Identifying Cluster Subtypes for Acquisition of Sun Protection Habits Within Stages of Change an Exploratory Analysis

Data from 1,042 participants in a home-based expert system intervention were analyzed to explore subgroup profiles based on the measures from the Transtheoretical Model of behavior change : (1) Pros , (2) Cons, and (3) Self-Efficacy measures of attitudes towards sun protection habits . Independent studies were performed within the first three stages of change: Study 1Precontemplation (N = 570); Study 2Contemplation (N = 213) ; and Study 3Preparation (N = 259) . Replication across a series of randomly drawn data sets from the same general population was conducted for each stage . Variables external to the initia l analysis were used to evaluate agreement among solutions . Study 1: A four-cluster solution replicated well across subsamples , and was retained for the analysis of the Precontempl3:tion. Significant differences among clusters on the nine Processes of Change , and on all behavioral measures (Sun Protection Behavior Scale, use of sunscreen , SPF of sunscreen, and use of tanning booth/sun lamps) were found. Study 2: A four-cluster solution replicated well across subsamples , and was retained for the analysis of the Contemp lation stage. Significant differences among clusters on eight of the nine Processes of Change , and on three of four behavioral measures were found. Study 3: A four-cluster solution replicated well across subsamples , and was retained for the analysis of the Preparation stage. Significant differences among clusters on eight of nine Processes of Change, and on two of four behavioral measures were found . The subtypes for sun protection habits within three of the stages of behavior change closely replicated the subtypes found for smoking cessation and strong evidence of externa l validity was found . Acknowledgements Special thanks to James 0. Prochaska , Ph . D., for allowing access to the data, collected under a National Cancer Institute grant. I must thank my advisor Wayne F . Velicer, Ph . D ., under whose supervision I chose this topic, for providing me invaluable insights and timely encouragement to begin and finalize this work. Thanks also to thesis committee members Colleen Redding , Ph .D., and Bryan Blissmer , Ph.D . I cannot end without thanking my famil y, on whose constant support and love I have relied throughout my time here . And to my husband for making everything worthwhile, and for always being there but never asking why it was taking so long .. . I


Introduction
The Transtheoretical Model of behavior change has been applied to the study and modification of a range of health-related behaviors , such as smoking , high-risk alcohol consumption , physical inactivity , sun exposure , and diet, among others (Burbank , Reihe , Padula, & Nigg , 2002;Prochaska et al., 2005;Prochaska et al., 1994;Prochaska et al., 2004) . Stage of change is the central organizing construct of the model. Stage of change represents the temporal dimension , and serves to organize the processes people use to quit problem behaviors or acquire healthy behaviors Velicer , Prochaska , Fava , Rossi , Redding , & Laforge , 2000) . The five stages are Precontemplation , Contemplation , Preparation , Action , and Maintenance.
The incidence and mortality of skin cancers have increased rapidly in the past few decades (Jemal, Devesa , Hartge , & Tucker , 2001 ;McLaughlin , et al., 2005) . Few prevention studies have been effective at reducing sun exposure behaviors (Rossi, Blais , Redding , & Weinstock, 1995). Rossi et al. (1995) described that one of the strong point of applying the Transtheoretical model is that it not only provides customi zed advice to those who are ready to change , but also to the vast majority of people who are neither prepared or motivated to change . Individualized and stage based interventions have been developed to personalize participants ' awareness and evaluation of their overexposure to solar ultraviolet radiation and their risk of skin cancer. Prochaska et al. (2004) , and Prochaska et al. (2005) demonstrated , in multiple behavior interventions implemented on a population-basis (with a population of parents and primary care patients , respectivel y), that stage-tailored expert system interventions 1 produced significant improvement in sun protective behaviors. The intervention group performed sisnificantly better on the use of sunscreens and avoiding sun exposure than the group of comparison. Weins tock, Rossi, Redding , and Maddock (2002) also examined the effect of a stage-tailored expert system intervention to increase sun protection among beachgoers . They demonstrated that in the treatment group , tho se who were in preaction stages at baseline were more likely than control to reduce unprotected sun exposure , to use sunscreen, and to progress to more advanced stages of change at 12month and 24-month assessments.

