Introduction

Adolescent sexual health and wellbeing continue to be significant challenges worldwide. The burden of curable sexually transmissible infections (STIs) is estimated at 376 million new infections each year (Rowley et al., 2019; World Health Organization, 2018) with adolescents accounting for the majority of infections (Dehne & Riedner, 2001). Similarly, young people in Australia are disproportionately affected by STIs. In 2017, young people aged 15–29 accounted for 73% of chlamydia notifications and 53% of gonorrhoea notifications, both having significantly increased in the last 10 years (Kirby Institute, 2018). STIs result in significant burdens to the health system in terms of direct medical costs of ~ AUD$350 million per year excluding HIV (Chesson et al., 2017) and, if left untreated, can result in longer-term sexual and reproductive health issues.

While there are many possible factors affecting the sexual practices of young people such as interpersonal influence (from peers and family), media and demographics, prevention efforts to reduce the STI burden among adolescents have primarily focused on providing sexual health information through school-based comprehensive sexuality education. There is some evidence that providing sexual health information in school settings reduces HIV-related risk (Fonner et al., 2014) and increases condom use (Kirby et al., 2007). Interventions focusing on improving conceptual mediators, such as self-efficacy, knowledge and intentions, were found to be most successful at improving sexual practices (Poobalan et al., 2009). The vast majority Australian adolescents have access to a school-based program (Fisher et al., 2019b), though the quality, timing and relevance of the content remain insufficient (Ezer et al., 2020). Australian adolescents have access to a wide array of sources of sexual health information beyond school-based such as social media campaigns (e.g. Gabarron & Wynn, 2016), as well as more informal sources such as the internet (e.g. Buhi et al., 2009; Santor et al., 2007), friends and family (e.g. Chang, 2014; Giordano, 2003; Treboux & Busch-Rossnagel, 1995) and community organisations (e.g. Fisher et al., 2012). The efficacy of these sources of information is yet to be determined.

Social-cognitive theories of behavioural change predict that young people need better sexual health knowledge if they are to improve their sexual health practices (Afifi & Weiner, 2006). Sexual health information seeking consists of looking for answers to specific questions or conditions; examples include a Google search on where to get condoms, asking a friend, peer or other confidant about their experiences of using a condom and seeking expert advice on how to use a condom properly (e.g. from a healthcare provider). Young people increasingly cite the internet as a source of sexual health information together with a wide variety of other sources such as peers and parents (Buhi et al., 2009; Fisher & Kauer, 2019; Sutton & Walsh-Buhi, 2017), seeking information about sexual health more frequently than about other health topics (Buhi et al., 2009). While healthcare providers are important sexual health information sources (Sutton & Walsh-Buhi, 2017), information received from peers has a strong effect on health-risk behaviour (Beal et al., 2001) and is perceived as more useful (Kallen et al., 1983). Young people are more comfortable talking to peers about sexual health than to parents (Diiorio et al., 1999). In short, the act of seeking sexual health information and its related positive outcomes has been well researched. That said, less is known about those socio-cultural predictive factors that dispose young people to seek out information.

Seeking information to gain knowledge and make informed decisions is a complex process (Gray et al., 2005a). A large number of theories and models have been developed to better understand seeking intention, though often with limited or mixed results (Kahlor, 2010). A robust model, the Planned Risk Information Seeking Model (PRISM; Kahlor, 2010), successfully combined a number of theories and out-performed the individual theories, explaining 59% of the variance in general health information-seeking intent. Briefly, PRISM identified several theoretical constructs which predict a person’s intention to seek out health information. Constructs pulled together from various theories (e.g. theory of planned behaviour) include subjective norms, perceived control, attitude toward the act, risk perception and affect toward risk, and perceived current and insufficiency of knowledge. While subsequently validated in other fields, only a handful of studies have used PRISM to understand sexual health-seeking intent. A US study explained 29% of the variance in sexual health information-seeking intent (Willoughby & Myrick, 2016), while a modified model, incorporating beliefs about information sources or ‘channel beliefs’, explained 71% of the variance in online contraceptive information-seeking intent among Chinese young people (Jiang & Ha, 2019).

