using constructs a nd interrelationships derived from the tripartite theory of body image (Thompson et al., 1999), as well as those from the theory of reasoned action (Ajzen & Fishbein, 1980), hea lth belief model (Rosenstock, 1974), revised protection motivation theory (R ogers, 1983), and a proposed integration of several health behavior models (Fishbein, 2000). The results generally support a model in which intentions mediate the relationship between appearance attitudes and tanning behaviors, appearance reasons to tan and intentions mediate the relationship between sociocultural influences and tanning behaviors, and appear ance reasons not to tan and intentions mediate the role of perceived threat on behaviors. The implications of these findings yield important information releva nt to the understanding of motives to UV expose, which can useful to the development of novel prevention and early intervention programs geared toward the re duction of skin cancer risk.

PAGE 6

1 Introduction The incidence of skin cancers has reached epidemic proportions in the United States with more than 1 million cases of basal and squamous cell carcinoma and 59,940 cases of malignant melanoma expected to be dia gnosed in 2007 (8,110 deat hs projected from melanoma; American Cancer Society, 2007). Re search suggests that ultraviolet (UV) radiation through sun and sunbed/sunlamp expos ure is a central risk factor for the development of skin cancers (e.g., U.S. Department of Health and Human Services, 2002). In the aims of reducing skin cancer ri sk, researchers have attempted to better understand motives for UV exposure. The fram ework for understand ing sunbathing and indoor tanning salon use has been primarily influenced by th eories of health behavior (e.g., theory of planned behavior; Ajzen, 1985; for an application see: Hillhouse, Adler, Drinnon, & Turrisi, 1997). Althoug h empirical models based on these theories have been informative, incorporating constructs and hypothesized relationshi ps based on theory from the body image field can be useful in understanding motives to UV expose because UV exposure is consistently related to wanting to look tan (e.g., Hillhouse, Turrisi, & Kastner, 2000; Wichstrom, 1994), and appearan ce-focused interventi ons have been found to be efficacious (Gibbons et al., 2005; Hillhouse & Turrisi, 2002; Jackson & Aiken, 2006; Jones & Leary, 1994; Mahler et al., 2003; Mahler et al ., 2005). Preliminary application of one body image theory, the trip artite theory (Thompson Heinberg, Altabe, & Tantleff-Dunn, 1999), to the area of tanning behavior has id entified constructs (Cafri, Thompson, Roehrig et al., 2006; Cafri et al., in press) and relationships among constructs (Cafri, Thompson, & Jacobsen, 2006) that ha ve further elucidated the reasons for UV exposure. The goal of the current paper is to exam ine the extent to which biopsychosocial factors predict sunbathing and indoor tanning behaviors prospectively (at six month follow-up), using constructs primarily deri ved from the body image field, but selected constructs from theories of health behavi or as well. Relationships hypothesized by the theory of reasoned action (Ajzen & Fishbe in, 1980), health belief model (Rosenstock, 1974), revised protection motiva tion theory (Rogers, 1983), a proposed integration of several health behavior models (Fishbein, 2000 ), and the tripartite theory of body image (Thompson et al., 1999) are considered, and asp ects of these theories are tested using structural equation modeling. The theory of reasoned action (Ajzen & Fishbein, 1980) is a social psychology theory used for the study of many health behaviors. The theory posits that attitudes toward a behavior and subjec tive norms predict intentions to behave, which in turn predict the particular behavior The theory of planned behavi or (Ajzen, 1985) adapts this theory to include behavioral control. The a pplication of the theori es of reasoned action and planned behavior to the study of UV expos ure indicates that intentions to sunbathe and indoor tan are positively associated with their respective behaviors (Hillhouse et al., 1997; Jackson & Aiken, 2000). More over, attitudes, subjecti ve norms, and perceived

PAGE 7

2 behavioral control were found to be positiv ely associated with sunbathing and indoor tanning intentions (Hillhouse et al., 1997; H illhouse et al., 2000; Jackson & Aiken 2000). One important limitation of past research is that models examining the association between intentions and their predictors are fit without including the corresponding behavior, or if the behavior is included, its association w ith intentions is evaluated separately from predictors of intentions (e.g., Jackson & Aiken, 2000; Hillhouse et al., 1997). This strategy generally precludes a compre hensive test of the theory of planned behavior as applied to UV exposure, and spec ifically limits evalua tion of the indirect effects of attitudes/subjective norms on the behavior of interest. A final limitation to consider is in many research studies either sunbathing or tanning salon use is assessed (e.g., Jackson & Aiken, 2000), however, both shoul d be evaluated because both are risk factors for the development of skin cance r (U.S. Department of Health and Human Services, 2002). The health belief model (Rosenstock, 1974) proposes that i ndividuals adopt a protective behavior to the extent that they pe rceive themselves to be susceptible, perceive the outcome to be severe, benefits of the be havior are protective against the threat, and barriers to the protective beha vior can be overcome. Revise d protection motivation theory (Rogers, 1983) builds upon the health belief model by argui ng that perceived threat (perceived susceptibility + perceived severi ty) elicits a fear res ponse that reduces the probability of maladaptive responses (e.g., UV exposure), suggesting a direct effect on intentions/behaviors. In contrast, Fishbein ( 2000) has argued that perceived threat only has an indirect influence on intentions/behav iors, mediated through major constructs of the theory of planned behavior (norms, attitu des, behavioral control). If perceived susceptibility is viewed as a proxy for pe rceived threat, one st udy found support for both revised protection motivation theory and Fishbe in’s (2000) view, such that susceptibility was associated with intenti ons to sunbathe and advantages of sunbathing mediated the relationship between these two variables (J ackson & Aiken, 2000). In the same study, the authors hypothesized and found evidence that perceived susceptibility mediates the relationship between skin cancer risk (based on skin type and family history of skin cancer) and intentions to sunbathe. Skin can cer risk based on physi cal indicators (e.g., skin type) is often considered a substantive demographic char acteristic or a confounding variable in models of UV exposure, therefore its direct relationship to outcome variables is typically considered. However, Jackson and Aiken (2000) found no evidence for the direct relationship on intentions beyond the indirect effect discussed above. The tripartite theory of body imag e (Thompson et al., 1999) posits that appearance-based sociocultural influences (peers, parents, media) lead to body dissatisfaction, which in turn lead to eati ng disorder symptoms. Several studies have already identified a significant relations hip between appearance motives and intentions/behaviors to UV expose (Hillhouse, Stair, & Adler, 1996; Hillhouse et al., 1997; Hillhouse, et al., 2000; Jackson & Aike n, 2000; Wichstrom, 1994). However, the manner in which appearance is assessed is not always specific to tanning, scales can be confounded with indicators that are not appearance-based, a nd the scales have limited evidence of construct validity. Consistent with the tripartite theory, two studies recently demonstrated the multidimensional nature of appearance motives (Cafri, Thompson, Roehrig et al., 2006; Cafri et al., in press), su ch that there are three higher-order factors:

PAGE 8

3 sociocultural influences to tan, appearance r easons to tan, and appearance reasons not to tan. Also consistent with the tripartite theor y, a series of mediation models demonstrated that general attractiveness reasons for tanni ng mediated the relationship between media influence and intentions/beh aviors to UV expose (Cafri, Thompson, & Jacobsen, 2006). A more comprehensive model, including all facets of the abovementioned higher-order constructs, as well as other relevant variables, has not been tested. A model illustrating hypothesized relationships based on the aforementioned theories can be found in Figure 1. Based on the theory of reasoned action, it is hypothesized that intentions wi ll mediate the relationship be tween each of the following and UV exposure behaviors (evaluated at si x month follow-up): appearance reasons to tan, appearance reasons not to tan, and percei ved threat. Based on the tripartite theory, appearance reasons to tan are predicted to have a positive relationship with intentions to sunbathe, appearance reasons not to tan will have a negative relationship, and appearance reasons to tan will mediate the relations hip between sociocultural influence and intentions. Based on an extension of the h ealth belief model (Jackson & Aiken, 2000) and revised protection motivation theory, percei ved threat is predicted to mediate the relationship between skin cancer risk and inte ntions to tan, such that risk will be positively associated with threat and threat wi ll be negatively associated with intentions to tan. Based on Fishbein’s (2000) integrative model and the empirical results of Jackson and Aiken (2000), an inverse relationship between perceived threat and appearance reasons to tan, and a positive relationship between perceived threat and appearance reasons not to tan are expected. Theory from the areas of body image and health psychology offer a useful framework for hypothesizing relationships among variables that can be used to predict UV exposure behaviors. A series of compe ting hypotheses about the relationships among the variables can be constructe d based on these theories. Hypot heses will be tested using structural equation models. The model that is most consistent with the data will be interpreted in terms of the magnitude and statistic al significance of the path coefficients.