Stages of Change and Profiles of Subtypes
Precont emplation is the stage in which people are not intending to change or modify their behavior in the near future , usually measured as the next six months (V elicer et al., 2000). People in this stage tend to avoid reading, talking, or thinking about their high risk behaviors. In relation to the key constructs of the Transtheoretical Model , it has been found that, among people in the Precontemplation stage, the Cons of changing the problem behavior ( e.g. smoking cessation) outweighed the Pros of modifying the behavior (Prochaska et al., 1994;Velicer , Rossi, Prochaska , & DiClemente , 1996). In this stage, for the acquisition of a healthy behavior, the Pros scores are a standard deviation 2 below average, and the Cons are a standard deviation above average. Studies suggest that to help the people who are in the Precontemplation stage to make observable modifications in their life , interventions should be developed to increase the Pros of a healthy behavior change by about one standard deviation (Prochaska, 1994;Velicer , Norman , Fava & Prochaska , 1999). In terms of the level of temptation to continue smoking, it w_as higher at the Precontemplation stage when compared to people in the Contemplation and Preparation stages (DiClemente et al., 1991. A limited number of distinct subtypes have been found and identified among smokers in the Precontemplation stage and the results have been replicated across multiple samples (Anatchkova et al., 2006a;Norman et al., 2000;Velicer et al., 1995;Velicer et al., 2007). High scatter and a pattern of very high scores on the Pros of smoking and the Temptations to smoke characterized one of the subtypes , which was labeled as the Jmmotive group. Below average scores were reported for the Cons of smoking (Norman et al., 2000 ;Velicer et al., 1995) . A second subtype , labeled as the Progressing group , showed a profile very similar to what would be expected of participants in the subsequent stage, Contemplation . The scores were above average across all measures . A third subtype, labeled as the Disengaged group , had a relative ly low and undifferentiated profile . This cluster was characterized by having scores average or below average on the Pros and Cons of smoking , and on the Temptations to smoke across a variety of situations. Norman et al. (2000) , and Anatchkova et al. (2006a) identified an additional cluster (labeled as Disengaged 2 cluster). The profile for this group showed an even lower level than the previousl y described Disenga ged group (the scores on all scales were about a standard deviation belo w average) . 3 Contemplation is the stage in which people are intending to change in the next six months , or twelve months (depending on the targeted behavior) .
People in this· stage are more aware of both the benefits and costs of changing (Velicer et al., 1996) . The scores on the Pros and Cons for the acquisition of a healthy behavior are both high (above average) and about equal in the Contemplation stage . This shows that the Pros of changing the problem behavior are higher for participants in this stage than for those in the Precontemplation stage, suggesting that the progress from the earlier stage to Contemplation involves an increase in the evaluation of the benefits of changing (Prochaska et al., 1994).
When comparing participants that are in the Precontemplation , Contemplation , and Preparation stages of change to evaluate the process of smoking cessation , those in the later stage reported significantly higher levels of confidence to stop or maintain nonsmoking and efficacy to abstain from smoking across various cues to smoke (less tempted) , followed by Contemplators and Precontemplators , representing a monotonically increasing function of self-efficacy across the stages (Di Clemente et al., 1991;Velicer et al., 2000).
A limited number of distinct subtypes have been found among smokers in the Contemplation stage (Anatchkova et al., 2005 ;Norman et al., 2000 ;Velicer et al., 1995).
One cluster corresponded to the expected profile of the stage (labeled as the Classic Contemplators group) . The scores of the Pros and Cons of smoking were about equal for this cluster. A second subtype (labeled as the Progr essing cluster) showed a profile very similar to what would be expected of participants in the subsequent stage , which is characterized by a cognitive shift in the balance in the Pros and Cons of smokin g: the disadvantages of smoking outweigh the benefits of smoking. A third subtype , labeled as the Disengag ed group, had a relatively low and undifferentiated profile. This cluster was characterized by having scores average or below average on the Pros and Cons of smoking , and on the Temptations to smoke across a variety of situations . Different from the solution described for the Precontemplation stage, an additional cluster represented a profile that resembled the previous stage (Anatchkova et al., 2005;Norman et al, 2000;Velicer et al .,-1995). This subtype was labeled as the Early Contemplators group . This finding shows that a subtype may be at risk for relapsing to an earlier stage (the perceived benefits of smoking outweigh the negative effects of doing so). After comparing clusters , it was found that participants in the Classic Contemplat ors subtype reported higher number of cigarettes per day than participants in the Progressing cluster (Anatchkova et al., 2005 ;Norman et al., 2000) . This finding suggests that even within stages, differences in the pattern of the Pros , the Cons, confidence levels and behaviors could be reported , reflecting variability within stage .
Pr epat·ation is the stage in which people are intending to take action in the immedia te future, usually measured as the next month . This suggests that individuals have typicall y taken some significant initiative to modify their behavior in the past year . Within the Preparation stage, the Pros of quitting an unhealthy behavior (e.g., smoking) or acquiring a healthy one (e.g., exercising) outweigh the Cons of doing so . Prochaska and colleague (1994) found that in 7 of 12 problem behaviors , a crossover between the Pros and the Cons occurred in early stages, but with behaviors like sunscreen use, high fat diets, and mammography screening , the crossover was evident during more advanced stage s. Among smokers , it has been found that the Preparation group was the least addicted when compared to the previous stages (DiClemente et al., 1991;Fava et al., 1995). Smokers in this stage were more actively modifying their behavior , reporting more quit attempts in the year prior to the study and in lifetime (DiClemente et al., 1991;Fava et al., 1995). Those in the Preparation stage also reported higher levels of confidence to stop or maintain nonsmoking and efficacy to abstain from smoking across various cues to smoke (less tempted) than individuals in the previous stages , DiClemente , 1991.
A limited number of distinct subtypes have been found among smokers in the Preparation stage (Anatchkova et al., 2006a ;Norman et al., 2000;Velicer et al., 1995).
One cluster corresponded to the expected profile of that stage, with low scatter and high level of elevation (labeled as the Classic Preparation group) . This cluster was characterized by above average scores across the Pros and Cons of smoking, and on the level of temptations to smoke . A second subtype showed a profile very similar to what would be expected of participants in a more advanced stage (labeled as the Progressing group). This profile was characterized by below average scores on the Pros and on the Temptations to smoke, and above average scores on the Cons of smoking (Norman et al., 2000 ;Velicer et al., 1995). This pattern is expected of individuals in more advanced stages for the cessation of unhealthy behaviors (Action, Maintenance). Participants in the Classic Preparation subtype, with a profile correspondent to the expected profile of the Preparation stage, reported smoking higher number of cigarettes per day than participants in the Progressing group (Anatchkova et al., 2005;Norman et al., 2000). This finding indicates that people in progressing clusters would need different strategies to promote change of behavior as their profile suggests a readiness to quit smoking that is not 6 apparent among participants in other clusters . A third cluster (labeled as the Disengaged group) was characterized by well below average scores on all the measures (Anatchkova et al., 2006a;Norman et al., 2000;Velicer et al., 1995). A fourth cluster (labeled as the Early Preparation group) represented a profile that resembles previous stages. This subtype is characterized by higher scores on the Pros compared to the Cons of smoking (a "V ' shaped profile) , and above average scores for the temptations to smoke (Norman et al., 2000 ;Veliceretal. , 1995).
Norman et al. (2000) described a second Disengaged group (labeled as the Disengaged 2 cluster; a five cluster solution was reported for the study). Different from the previous Disengaged subtype, the second group shows a shift between the scores for the Cons and the Pros, and the scores for the temptations are far lower than in the first subtype .

Stages of Change and Sun Protection Habits
Participants in the Precontemplation stage for sun protection habits are currently not exhibiting sun protection habits, and are not thinking about acquiring those habits in the next 12 months . In this stage, it is expected for participants to display a profile with below-average scores on the Pros of acquiring sun protection habits and the Self-efficacy measure , and high scores on the Cons of acquiring sun protection habits (both constructs from the Decisional Balance measure.
Participants in the Contemplation stage are currently not exhibiting sun protection habits , but are seriously thinking about starting to do so within the next 12 months. In this stage, it is expected for individuals to display a profile with about equal and above average scores with regard to the Pros and Cons of acquiring sun protection habits , but below average scores with regard to the level of Self-efficacy .
Preparation is the stage in which people intend to acquire sun protection habits in the next month . In this stage, participants report above average (but not equal) scores with regard to the Pros and Cons of acquiring sun protection habits, and below average scores with regard to the level of confidence to cope with high-risk situations without relapsing to unhealth y habits . The increase of the Pros of acquiring sun protection habits , when compared to the scores from the Cons measure , suggests readiness of participant s to change , but action may be delayed by a reduced engagement in health y behaviors when in the midst of difficult situations.

Cluster Analysis
Clustering methods have been recognized as a multi varia te statistical technique widely used within the socia l sciences . It starts with a data set containing information about a sample of individuals and attempts to classif y individuals that share certain properties into relati vely homogeneous subgroups that are different in some respects from the individuals in other subgroups (Aldenderfer & Blashfield , 1984;Everitt , Landau , & Leese, 2001).

Aims of the Study
Aim 1. To explore the existence of distinctive , interpretable, internall y consistent and externally valid subtypes across a series of data sets drawn from samples representing three of the theoretically identified stages of behavior change of sun protection habits : Studyl-Precontemplation ; Study 2-Contemplation ; and Study 3-Preparation. This will be followed in order to conduct a critical assessment of the model 8 and to evaluate its application to different behaviors . Analysis of subgroup profile will be based on (1) the Pros , (2) the Cons , and (3) Self-Efficacy measures of sun protection habits .

Aim 2.
To explore the existence of distinctive , interpretable , internally consistent and externally valid subtypes across general samples that represent three of the theoretically identified stages of behavior change of sun protection habits . Analysis of subgroup profile will be based on (1) the Pros , (2) the Cons, and (3) Self-Efficacy measures of sun protection habits. It is expected that , given the relevance of this model in the study of health y and unhealth y behaviors , and given that it has been applied in the sun exposure and protection context , the clustering solution will be consistent with the model.