While models like PRISM may predict those factors that inform health-seeking behaviours, they do not identify or describe the different types of sexual health information seekers. Other research demonstrates that people with distinct information-seeking profiles or types varied in how likely they were to seek out information (Wei, 2016). Different types of information seekers can be represented using latent class analysis (LCA) with the nuance provided by delineating such types may further enhance PRISM and illuminate the most powerful pathways young people take to become better informed, thus allowing for targeted interventions to improve sexual health information-seeking intent.

Using latent class analysis, the present study aims to identify and characterise types of sexual health information-seeking adolescents in Australia and examine the relationship between those types and sexual health outcomes as different types of information seekers may ultimately result in different outcomes. Such findings will further expand scientific understandings of the characteristics which may influence sexual health information-seeking intent and their related health outcomes allowing future sexual health information design (e.g. websites, apps, health promotion materials, sexuality education curriculum) to respond to the diverse needs of adolescents. LCA is increasingly finding favour in the social and health sciences (Collins & Lanza, 2010) because it is a powerful tool for identifying clusters of beliefs, behaviours or values in a given population. Once a set of types, or classes, is identified through the LCA analysis, it is possible to then use multivariate analysis to predict membership in these classes. Factors such as gender, sexual orientation, geographic or cultural factors may dispose a person to membership in one class more than another. For example, rural or regional adolescents may experience a cultural context which encourages greater perceived information-seeking control resulting in greater confidence to seek sexual health information and from more sources (Sarbazi et al., 2019).

Materials and Methods

Sample

The current study used data from the online National Survey of Australian Secondary Students and Adolescent Sexual Health survey, a large national sample of 14–18-year-olds in Australia conducted from April to May 2018. A full account of methods and measures has been published elsewhere (Fisher et al., 2019a). Briefly, participants were recruited through Facebook advertising using minimum quota sampling; consent involved reading a participant information sheet and checking ‘I AGREE’ to participate. Participants (N = 6,929) answered questions about their sexual health knowledge, behaviours and educational experiences including on their sexual health information-seeking practices (see Fisher et al., 2019a for details). All study protocols were approved by the La Trobe University Human Ethics Committee (HEC18030).

Measures

Demographic Characteristics

Measures included gender (female, male or transgender and gender diverse), age (14 to 18), school type attended (government, Catholic, independent or not in school), remoteness recoded to major city or rural (inner/outer regional, remote and very remote; Statistics, 2018), self-identification as Aboriginal and/or Torres Strait Islander (i.e. Indigenous) and self-identified sexual orientation (straight/heterosexual or lesbian/gay, bisexual or not sure/questioning [LGBQ]).

Sources of Information

Participants were asked what sources of information they had ever used for advice about sexual health including the following 13 sources: doctor/GP, school counsellor, school nurse, teacher, youth worker, mother/female guardian/step-parent, father/male guardian/step-parent, female friend, male friend, older brother/sister, internet, school programs and community health services. Answers were dichotomous (yes/no); participants were also given the choice to not respond to the question by answering ‘prefer not to answer’ (treated as missing in the analysis).

Sexual Health Knowledge

The fifty-one items about sexual health included questions about the transmission, prevention, symptoms, impacts of STIs, HIV, hepatitis A, B and C, and HPV (see Fisher et al., 2019a for details). The items were summed together to form a single scale and found to be approximately normally distributed (skewness = − 0.27).

Scores ranged from 0 to 49 (m = 28.96, sd = 8.47) and were converted into three sexual health knowledge categories: low (less than 50% correct responses; 33%, n = 2,176), medium (50–79% correct responses; 62%, n = 4,153) and high (80% or more correct responses; 5%, n = 335). Low, medium and high categories were mapped to standard grading scales in Australia where less than 50% is generally considered as ‘not passing’ and 80% or more is considered very good or excellent.