PAGE 9

4 Method Participants The primary sample consists of 589 female s from the University of South Florida that were reported on in a prev ious study (Cafri et al., in press). We chose to focus on females because as a group they are more lik ely than males to use indoor tanning salons and engage in outdoor tanning behaviors (D avis, Cokkinides, Weinstock, O’Connell, & Wingo, 2002; Demko, Borawski, Debanne, Coope r, & Stange, 2003; Lazovich et al., 2004), and thus are at greater risk for devel oping skin cancer. Furthermore, the rates of tanning behavior, in particular tanning salon use, were extr emely low for a sample of university males (Cafri et al., in press). A dditional criteria included being between 18 and 26 (M = 19.88, SD = 1.84) in order for the samp le to be representative of young adults, as well as having a skin type between I and IV (Fitzpatrick, 1988) because individuals with skin types V and VI (i.e., brown or black skin color) are at a significantly reduced risk for developing skin cancer (Goldsmith, 1987). The racial distribution of participants was 80% White/Caucasian, 5 % Black/African Amer ican, 3% Asian/Paci fic Islander, >1% American Indian/ Alaska Native, and 11% Othe r. Percentages of participant skin types were: 12% Type I, 26% Type II, 36% Type II I, and 26% Type IV. In order to measure the temporal stability of the measures, an independent sample of 80 females resembling the primary sample in demographic characteri stics, was evaluated during the months of April and May by measuring same individuals twice over the course of one week. Procedure Participants were recruited from intr oductory psychology classes. Questionnaires were completed online. Time 1 data on biopsyc hosocial variables and intentions to UV expose were collected between the months of October and November. Time 2 data on UV exposure behaviors were collected six mont hs later between the months of April and May. In order to minimize attrition, participan ts were contacted a minimum of five times by email and phone before they were c onsidered non-responders. A total of 311 participants (52%) completed the survey at Ti me 2. Participants were given course credit for their participation at Time 1 and given th e option of course credit or $15 at Time 2. Measures Appearance Factors. Three appearance factors deve loped in previous studies (Cafri, Thompson, Roehrig et al., 2006; Cafri et al., in press) are used in the current study (see Figure 1): sociocultural influences to tan, appearance reasons to tan, and appearance reasons not to tan. In the current study, each of these three factors is specified as a higherorder factor, with lower-order factor indicators, with each of these lower-order factors in turn having individual items serving as indi cators (for exact specifications of the measurement model see: Cafri et al., in press) Each item is measured on a 5-point Likert scale. The sociocultural influences to tan f actor consists of four lower-order factors: media, friends, family, and significant others. Appearance reasons to tan factor includes three lower-order factors: general appearan ce enhancement, reducing the appearance of

PAGE 10

5 acne, and enhancement of body shape. Appearan ce reasons not to tan consists of two lower-order factors: sk in aging and immediate skin damage. Evidence for validity of the scales includes: item construction based on previous theory/resear ch in the body image field and focus groups with people who tan, expl oratory and confirmatory factor analytic models, and convergence of factors with UV exposure and protection outcomes (Cafri, Thompson, Roehrig et al., 2006; Cafri et al., in press). 1 Perceived Threat-Skin Cancer Perceived threat is sp ecified as a higher-order factor, with perceived suscep tibility and perceived severi ty as lower-order factor indicators, with each of these lower-order fact ors in turn having individual items serving as indicators. Perceived suscep tibility is measured using a four item measure, each item with a 6-point Likert response format (Jack son & Aiken, 2000). Based on the results of a previous study there is evidence of validity ba sed on confirmatory factor analysis, as well as reliability (test-retest r = .72; Coefficient = .63, .71; Jackson and Aiken, 2000). Perceived severity is measured using a three item measure, ea ch item with a 6-point Likert response format (Jackson & Aiken, 2000) Based on the results of a previous study there is evidence of validity based on confirmatory factor analysis, as well as internal consistency reliability, but not test -retest reliability (test-retest r = .46; Coefficient = .64, .77; Jackson and Aiken, 2000). Notably, test-re test reliability estimates are higher in the current sample (see Table 1). Skin Cancer Risk. Risk for skin cancer was assessed using two items, skin type (6point Likert scale; Fitzpatr ick, 1988) and untanned skin colo r (3-point Likert scale; Weinstock, 1992). These items had a statistic ally significant association with sunsensitivity in a previous study, as measur ed by minimal erythema dose, the dose of ultraviolet B light required to produce visibl e redness of the skin (Weinstock, 1992). Sun sensitivity is a major risk factor for melanom a, therefore these findings reflect criterionrelated validity. Temporal stability of the skin type measure was adequate in a previous study (test-retest r = .82; Jackson & Aiken, 2000). UV Exposure Behaviors at Six Month Follow-Up. Indoor tanning was assessed with a single item that asks about use over the past six months ("Please give me your best estimate of how many times you have i ndoor tanned in the last 6 months ; Hillhouse, Turrisi, Holwiski & McVeigh, 1999). Particip ants respond to the item by checking the box that best approximates the range of times they indoor tan on a seven-point scale (0, 1-10, 11-20, 21-30, etc.). Sunbath ing behavior was assessed using a single item modeled after the indoor tanning item ("Please give me your best estimate of how many times you have sunbathed in the last 6 months ") with an identical scoring method. Test-retest reliability of these items over a 7-10 day pe riod was adequate in a previous sample (indoor r = .84, sunbathing r = .78; Cafri, Thompson, & Jacobsen, 2006). UV Exposure Intentions. Indoor tanning intentions were assed with a single item that asks participants to provide a six mont h estimate of times they plan to go indoor tanning (“Please give me your best estim ate of how many times you plan to use an indoor tanning salon in the next 6 months”; Hillhouse & Turrisi, 2002). The scoring method is identical to the UV exposure behavior items. Sunbathing intentions were assessed using an item similar to the indoor tanning item (" Please give me your be st estimate of how many times you plan to sunbathe in the next 6 months ") with an identical scoring format.

PAGE 11

6 Test-retest reliability of these items over a 7-10 day period was adequate in a previous sample (indoor r = .88, sunbathing r = .82; Cafri, Thompson, & Jacobsen, 2006). Missing Data Three hundred and eleven participan ts had complete data. Among the 278 participants that did not have complete data (i.e., could not be followed-up at Time 2), information on all measured variables ex cept UV exposure behaviors was available. Missing data for these individuals was ha ndled through maximum likelihood estimation of the raw data (i.e., full information ma ximum likelihood; Aurbuckle, 1996, Muthen, Kaplan, & Hollis, 1987). In order for this meth od to have favorable statistical properties it at least requires the untestable assumpti on that the data are missing at random (MAR) 2 and multivariate normality, but it has been suggested that maximum likelihood is somewhat robust to violations of this latter assumption (Allison, 2001). Planned Analyses Study hypotheses are evaluated using st ructural equation modeling with maximum likelihood estimates of parameters (AMOS 6.0; Arbuckle, 2005). Specification of the “structural” por tion of the models evaluated in th is study is detailed throughout, but the “measurement” portion is not. Genera lly, the measurement portion consists of individual items serving as indicators of th eir respective factors because this leads to parameter estimation using optimal weights (cf. scale composites using unit weights for items; Bollen & Lennox, 1991). When only one item is used as an indicator of a latent variable (i.e., UV exposure variables), a va lue for the error vari ance is specified by multiplying the variance of the variable in the current sample by its estimate of unreliability.4 When testing competing hypotheses about the relationships among the variables, nested models are compared using the likelihood ratio test (i.e.,2; difference between chi-square values), and non-nested models are compared using the Akaike information criteria (AIC) and th e Browne-Cudek criteria (BCC; Browne & Cudek, 1989). AIC and BCC are indexes based on the extent to which the model fit the data and also incorporate a penalty for mode l complexity (they are also interpreted as cross-validation coefficients). When the aim is to evaluate a final model for fit, the point estimate of the Root Mean Square E rror of Approximation (RMSEA) and a 90% confidence interval, Comparative Fit Index (C FI), and the Non-Normed Fit Index (NNFI) are used because of their relatively good pe rformance in simulation studies (e.g., Hu & Bentler, 1998). Several cut-off values are used to judge model-data fit: RMSEA < 0.05 suggest good fit, 0.05-0.08 suggest marginal f it, and > 0.10 suggest questionable fit, and a CFI > .95 indicates good f it (Hu & Bentler, 1998, 1999; MacCallum, Browne & Sugawara, 1996). Mediation/indirect effect s are based on the product of regression coefficients, with standard errors calculated based on the first-order delta method formula (Sobel, 1982) when there is one mediator, and the multivariate delta method (Taylor, Mackinnon, & Tein, in press) wh en there are two mediators.