Method Participants
The sample used in this secondary data analysis is a portion of a sample collected from a larger, multiple behavior intervention study to guide a population of primary care patients to quit smoking, eat healthier , prevent skin cancer, an receive regular mammograms. Outcome analyses conducted with the sample from the larger home-based expert system intervention can be found elsewhere  . A total of 123 84 respondents were contacted by phone. A total of 3 820 patients refused to participate. A total of 8564 subjects agreed to participate , but 3157 were screened out because they did not have at least one of the three health risk behaviors (sun, diet, & smoking) targeted for the intervention. At baseline, the larger sample included 5407 participants with complete data. The participants were then randomly assigned to either the home-based expert system intervention or comparison condition for control. The expert system treatment participants were mailed intervention materials for each of their at-risk behaviors . Mailed materials included the baseline feedback report and an integrated multiple risk behavior stage-matched self-help guide.
The reply to the phone survey produced the expert system report for the interventio n group . As part of the larger intervention , participants were assessed at 6, 12, and 24 months after the initial assessment.
The University's Institutional Review Board approved this study as it met the University and Federal guidelines for research involving human subjects .
Eligibility for this analysis included being at risk for sun exposure . The measures required for this analysis were only collected from the treatment group . Only baseline information from the sun protection segment of the intervention was included in the analysis , which was conducted on only the treatment group . The participants were in either the Precontemplation , Contemplation, or Preparation stage of change for sun exposure at baseline . The number of participants for this study is 1,042 subjects. The stage distribution of the sample was Precontemplation (PC), 570 (54.70%) ; Contemplation (C), 213 (20.44%) ; and Preparation (PR), 259 (24.86%).
In the Contemplation stage, most subjects were female (71.2%) and were White (98.8%). The subjects' mean age was 46.68 (SD = 12.74). Most subjects reported a good health status (43.0%) , and 67.3% were married .

Mater;als
Stages of Change. The stages of change for general sun exposure and for sunscreen use were evaluated, using algorithms developed to assess intentions and behaviors for reducing sun exposure (Rossi, Blais , Redding & Weinstock , 1995;Maddock , Redding , Rossi & Weinstock , 2005;Prochaska et al., 2005). The general sun protection algorithm classified subjects by stage based on questions that measured their behaviors , and intentions to protect themselves by avoiding sun exposure , using sunscreen , and by wearing protective gear whenever they know they would be out in the sun for a prolonged period of time during the summer.

11
A 4-item algorithm assessing baseline intentions and actions was used to evaluate the classification of the participants . If participants report that they do not protect themselves from exposure to the sun consistently (that is, whenever they know they will be out in the sun for more than 15 minutes) , and they haven 't done so in the past 12 months , they were included in the staging procedure. Subjects were in : (1) the Precontemplation stage if they were not intending to protect themselves from exposure to the sun in the next 12 months ; (2) the Contemplation stage if there were intending to protect themselves from the sun in the next 12 months ; and (3) the Preparation stage if they were intending to protect themselves in the following month . Items applied for the staging algorithm and other measures were included as Appendix A.
Decisional Balance . The Decisional Balance construct derives from Janis and Mann ' s model of decision -making (Janis & Mann , 1977). This construct includes categories of Pros , or advantages related to the behavior . The categories include questions about instrumental gains for self and others that can be identified from modifying the targeted behavior-or from keeping the behavior-and approval from self and others because of the behavior or the modification of the behavior. Also, it includes for categories of Cons, or disadvantages related to the behavior. The categories include questions about instrumental costs for self and others that can be identified from modifying the same behavior-or in some cases, from keeping the behavior -, and disapproval f~om self and others because of the behavior or modifications in it (terminology can reverse direction depending whether the problem involves cessation or acquisition). It has been found that this instrument reliably differentiated participants in different stages of change after being modified and applied within the framework of the Transtheoretical Model of behavior change (Velicer , Di Clemente , Prochaska , & Brandenburg , 1985). Its reliability for sun protection and exposure habits has been assessed in previous studies (Maddock et al, 2005;Prochaska et al., 1994). The decisional balance instrument that will be used in this study consists of 4 items assessing the pros of sun protection (a = 0.78) and 4 items assessing the cons of sun protection (a=0.74). The instrument asked participants to rate how important each item is in deciding whether or not to protect themselves from too much sun exposure on a 5-point Likert scale from non important (= 1) to extremely important (=5) .
Self-Efficacy. The Self-efficac y construct is related to the way the individual will cope with difficult situations in where the risk of relapsing or engaging in the unhe althy behavior is higher. The self-efficacy instrument consists of 7 items measuring confidence to use sun protection (a= 0.84) . The participants were asked to rate how confident they were in protecting themselves from the sun in a variety of situations on a 5-point Likert scale from not at all confident (=1) to extremely confident (=5). It has been applied in previous research , and its predicti ve ability has been evaluated (Maddock et al, 2005).
Processes of Change . Processes of change are cognitive , emotional , and behavioral strategies that people use to change. These processes are independent variables that people need to apply, or be engaged in, to move from stage to stage  . Nine of the 10 traditional processes of change were assessed using 2 items each. The processes included counter conditioning , consciousness raising , dramatic relief , environmental reevaluation , he! ping relationships , reinforcement management , self reevaluation , social liberation, and self-liberation. Respondents were asked to rate how often they used the processes of change in the past 30 days on a 5-13 point Likert scale ranging from never (= l) to always (=5). Coefficient alphas ranged from 0. 71 to 0.81 for the processes of change scales for sun protection habits in previous studies (Maddock et al, 2005).
Behavioral Measures. Behavioral and behavioral history measures have been previously id~ntified as differentiating factors between stages of change (Di Clemente et al., 1991;Fava et al., 1995) and between subtypes within stages (Anatchkova et al., 2006a ;Norman et al., 2000 ;Velicer et al., 1995). The association between tan attractiveness and skin protection , and stages of change has been assessed (Maddock et al., 2005). Findings suggest that attitudes towards tan attractiveness decreased across stages while attitudes towards skin protection increased. Variables that were emplo yed as validity measures for clustering solutions are: use of sunscreen and its SPF, and tanning booth use history -in the past ("have you ever used a tanning booth or sunlamp "), and in the previous year (" have you used a tanning booth or sunlamp in the past year") .
The need to promote specific approaches to reduce sun exposure, as limiting the time under the sunlight, specially during the midday hours, covering up exposed skin area through the use of protective clothing (long sleeves, long pants, or wide-brimmed hats), and using waterproof sunscreen with a sun protection factor of at least 15, has been introduced in previous research (Rossi et al, 1995). The Sun Protection Behavior Scale (SPBS , a.= 0.82) is a brief inventory with three sub-scales that include some of these measures of sun protection (Weinstock , Rossi, Redding , Maddock , & Cottrill , 2000) .
Two of the three subscales were used : Sun Screen Use (a.= 0.86), and Sun Screen Avoidance (a.= 0.82). It is a 7-item scale; each item is a self-report of behavior on a 5point Likert scale of frequenc y (never, rarely, sometimes , often, and always) "when in the sun for more than about 15 minute s during the summer" (Weinstock et al., 2000) . The composite score of this scale (minimum possible score= 7; maximum possi ble score= 35) was used.
Demograp hic Variables. Data regarding demographic variables (age, gender , health status, race and ethnicity , and marital status) were collected and the association between these variables and cluster subtypes was evaluated.
Procedure Sample Selectio n. Only baseline information from the sun protection segment of the intervention was included in the analysis, which was conducted only on the treatment group. The participants were classified in the Precontemplation, Contemplation , or Preparation stage of change for sun exposure at baseline after conducting the staging algorithm . Independent clustering studies were conducted for each stage .

