Influence of Social and Psychosocial Factors on Summer Vacationers’ Sun Protection Behaviors, the PRISME Study, France

Objectives: Summer intermittent sun exposure is a major risk factor for melanoma. Socioeconomic position, cognitive and psychosocial factors play a role in sun protection behaviors but the underlying mechanisms are unknown. This study aimed to measure the influence of educational level on sun protection behaviors in French summer vacationers on the Mediterranean coastline, and to identify the mediating psychosocial factors in this pathway. Methods: In summer 2019, French vacationers aged 12–55 staying in coastline campsites were asked about their holiday sun protection behaviors, their knowledge, attitudes, perceived control, and social norm relative to sun protection. A structural equation model measured the direct and indirect effects of educational level on protection behaviors via cognitive and psychosocial factors. Results: Sun protection during vacation increased with educational level. Theoretical knowledge partially mediated this association, from 22% to 86%, particularly for intermediate educational levels. Conclusion: Our results highlight the importance of implementing suitable sun prevention interventions for vacationers, especially those with a lower socioeconomic position. Improving theoretical knowledge around sun protection may be an important part of broader efforts to encouraging improved preventive behaviors.


Introduction
Skin sensitivity is an important physical characteristic to take into account when studying the sun protection of individuals. In order to evaluate it, six physical characteristics items were collected in our questionnaire: skin color, eye color, hair color, presence of moles, tendency to sunburn the day after critical exposure, and tendency to tan one week after critical exposure.
The most commonly used classification, especially during clinical evaluation in dermatology, is the Fitzpatrick phototype (1): • Phototype I: very white / fair skin with often freckles, blond or red hair, blue or green eyes -Never tans and gets sunburns very easily. • Phototype II: fair skin and sometimes has freckles, blond, red or brown hair, green / brown eyes -Barely tans and easily gets sunburns. • Phototype III: medium light skin, chestnut or brown hair, brown eyes -Gradually tans and occasionally gets sunburns. • Phototype IV: dark skin tone, brown / black hair, brown / black eyes -Tans easily and rarely gets sunburns. • Phototype V: very dark skin, black hair, black eyes -Tans quickly and a lot, very rarely gets sunburns. • Phototype VI: black skin, black hair, black eyes -Never gets sunburns.
From the answers to the six questionnaire items, it would have been difficult to categorize the participants according to this classification because few would perfectly correspond to all the physical characteristics of a phototype. To do that, it would be necessary to decide on a prioritization between these six characteristics, what we did not want to do because it led to a part of subjectivity in the method (2,3).
In order to create the most homogeneous possible classes for these six characteristics using a statistical approach, we conducted a multiple correspondence analysis (MCA) followed by a hierarchical ascendant classification (HAC) using R-Studio version 1.3 (FactomineR).
The resulting skin sensitivity variable created was introduced as an explanatory and adjustment variable in the various models constructed in the PRISME study.

Multiple correspondence analysis (MCA)
The sample included 1355 individuals.
The MCA took into account six active variables containing 29 answer modalities (Table S1.1). The "extreme" individuals corresponding to darker to black skins are at the top right of the graph of individuals (Graph S1.1a). As is often the case in MCA, we observed the "Gutmann effect" on this graph in the form of a horseshoe. On axis 1 (Dimension 1), we observe an opposition between, on the left side of the graph, people with red/blond hair, light eyes and a very light skin color, and on the right side of the graph, people with dark skin, black eyes and black hair (Graph S1.1b).

Correlation (Eta coefficient) between active variables and the first three dimensions of the MCA-PRISME, France, 2019
The presence of moles (q4) has less influence than the other variables in the construction of the first three axes (Graph S1.2, Table S1.2). The skin, eye and hair colors on the other hand have a strong influence on the first two axes.  Test values highlighted in red represent a significantly positive association and blue values a significantly negative association between the modalities of answers and the associated dimension.
The first axis represents the skin sensitivity gradient, while the second separates extreme sensitivity from average sensitivity (Table S1.3).  With a total variance of 3.83 (29 modalities of answers / 6 variables -1), the first axis represents 10.2% of the total variance and the second axis 6.9%. The variance decreases rapidly after axis 2 (Graph S1.3). The average variance is 0.166, so the first eight dimensions have a higher variance than the average.

Hierarchical ascendant classification
By varying the number of axes from 3 to 19, we see that over 10 dimensions the histogram of inertia gain is strongly modified with a jump in inertia between 2 and 3 classes that does not allow for an efficient class division (Graph S1.4). In order to minimize the loss of inter-class inertia, the choice of an 8-dimensional MCA allowing the creation of 4 classes during the HAC seems the most relevant. The results of the HAC with these parameters were: Graph S1.5. Cluster dendrogram of the HAC with four classes (after MCA including 8 dimensions) -PRISME, France, 2019 Graph S1.6. Factorial plan of the HAC with four classes (after MCA including 8 dimensions) on dimensions 1 and 2 -PRISME, France, 2019 Inertia gain Table S1. 4

. Association between the answers of the active variables included in the MCA (8 dimensions) and the four classes of skin sensitivity created by the HAC -PRISME, France, 2019
Test values measure the discriminating power of each modality in the class. Values highlighted in red represent a significantly positive association (>=1.96 ) and blue values a significantly negative association (<=1.96) between the answer modalities and the class, which means that the proportion of individuals in this modality is higher or lower in this class than in the total population.

Conclusion
The associations found allowed us to describe the characteristics of the four skin sensitivity classes ( When comparing these characteristics with Fitzpatrick's phototype classification, class 1 appears to correspond to phototypes 1 and 2, class 2 to phototype 3, class 3 to phototype 4 and class 4 to phototypes 5-6. In our sample of 1 355 participants, 34% belonged to class 1, 46% to class 2, 17% to class 3 and 4% to class 4. The proportions found are consistent with the distribution of phototypes described in the French Cancer Barometer of 2015 (phototype 1-2=32%; phototype 3=40%; phototype 4=24%; phototypes 5-6=4%) (4). Moreover, participants of the PRISME study also had a measurement of their skin color on the inner side of the arm at inclusion (T0) with a colorimeter (5). If we describe this colorimetry measure (inside arm at t0) according to the four classes of skin sensitivity, we obtain values consistent with this classification (Table S1.6).