Ultraviolet radiation from the sun or indoor tanning is the cause of most skin cancers [1]. Although indoor tanning has decreased in recent years, it remains most common among adolescents and young adults [2,3], whose skin is particularly vulnerable to long-term damage [1]. US states have adopted several types of legislation to minimize indoor tanning among minors: a ban on indoor tanning among all minors, a partial minor ban by age (eg, <14 years), or the requirement of parental consent or accompaniment during indoor tanning. Currently, only 6 US states (Alaska, Colorado, Iowa, Montana, New Mexico, and South Dakota) have no indoor tanning legislation for minors [4]. The increase in indoor tanning restrictions may explain reductions in the number of indoor tanning providers, consumer spending on indoor tanning [5], and past-year indoor tanning among girls (24.1% in 2009 and 9.5% in 2015) and boys (5.7% in 2009 and 3.3% in 2015) attending high school and young adults aged 18-34 years (14% in 2007 and 4% in 2018) in the United States in recent years [3,6]. More stringent regulations have been associated with greater reductions in indoor tanning behavior and have been estimated to have a greater impact on melanoma incidence, mortality, and cost [7-10].

Internet search trends indicate public interest in a topic and are associated with actual health-related events such as influenza and COVID-19 outbreaks [11,12], medication use [13,14], melanoma mortality rates [15], and tobacco- and alcohol-related policy changes [16,17]. This study investigated whether internet searches (as an indicator of interest) related to indoor tanning varied across US states by the type of indoor tanning legislation, using free, publicly available data from Google Trends from 2006 to 2019. We hypothesized that the reduction in search rates over time would be associated with stricter indoor tanning regulations (eg, a ban on indoor tanning among all minors).

Data were downloaded from Google Trends [18]; these data reflect how many searches have been conducted on a specified topic relative to the total number of searches on Google within the selected time frame and geographic location. Search volume indices range from 0 (no searches) to 100 (peak number of searches). We selected the topic “indoor tanning,” which includes related search terms (eg, “tanning bed”). The indoor tanning time series consists of search volume indices from January 2006 to October 2019 for each state, along with the District of Columbia, and the United States as a whole. Google Trends data were available for 2004 and 2005, but state data were not sufficient for analysis.

In order to study longitudinal trends, seasonal effects were first excluded from the time series, since indoor tanning is most popular during spring in many parts of the United States [19]. We fitted 2 linear models for each state to evaluate the Google Trends data on indoor tanning and their association with the legislation type (a ban of all minors [n=22], a partial ban [n=10], requirement of parental consent [n=13], and no legislation [n=6] as of October 2019) as documented by the National Conference of State Legislatures [4]. Model 1 is a change-point model with the date of legislation enactment as the change point and as an outlier, since we observed an additional peak for some states on the date of legislation enactment. The first legislation was enacted in Wisconsin in 1991 (a partial ban), and the latest legislation included in the analyses were those enacted in Maine and Maryland (complete bans) in September and October of 2019, respectively. Model 2 is a model without any change points. To account for the correlations among adjacent time points in both models, an autoregressive moving average error structure was used. The fitted slopes (change rates) for all states in both models were calculated. The association between legislation type and the fitted slopes was assessed using a heterogenous variance model owing to unequal variations in slopes among legislation types. For Model 1, the difference in slope before and after the date of legislation enactment was first evaluated to determine whether Model 2 was sufficient for comparisons among legislation types. For multiple comparisons, P values were adjusted on the basis of the Tukey method for multiplicity adjustment. A P value less than .05 was considered significant. Statistical analyses were conducted using R (version 4.0.1, The R Foundation) [20] and SAS (version 9.4, The SAS Institute) [21].

Figure 1 shows the trend in Google searches related to indoor tanning for the United States overall with the fitted regression line for the no-change-point model. The slope (change rate for tanning search trends) decreased over time. When individual states (Multimedia Appendix 1) were grouped by legislation type, the decreasing slope rates differed. These decreasing rates were greater for states that imposed a ban on all minors than for states with other types of legislation. We calculated P values to compare slopes before and after the date of legislation enactment, based on change-point Model 1, for states that imposed a minor ban. Before multiplicity adjustment, P values were significant for only the District of Columbia (2015), Delaware (2015), and Nebraska (2014) (P=.02-.04). After multiplicity adjustment, all these P values were not significant (P≥.44). In addition, the slope differences before and after the enactment of the legislation by legislation type were not significant (P=.84). Hence, we only compared the legislation types in accordance with Model 2.

As indicated in Model 2, the legislation type was significantly associated with a reduction in search rates (slopes) (P=.01). Table 1 shows the results of pairwise comparisons in the reduction in search rates by legislation type. We found that the reduction in the indoor tanning search trend was more marked for states that imposed a ban on all minors (0.6% per month smaller search volume) than for those that imposed a partial ban (P=.009, adjusted P=.04). Furthermore, this rate reduction was more marked for states that imposed a ban on all minors (0.5% per month smaller search volume) than for those that required parental consent (P=.005, adjusted P=.02). Moreover, this rate reduction was more marked for states that imposed a ban on all minors (0.2% per month smaller search volume) than for those with no legislation; these values were borderline significant before adjustment but did not significantly differ after multiplicity adjustment (P=.08, adjusted P=.28). The rate reduction was more marked for states with no legislation than for those requiring parental consent and those that imposed a partial ban, but these values did not significantly differ after multiplicity adjustment (adjusted P≥.22). Finally, rate reductions between states requiring parental consent and those that imposed a partial ban did not significantly differ (P=.22, adjusted P=.98).