Development and Iden tification of Subtypes. Cluster analysis attempts to classify
individuals that share certain properties into relatively homogeneous subgroups that are different in some respects from the individuals in other groups (Aldenderfer & Blashfield , 1984; Everi tt, Landau , & Leese, 200 1). The choice of the variables to be used with the clustering procedure is one of the most important steps of the analysis (Aldenderfer & Bla shfield, 1984). Once the participant s were assigned to the stages, the measures included in the Transtheoretical Model of behavior change were used for the development of subtypes within each stage. The Decisional Balance (Pros and Cons) and Self-efficacy measures were initially applied in the classification procedure . Processes of change, variables regardin g skin protection and sun exposu re habits, and demographic variabl es were used to conduct the validation procedure .
Random subsamples were drawn from the general sample to determine whether a cluster structure could be replicated across samples . The cluster analysis was performed independentl y on each subset (Milligan & Cooper, 1987). The consensus between solutions was then evaluated . If the same cluster solution is found across different samples , it is plausible that the discovere d solution has a high level of generalizability , and could be found in any general sample (Everitt , Landau , & Leese , 2001).

Standardization of Variables.
Variables are routinely standardized to a comparable metric prior to any analysis to equalize the contribution of each variable to the outcome of each study (Aldenderfer & Blasfield , 1984). This procedure has been used in comparable cluster analysis studies (Anatchkova , et al., 2005(Anatchkova , et al., , 2006a(Anatchkova , et al., , 2006bNorman et al., 2000 ;Velicer et al., 1995). For our study, all the variables to be included in cluster identification procedures (Decisional Balance -Pros and Cons-, and Self efficacy) were standardized to Tscores (M = 50, SD= 10) for the general sample from each stage , and for each independent subsample drawn from each stage.

Data Analysis
Similarity measures have been developed to estimate the leve l of proximity between the individuals. The most commonly used is the squared Euclidean distance (Everitt , Landau, & Leese , 2001) , which was employed in this study. The squared Euclidean distance metric was calculated on the three standardized var iables : Pros , Cons, and Self-efficacy.
Ward ' s minimum variance method (Ward , 1963) was employed in this study .
Several indices were used to determine the number of clusters : the cubic clustering criterion (CCC), the pseudo F test (Calinski & Harabasz , 1974) , and the pseudo t 2 test.
Values of the CCC greater than 2 or 3 indicate good clusters. A local peak, followed by a drop in the value of the CCC , indicates an appropriate number of clusters for the data (SAS Institute Inc ., 1999). The pseudo F statistic s, where large values indicate a stopping point , and the pseudo t2 (SAS Institute Inc ., 1999) were also used. Following the results from the three indices , visual inspection of the cluster profiles was also performed with a focus on the shape (configuration of the scores ; pattern of dip and rises) , level (the mean score of the case over all the variables) , and scatter (the standard deviation ; dispersion of the scores around own average) of the profiles. The profiles for each of the solutions were evaluated to determine the similarities of solutions across samples .
External Validation. One of the basic steps that characterize all cluster analysis studies is the validation of the resulting cluster solution (Milligan & Cooper , 1987).
There are different ways to follow this basic rule , including replication across a series of data sets from the same general population , and Monte Carlo procedures . Also , significance tests on variables external to the creation of the clusters are among the most appropriate ways to validate a clustering solution (Aldenderfer & Blashfield , 1984 ).

Processes of Change.
For each stage, multivariate analysis of variance was performed using the cluster membership as grouping variable , and the 9 Processes of Change meas~res as dependent variables . The most common macro summary index for MANO VA is Wilks' lambda, and it' s associated Ftest (Harlow , 2005) . A common multivariate effect size for MANOV A is eta-squared (11 2 ), which represents the proportion of linear combination of dependent variables that is explained by the grouping variables. An effect size is a quantity that measures the size of an effect in a way that is independent of certain details of the study, as sample size and normality of the

Results
Solutions with three to five clusters were found that described the data best. A four cluster solution replicated well across the three samples . The cluster profiles for Sample 1 (N= 182; 8 participants were not clustered) , after applying the four-cluster solution , are presented in Figure 1. The cluster profiles for Sample 2 (N= 187; 3 participants were not clustered) are presented in Figure 2. The cluster profiles for Sample 3 (N= 186; 4 participants were not clustered) are presented in Figure 3. These 15 participant s were excluded because they reported incomplete information . A cluster analysis was then performed on the total sample of 555 participants , and this analy sis served as the basis for the external validity . The cluster profiles for the general samp le are presented in Figure 4.
Cluster 1: The first cluster (N = 142) was labeled Immotive , and was characterized by an inverted "V" shape with medium scatter and average level. This profile corresponds to what is expected for people in the Precontemplation stage . The means for Pros and Self-efficacy scores were below average , whi le the mean for the Cons scale was above average . This pattern is expected among people not considering modifying their behavior given the importance assigned to the disadvantages of consistently protect their skin from sun damage . Participants in this group consider the Cons of consistently protecting themselves from sun exposure more important than the Pros of doing so.
Cluster 2: The second cluster (N = 197) was labeled Progressing , and had a " V ' shape with high scatter and above average level. This subgroup had average scores on the Cons scale , and above average scores on the Pros and Self-efficacy scales . This 20 pattern seems to indicate that these individuals may be ready to progress to a more advanced stage. They are considering both the benefits and the disadvantages of acquiring a h~althier lifestyle regarding sun exposure . Self-efficacy scores are higher than what would be expected from people that are not currently performing skin protection habits .
Cluster 3: The third cluster (N = 57) was labeled Disengaged and had a slightly inverted " V ' shape , low to medium scatter , and low level. This subtype consists of participants with Pros , Cons , and Self-efficacy scores all around a standard deviation belo w average . This was the smallest group within the overall sample.
Cluster 4: The fourth cluster (N = 159) was labeled Early Progressing and had a shallow "V" shape (almost a flat shape) with medium-to-high scatter , and low level. The mean for the Self-efficacy scale is around average , while the scores for the Pros and Cons scales are below average. ANOVAs were performed on each of the dependent measures. The Tukey HSD post -hoc test was employed as the follow-up procedure for tests where significant main effects were found. Effect sizes (partial eta-squared) were also calculated to determine and compare differences in the use of specific Processes of Change between clusters within each of the stages. Guidelines for univariate effects apply here : 0. 01 for a small effect , 0.06 for a medium effect , and 0.13 or more for a large effect (Cohen, 1977). The effect sizes ranged from small-to medium to large. Across all processes, significance follow up Tukey tests revealed that participants in the Progressin g cluster are using the Processes of Change significantly more than the members of the other clusters , while participants in the Di sengaged cluster are typically using the processes the least.
Also, even when some variation in the patterns was detected across processes , the profiles displayed by the Immoti ve and the Early Progressin g clusters were similar .
Effect sizes were larger across experiential Processes of Change when compared to behavioral processes .  Table 2. The effect size for this measure was in the large range . The follow up Tukey test revealed that participants in the Progressing cluster reported protecting their skin from exposure damage more than the participants assigned to the other clusters ; while members of the Immotive and Disengaged clusters reported doing the least to avoid damaging their skin from the effect of the sunlight.
Chi-square analyses were conducted on various behavioral variables to examine the differences in cluster membership among these measures. These descripti ve statistics , chi square values and percentage patterns for the clusters are presented in Table 3.
Significant differences between clusters were found for the sunscreen use and SPF of the sunscreen measure , x2 (15) = 107.58, p <. 001; Cramer's <1> 2 = .258. A higher percentage of participants in the Disenga ged cluster (75%) reported they never use sunscreen , while a lower percentage of participants in the Progressing subtype (11 .6%) reported the same behavior. Members of the Immotive cluster reported similar behavioral profile as the Early Progressing cluster. In terms of SPF of sunscreen, among those subjects that use this article , a higher percentage of members of the Progressing and Earl y Progressing clusters (43.2% and 34.8%, respectively) reported using sunscreen with a protection factor of 15 to 29, in comparison to participants in the Immotive and Disen gaged clusters (20% and 7.7%, respectively) .
The use of tanning/sun lamps was also included as a behav ioral measure to assess the difference between clusters . The respo ndents were asked to report if they ever used a tanning /sun lamp in the past, and if yes, to report the use of this item during the previous year . Chi-square analyses were conducted on these two variables. These descripti ve statistics, chi-square values and percentage patterns for the clusters are also presented in