PAGE 12

7 Results Attrition It is often of interest to compare participants that dropout with those that were retained in order to determine the extent to which the retained participants are representativeness of the full sample. Base d on the results in Table 1 there is an indication that the groups differ on several variables, and on at least one of the variables, general appearance, even after a strict correction of the nominal alpha based on multiple comparisons (Bonferroni adjustment of th e nominal alpha level =.003). This also suggests that the data may not be missing completely at random (MCAR)2 with respect to the means. More general tests of the MCAR assumption have been suggested (Kim & Bentler, 2002; Little, 19 88; Muthen et al., 1987).3 Given the simple pattern of missing data, a multi-group structural equation m odel was in used, in which unstructured variances, covariances, and m eans of all measured variables (excluding UV exposure behaviors) are constrai ned equal across completers and non-completers, 2(1595) = 2232.58, p < .05, CFI= .97, NNFI=.95, RMSEA=.026 (.023, .029). The result suggests that the constraint were reasonable, indicating the tw o groups come from a single population (Kim & Bentler, 2002), and in tu rn that the MCAR assumption cannot be rejected (but may still be fa lse). However, whether or not the MCAR assumption holds is somewhat immaterial because the approach ta ken to handling the missing data presumes the less stricter assumption that the data are missing at random (MAR). Assuming the data are MAR, parameter estimates will be consistent, asymptotically efficient, and asymptotically normal (Allison, 2001, 2003; Muthen et al., 1987). Outliers and Normality Initially, the data were screened for outli ers. Five outliers were identified, and analyses with and without these cases indicate d no substantive differences (cf. Tables 3 vs. 5, Tables 6 vs. 8, Figures 6 vs. 8) Mardia’s index of multivar iate kurtosis was 306.86, and the critical ratio was 34.62 for the model tested, suggesting the presence of multivariate non-normality.5 Examination of univariate skewness (SK) and kurtosis (KU) for the individual items associated with all factors except UV exposure variables indicated slight deviations from norma lity: SK-range .03-1.29, M = .51, SD = .31; KUrange .01-1.37, M = .84, SD =.34. Univariate indexes suggested more moderate deviations from normality for sunbathing intentions, SK = 1.24, KU = 1.47, sunbathing behaviors, SK = 1.34, KU = 2.48, tanning in tentions, SK 1.73, KU = 2.73, and indoor tanning behaviors, SK = 1.97, KU = 3.42. Given c oncerns related to normality, the UV exposure variables were transformed using a na tural log transformati on, which resulted in more normal distributions (all SK < 1.31 and KU < .69). Based on simulations studies (West, Finch, & Curran, 1995; Curran, We st, & Finch, 1996), the impact of nonnormality on the X2 statistic, fit indexes, and estimation of standard errors was deemed to be little. The means, standard deviations, internal consistency, and temporal stability estimates of the scales are reported in Ta ble 1. Correlations among unit weighted scale

PAGE 13

8 scores are provided in Table 2. A measurem ent model was estimated with all possible correlations among the factors presented in Figu re 1. The results suggest adequate fit (see Table 4), as well as provide an upper bound of f it for the subsequent models evaluated. Model Comparisons A central tenet of the theory of r easoned action adopted in the hypothesized model is that intentions mediate the relationship between appearance attitudes/sociocultural influence and UV expos ure behaviors. Alternatively, it might be posited that there is no mediating effect, and instead only a direct influence on behaviors (see Figure 2). A saturated model in which both sets of paths are pres ent (Table 4; Model 1a) was compared to a more constrained model in which intentions mediate the relationship between appearan ce attitudes/sociocultural influence and UV exposure behaviors (Model 1b). The non-sign ificant difference suggests that the constraints were reasonable. In contrast, there was a signifi cant difference between the saturated model (Model 1a) and two alternatives (Models 1c and 1d), in which the mediating role of intentions between appearance at titudes/sociocultural influen ce and intentions is ignored, suggesting that the constraint s were not reasonable. Anot her indication that treating intentions as mediators was more appropria te than modeling the direct influence on behaviors is suggested by AI C and BCC values that are substantially lower for model 1b versus either model 1c or 1d. The tripartite theory posits that appearance reasons to tan will mediate the relationship between sociocul tural influence and intentions. The alternative model considered here is based on the theory of r easoned action. If sociocul tural influences are viewed as subjective norms and appearance r easons to tan and not tan are regarded as attitudes toward the behavior, the theory of reasoned action would predict the same direct relationships with intentions to UV expose as the tripartite th eory, but instead of sociocultural influences having an indirect effect, it would be predicted to have a direct effect (see Figure 3). A saturated model in whic h both sets of paths are present (Table 4; Model 2a) was compared to a model that constr ained the direct influe nce of sociocultural influence on intentions to zero (Model 2b), with the non-si gnificant difference suggesting that the constraints were reasonable. In contrast, there was a significant difference between the saturated model (Model 2a) and an alternative (Models 2c) that constrains to zero the direct influence of sociocultural influence on appearance reasons to tan, suggesting that the constrai nts were not reasonable. Another indication of the appropriateness of only modeling the indirect influence of sociocultural influence on intentions to tan is suggested by AIC and BCC values that are s ubstantially lower for model 2b than model 2c. A direct relationship between perceived threat and UV intentions is hypothesized based on revised protection motivation theory (Rogers, 1983), as well as an indirect effect through appearance reasons to tan and appearance reasons not to tan, based on the model proposed by Fishbein (2000) and the re sults of a previous study (Jackson & Aiken, 2000) (see Figure 4). The hypothesized model is a saturated model in which both sets of paths are present (Table 4; Model 3a), whic h was compared to a model that constrained the direct influence of perceived threat on appearance reasons to tan and not tan to zero (Model 3b), with a significant difference indicating that the constraints were not appropriate. The non-significan t difference between the saturated model (Model 3a) and

