Participants were 6516 adult guests at 26 western U.S. and Canadian ski areas, who were recruited, consented, and interviewed on chairlifts from January to April 2001 (n=2991; 99.3% completion rate; 0.7% refused [n=23]) and January to March 2002 (n=3525; 99.1% completion rate; 0.9% refused [n=33]). Across both years, 382 guests (n=201 in 2001; n=181 in 2002) were ineligible because they were aged <18 (n=60); ski-area employees (n=198); previously interviewed (n=99); or not English speakers (n=25).
Located in Alaska, California, Colorado, Idaho, Montana, New Mexico, Nevada, Oregon, Utah, and British Columbia, the ski areas varied in size, lift-ticket prices, and guest demographics. All ski areas were National Ski Area Association (NSAA) members and had at least two aerial chairlifts.22
See Buller et al.22
for ski-area recruitment procedures.
Diffusion of innovations theory23
provided the theoretical framework for Go Sun Smart. Diffusion of innovations theory predicts that preventive behavior changes when persuasive messages are disseminated via multiple channels to inform a population about an innovative practice and to influence adoption of it—in this case, sun safety (i.e., applying sunscreen and lip balm and wearing a hat to avoid sunburn, as well as wearing eyewear to prevent vision damage).
Go Sun Smart, created by this study’s researchers, consisted of print, electronic, and interpersonal messages. Employees were the primary audience, but some employee-targeted messages were simultaneously communicated to guests. Guest materials included posters and brochures for ski and snowboard schools, signage at the base of chairlifts and on chairlift poles, electronic signs and grooming reports, brochures, and table tents and posters in lodges. An employee-training program advocated that employees advise guests against excessive sun exposure. The Go Sun Smart logo branded all materials, and the mention of three key behaviors appeared in all messages: wear sunscreen, sunglasses, and a hat. See Buller et al.22
and Walkosz et al.24
for a detailed description of the theoretical model and a complete program description, respectively.
Ski-area contact personnel received three sets of program materials at intervention areas (see experimental design below) from late December to early March to rotate messages and to address the increased UVR in spring. Contact personnel met with investigators in August 2001 and received Go Sun Smart program guides. Investigators visited contact personnel in November and December 2001 to review the program implementation protocol, and Go Sun Smart was implemented from January to April 2002.
Experimental Design and Survey Procedures
A pair-matched, group-randomized, pre-test/post-test, quasi-experimental design was performed.25
Independent samples of guests were recruited at baseline (2001) and post-test (2002) within a cohort of ski areas present in both survey periods and randomly assigned to either intervention or nontreatment control groups. Randomization of multiple groups creates a strong design, avoiding many common threats to internal validity and reducing contamination (while 22.7% of control guests reported visiting an intervention ski area and 45.8% of them recalled sun-safety messages, there was no evidence of contamination [visit ×condition and message recall ×condition, p
>0.05]). Cross-sectional samples reduced testing threats.25
Size and location (a surrogate for climate) of ski areas were the primary matching variables; ownership structure and the proportion of female employees were also included in the matching.22
Trained staff interviewed guests on chairlifts with a minimum run time of 4 minutes during 3-day periods (1 weekend day and 2 weekdays) to obtain pre-intervention data from January to April 2001 and post-intervention data from January to March 2002. Pairs of ski areas were visited during the same week in both years. Sample sizes per area varied from 57 to 220 participants determined by the number of guests on the mountain.
After boarding the chairlift, interviewers consented guests as approved by the IRBs at the authors’ organizations. The interviewer sat at one end of the chair and recruited the guest immediately next to him or her (if seated in the middle, the guest to the right). If the initial guest refused or was ineligible, another guest on the chair was recruited, but only one 4-minute interview was completed per lift ride. Answers were recorded in a spiral-bound survey booklet. Respondents were given a sunscreen lip balm after completing the survey. Interviewers completed 12–20 surveys per day. Interviews were completed on all eligible chairlifts, but the main lifts accessing large areas of the mountain were over-sampled.
Sun-protection behaviors were ascertained by asking if the guest was wearing sunscreen (yes/no or don’t know; and if so, the sun-protection factor [SPF], the parts of the body on which it had been applied, the time it had been applied, and whether it had been reapplied that day) and sunscreen lip balm (yes/no or don’t know; and if so, the SPF) and observing if the guest wore a head cover, neck cover, face cover, gloves, and eyewear. Two unweighted summed composite scores were created: (1) sunscreen SPF 15+ and lip balm SPF 15+ (range=0–2); and (2) sunscreen SPF 15+; lip balm SPF 15+; goggles; gloves; face cover; neck cover; and head cover (range=0–7).