Discussion
One of the findings of this exploratory study is the identification and replication of an initial four-cluster solution among three different samp les that were drawn from the same initial sample of participants in the Precontemplation stage . Data were random ly divided into subsets and analyses of the profiles were performed on each subset.
Consensus on the subgrou ps profile s based on the scores from the Pros , Cons, and Selfefficacy scores was evaluated , resulting a cluster structure that could be similarly evaluated in more than one sample . These groups are of special interest because traditionall y individual s in the Precontemplation stage of beha vior change present the most serious challenge for the promotion of healthy behaviors, and for the prevention of risk beha viors.
Description of the clusters . The initial solution was applied to the general sample , and four clusters were labeled on the base of the shape, level and scatter of the profiles based on the scores from the Pros , Cons , and Self-efficacy measures . The evaluation of the clusters was supported by external validity information.
The Immotive cluster (25.59% of participants) was the subtype that most resembled the profile of classic Precontemplators , with high scores on the Cons of sun protection , and low scores on the Pros of sun protection and Self-efficacy measure . An year . Also, in comparison with other clusters, most of the members of this subtype are younger (25.4% are 21-33 years old, and 64% are less than 50 years old).
The largest subtype was the Progressing (35.49% of participants) , which had almost the opposite profile of the previous subtype, and was characterized by a V shape .
Compared to the Immotive group, this cluster was less stable across samples . This pattern , with yery high scores on the Pros of sun protection and the Self-efficacy processes . This cluster reported a relatively high score in the Sun Protection Behavior Scale (SPBS) , showing that this group is moderately effective in the implementation of specific strategies that reduce unprotected sun exposure . One third of the participants categorized in this cluster never use sunscreen , and also one third reported using sunscreen with a SPF 15-29. Most of the members of this subtype had never used a tanning booth or sun lamp . Among participants you ' ve used one or both of these items, most of them had not used one in the past year. Also , most of the older participants were members of t~is cluster (age range 50+).
The smallest cluster was the Disengaged cluster (10 .27% of all participants) , which was characterized by a V shape similar to the Immotive cluster. All three scores were below average , but the scores of the Cons of sun protection behavior are higher than the scores from the other measures . This may represent that even when neither the benefits nor disadvantages of acquiring healthy habits, or feeling convinced that sun protection would be used in a variety of challenging situations , are perceived as important processes , a need to clarify the significance of these concepts for the prevention of skin cancer shou ld be addressed . This group was stable across samples . When the Processes of Change were analyzed , results indicated that people in the Disengaged cluster reported using all the process the least. In regard to the behavioral variables , this cluster reported the lowest score on the SPBS (similar pattern as the Immotive subtype) , showing that this group is also the least effective in the implementation of specific strategies that reduce unprotected sun exposure . At the same time , in comparison to the other subtypes, more participants in this group are not using sunscreen at all. Only 12% of the members of this cluster reported previous use of a tanning booth/sun lamp . Among these participants , most of them had not used one in the past year . Also , most of participants in this cluster were young adults (age range 34-49) .

Conclusion
This study provides additional evidence for the presence of within-stage differences for the Precontemplation stage of change across different behaviors (Anatchkova et al., 2006a ;Norman et al., 2000 ;Velicer et al., 1995;Velicer et al., 2007).
The Immotive cluster was the group that most clear ly exemplified the Precontemplation stage , and the Progressing group exhibits a profile that resembles more advanced stages .
These two groups had average , and above average level. The Disengaged subtype (which resembles the Immoti ve cluster , but with a belo w average level) was the one that reported 28 more risk behaviors, and was the least effective in the implementation of specific strategies that reduce unprotected sun exposure. The Disengage d and the Early Progressing clusters groups had a below average level. The Early Progressing subtype was intermediate between the Progressing cluster and the Immoti ve cluster. The results of this study add to our understanding of adults that are not thinking in acquiring sun protection habits in the next year.

Study 2. Contemplation Results
Three to five clusters were found to describe the data best for Sample 1 (N = 103; 3 participants were not clustered) and Sample 2 (N = 103; 4 participants were not clustered). The values for the Pros, the Cons, and the Self-efficacy measures for sun protection habits were plotted and evaluated in detail for each solution . A four-cluster solution replicated well across the two subsamples . This solution was retained for the analysis of the general sample . The cluster profiles for Sample 1 are presented in Figure   5. The cluster profiles for Sample 2 are presented in Figure 6. Eight participants were excluded becau se they reported incomplete information . A cluster analysis was then performed on the total sample of 205 participants , and this analysis served as the basis for the external validity. The cluster profiles for the genera l sample are presented in Figure 7.
Cluster 1: The first cluster (N= 59) was labeled Progressing. It is characterized by scores above average on the Pros scale, scores a standard deviation above average on the Self-efficacy measure , and scores about a standard deviation below average on the Cons scale. The mean on the Cons measure are lower than what would be expected for participants in the Contemplation stage . Also, the mean on the Self-efficacy scale is higher than would be expected of subjects in the same stage; this profile is more similar to stages where actual skin protection behavior is taking place . The overall elevation of the gro up is average . The shape is a "V" expected of participants progressing to a more advanced stage . The scatter was medium-to-hi gh.