PAGE 14

9 the alternative (Models 3c) that constrains the direct influence of perceived threat on intentions to zero, suggests th at the constraints were ap propriate. Lower AIC and BCC values for model 3c than mode l 3b is another indication that modeling only the indirect influence of perceived threat on inte ntions to tan is appropriate. Based on an extension of the health belief model (Jackson & Aiken, 2000), perceived susceptibility is hypothesized to mediate the relationship between skin cancer risk and intentions to sunbathe Alternatively, there may only be a direct influence of skin cancer risk on intentions to tan (see Figure 5) Whether skin cancer risk has a direct or indirect influence, or both, will be tested by comparing a saturated model in which both sets of paths are present to more constraine d models in which only one set of paths is present. A saturated model in which both sets of paths are present (Table 4; Model 4a) was compared to a model that constrained th e direct influence of skin cancer risk on intentions to zero (Model 4b) with the non-significant di fference suggesting that the constraints were reasonable. In contrast, th ere was a significant difference between the saturated model (Model 4a) and an alternative (Models 4c) th at constrains to zero the direct influence of skin cancer risk on pe rceived threat, suggesti ng that the constraints were not reasonable. Anothe r indication of the appropria teness of only modeling the indirect influence of skin cancer risk on in tentions to tan is suggested by AIC and BCC values that are substantially lowe r for model 4b than model 4c. Interpretation of the Final Model The final model selected was 4b. Standa rdized regression coefficients and significance tests of individual paths are presented in Figur e 7, with only one path not statistically significant. With the exceptions of direct influences of perceived threat on intentions (ruled out based on tests of alternative models) and the non-significant association between perceived threat and appe arance reasons to tan, the relationships are generally consistent with what was originally hypothesized (Figure 1). In this model, the R2 for sunbathing intentions is .21 and .55 for behaviors, while for indoor tanning intentions it is .24 and .46 for behaviors. Specific indirect e ffects in the model are also estimated and tested for significance based on th is model (Table 7). The results of these analyses suggest support for the mediating pa thways posited by the tripartite theory of body image and theory of reasoned action (entries 1, 4, 5, 6, 9, and 10), and partial support for mediating mechanisms posited by Fi shbein’s (2000) inte grative model (i.e., entries 3 and 8, but not 2 and 7). 6

PAGE 15

10 Discussion This study examined models that pros pectively predict s unbathing and indoor tanning behaviors using construc ts and hypothesized relationships derived from theories of body image and health behavior. The re sults generally support a model in which intentions mediate the relati onship between appearance attitu des and tanning behaviors, appearance reasons to tan and intentions me diate the relationship between sociocultural influences and tanning behaviors, and appear ance reasons not to tan and intentions mediate the role of perceived threat on beha viors. The implications of these findings on research designed to identify risk variable s and interventions ar e considered below. Implications on Theory The results suggest that intentions medi ate the relationship between appearance attitudes and tanning behaviors, which is cons istent with the theory of reasoned action (Ajzen & Fishbein, 1980). Although previous studies have identified a significant univariate association between intentions and behavior s (e.g., Jackson & Aiken, 2000; Hillhouse et al., 1997), such anal yses are unable to rule out alternative models in which attitudes have a direct influe nce on behaviors, precluding a more rigorous test of the theory of reasoned action undertaken in this study. Moreover, the simultaneous modeling of intentions and behaviors enables estimates a nd tests of the indirect effects of attitudes and social norms on UV exposure behaviors. A relatively strong direct relationship between appearance reasons to tan and intentions was observed, as well as an indir ect effect with behaviors to UV expose. This result is consistent with past research (Cafri, Thompson, Roehrig et al., 2006; Hillhouse et al., 1996; Hillhouse et al., 1997; Hillhous e, et al., 2000; Jackson & Aiken, 2000; Wichstrom, 1994), but as noted above, estima tion of the indirect effect of appearance reasons to tan on UV exposure be haviors (via intentions) repr esents a unique contribution to the existing literature. Mo reover, appearance reasons to tan and intentions were found to mediate the relationship between socioc ultural influences and tanning behaviors, which is consistent with the tripartite theo ry of body image (Thompson et al., 1999). This outcome is also in line with the resu lts of a previous study (Cafri, Thompson, & Jacobsen, 2006), but use of a more comprehens ive model, indicates a result with greater validity. A modest direct relationship betwee n appearance reasons not to tan and intentions was observed, as well as an indi rect effect with behaviors. The direct relationship with intentions is consistent with the resu lts of the only other study to examine appearance reasons not to tan (Cafri, Thompson, Roehrig et al., 2006), and estimation of the indirect effect of a ppearance reasons not to tan on UV exposure behaviors (via intenti ons) is a novel result. Furthermore, appearance reasons not to tan and intentions were found to me diate the role of perceived th reat on behaviors. This is only partially supports the appli cation of Fishbein’s (2000) mode l in this context, because the appearance reasons to tan factor was not found to be a mediator.

PAGE 16

11 Implications for the Design of Interventions One of the major implications of the resu lts on the design of interventions is that components should be designed to target a ppearance reasons to tan and sociocultural influences because existing appearance-based interventions target only appearance reasons not to tan (Gibbons et al., 2005; Hillhouse & Turrisi, 2002; Jackson & Aiken, 2006; Jones & Leary, 1994; Mahler et al., 2003; Mahler et al., 2005). For instance, reduction of the positive valuation of a tan appearance could be achieved through a cognitive dissonance approach in which pe ople who tan are asked to challenge their idealization of a tan appearan ce. An indication of the pros pective utility of such an approach in decreasing UV expos ure behaviors is that dissona nce interventions have been found in several controlled investigations to reduce body dissatisfaction and eating disturbances in samples at-risk for developi ng an eating disorder (e.g., Stice, Trost, & Chase, 2003). Notably, one multi-component intervention has manipulated perceived media influence through emphasizing a growing trend toward untanned skin tone in the media, with evidence of program efficacy (Jackson & Aiken, 2006). Certainly, other sociocultural influences, such as peer nor ms for a tan appearance, should also be considered as targets in future interventions In light of the subs tantial direct effect sociocultural influences have on appearance reasons to tan and indirect effects on behaviors to UV expose, and considering the significant role chil dhood/ adolescence play in attitude formation, attempting to change perceived sociocultural norms at a young age as part of an early inte rvention program could substantially reduce UV exposure behaviors later on in life. A slightly different approach to the design of interventions than the one considered above, is to focus on behavioral al ternatives to UV expose (Turrisi, Hillhouse & Gebert, 1998), such as the use of sunl ess tanning products. One study distributed sunless tanning samplers as an adjunc t to an existing intervention, but found no significant differences with an intervention alone conditi on, although these significance tests were characterized by low statistical power (Mahler et al., 2005). Moreover, caution should be exercised because the use of s unless tanning products may perpetuate an appearance norm that in the long run leads to more UV exposure. The observed associations between appearance reasons not to tan and intentions/behaviors to UV expose are consiste nt with the efficacy of interventions that target this construct (Gibbons et al., 2005; Hillhouse & Turrisi, 2002; Jackson & Aiken, 2006; Jones & Leary, 1994; Mahler et al., 2003 ; Mahler et al., 2005). Moreover, given the moderate indirect effect of perceived th reat on tanning behavior s, existing and future interventions should consider coupling a comp onent that manipulates perceived threat (i.e., susceptibility and severity) with a ppearance reasons not to tan. At least one intervention has done this, with evidence of program efficacy (Jackson & Aiken, 2006). Consistent with earlier arguments regarding the targeting of perceived sociocultural influences as part of an early intervention program, emphasizing the threat of skin cancer at a young age may lead to growth in appearan ce reasons not to tan, which in turn might reduce UV exposure behaviors later on in life. Several limitations of this study should be considered. First, with respect to external validity, the exclus ive use of female college students who are predominantly Caucasian limits the extent to which these findings can be generalized. It is however

PAGE 17

12 important to recognize that adolescent a nd young adult Caucasian females represent a very high-risk group, both in terms of skin-type susceptibility and use of behaviors that lead to skin cancer (Davis et al., 2002; Demko et al ., 2003; Lazovich et al., 2004). Nevertheless, it would be important for future studies to utilize sampling procedures that are more inclusive of gender, ethnicity, ag e, and level of edu cation. A second limitation to consider is that only a limited number of constructs were evaluated from health behavior theories. For instance, perceived be havioral control was not evaluated, which is a construct that is part of an updated versi on of the theory of r easoned action, the theory of planned behavior (Ajzen, 1985) However, such a variable a ppears to be more relevant to protective health behaviors, therefore its substantive contribution to examining models of UV exposure and the design of interventions is not entirely clear. Other variables may be more relevant, such as barriers to UV exposure (health belief model; Rosenstock, 1974). A third limitation to consider is that the relationships among c onstructs posited in the tested models are based on presumed cau sal relationships. An important area for future research is experimental studies investigating relations hips among psychosocial constructs and UV exposure variables, such as single component interventions, which would not only provide evidence of causality, bu t also efficacy of individual components. A fourth limitation is that skin cancer risk wa s modeled in term of frequency of exposure. To the extent that risk depends on other factors, such as length of exposure and protection, the extent of the risk will be in accurate. Moreover, the use of self-report as opposed to more objective measures of UV exposu re, such as skin reflectance or personal dosimetry (Glanz & Mayer, 2005), should also be considered a limitation to quantifying risk. Finally, missing data due to attrition should be considered a potential limitation. If the missing data do not meet the MAR assumpti on presumed in this study, bias parameter estimates are likely. However, it has been sugge sted that the amount of bias will be less under direct maximum likelihood (the approach taken in this study) than other missing data treatments, such as listwise deletion (M uthen et al., 1987). Fu ture research should work towards developing a better understandi ng and prevention of behaviors that place people at risk for developing skin-cancer.