Sunburning was measured by asking if the guest had ever been sunburned while skiing or snowboarding (yes/no or don’t know; if so, whether the guest had been sunburned that winter [yes/no or don’t know]). It has been recommended that sunburn be measured in program-evaluation studies when measurement time is limited. Sunburn was defined as skin that was red or painful, or both, from sun exposure but not exposure to wind or cold.26,27
The period was shortened to the winter season (rather than a year27
) to focus on the intervention period, but it was believed to be sufficiently long enough to capture this somewhat rare event. A similar sunburn measure was validated against a diary 6 months later,28
and a measure of ever-sunburned showed high test–retest reliability.29
Likert-type items (strongly agree –strongly disagree ) measured attitudes toward sun protection, self-efficacy expectations, sensation-seeking, and skepticism.
Several questions measured guests’ exposure to sun-protection messages. Guests were asked whether they recalled any messages about protecting their skin, lips, or eyes from the sun while at the resort (yes/no or don’t know; if so, whether it was on a poster, sign, brochure, trail map, or website). Guests also were asked if they had been told by anyone at the resort to protect their skin, lips, or eyes from the sun (yes/no or don’t know; if so, whether by a lift operator, ski/snowboard instructor, ski patroller, or someone else). In the follow-up survey, guests were asked if they had seen the Go Sun Smart logo (yes/no or don’t know; and if so, where). Potential contamination was determined by asking guests what other ski areas they had visited that winter (since November 1) and noting whether those who had visited other ski area(s) were randomly assigned to the intervention.
The guests were queried about their demographics (age, race, Hispanic ethnicity, education, gender); skin sun-sensitivity (always burn/unable to tan; usually burn/can tan if work at it; sometimes mildly burn/tan easily; rarely burn/tan easily) 20
; skiing/snowboarding expertise (beginner, intermediate, expert); number of days spent skiing/snowboarding during that winter (since November 1); time started skiing/snowboarding that day; and home ZIP code. The ZIP code was used to determine the distance from a guest’s home to the ski area (those living more than 200 miles away were classified as destination rather than local guests). Interviewers observed whether guests used skis or a snowboard (or other equipment). Weather conditions (i.e., cloud cover [sunny, partly cloudy, cloudy]; wind [none, light, moderate/strong]; precipitation [none, flurries/light snow, heavy snow, other]; temperature; and the time, the date, and the chairlift name were recorded.
Hypothesis A (guests at ski areas assigned to implement Go Sun Smart would engage in more sun protection than guests at ski areas in the control group) was tested by comparing the intervention and the control groups on the sun-protection outcome measures at the individual level nested within ski areas. Hypothesis B (guests with more exposure to Go Sun Smart would (1) engage in more sun protection and (2) express more favorable attitudes toward sun safety than guests with less exposure to Go Sun Smart) was first tested by examining the association of implementation scores with sun-protection outcomes.
Next, guests’ recall of sun-protection messages and sun-protection outcome measures was examined within both experimental conditions by evaluating the interaction between the experimental group and message recall. Comparisons were examined using a mixed-effects model,24,30
which is a variation of hierarchical linear modeling,31
and adjusting for the association among guests within ski areas (measured by the intraclass correlation)—which, when non-zero, inflates the variance among individuals beyond that estimated by traditional analysis methods and Type-I error (i.e., tests of significance are too liberal).32,33
The Statistical Analysis System (SAS), using the PROC MIXED program, was employed.34
Analyses included pairs to account for the effect of matching. Eighteen covariates were analyzed (days since November 1; time started skiing; cloud cover; wind; precipitation; temperature; proportion of days skied this season; expertise; local or destination guest; equipment; contamination [visit to intervention ski area]; skin sun-sensitivity; race—white/all others; sensation-seeking; skepticism; age; education; and gender), using backwards stepwise elimination (p
<0.05 criterion for retention). Significant covariates related to each outcome were included in the regression analysis. Outcome data were not transformed prior to analysis. Models employed casewise deletion of missing values on the outcome or covariates (due to item non-response). A p
-value of 0.05 was used to evaluate models, unadjusted for multiple comparisons.
Guests were a secondary target population; Go Sun Smart was expected to have a weaker effect on them than on employees (the primary target), and the study was not primarily designed to analyze guests. Hence, adjustments that would reduce statistical power (e.g., Bonferroni correction) were avoided, because there is an increased risk of Type-I error, and p-values were stated to permit readers to assess differences with their preferred correction methods. Adjusted proportions and means are reported.