Cluster 2: The second cluster (N = 47) was labeled Classic Contemplators.
Members within thi s cluster are characterized by a profile that is similar to the profile expected for people in the Contemplation stage. It was generally flat with low scatter and high level. The scores are almost a standard deviation above the mean on Pros and Cons scales, showing the evaluation of both the benefits and the negative aspects of acquiring the targeted behavior. This represents a possible conflict between current sun exposure status and the harmful effects of this habit. The mean score on the Self-efficacy scale is also above averag e, something expected of subjects in more advanced stages.
Cluster 3: The third cluster (N = 78) was labeled Early Contemplator s. Subjects in this cluster_ are characterized as having average scores on the Cons of acquiring skin protection habits, and below average scores regarding the Pros and Self-efficacy measures . This was the largest group within the overall sample . An inverted " V' shape with medium scatter and average level distinguished this subtype. This shape resembles the profile of Precontemplators more than that of Contemplators . Participants in this group still consider the Cons of constantly protect themselves from sun exposure significantly more than the Pros of doing so.
Cluster -I: The fourth cluster (N = 21) was labeled Disengaged. An inverted "V" shape, high scatter and low-level pattern characterized this cluster . The scores for the Cons scale are less than a standard deviation below average , while the scores for the Pros and Self-efficac y measures are more than a standard deviation below average . This group was the smallest group within the sample.  Table 5. Following a significant difference for group on the MANOVA, separate ANOVAs were performed on each of the dependent measures . The Tukey HSD post-hoc test was employed as the follow-up procedure for tests where significant main effects were found. Effect sizes (partial eta-squared) were also calculated to determine and compare differences in the use of specific Proce sses of Change between clusters within each of the stages . Guidelines for univariate effects apply here : 0.01 for a small effect , 0.06 for a medium effect , and 0.13 or more for a large effect (Cohen , 1992).  Table 6. The effect size for this measure was in the large range . The follow up Tukey test revealed differences between clusters in the effectiveness of the implementation of specific strategies to reduce unprotected sun exposure.
Chi-square analyses were conducted on additional behavioral variab les to examine the differences in cluster membership among these measures . Descriptive statistics , chi-square values and percentage patterns for the clusters are presented in Table   7. Significant-differences between clusters were found for sunscreen use and SPF of sunscreen , x2(15) = 32.34, p < .01; phi = .398; Cramer's <1> 2 = .230. The use of tanning/sun lamps was also included as a behavioral measure to assess the difference between clusters. The respondents were asked to report if they ever used a tannin g/sun lamp, and if yes , to report the use of this item during the previous year. Chi-square analyses were conducted on these two variables . Significant differences in cluster membership were only found for the use of a tanning booth/sun lamp in the past year (this analysis was conducted among those who reported using this item at least once prior to the previo~s year),/ (3) = 9.10, p < .05; Cramer ' s <1> 2 = .335.
External Validation: Demographic Variables. Chi-square analyses were conducted on various demographic variables to examine the association of these variables with cluster membership : gender , age group, income , health status, employment status, and marital status . Descriptive statistics , chi square values and percentage patterns for the clusters are presented in Table 8. None of the chi-square analyses based on the four subtypes was significant.

Discussion
One of the findings of this exploratory study is the identification and replication of an initial four-cluster solution among two different samples that were randomly drawn from the same initial sample of participants in the Contemplation stage. Data was randomly divided into subsets and analyses of the profiles were performed on each subset. Consensus on the subgroups profiles based on the scores from the Pros , Cons, and Self-efficacy scores was evaluated, resulting a cluster structure that could be similarly evaluated in more than one sample. The understanding of cluster subtypes in this stage has the potential to guide the development of tailored interventions for the promotion of sun protection habits . Participants in the Contemplation stage are of special interest because it is in this stage where individuals start considering the benefits of change , but at the same time, they are still highly considering the disadvantages of modifying their behavior .
Description of the clusters. The initial solution was applied to the general sample, and four clusters were labeled on the base of the shape, level , and scatter of the profiles from the Pros , Cons, and Self-efficacy measures . The evaluation of the clusters was supported by information external to the initial analysis.
The Progressing cluster (28 . 78% of all participants) showed a profile that resembles the one for people in more advance stages of change. This cluster was stable across samples . The pattern followed by this group, with above average scores on the Pros of sun protection and the Self-efficacy measure , but below average score on the Cons, describes a group of individuals that are ready to move to the next stage (it was characterized by a V shape) . This profile may represent subjects that are cautious about declaring therpselves members of the next stage . They may not be sure they are taking , or might consider taking , all the precautions necessary to protect themselves from sun exposure . But they are already thinking about taking those precautions . When the Proce sses of change were analyzed, results showed that participants in the Progressing clusters are using 4 of 8 processes more than members of the Early Contemplators and Disengaged clusters . In addition , the Progressing cluster is applying the Counter Conditioning process more than the Classic Contemplators group . In term of behavioral variab le s, participants in the Progressing cluster reported protecting their skin from sun damage more ·than the participants classified in the Early Contemplators and Disengaged subtypes . In terms of SPF of sunscreen , more than half of members of the Progressing cluster reported using sunscreen with a SPF of 15-29 . Most of the participants reported they have never use a tanning booth/sun lamp , and none of them used it during the previous yea r. More participants in the Progressing cluster reported using sunscreen wit h SPF of 30 or more than members of the other subtypes . Subjects in this subtype could potentially benefit from interventions designed for the next stage rather than their currently stage classification .
The Classic Contemplators cluster (22. 93% of all participants) was the subt ype that most resembled the profile of individuals in the Contemplation stage, with above average scores on the Pros and Cons of sun protection . This profil e was les s stable than the Progressing group . This pattern describes a group of individuals that considers both the importance and disad vantages of constantly protect themselves of exposure to the sun . Scores from the Self-efficacy measure are higher than what would be expected of participants that are not curre ntly considering protecting their skin in a regular basis in 35 the immediate future . Results showed that participants in the Progressing clusters are similarly using 4 of 8 processes more than members of the Early Contemplators and Disengaged clusters . In addition , this subtype is applying the Reinforcement Management process more than participants in the Disengaged cluster. Similarly to the Progressing cluster , this group reported higher scores in the SPBS measurement. All of the participants in the Classic Contemplators subtype reported using sunscreen . Most of them used sunscreen with SPF of 15-29. A little less than half of the members of this group have used a tanning booth/ sun lamp .
The largest subtype was the Early Contemplators cluster (38.05% of all participants) . This cluster , which had the opposite profile of the Progressing subtype , and was characterized by an inverted V shape, was stable across samples . This profile , with scores above average on the Cons, and below average on the Pros and Self-efficacy scales, resembled the pattern typically followed by individuals in the Precontemplation stage . The main intervention approach for these participants should aim to increase the Pros of sun protection , and to increase their level of confidence in the ability to maintain healthier habits across difficult situations . When the Processes of Change were analyzed , results indicated that this group used 4 of 8 processes less than participants in the Classic Contemplation and Progressing clusters. This group reported a lower score in the Sun Protection Behavior Scale . Most of the participants in this cluster used sunscreen with a SPF 15-29, and were the second largest group that never uses sunscreen . At the same time , a highe~ proportion of people that reported they ' ve used a tanning booth/sun lamp before , also reported they ' ve used it at least once in the past year .
The smallest subtype was the Disengaged cluster (10.24% of all participants) . The pattern followed by this group was stable across samples . All the three scores from the scales were below average . This may represent that neither the benefits nor the disadvantages of acquiring healthy habits, or feeling convinced that sun protection would be used in challenging situations , are perceived as important measures. A need to address the significance of these concepts for the prevention of skin cancer should be emphasized. Regarding the Processes of Change , participants in the Disengaged group used the processes the least. At the same time , similar to the Early Contemplators subtype , this group reported a lower score in the Sun Protection Behavior Scale, showing that both groups are the least effective in the implementation of specific strategies that reduce unprotected sun exposure . When additional behavioral variables were analyzed , it was found that more members of this cluster reported never using sunscreen , in comparison to the other clusters . Less members of this cluster use sunscreen of SPF 15-29, or 30 or more. At the same time, this group used a tanning booth/sun lamp in the previous year less than the Early Contemplators and Classic Contemplators groups . This may suggest that these individuals are less exposed to situations that may increase their awareness of skin protection or tanning .