PAGE 18

13 Footnotes 1One reviewer raised the issue of item bias in the subscales related to evaluating a tan appearance, such that item wordings cont aminated the construct of interest with intentions, leading to a possibl e upward bias in the association between these subscales and intentions. To evaluate this possibilit y, new subscales were constructed by deleting items with potential bias (6 items delete d from the general factor: 2,3,4,6,8,9; 1 item deleted from the acne factor: 13; 2 items de leted from the aging factor: 31, 32; 2 items deleted from the media factor: 34, 39; 1 item de leted from the family factor: 47) (for the items see appendix). Next, the correlations be tween each revised subscale and intentions to sunbathe and indoor tan were evaluated. Thes e were compared to the same correlations using the subscales without deleted items. The differences between these correlation coefficients for sunbathing (left of slash) and indoor tanning (ri ght of slash) were: .039*/.019, .019*/.019*, .015/.032, .002/.013*, and .005/.002 for general, acne, aging, media, and family, respectively (* indicates p < .05 for a test of difference of dependent correlation coefficients; Chen & Popovich, 2002). Despite the significance in some cases, the magnitude of these differences is quite sma ll, suggesting little bias if bias is indeed the cause of the difference in the correlations (smaller co rrelations could be due to decreased validity resulting from item deletio n). Nevertheless, anal yses were conducted both including and excluding the specified item s, with virtually identical results for comparable indexes and results of significan ce tests (cf. Tables 3 vs. 4, Tables 6 vs. 7, Figures 6 vs. 7). Interpretation of results is based on results excludi ng the specified items. 2 Missing at random (MAR) is an assumption that the probability of missing data on the variable of interest is not re lated to scores on that variable controlling for other variables in the model (Little & Rubin, 1987). Missing completely at ra ndom (MCAR) is a stricter assumption that requires the probability of missi ng data to be unrelat ed to any variables in the model (Little & Rubin, 1987). 3 Strictly speaking, these are not tests of the MCAR assu mption, they are tests of homogeneity, which can be used to falsif y the MCAR assumption (Kim & Bentler, 2002). That is, if the results s uggest that the groups are not homogeneous, this implies a rejection of MCAR. However, acceptance of homogeneity does not prove that the data are MCAR. 4 Unreliability was based on one minus the estimat e of its test-retest reliability (see Table 1). 5 Results are based on li stwise deleted data. 6 If a Bonferroni adjustment to the nominal alpha is based on the number of tests for both the direct and indirect ef fects (.05/20 = .0025), the deci sions regarding statistical significance are identical. If a B onferroni adjustment to the nominal alpha is based on the number of tests for the alternative models direct, and indirect effects (.05/29 = .0017), the decisions regarding statistical significan ce are the same except entry 3 in Table 7. However, a Bonferroni adjustment is overly strict when the tests are dependent. If even a

PAGE 19

14 relatively small amount of dependency is assu med (average r = .20) when calculating the nominal alpha for the 29 test s, the result would be .003 us ing the Dubey and ArmitageParmar method (Sankoh, Huque, & Dubey, 1997) and entry 3 in Table 7 would be significant.

PAGE 20

15 References Allison, P.D. (2001). Missing data Thousand Oaks, CA: Sage Publications. Allison, P.D.(2003). Missing data techni ques for structural equation modeling. Journal of Abnormal Psychology, 112 545-557. Ajzen, I. (1985). From intentions to actions: A theory of planned behavior. In J. Kuhl & J. Beckman (Eds.), Action-control: From cognition to behavior (pp. 1139). Heidelberg, Germany: Springer. Ajzen I, Fishbein M. (1980). Understanding Attitudes and Pr edicting Social Behavior Englewood Cliffs, NJ: Prentice-Hall. American Cancer Society (2007). Cancer facts and figures Atlanta, GA: Author. Arbuckle, J.L. (2005). Amos 6.0 User’s Guide Spring House, PA; Amos Development Corporation. Arbuckle, J.L. (1996). Full information estimatio n in the presence of incomplete data. In G.A. Marcoulides and R.E. Schumacker (Ed.s), Advanced Structural Equation Modeling (pp. 243277). Mahweh, NJ: Lawrence Erlbaum Associates. Bollen, K., & Lennox, R. (1991). Conventional wisdom on measurement: A structural equation perspective. Psychological Bulletin, 110 305-314. Browne, M.W. & Cudeck, R. (1987). Single sample cross-validation indices of covariance structures. Multivariate Behavioral Research, 24 445-455. Cafri, G., Thompson, J. K., & Jacobsen, P. (2006). Appearance reasons for tanning mediate the relationship between me dia influence and UV exposure and protection. Archives of Dermatology, 142, 1067-1069 Cafri, G., Thompson, J.K., Roehrig, M., Rojas, A., Sperry, S., Jacobsen, P., Hillhouse, J.J. (in press). Appearance Motives to Tan and Not Tan: Evidence for Validity and Reliability of A New Scale. Annals of Behavioral Medicine

PAGE 21

16 Cafri, G., Thompson, J.K., Roehrig, M., van den Berg, P., Jacobsen, P.B. & Stark, S. (2006). An Investigation of Appearance Motives for Tanning: The Development and Evaluation of the Physical Appear ance Reasons for Tanning Scale (PARTS) and Its Relation to Sunbathing and Indoor Tanning Intentions. Body Image 3 199-209. Chen, P.Y., & Popovich, P.M. (2002). Corr elation: Parametric and nonparametric measures. Thousand Oaks, CA: Sage. Curran, P.J., West, S.G., & Finch, J.F. (1996) The robustness of test statistics to nonnormality and specification error in confirmatory factor analysis. Psychological Methods, 1 16-29. Davis, K.J., Cokkinides, V.E., Wein stock, M.A., O’Connell, M.C., & Wingo, P.A. (2002). Summer sunburn and sun expos ure among U.S. youths ages 11 to 18: National prevalence and associated factors. Pediatrics, 110 27-35. Demko, C.A.; Borawski, E.A., Debanne, S.M., Cooper K.D., Stange, K.C. (2003). Use of indoor tanning facilities by white adolescents in the United States. Archives of Pediatric and Adolescent Medicine, 157 854-860. Fishbein, M. (2000). The role of theory in HIV prevention. Aids Care, 12 273-278. Fitzpatrick, T.B. (1988). The validity and practic ality of sun-reactive sk in types I through VI. Archives of Dermatology 124, 869-871. Gibbons, F.X., Gerrard, M., Lane, D.J., Mahler H.I.M., & Kulik, J.A. (2005). Using UV photography to reduce use of tanning boot hs: A test of cognitive mediation Health Psychology, 24, 358 363 Glanz, K. & Mayer, J.A. (2005). Reducing ultr aviolet radiation exposur e to prevent skin cancer: Methodology and measurement. Am erican Journal of Preventive Medicine, 29, 131-142. Goldsmith, M.F. (1987). Paler is be tter, say skin cancer fighters. Journal of the American Medical Association, 257 893–894. Hillhouse, J.J., Adler, C.M., Drinnon, J., & Tu rrisi, R. (1997). App lication of Azjen’s theory of planned behavior to pred ict sunbathing, tanning salon use, and sunscreen use intentions and behaviors. Journal of Behavioral Medicine, 20 365378. Hillhouse, J.J., Stair, A.W., & Adler, C. M. (1996). Predictors of sunbathing and sunscreen use in college undergraduates. Journal of Behavioral Medicine, 19 543-560.

