Background
Internet addiction (IA) is defined as an impulse-control disorder of Internet use that has negative impacts on daily life function, family relationships, and emotional stability [
1‐
3]. Contemporary adolescents grow up in an environment that Internet use has become increasingly important in multi-facets of life [
4]. Adolescents tend to have lower self-control for online activities and are more likely than adults to develop IA [
5]. The prevalence of IA among adolescents varies across countries [
2], ranging from 1.9 to 8.2 % in European countries [
6‐
9], from 2.3 to 20.7 % in Asian countries [
10‐
15], and from 6.7 to 26.7 % in Hong Kong [
16‐
18]. The variation is possibly due to different methodologies (e.g., measurement tools, mode of data collection, sampling methods) and time-frames used in these studies [
18]. Among adolescents, IA has been associated with psychosocial problems, including poor academic performance, poor parental-child relationships, physical and mental health problems, withdrawal from daily life, hostility, and aggressiveness [
2,
5,
17,
19‐
21]. Previous studies conducted in Asia (e.g., Taiwan [
20‐
25], Hong Kong [
16‐
18], South Korea [
11,
26,
27], and mainland China [
13,
28]) reported significant risk factors of IA, including sexual intercourse experience, family problems, low self-esteem, social isolation, impulsivity, Internet accessibility and other risk behaviors [
12,
27,
29‐
32].
As theory-based interventions are more effective than non-theory-based ones [
33], it is warranted to apply such theories to understand IA. The Health Belief Model (HBM), a commonly used theory, has been applied to explain various health-related behaviors (e.g., condom use [
34], tooth brushing [
35], and motorcycle helmet use [
36]) and to design related interventions (e.g., influenza vaccination [
37] and helmet promotion program [
38]) among adolescents. It consists of six constructs: perceived susceptibility, perceived severity, perceived benefits, perceived barriers, cue to action and self-efficacy [
39,
40]. According to the model, adolescents who are not threatened by IA (those who perceive low susceptibility and low severity), see good reasons to use the Internet (high perceived benefits), expect adverse consequences from reducing Internet use (high perceived barriers), find it difficult to reduce Internet use (low perceived self-efficacy to reduce Internet use), and are seldom reminded by others to reduce Internet use (absence of cue to action), are more likely than others to develop IA. This important theory has however, not been applied to understand IA among both adolescents and adults.
Furthermore, peer influence is one of the strongest determinants of risk behaviors among adolescents [
41‐
45], as they tend to respond strongly to social reward and conformity [
46]. According to the Theory of Planned Behavior, supportive subjective norm, which is defined as significant others’ approval of a particular risk behavior, is an important determinant of the risk behavior [
47]. In addition, affiliation with peers engaging in a particular risk behavior is a risk factor of that behavior (e.g., substance abuse, smoking, and alcohol drinking [
41‐
45,
48]). Such an association can be explained by processes such as social pressure [
47], social selection [
41,
49,
50] and social learning [
51,
52]. Previous studies have reported that affiliations with peers who use alcohol, illicit drugs, or have other deviant behaviors were risk factors of IA among adolescents [
19,
53], but the impact of the presence of peers with IA remains unknown. We contend that affiliation with peers who have IA is a potential risk factor of IA among adolescents. While the aforementioned types of risk behaviors are regarded by the public as harmful, undesirable misconduct, or even illegal, Internet use by itself could be neutral or even seen as beneficial in the eyes of some adolescents. Hence, peer influences on IA, which have not been well-studied, may be different from those of other risk behaviors. Research is warranted to clarify the relationship between peer influence and IA.
It is important to investigate the underlying mechanism behind the association between affiliation with peers with IA and IA among adolescents. According to the Peer Cluster Theory [
54], peers are strong socializing agents who actively shape other adolescents’ behaviors by changing their corresponding beliefs and attitudes. Peers with behavioral problems may serve as negative role models [
55], hence adolescents may directly learn and adopt favorable beliefs of Internet use and IA (e.g., low perceived severity of IA and high perceived benefits of Internet use) from peers with IA [
56]. Moreover, observing peers with IA who have failed to control their Internet use may lower one’s perceived self-efficacy on controlling Internet use through vicarious learning [
57], thus heightening his/her risk of developing IA. Therefore, cognitions related to the HBM are potential mediators of the association between affiliation with peers with IA and IA among adolescents.