Conclusion
This study supports the presence of important differences within the Contemplation stage of change across different behaviors (Anatchkova et al., 2006a ;Norman et al., 2000 ;Velicer et al., 1995). The Classic Contemplators cluster was the group that most clearly exemplified the Contemplation stage, and the Progre ssing cluster exhibits a profile that resembles more advanced stages. The Early Contemplators group 37 was more similar to the previous stage. The Disenga ged cluster , which resembles the Early Contemplators cluster but with a below average level , was the one that reported more risk behaviors across most of the varia bles included in the analysis. The results of this study add to our understandin g of adults that are considering changing their sun protection behavior in the future , but haven ' t successfully taken action in the past.

Results
Three to five clusters were found to describe the data best for Sample 1 (N = 128; 2 participants were not clustered) and Sample 2 (N_= 128; 1 participant wasn ' t clustered) .
The va lues for the Pros , the Cons , and the Self-efficacy measures for sun protection habits were plotted and evaluated in detail for each solution . A four-cluster soluti on replicated well across the two subsampl es. This solution was retained for the ana lysis of the general sample. The cluster profiles for Sample 1 are presented in Figure 8. The cluster profiles for Sample 2 are presented in Figure 9. Eight participants were excluded because they reported incomplete information . A cluster analysis was then performed on the total sample of 256 participants , and this analysis served as the basis for the external validity . The cluster profiles for the genera l sample are presented in Figure 10.
Cluster 1: A shallow inverted " V " shape (almost a flat shape) with moderate to high scatter and high level characterized this cluster (N = 78), which was labeled Early Preparation . The mean for the Pros and Self-efficacy scales are slightl y above average whi le the scores for the Cons are one standard deviation above average . This pattern is expected among participants that could consider or attempt to acquire sun protection habits, but the Cons of acquiring a healthi er behavior still outweigh the benefits of doing so.
Cluster 2: The second cluster (N = 39) was labeled Progressing and had a "V" shape w ith medium to high level and medium scatter. The means for the Pros scores are one standard deviation above average , showing that the switch in the importance acknowledged to the benefits of acquiring healthier beha viors is important in the progress 39 to more advanced stages . Scores for the Self-efficac y measure are also one standard deviation above average . But the Cons scores are one standard deviation below average , which is lower than what would be expected from people that are not consistently protecting themse lves from the sun in the present.
Cluster 3: The third cluster (N = 115) was labeled Level Preparation . It was characterized by low level, around average on the Pros and Cons measures , and below average on the Self-efficacy scale . The shape was flat with medium to high scatter. This profile is consistent with participants that may not be ready to acquire sun protection habits in the immediate future , even when they report that they feel read y to do so. This was the largest group within the overall sample.
Cluster./: The fourth cluster (N = 24) was labeled Disengaged , and was characterized by scores around one standard deviation below average with regard to the Cons and Self-efficacy scales , and two standard deviations below average with regard to Pros measure . This group was the smallest group within the overall sample. The level of this group wa·s well below average . The shape was relativel y flat with a slight linear increase for Pros scores to Self-efficac y scores.  Table 9. One of the participants previously clustered didn't report enough information on the variables included in this analysis . Following a significant difference for group on the MANOV A, separate ANOVAs were performed on each of the dependent measures . The Tukey HSD post-hoc test was employed as the follow-up procedure for tests where significant main effects were found. Effect sizes (partial eta-squared) were also calculated to determine and compare differences in the use of specific Processes of Change between clusters within each of the stages. Guidelines for univariate effects apply here : 0.01 for a small effect, 0.06 for a medium effect, and 0.13 or more for a large effect (Cohen , 1977). The effect sizes ranged from small to large . Across 6 of 8 significant processe s, significance follow up Tukey tests revealed that participants in the Pro gressing cluster are usin g the proces ses more than the members of other clusters. Participants in Diseng aged 2 cluster are typically using the processes the least. Even when some variation in the patterns was detected across Processes of Change , the profiles displayed by the Early Preparation and Level Preparation clu sters were similar in some aspects. Effect sizes were generally larger across experiential processes when compared to behavioral processes . The larger effect sized were reported for the Environme ntal Reevaluation and 41 Self-Reevaluation processes of change (0.151 and 0.108, respectively) . Differences between clusters were not reported for the Reinforcement Management factor.
External Validation: Behavioral Variables. A significant effect was found on scores of the Sun Protection Behavior Scale (SPBS) using ANOVA (F (3,252) = 18.71, p < . 001, 11 2 = .182), indicating differences between clusters in this measure of sun exposure behavior . Descriptive statistics, F value, and Tukey patterns for the clusters are presented in Table 10. The effect size for this measure was in the medium range . The follow up Tukey test revealed that participants in the Progressing subtype reported protecting themselves from sun exposure more than participants in the Early Preparation ,

Level Preparation , and Disengaged clusters
Chi-square analyses were conducted on additional beha vioral variables to examine the differences in cluster membership among these measures . These descripti ve statistics , chi-square values and percentage patterns for clusters are presented in Table 11.
Significant differences between clusters were only found for the sunscreen use and SPF of sunscreen measure , x2 (15) =45.33 , p < .001; Cramer's <!> 2 = .248). Almost half of the participants classified as in the Disengaged cluster (45.5%) reported never using sunscreen, while a lower percentage of participants in the Progressing cluster (5.1 %) reported the same behavior. In terms of SPF of sunscreen , more than half of the participants classified as in the Progressing group (53.9%) reported using sunscreen with a protection factor of 15-29, and 28.2% of participants classified as in the same subtype reported using sunscreen with a protection factor of 30 or more. A lower percentage (9%) of members of the Disengaged reported using sunscreen with a protection factor of 30 or more .

42
The use of tanning booths or sun lamps was also included as a beha vioral measure to assess the difference between clusters. The respondents were asked to report if they ever used a tanning booth or a sun lamp in the past, and if ye s, to report the use of this item during the previous year . Chi-square analyses revealed no significant differences in cluster membership by the use of tanning booths /s. un lamps in lifetime or during the previous year .
External Validation: Demographi c Variables. Chi-square anal yses were conducted on various demographic variables to examine the association between these variables and cluster membership for the Preparation stage: gender , age group , health status , race , ethnicity , and marital status . Table 12 presents frequencies, descriptive statistics , chi-square values, and percentage patterns for clusters .
Only one chi-square analysis based on the four cluster groups was significant.
There was an association between cluster membership and marital status , x2 (15) = 25 .34, p<. 05; Cramer ' s <1> 2 = .198. A large percenta ge of people in the Progressing cluster (26 .7%) were not married when compared to the Earl y Progressing (9.7%), Le vel Preparation (9.6%) , and Disengaged (0%) subtypes . Also , a smaller percentage of people in the Progressing group (60%) were married or divorced (0%) when compared to the rest of the clusters .

