We sampled 4017 youth aged 15–25 years from an existing online panel designed to be representative of the US population.²³ Members of the panel completed Internet-based surveys using either their own computers or computers provided by the company operating the panel. The primary purpose for this survey was to measure the reach and correlates of a sexual health-oriented campaign (Get Yourself Tested: GYT).²⁴ Potential respondents who were 18 years or older were solicited directly; 51% of those solicited completed the survey. For those aged 15–17 years, we requested parental assent; 53% of respondents solicited in this manner completed the survey. An institutional review board of the CDC provided an ethical review of procedures and content.

Respondents provided age, gender, educational status, household characteristics and chose categories of racial and ethnic identity, as well as sexual orientation or identity (Table 1). The primary survey question on sexual health (see Box 1 for the full wording) captures attitudes towards seven dimensions that we abstracted from global definitions of sexual health;^1–3 that is, six of the seven dimensions were named in the WHO definition and generally in at least one of the other two definitions (we added spirituality from the CDC/HRSA Advisory Committee definition). The items were measured on four-point scales (very important to me – not at all important to me).

In assessing potential correlates of sexual health attitudes, we chose three main content areas: (1) sexual health-related discussions with sex or romantic partners; (2) sexual risk and protective behaviours; and (3) health-seeking behaviours. Each domain is a common feature of sexual health research,^4,22 without essentially repeating the content of global definitions. Within these domains, we constructed variables that would reflect pro-health behaviours (e.g. going for a sexual health check-up) and others that would not (e.g. not using a condom due to being drunk or high). For talking to partners about sexual health, we asked if respondents had discussed HIV, STD or pregnancy with their most recent sex partners. For sexual behaviour, we asked about sexual experience (any anal or vaginal sex), sex without a condom and reasons for sex without a condom (drunk or high, discussed risk). Finally, we asked if respondents had sought sexual health care (seen a doctor or nurse for a sexual health check-up) and if they had been tested for HIV or other sexually transmitted diseases (STDs).

The sample was weighted against the 2010 US Census for sociodemographic factors (e.g. race, age, household income), as well as for Internet access and non-response rates. We present basic descriptive data via weighted frequencies. Except as noted, all calculations were performed via the Complex Samples function of SPSS v21 (Chicago, IL, USA).

We ran a principal components analysis (performed in R to account for the sample weights) to test whether attitudes about the seven dimensions cohered with one another. Results yielded a single component that accounted for 77.6% of the total variance among the seven items. We then scaled the seven items, which had a weighted mean = 16.26 (95% CI = 16.07–16.44) and composite reliability = 0.91 (0.70 is a general standard for adequate reliability).²⁵ We analysed overall scores for the seven dimensions using the General Linear Model function. Outcomes include F tests and P values, along with estimated means and 95% confidence intervals for subgroups.

We retained analyses of individual items because each item has unique conceptual meaning. To look at associations between sexual health attitudes and other measures, we ran ordinal regressions of the weighted data. The seven dimensions served as outcome variables in separate equations with partner discussions and behaviours as predictors. We also assessed the associations between sexual health attitudes and demographic variables via odds ratios and used demographics as covariates in other equations. Finally, the nature of the exploratory matrix (i.e. seven dimensions by three domains, each with sub-variables) raises the issues of α inflation from multiple comparisons. We therefore emphasised patterns of findings in the results over individual associations.

Respondents generally reported that they found the dimensions to be important or very important (Fig. 1). The emotional dimension (68.1%, 95% CI = 66.0–70.2%) was most likely to be rated very important, followed by pleasure (57.8%, 95% CI = 55.5–60.0%). Mental (49.7%, 95% CI = 47.5–52.0%) and mutual (48.1%, 95% CI = 45.9–50.4%) dimensions were next, then the physical (39.8%, 95% CI = 37.6–42.1%) and spiritual (39.2%, 95% CI = 37.0–41.5%) dimensions. The social dimension was the least likely to be rated very important (34.3%, 95% CI = 32.1–36.5%). Less than 10% of the sample rated any dimension as not at all important.