Sexual Behaviours

This analysis focused on five dichotomous variables. As above, participants could also select ‘prefer not to answer’ which was coded as missing data for this analysis. All participants were asked if they had ever had vaginal or anal sex. Those who answered yes were also asked if they used a condom the first time they had sex, the last time they had sex, whether they were using oral contraceptive the last time they had sex and whether they had ever had sex resulting in a pregnancy. Participants were also asked about other types of contraception such as IUDs, implants, diaphragm and injections but uptake was very low, so these were excluded from the analysis.

Data Analysis

The open-source statistical program RStudio (version 1.2.5019) was used for the data analysis. The R code used to run the analyses was adapted from Mayerl et al. (2017). The code was changed for use with the current dataset and revised for the specific analyses needed.

Latent Class Analysis of Sources of Information

First, an exploratory latent class analysis was conducted using the poLCA function in R (Team, 2020) for polytomous variables to create different models of information sources that were used for advice about sexual health. Models with one to seven classes were estimated; the number of classes estimated was no more than seven to facilitate interpretation and ensure an adequate number of participants within each class for statistical analysis. Several goodness of fit indices of the models were evaluated to determine the best relative model fit including the Akaike information criteria (AIC) and the Bayes information criterion (BIC) and the difference in AIC and BIC between models, models with lower AIC and BIC are preferred, with BIC providing a better goodness-of-fit than AIC. The log-likelihood function was used to examine the relative improvement between model fit; the p value was used here to determine if there was a significant difference between models. We also considered relative entropy ranging from 0 to 1 to determine classification quality with values of 0.80 and above indicating a good separation between classes. The statistical model fit assessment was also accompanied by considerations about the usefulness of the model, particularly in relation to the proportion of participants belonging to each class and whether the groupings of posterior item probabilities could be interpreted and made sense.

Once a model was selected, the conditional item probability of each source of information belonging to a class was graphed and the likelihood of each participant belonging to a class was determined using posterior probabilities for membership of each latent class.

Multinomial Logistical Regression Model

The distribution of the latent class of information usage over demographic characteristics, sexual health knowledge categories and sexual behaviours was then examined using multinomial logistic regression using the R-package nnet function. Two multinomial logistic regression models were estimated with the latent classes of information usage regressed on (1) knowledge, age, gender, school type, location, sexual orientation and sexual activity, and (2) condom use at first and last sexual experience, use of the contraceptive pill at last sex and sex resulting from pregnancy for only sexually active participants, controlling for demographic characteristics by adding gender, age, school type, location and sexual orientation as covariates.

Results

Sources of Sexual Health Information Classes

Several latent class analyses were modelled with one to seven classes using the thirteen sources of sexual health information (see Table 1). There were no significant differences in the log-likelihood ratio between models with entropy only acceptable for the four-class model. The AIC and BIC improved with increasing number of classes; plots of these measures indicated that the point of diminishing returns was around three classes. The plausible value also points to the four-class model with increases of less than 1 after the four-class model.

Table 1 Model selection criteria of the seven models
Full size table

The probabilities of each item belonging to the four classes (the conditional item probabilities) are presented in Table 2.

Table 2 Conditional counts and probabilities of utilisation of sources for sexual health advice and information by latent classes
Full size table

Utilisation of sources with conditional item probabilities above 50% (Wilkerson et al., 2010) was included for class descriptions. The class with the largest estimated population sample (42.5%), labelled ‘Everyday seekers’, were highest for using mother, female friends, the internet, male friends and school programs as sources of sexual health information. The second-largest class, ‘Peer seekers’ (28.1%) consisted of using the internet, female and male friends. The third-largest class, the ‘Non-seekers’ class (17.6%), used very few, if any, information sources, with no source above the 50% mark. The fourth and smallest class, ‘Variety seekers’ (11.9%), used a large variety of the listed sources of information.

Demographics and Sexual Behaviours by Sources of Information Classes

Table 3 presents demographic characteristics, sexual health knowledge and sexual behaviours across the four classes. To illustrate class descriptions, the likelihood of class membership across the same variables were examined with a multinomial logistic regression (see Table 4).