PAGE 22

17 Hillhouse, J.J. & Turrisi, R. (2002). An examination of the efficacy of an appearancefocused intervention to reduce UV exposure. Journal of Behavioral Medicine, 25 395-409. Hillhouse, Turrisi, Holwiski & McVeigh, ( 1999). An examination of psychological variables relevant to arti ficial tanning tendencies. Journal of Health Psychology, 4 507-516. Hillhouse, J.J., Turrisi, R., Kastner, M. (2000). Modeling Tanning Salon Behavioral Tendencies Using Appearance Motivation, Self-Monitoring, and the Theory of Planned Behavior. Health Education Review, 15 405– 414. Hu, L. & Bentler, P. M. (1998) Fit indices in cova riance structure modeling: Sensitivity to underparameterized model misspecification. Psychological Methods, 3 424453. Hu, L., & Bentler, P. M. (1999). Cutoff criteri a for fit indexes in c ovariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling 6 1-55. Jackson, K.M. & Aiken, L. S. (2000). A psychosocial model of sun protection and sunbathing in young women: The impact of health beliefs, attitudes, norms, and self-efficacy for sun protection. Health Psychology, 19 469-478. Jones, J.L., & Leary, M.R. (1994). Effects of appearance-based admonitions against sun exposure on tanning intentions in young adults Health Psychology, 13 86-90. Kim, K.H., Bentler, P.M. (2002). Tests of hom ogeneity of means and covariance matrices for multivariate incomplete data. Psychometrika, 67 609-624. Lazovich, D., Forster, J., Sore nsen, G., Emmons, K., Stryker, J., Demierre, M.F., Hickle, A.H., Remba, N. (2004). Char acteristics associated with use or intention to use indoor tanning among adolescents. Archives of Pediatrics & Adolescent Medicine, 158 918-924. Little, R.J.A. (1987). A test of missing comple tely at random for multivariate data with missing values. Journal of the American St atistical Association, 83 1198-1202. Little, R.J.A., Rubin, D.B. (1987). Statistical Analysis with Missing Data New York: Wiley & Sons. MacCallum, R.C., Browne, M.W., & Sugawara, H.M. (1996). Power analysis and determination of sample size fo r covariance structural modeling. Psychological Methods, 1 130-149.

PAGE 23

18 Mahler, H.I., Kulik, J.A., Gibbons, F.X., Gerrard M. & Harrell, J. (2003). Effects of appearance-based interventi ons on sun protection inte ntions and self-reported behaviors. Health Psychology, 22 199-209. Mahler, H.I.M, Kulik, J.A, Harrell, J., Correa, A., Gibbons, F. X. & Gerrard, M. (2005). Effects of UV photographs, photoaging info rmation, and use of sunless tanning lotion on sun protection behaviors. Archives of Dermatology 141, 373-380. Muthen, B., Kaplan, D., & Hollis, M. (1987). On structural equation modeling with data that are not missing completely at random. Psychometrika, 52 431-462. Rogers, R.W. (1983) A protection motivation theory of fear a ppeals and attitude change: A revised theory of protection motivation. In J.R. Cacioppo & R.E. Petty (Eds.) Social psychology: A sourcebook (p p.153-176). New York: Guilford press. Rosenstock, I.M. (1974). The health belief model and preventive health behavior. Health education monographs, 2 354-386. Sankoh, A.J., Huque, M.F., & Dubey, S.D. (1997). Some comments on frequently used multiple endpoint adjustment methods in clinical trials. Statistics in Medicine, 16 2529-2542. Sobel, M.E. (1982). Asymptotic confidence inte rvals for indirect eff ects in structural equation models. In S. Leinhardt (Ed.), Sociological Methodology (pp.290-312). Washington, DC: American Sociological Association. Stice, E., Trost, A., & Chase, A. (2003). Healthy weight control and dissonance-based eating disorder prevention programs: Results from a controlled trial. International Journal of Eating Disorders, 33 10-21. Taylor, A.B., Mackinnon, D.P., Tein, J. (in pr ess). Tests of the three-path mediated effect. Organizational Research Methods. Thompson, J.K., Heinberg, L., Altabe, M., & Tantleff-Dunn, S. (1999). Exacting Beauty Washington, DC; American Psychological Association. Turrisi, R. Hillhouse, J. & Gebert, C. (1998). Ex amination of cognitive variables relevant to sunbathing. Journal of Behavioral Medicine, 21 299-313. U.S. Department of Health and Human Servic es. Ultraviolet radiati on-related exposures: broad spectrum ultraviolet (UV) radia tion, UVA, UVB, UVC, solar radiation, and exposure to sunlamps and sunbeds. In the Tenth Annual Report on Carcinogens (2002) Retrieved August 27th 2004 from http://ehis.niehs.nih.gov/roc/ tenth/profiles/s183uvrr.pdf

PAGE 24

19 Weinstock, M.A. (1992). Assessment of sun sensitivity by questionnaire: Validity of items and formulation of a prediction rule. Journal of Clinical Epidemiology, 45 547-552. West, S.G., Finch, J.F., & Curran, P.J. ( 1995). Structural equation modeling with nonnormal variables: Problems and remedies. In R.H. Hoyle (Ed.). Structural equation modeling: Concepts issues and applications (pp.56-75). Thousand Oaks, CA: Sage. Wichstrom, L. (1994). Predictors of Norweg ian adolescents’ sunbathing and use of sunscreen. Health Psychology 13, 412-4.

PAGE 25

20 Appendices

PAGE 26

21 Appendix A: Scale Items APPEARANCE REASONS TO TAN General 1. Having a tan gives me more sex appeal. 2. I tan because it makes me more attractive. 3. I tan because it makes me look better. 4. I tan because it makes me more confident in my appearance. 5. I feel more confident in my appearance when I am tan. 6. I tan before a big social event becaus e it makes me feel more attractive. 7. The tanner I am, the more attractive I feel. 8. I tan to avoid looking pale. 9. I tan because it adds a ni ce glow to my appearance. Acne 11. When I am tan, I feel less concer ned about the appearance of acne. 12. The less tan I am the more I’m worried about my acne showing. 13. I tan because it helps reduce the amount of acne on my face and body. 14. Tan skin helps me cover up acne-related scars. Body shape 16. I look like I have less fat on my body when I am tan. 17. The more tan I am the more physically fit I look. 19. A tan gives my body the appearan ce of having more muscle tone. 20. A tan helps me look like I’m in good physical shape. 23. I look slimmer with a tan. 24. Being tan conceals my app earance of stretch marks. APPEARANCE REASONS NOT TO TAN Immediate Skin Damage 25. I’m concerned about getting blemis hed skin as a result of tanning. 26. I’m concerned about freckling from tanning. 27. The appearance of a sunburn makes me look unattractive. 28. Getting sunspots worries me. 29. I’m concerned about my skin peeling after too much tanning. 30. I’m concerned about the appearance of rough or leathery skin from tanning. Aging 31. I don’t tan as much as I would like because I’m worried about premature skin aging. 32. I don’t tan because it will age my skin quicker. 33. I’m hesitant to tan because it will wrinkle my skin.

PAGE 27

22 Appendix A: (Continued) SOCIOCULTURAL INFLUENCES Media 34. I try to have a tan like fam ous people I see in magazines. 35. I wish I was as tan as celebrities in the media. 36. I want to be as tan as TV stars. 37. I wish I had a tan like people on TV. 38. I want to be as tan as people in magazines. 39. I try to be as tan as people in movies. 41. I would like my skin tone to be darker like people in TV and movies. Friends 42. I like to be as tan as my friends. 43. Positive appearance comments from my friends make me want to tan more. 44. I receive negative appearance comments from my friends when I am not tan. 45. My friends say I look good when I am tan. Family 46. I want a tan because people in my family think it makes my skin look nice. 47. I try to get a tan because my fa mily members say it is attractive. 48. I want to be tan because my family me mbers think it makes me look healthier. Significant Other 49. My boyfriend/girlfriend likes the way I look when I am tan. 50. Comments about my appearance from my boyfriend/girlfriend encourage me to tan. PERCEIVED THREAT Perceived Susceptibility 51. If you DON'T use sun protection, how sus ceptible do you feel you are to skin cancer? 52. The possibility of skin cancer worries me. 53. Whenever I hear of a friend or relative (o r public figure) getting skin cancer, it makes me realize that I could get it too. 54. I don't need to worry about getting skin cancer until I am much older. Perceived Severity 55. It would be terrible to get a malignant tumor on my skin. 56. Getting skin cancer would severely affect my life. 57. It would be terrible to have skin cancer. TANNING INTENTIONS AND BEHAVIORS 58. Please give me your best estimate of how many times you have indoor tanned in the last 6 months 59. Please give me your best estimate of how many times you have sunbathed in the last 6 months 60. Please give me your best estimate of how many times you plan to use an indoor tanning salon in the next 6 months

PAGE 28

23 Appendix A: (Continued) 61. Please give me your best estim ate of how many times you plan to sunbathe in the next 6 months SKIN CANCER RISK (SKIN TYPE) 62. If you were to lie in the sun for one hour UNPROTECTED (no sunscreen, no protective clothing, etc.) in the early su mmer when you had NO tan, your skin would: (mark the best answer) 63. What is the color of your untanned skin?