We assessed age, gender, sexual orientation (heterosexual vs other), Hispanic/Latino ethnicity and race identification as demographic covariates of sexual health attitudes (Table 2). For the purposes of these analyses, we derived a composite race and ethnicity variable to enter into the ordinal regressions. Respondent gender (male vs female) and age were consistently related to sexual health attitudes; for example, female respondents and older respondents were more likely to report that dimensions of sexual health were important to them. In contrast, neither sexual orientation nor race/ethnicity was associated with sexual health attitudes, except for the spiritual dimension. Here, Hispanic and Black (non-Hispanic) respondents were more likely than White respondents to rate spirituality as important, which was associated with the importance of religion to respondents (OR = 11.56, 95% CI = 8.59–15.55). In subsequent regression analyses, we used gender and age as covariates. We did not use race or ethnicity because adding this as a covariate to spirituality as an outcome did not affect the findings.

The associations between attitudes to dimensions and the collection of discussions and behaviours we assessed were complex, but two major patterns are visible in the results. We also noted some examples of where overall scores could mask differences among the individual dimensions.

For the first pattern, and consistent with previous youth sexual health research,²² we found that the attitudes to the pleasure and physical dimensions were positively associated with discussing condom use, birth control and how to manage pregnancy. These discussions might reflect the importance of pregnancy relative to STD/HIV prevention for many youth, especially in longer-term relationships.^26,27 We also found that the overall scores were associated with increased likelihood of sexual experience and increased likelihood of sex without a condom. The importance of pleasure and physical dimensions, however, were associated with increased likelihood of sexual experience and decreased likelihood of sex without a condom. Moreover, among those who had had sex without a condom, those who rated pleasure highly were more likely than others to have discussed the risk of doing so ahead of time. These findings suggest that, for those who rate overall pleasure highly and for whom sexual intimacy is important, sexual activity is thoughtful. That is, respondents reported sexual experience, but were more likely than others to mitigate risk with condom use. Even for ‘risky’ sexual experience (sex without condoms), they tended to discuss the prospects ahead of time. These findings were controlled for age and gender, and show the value of looking at ratings for both overall and individual items.

In contrast, those who rated spirituality as important were least likely to discuss condom use or birth control, although they did discuss pregnancy. These respondents were also less likely to be sexually experienced than others. We lack the data to make firm inferences, but the pattern of results is consistent with a subset of respondents who value the dimensions of sexual health and who also value abstinence from intercourse. This subset, however, was equally likely overall to have had sex without a condom, and more likely to report being drunk as a reason for having sex without a condom. This pattern of results deserves further exploration; evidence is consistent with respect to reduced sexual activity among youth for whom religion is important,^28,29 but some researchers have also found that condom use is reduced,³⁰ while others have not.²⁹

The second pattern is a function of which discussion topics and behaviours were most consistently associated with sexual health attitudes. For sexual health discussions, the most notable point is that birth control and pregnancy as topics very much outweighed HIV and STD, both in absolute terms (i.e. a larger proportion of respondents discussed the topics with partners) and in terms of being correlated with sexual health attitudes. Aside from the possibility that youth might put considerations about pregnancy ahead of considerations about infection,^26,27 youth might simply not construe STD as a salient factor in sexual health, akin to the WHO definition of sexual health as ‘…not merely the absence of disease…’.¹

The three behaviours that were most consistently correlated with sexual health attitudes were sexual experience, discussing risk before sex without a condom and going for a sexual health check-up. We note here that discussing risk before taking it in the form of sex without a condom is one of the more complex discussions in a relationship, and that going for a sexual health check-up is one of the more complex and resource-consuming behaviours. Going for a sexual health check-up was also associated with four dimensions, as well as with overall scores. Sexual health attitudes were not generally associated with STD/HIV testing, possibly because testing is less dependent than is seeking care on individual action (e.g. as part of routine screening), and possibly because STD testing rates in the US are not especially high.³¹ In summary, the more complex and involved the sexual health behaviours, the more relevant were attitudes to sexual health dimensions.

The data are susceptible to socially desirable responding, recall bias for behaviours and other limitations of survey data. Although the response rate of ~52% is not unusually low for online surveys,³² youth interested in the topic might have been more likely to complete the survey. Particular to this paper, we lacked some nuance in relationship quality measures, being limited to sexual health discussions only. This limitation made it difficult to disentangle interpretations of sexual health attitudes from the possible overriding topic of pregnancy in relationships. We also used a relatively crude comparison for sexual orientation or affiliation (heterosexual vs other) to generate analytic power.