Table 3 Proportions for various demographic and sex-related factors by sources of information class
Full size table
Table 4 Multinomial logistic regression model for sources of information classes
Full size table

Class Descriptions

Characteristics of the ‘Non-seekers’ Class

The ‘Non-seeker’ class (17.6%) were those who tended not to use any sources of sexual health information. They were more likely to be male and younger than the other classes as well as have lower sexual health knowledge and not yet have had sex. They, or their partner, were also less likely than the other classes to have been using a contraceptive pill at last sexual event. Non-seekers served as the reference group for the multinomial logistic regression model given the intent of the study was to understand information seeking, ideally in comparison to non-seeking.

Characteristics of the ‘Peer Seekers’ Class

As the name suggests, this class turned to friends as the most important source of information. The ‘Peer seekers’ class (28.1%) were more likely to be female, slightly older and attend Catholic schools than non-seekers. The odds of having medium, but not high, sexual health knowledge for peer seekers was 1.31 times higher than non-seekers. Peer seekers were 2.71 times more likely to have ever had sex but no more likely than non-seekers to have used a condom or contraceptive pill.

Characteristics of the ‘Everyday Seekers’ Class

The ‘Everyday seekers’ class (42.5%) are notable for their reliance on their mothers as a source of information (99.8% of this class) and to a lesser extent, fathers (40.5%). Other sources of information are important, but this group clearly trusts information that is from within the family unit—an everyday, convenient source of information. This group is more likely to be female and attend a Catholic school than non-seekers. The odds of having medium or high sexual health knowledge for everyday seekers was 1.42 and 1.93 times higher than non-seekers. Everyday seekers were 2.59 times more likely than non-seekers to have ever had sex and 1.76 times more likely to have used a contraceptive pill but no more likely than non-seekers to have used a condom.

Characteristics of the ‘Variety Seekers’ Class

This group gets their information from a variety of sources: friends, peers, school programs and parents (73% have asked their mother; 49% their father). Unlike the everyday seeker class, this group casts around more widely for information. The ‘Variety seekers’ class (11.9%) were less likely to be male compared to non-seekers. Variety seekers were 1.7 times more likely to be from non-urban areas and 2.19 times more likely to self-identify with an Indigenous background than non-seekers. Variety seekers were not different from non-seekers with regards to school type, though their odds of having medium or high sexual health knowledge was 1.91 and 2.76 times higher. Variety seekers were 4.78 times more likely to have ever had sex and 1.58 times more likely to have used a contraceptive pill but no more likely than non-seekers to have used a condom.

Discussion

The current study sought to identify and characterise types of sexual health information-seeking adolescents in Australia and their related sexual health outcomes. Four latent classes, based on past use or non-use of various information sources were identified and described as non-seekers, everyday seekers, peer seekers and variety seekers.

Adolescent females were consistently more likely than males to seek sexual health information compared to non-seekers. Previous research similarly notes females are more likely to seek out information, often mediated by age and other contextual factors (Hallyburton & Evarts, 2014). Interestingly, trans and gender diverse young people were proportionally either non-seekers or variety seekers, which may be a factor of low perceived risk (e.g. nature of sexual activity and partner, not yet sexually active; Magee et al., 2012). Lack of or perceived lack of access to services and information may also have been a factor for trans and gender diverse young people. No difference in class membership was found for lesbian, gay, bisexual and not sure/questioning participants, similar to other research on LGBQ information-seeking behaviour (Jabson et al., 2017). Findings suggest that gender may be an important factor impacting on sexual health information seeking.