PAGE 29

24 Appendix B: Tables and Figures Table 1. Means, Standard Deviations, and Reliability of Scales

PAGE 30

25 Appendix B: (Continued) Attrition Analyses Completers vs. Non-Completersa Factor Items Mean SD TestRetest r tvalue pvalue d Appearance Reasons to Tan 19 3.30 .96 .96 .91 General Attractiveness 9 3.60 1.06 .95 .94 3.47 .001 .28 Acne 4 2.77 1.18 .91 .80 1.47 .143 .12 Body shape 6 3.23 1.11 .92 .88 .884 .377 .07 Appearance Reasons Not to Tan 9 3.14 .81 .82 .86 Immediate Skin Damage 6 3.19 .83 .83 .87 1.19 .233 .10 Skin Aging 3 3.05 1.06 .73 .75 .036 .972 .00 Sociocultural Influences to Tan 16 2.58 .91 .94 .93 Media 7 2.31 1.14 .97 .89 1.14 .253 .09 Family 3 2.39 1.15 .77 .81 2.12 .035 .17 Friends 4 2.96 0.95 .91 .88 2.49 .013 .21 Significant Other 2 3.10 1.19 .86 .85 1.94 .053 .16 Perceived Susceptibility-Skin Cancer 4 3.46 .50 .71 .82 1.15 .251 .09 Perceived SeveritySkin Cancer 3 4.62 .66 .89 .79 2.82 .005 .23 Skin Cancer RiskFitzpatrick Skin Type b 1 2.76 .97 .92 1.28 .202 .11 Skin Cancer RiskUntanned Skin b 1 1.47 .53 .68 Indoor Tanning Intentions (6 month frequency)c 1 1.99 / 7.86 1.33 / 15.31 .92 .57 .570 .05 Sunbathing Intentions (6 month frequency) c 1 2.40 / 10.43 1.17 / 11.07 .89 .61 .541 .05 Indoor Tanning Behaviors (6 month frequency) c 1 1.83 / 6.57 1.37 / 12.65 .96 Sunbathing Behaviors (6 month frequency) c 1 2.38 / 9.93 1.02 / 9.49 .80

PAGE 31

26 Appendix B: (Continued) Note: Means and standard devia tion values for multi-item Like rt scales expressed as item averages. a The direction of the results for all but indoor tanning intentions is such that completers had higher scores than non-completers. b Internal consistency for the two skin cancer risk items is .62. Attrition analyses ba sed on a single variable consisting of the two skin cancer risk items recoded (such that highe r scores indicate more risk), standardized, then summed. c For the UV exposure items, values to th e left of the slas h are based on the response scale, values to the right are in frequency units. Fr equency units were obtained by using the midpoint of the interval specif ied for each response value. For sunbathing intentions, 82.7% of the sample planned to sunbathe and 52. 8 % planned to indoor tan at least once during the upcoming 6 month period. Among the individuals followed up after the 6 months, 86.7% of the sample repo rted sunbathing and 39.3 % reported indoor tanning at least once.

PAGE 32

27 Appendix B: (Continued) Table 2. Correlations Among Scales Note: Skin cancer risk variable is as describe d in the Note in Table 1. Perceived threat is a composite based on unit weights for perceive d susceptibility and severity subscales. 1 2 3 4 5 6 7 8 9 1. Appearance Reasons to Tan 2. Appearance Reasons Not to Tan .08 3. Sociocultural Influences to Tan .74 .09 4. Perceived Threat .12 .44 .01 5. Skin Cancer Risk -.01 .24 .04 .17 6. Indoor Tanning Intentions .35 -.13 .30 -.08 -.04 7. Sunbathing Intentions .32 -.15 .21 .04 -.12 .20 8. Indoor Tanning Behaviors .26 -.14 .22 -.10 -.09 .61 .21 9. Sunbathing Behaviors .23 -.15 .12 -.01 -.09 .09 .62 .14

PAGE 33

28 Appendix B: (Continued) Table 3. Tests of Alternative Models (without item deletion) Model Description 2 df 2a df a AIC BCC Measurement Model 3294.61 1499 3716.613762.19 Role of Intentions and Behaviors 1a Figure 1 + Figure 2 b 3344.51 1517 3730.513772.20 1b Figure 1 3350.32 1523 5.81 6 3724.323764.71 1c Figure 2 (excluding dashed paths) 3819.64 1525 475.13* 8 4189.644229.60 1d Figure 2 (including dashed paths) 3587.67 1523 243.16* 6 3961.674002.07 Role of Sociocultural Influences 2a Figure 1+ Figure 3 b 3345.98 1521 3723.983764.80 2b Figure 1 3350.32 1523 4.34 2 3724.323764.71 2c Figure 3 4036.35 1522 686.03* 1 4412.354452.96 Role of Perceived Susceptibility 3a Figure 1 3350.32 1523 3724.323764.71 3b Figure 4 (excluding light dashed paths) 3451.41 1525 101.09* 2 3821.413861.37 3c Figure 4 (excluding heavy dashed paths) 3359.32 1525 9.00 2 3729.323769.30 Role of Skin Cancer Risk 4a Model 3c + Figure 5 b 3358.38 1523 3732.383772.78 4b Model 3c 3359.32 1525 .94 2 3729.323769.30 4c Figure 5 3389.23 1524 30.85* 1 3761.233801.41 Indicates the value is significant at p < .0056 (Bonferroni adjustment = .05/9). AIC= Akaike information criterion. BCC= Browne-Cudeck criterion. a All comparisons of nested models are based on comparing satu rated models, “a” model in a given number class, with more constrained models “b-d” in the same number class. bIn 1a, 2a and 4a, the models consist of the model to left of the plus sign and the non-overlapping paths from the model to the right of the plus sign.

PAGE 34

29 Appendix B: (Continued) Table 4. Tests of Alternative Models (with item deletion) Indicates the value is significant at p < .0056 (Bonferroni adjustment = .05/9). AIC= Akaike information criterion. BCC= Browne-Cudeck criterion. a All comparisons of nested models are based on comparing satu rated models, “a” model in a given number class, with more constrained models “b-d” in the same number class. bIn 1a, 2a and 4a, the models consist of the model to left of the plus sign and the non-overlapping paths from the model to the right of the plus sign. Model Description 2 df 2a df a AIC BCC Measurement Model 2099.99 907 2445.992474.99 Role of Intentions and Behaviors 1a Figure 1 + Figure 2 b 2146.55 924 2458.552484.69 1b Figure 1 2152.17 930 5.62 6 2452.172477.31 1c Figure 2 (excluding dashed paths) 2607.88 932 461.33*8 2903.882928.68 1d Figure 2 (including dashed paths) 2369.54 930 222.99*6 2669.542694.67 Role of Sociocultural Influences 2a Figure 1+ Figure 3 b 2149.56 928 2453.562479.04 2b Figure 1 2152.17 930 2.61 2 2452.172477.31 2c Figure 3 2810.03 929 660.47*1 3112.033137.33 Role of Perceived Susceptibility 3a Figure 1 2152.17 930 2452.172477.31 3b Figure 4 (excluding light dashed paths) 2239.07 932 86.9* 2 2535.072559.88 3c Figure 4 (excluding heavy dashed paths) 2159.01 932 6.84 2 2455.012479.81 Role of Skin Cancer Risk 4a Model 3c + Figure 5 b 2158.01 930 2458.012483.14 4b Model 3c 2159.01 932 1 2 2455.012479.81 4c Figure 5 2184.46 931 26.45* 1 2482.462507.43