Discussion
One of the findings of this exploratory study is the identification and validation of an initial four-cluster solution among two samples that were drawn from the same initial sample of participants in the Preparation stage . Data wa s randomly divided into two subsets and an examination of the profile was performed on each subset. This examination was based on the scores from the Pros, Cons, and Self-efficacy scores. This process resulted on a cluster structure that could be similarly evaluated in more than one sample . Understanding of cluster subtypes within the Preparation stage is of special interest because individuals in this stage intend to take action in the next month , and it is stage where serious decision making processes take place. This stage is typically related to a higher frequency of attempts to acquire healthier habits in comparison to the Precontemplation and Contemplation stages.
Description of the clusters. The initial solution reached from the random subsamples was applied to the general sample, and four clusters were labeled corresponding to different patterns (shape, level, and scatter) based on the scores from the Pros , Cons, and Self-efficacy measures .
The profiles of the clusters suggest heterogeneity within the Preparation stage.
The Early Preparation cluster (30.47% of participants) was characterized by a shallow inverted V shape, and by above average scores on all the measures . This pattern indicates that even when participants consider acquire sun protection habits in the near future , there is a battle between the benefits and the inconvenience of doing so. Efforts should focus on the reduction of the Cons, or perceived negative aspects of the acquisition of healthy habits . The scores from the Self-efficacy measure are higher than what would be expected from individuals not intending to practice sun protective behaviors. This was the second largest group overall , and was stable across samples. When the Processes of Change were analyzed , results indicated that this group used these processes in the medium range . Participants in the Ear ly Preparation cluster are using 6 of 8 processes more than members of the Disengaged cluster. Fewer differences were reported in comparison to the Level Preparation subtype. Scores in the Sun Protection Behavior Scale (SPBS) were similar to the Level Preparation and Disengaged clusters . A small group of participants categorized in this cluster never use sunscreen, while a larger group (but less than half) reported using sunscreen with a SPF 15-29.
The Progressing cluster (15 .23% of all participants) showed a profile that resembles the one for people in the Action stage of change. This cluster was stable across samples. The pattern followed by this group , with above average scores on the Pros of sun protection and the Self-efficacy measure , but below average score on the Cons , describes a group of individuals that are ready to move to the next stage (it was characterized by a V shape). This profile may represent subjects that are cautions about declaring themselves members of the next stage. They may not be sure they are taking , or might consider taking , all the precautions necessary to protect themselves from sun exposure . When the Processes of Change were analyzed , results indicated that members of this subtype reported using 7 of 8 processes more than participants in the Level Preparation and Disengaged clusters. Also , they reported using 2 of the 8 processes more than members of the Early Preparation cluster. In relation to the SPBS , the Progressing cluster reported the highest score on this scale, showing that this group is the most effective for the implementation of specific strategies that reduce unprotected sun exposure. More than half of the members of this cluster are using sunscreen with a SPF 15-29, and more than one fourth of this group is using sunscreen with a SPF 30 or more .
Also , one fourth of this group is married , which is a higher proportion in comparison to the other clusters .
The largest subtype was the Disengaged cluster (44 .92% of all participants) , which was characterized by similar and around average scores on all the measures . Low level and low scatter characterized this profile . Compared to the previous subtypes , this cluster was less stable across subsamples. Individuals in this group may be minimizing the positive aspects of acquiring strategies to reduce unprotected sun exposure . When the Processes of Change were analyzed , results indicated that this group used these processes in the medium range . Scores in the Sun Protection Beha vior Scale (SPBS) were similar to the Early Preparation and Disengaged clusters . Most of the participants categorized in this group reported using sunscreen with SPF 15-29. In comparison to the Early Preparation and Progressing subtypes , a higher proportion of participants never use sunscreen . Also , in comparison to all the clusters , a higher proportion of participants categorized in this group are married .
The smallest subtype was the Disengaged cluster (9.38% of all participants) . The pattern followed by this group was the least stable across samples . All the three scores from the scales were below average . This may represent that neither the benefits nor the disadva ntages of acquiring healthy habits , or feeling convinced that sun protection would be used in challenging situations, are perceived as important measures . A need to address the significance of these concepts for the prevention of skin cancer should be emphasized . Regarding the Processes of Change, participants in the Disengaged group used the processes the least. Scores in the Sun Protection Behavior Scale (SPBS) were similar to the Early Preparation and Level Preparation clusters. When additional behavioral variables were analyzed , it was found that a higher proportion of members of this cluster reported never using sunscreen , in comparison to the other clusters. The smallest proportion of members of this cluster use sunscreen of SPF 15-29. None of the participants in the Disengaged subtype reported being married , which is a smaller proportion in comparison to the other clusters .

Conclusion
This study supports the presence of important differences within the Preparation stage of change across different beha viors . The Early Preparation group was more similar to the previous stage . The Progressing cluster resembled a profile characteristic of more advanced stages, since the Cons are below average, while scores from the Pros and Selfefficacy measures are well above average . The Disengaged cluster was the one that reported more risk behaviors across most of the variables included in the analysis .
Different from other studies , there was no cluster that most clearly exemplified the Preparation stage . Also, the Level Preparation cluster suggests that once people get closer to advanced stages (Action and Maintenance) , the balance between the Pros and Cons may be less comprehensible . The results of this study add to our understanding of adults that are considering changing their sun protection behavior in the near future.

Summary
Previous studies among smokers have found distinctive subtypes within stages based on scores from the Pros , Cons, and Temptation scales for smoking cessation and acquisition. Meaningful clusters were identified based on constructs of the Transtheoretical Model for behavior change, and these clusters differen tiated on other smoking-related variables external to the initial analysis (Anatchkova et al., 2005(Anatchkova et al., , 2006a(Anatchkova et al., , 2006bNorman et al., 2000;Velicer et al., 1995;Velicer et al., 2007). The pattern of results for the external validation analyses provided strong support for the model.
Differences between clusters in variables such as cigarettes per day and minutes until first cigarette wer~ consistent with the existing theory. Results in these studies were typically in the direction predicted by profile interpretation : clusters with profiles that resembled earlier stages, or that were interpreted as being less engaged with the targeted behavior, reported less healthy habits on most of the following analyses. The present study is an exploration of the application of a similar approach applied to a different behavior, the acquisition of sun protection habits . As in smoking acquisition and cessation, distinctive profiles were found. Differences between subtypes on other sun-related variables were in the direction predicted by profile interpretation. Groups with profiles that resembled earlier stages for the acquisition of sun protection habits also reported a higher level of high risk behaviors. These findings provide a strong support for the Transtheoretical Model and its application among different behaviors.
Clusters within stages for smoking have been demonstrated to be able to predict future behavior, which is very important for the development of interventi ons that apply the Transtheoretical Model for the promotion of sun protection habits. It was found for smoking cessation that the profiles within stages were described by patterns of stability It also should be noted that , even with the combined sample, the size of some subgroups remained very small, and thus these analyses may have limited statistical power.
The results of this study may not hold for other aspects of sun exposure , such as perceived suscept ibility for skin cancer , attitudes toward tanning , knowledge of the risk of exposure to artific ial ultraviolet light, and concern for appearance .
Fu ture research. The sample used in the present study is a portion of a sample collected from a larger , multiple behavior intervention study to guide a population of primary care patients to quit smoking, eat healthier , prevent skin cancer , an receive regular mammograms . Only information from the sun protection segment of the intervention was included in the analysis , which was conducted on only the treatment group (the control group wasn ' t includ ed in the study since they were not assessed on all the relevant variable s). As part of this intervention , participants were assessed at 6, 12, 50 and 24 months after the initial assessment. Additional studies could investigate how the identification of subtypes within stages can been applied for the prediction of future behavior for sun protection habits. Follow-up evaluations could be conducted to determine the proportion of participants that remains in the same cluster as in baseline is clustered in a subtype with a different profile . The measurement of over time progress and improvement in the target behavior can be measured by criteria as the percentage of participants that change to, clusters with profiles characteristic of more advanced stages.