Participants attending a Catholic school were more likely to have used interpersonal sources such as peers, mothers and/or school programs. Other research has identified the significant explanatory power of family background and related non-school factors in explaining Catholic versus public school differences (Xu & Kelly, 2020). Catholic schools are privately funded community-minded schools which teach Catholic values. It may be that parents who send their children to a Catholic school also promote tighter social and ‘trusted’ informational circles. Rural-based adolescents and Aboriginal and/or Torres Strait Islander (e.g. Indigenous) were more likely to use a large variety of sources. Around two-thirds of Indigenous Australians live in rural areas (Australian Bureau of Statistics, 2017) which may explain the similarity of this finding. Some research has suggested that rural dwellers and Indigenous persons seek information through a variety of sources (Islam & Ahmed, 2012), though no recent work appears to have examined the sexual health information-seeking behaviours of rural or Indigenous adolescents. Our work suggests that familial, geographic and cultural context may play an important role in sexual health information seeking.

The more sources used for sexual health information the more likely participants were to score higher on a sexual health knowledge scale. Internet usage among all types of seekers (peer, everyday and variety) was similarly very high (84–86%) supporting previous research on sexual health information seeking that young people may not be impacted by perceived seeking control, that is they feel capable of searching for and finding sexual health information online (Willoughby & Myrick, 2016). Our work expands upon this notion to suggest the more sources used beyond the internet, perhaps to corroborate online search results, the greater their sexual health knowledge. The high proportion of young people seeking sexual health information from the internet highlights the need for young people to develop good sexual health literacy to better navigate this information (Gray et al., 2005b). Everyday and variety seekers, those using more sources, were also more likely to report a contraceptive pill was being used by themselves or their partner at last sexual event, though no differences in condom use were found across classes. The strongest predictor of using more sources for sexual health information was having had sex. Previous research has found that sexually active university students were more likely to have perceived vulnerability to sexual health risks, particularly for pregnancy risks, and had a higher frequency of sexual health information seeking (Chang, 2014). Sexually active adolescents similarly may be seeking out more sources of information due to an increase in perceived vulnerability, particularly around pregnancy prevention, and thus be more likely to be using oral contraceptives. Similarly, it may be that being older, and thus more likely to have become sexually active, heightens a young persons’ perceived sexual health knowledge insufficiency (Kahlor, 2010) resulting in engaging in a wider search for information and greater uptake of contraceptive pill use instead of condoms.

The current study is not without limitations. Recruitment procedures resulted in a convenience sample. However, the large and diverse nature of the national sample provided the largest exploration of adolescent sexual health information seeking to date and was closely aligned with the national population of adolescents aged 14–18 in Australia (Fisher et al., 2019a, 2020); nearly 7,000 participants provided an excellent basis for expecting the observed differences between groups would be similar to the general population. The study occurred in a single high-income country; therefore, results may not translate to other cultural contexts. Finally, the study did not examine all constructs of PRISM in order to confirm the relationship of findings to the model. Rather, PRISM was used as an interpretive framework to understand the characteristics of the identified classes and suggest possible new pathways for future research on sexual health information seeking. Further mixed methods studies would be useful to explore these elements as well as further nature of the information each group sought as there may be differences in how each source was used between information user groups.

Findings support the constructs of perceived vulnerability to risks and perceived knowledge insufficiency in PRISM. The results also suggest that to more accurately predict sexual health information-seeking intent as well as seeking action, sociocultural factors such as gender, familial, geographic and cultural context should be considered as key mediators of the theoretical constructs of PRISM. Further, recognition of different types of information-seeking intent types, that is those who might use a few (e.g. peers) through to a large number (e.g. variety) of sources may more clearly illuminate distinct pathways for each type of information seeker in PRISM. Such nuances may in turn provide for more precise targeting of sexual health information design (e.g. websites, apps, health promotion materials, sexuality education curriculum) to respond to the diverse needs of adolescents. For example, given that more sources used in seeking out sexual health information was predictive of practices such as condom and contraceptive use, program designers looking to increase these practices could ensure a diversity of messaging across numerous platforms. Taken together, the results suggest that policy makers, particularly those focused on adolescent sexual health education and sexual health promotion, need to consider the variety of information-seeking types and ensure breadth and depth of both policy and funding streams to improve adolescent sexual health information seeking. Policy could explicitly call for a diverse portfolio of information sources and provide funding streams to support the implementation and further evaluation of a multitude of sexual health information sources for young people.