PAGE 35

30 Appendix B: (Continued) Table 5. Tests of Alternative Models (outliers deleted) Indicates the value is significant at p < .0056 (Bonferroni adjustment = .05/9). AIC= Akaike information criterion. BCC= Browne-Cudeck criterion. a All comparisons of nested models are based on comparing satu rated models, “a” model in a given number class, with more constrained models “b-d” in the same number class. bIn 1a, 2a and 4a, the models consist of the model to left of the plus sign and the non-overlapping paths from the model to the right of the plus sign. Model Description 2 df 2a df a AIC BCC Measurement Model 3284.65 1499 3706.65 3752.66 Role of Intentions and Behaviors 1a Figure 1 + Figure 2 b 3335.50 1517 3721.50 3763.59 1b Figure 1 3341.19 1523 5.69 6 3715.19 3755.97 1c Figure 2 (excluding dashed paths) 3805.87 1525 470.37 8 4175.87 4216.21 1d Figure 2 (including dashed paths) 3578.36 1523 242.86 6 3952.36 3993.14 Role of Sociocultural Influences 2a Figure 1+ Figure 3 b 3336.87 1521 3714.87 3756.08 2b Figure 1 3341.19 1523 4.32 2 3715.19 3755.97 2c Figure 3 4020.07 1522 683.2* 1 4396.07 4437.06 Role of Perceived Susceptibility 3a Figure 1 3341.19 1523 3715.19 3755.97 3b Figure 4 (excluding light dashed paths) 3439.13 1525 97.94* 2 3809.13 3849.47 3c Figure 4 (excluding heavy dashed paths) 3350.77 1525 9.58 2 3720.77 3761.11 Role of Skin Cancer Risk 4a Model 3c + Figure 5 b 3349.45 1523 3723.45 3764.22 4b Model 3c 3350.77 1525 1.32 2 3720.77 3761.11 4c Figure 5 3380.95 1524 31.5* 1 3752.95 3793.51

PAGE 36

31 Appendix B: (Continued) Table 6. Tests of Indirect E ffects (without item deletion) Origin Standardized Indirect Effect Unstandardized Indirect Effect SE Z p Sunbathing Behaviors 1. Sociocultural Influences to Tan (3 paths) 0.28 0.11 0.02 5.34 <.001* 2. Perceived Threat (3 paths) a 0.01 0.00 0.01 0.69 .315 3. Perceived Threat (3 paths) b 0.11 0.07 0.02 3.60 <.001* 4. Appearance Reasons to Tan (2 paths) 0.31 0.09 0.01 7.81 <.001* 5. Appearance Reasons Not to Tan (2 paths) 0.15 0.05 0.01 4.32 <.001* Indoor Tanning Behaviors 6. Sociocultural Influences to Tan (3 paths) 0.28 0.16 0.02 6.99 <.001* 7. Perceived Threat (3 paths) a 0.01 0.01 0.01 0.69 .314 8. Perceived Threat (3 paths) b 0.11 0.10 0.03 3.81 <.001* 9. Appearance Reasons to Tan (2 paths) 0.30 0.12 0.01 8.32 <.001* 10. Appearance Reasons Not to Tan (2 paths) 0.15 0.07 0.01 4.69 <.001* Significant after a Bonferroni adjustme nt to the nominal alpha (.05/10 = .005). a Through appearance reasons to tan. b Through appearance reasons not to tan.

PAGE 37

32 Appendix B: (Continued) Table 7. Tests of Indirect Effects (with item deletion) Origin Standardized Indirect Effect Unstandardized Indirect Effect SE Z p Sunbathing Behaviors 1. Sociocultural Influences to Tan (3 paths) 0.28 0.11 0.02 5.28 <.001* 2. Perceived Threat (3 paths) a 0.01 0.01 0.01 1.41 .147 3. Perceived Threat (3 paths) b 0.09 0.06 0.02 3.20 .002* 4. Appearance Reasons to Tan (2 paths) 0.30 0.08 0.01 7.67 <.001* 5. Appearance Reasons Not to Tan (2 paths) 0.14 0.07 0.02 3.86 <.001* Indoor Tanning Behaviors 6. Sociocultural Influences to Tan (3 paths) 0.27 0.15 0.02 6.86 <.001* 7. Perceived Threat (3 paths) a 0.01 0.01 0.01 1.41 .147 8. Perceived Threat (3 paths) b 0.10 0.09 0.03 3.43 .001* 9. Appearance Reasons to Tan (2 paths) 0.29 0.11 0.01 8.12 <.001* 10. Appearance Reasons Not to Tan (2 paths) 0.15 0.10 0.02 4.28 <.001* Significant after a Bonferroni adjustme nt to the nominal alpha (.05/10 = .005). a Through appearance reasons to tan. b Through appearance reasons not to tan.

PAGE 38

33 Appendix B: (Continued) Table 8. Tests of Indirect E ffects (with outliers deleted) Origin Standardized Indirect Effect Unstandardized Indirect Effect SE Z p Sunbathing Behaviors 1. Sociocultural Influences to Tan (3 paths) 0.29 0.11 0.02 5.35 <.001* 2. Perceived Threat (3 paths) a 0.01 0.00 0.01 0.73 .305 3. Perceived Threat (3 paths) b 0.11 0.07 0.02 3.55 <.001* 4. Appearance Reasons to Tan (2 paths) 0.31 0.09 0.01 7.86 <.001* 5. Appearance Reasons Not to Tan (2 paths) 0.15 0.05 0.01 4.25 <.001* Indoor Tanning Behaviors 6. Sociocultural Influences to Tan (3 paths) 0.27 0.16 0.02 6.92 <.001* 7. Perceived Threat (3 paths) a 0.01 0.01 0.01 0.73 .305 8. Perceived Threat (3 paths) b 0.11 0.10 0.03 3.80 <.001* 9. Appearance Reasons to Tan (2 paths) 0.29 0.12 0.01 8.22 <.001* 10. Appearance Reasons Not to Tan (2 paths) 0.15 0.07 0.01 4.71 <.001* Significant after a Bonferroni adjustme nt to the nominal alpha (.05/10 = .005). a Through appearance reasons to tan. b Through appearance reasons not to tan

PAGE 39

34 Appendix B: (Continued) Figure 1. Hypothesized Model

PAGE 40

35 Appendix B: (Continued) Figure 2. Alternative Models for the Role of Intentions and Behaviors

PAGE 41

36 Appendix B: (Continued) Figure 3. Alternative Model for the Role of Sociocultural Influence

PAGE 42

37 Appendix B: (Continued) Figure 4. Alternative Models for the Role of Perceived Threat

PAGE 43

38 Appendix B: (Continued) Figure 5. Alternative Model for th e Role of Skin Cancer Risk

PAGE 44

39 Appendix B: (Continued) Figure 6. Standardized Path Coefficients for Final Model (without item deletion) Significant after a Bonferroni adjustme nt to the nominal alpha (.05/10 = .005)

PAGE 45

40 Appendix B: (Continued) Figure 7. Standardized Path Coefficients for Final Model (with item deletion) Significant after a Bonferroni adjustme nt to the nominal alpha (.05/10 = .005)

PAGE 46

41 Appendix B: (Continued) Figure 8. Standardized Path Coefficients for Final Model (with outlier deletion) Significant after a Bonferroni adjustme nt to the nominal alpha (.05/10 = .005)

PAGE 47

About the Author Guy Cafri received his B.A. in Psychology from Macalester College and M.A. in clinical psychology from the University of S outh Florida. He has published in the areas of: body dysmorphic disorder, steroid use, eatin g disorders, body image, and skin cancer prevention. In addition to these areas, he also has an interest in applied statistics.