In the present study, we investigated the prevalence of IA among Chinese secondary school students in Hong Kong, China. Two categories of associated factors were investigated: 1) cognitive factors based on the HBM, and 2) perceived number of peers with IA (PNPIA). We hypothesized that the association between PNPIA and IA, if any, would be mediated by some HBM factors. This is the first study investigating the role of the HBM to understand IA, including its direct and mediating effects.
Discussion
About one sixth of the sampled students were classified as IA cases. The prevalence was higher than that in mainland China [
10,
13] (2.4 to 8.8 %, according to Young’s criteria) but was slightly lower than that in Taiwan (20.7 %, according to CIAS) [
12]. Another local study, based on the 10-item Young’s Chinese Internet Addiction scale, reported prevalence of IA of 19.1 % among primary and secondary schools in 2008 [
16]. The prevalence of mobile phones with Internet applications in the Hong Kong general population aged 15 and over was 62 % in 2012 [
66] and about 80 % in 2014 (unpublished data). The prevalence of ownership of such phones was 74 % among our participants. Such high prevalence may speed up the spread of IA among adolescents; a hypothesis to be tested in future studies.
Our study found a statistically significant but mild sex difference in IA prevalence, with more male students classified as IA than female students. Shek’s study conducted in Hong Kong reported that male and female students had similar risk of having IA [
18]. Similarly, the sex difference was also about 3 % (4.04 and 0.84 % for males and females) in China [
10] and about 2.5 to 4.2 % in some European studies (6.2 to 6.7 % for males and 2 to 4.2 % for females) [
7,
9]. Hence, the sex difference in IA prevalence detected in this and other studies was smaller than that of the prevalence of other risk behaviors among Asian students, such as smoking (28 and 3 % in Beijing [
67], and 16.9 and 7.0 % in Korea [
27]) and alcohol use (33.6 and 13.1 % in Guangdong [
68]). Risk behaviors often involve cultural connotations such as masculinity, braveness and adventure [
69,
70], but health professionals and parents should keep in mind that IA is not only confined to male secondary students. We hypothesized that Internet use would be less gender biased as compared to smoking and alcohol use, as Internet use is related to new technologies, daily functions, and applications. It is warranted to test this hypothesis in future studies.
It is seen that IA increases with school grade, suggesting the presence of an age effect. However, we did not measure age in this study. Importantly, the prevalence of IA among Secondary 1 (7
th year of formal education) students was already close to 10 %. Therefore, the age of onset of IA may be younger than that of other unhealthy behaviors, such as smoking [
70]. IA is associated with multi-dimensional problems such as depression and academic behaviors [
5,
20,
71]. Its negative impact on young adolescents requires special attention as lasting harms could result. Early detection and early intervention for IA among students are warranted.
We found that single parenthood was a risk factor of IA, especially for those living only with the father. In Chinese culture, maternal influences are stronger than that of paternal influences [
72]. Previous studies reported that adolescents of single-parent families showed higher risks of substance abuse, drinking and smoking, especially among those living with only the father [
73]. The differential impact of living with the father versus the mother may be mediated through the trend that mothers are more likely than fathers to exercise control on their children [
74]. Interventions on IA hence need to pay special attention to single-parent families, especially to those involving a single father. Maternal education level, as compared to paternal education level, also showed stronger association with IA, again suggesting the presence of stronger maternal influence over paternal influence on IA. Since interventions for prevention of IA may require family participation [
11,
75], the relative strength of maternal versus paternal influences is potentially important.
To our knowledge, this is the first study applying the HBM to explain IA. In this study, perceived susceptibility to IA in the next year among non-IA cases was noticeable. However, we found that perceived severity of consequences of IA was relatively low as half of the IA cases did not perceive that IA would be harmful to himself/herself. Furthermore, perceived susceptibility and perceived severity were found to be positively associated with IA, despite negative associations expected by the HBM. The discrepancy may be due to causality issues in the cross-sectional study design, which have been observed in other studies involving the HBM e.g., [
76]. That is, those classified as having IA might be likely to recognize or experience some symptoms related to IA and hence in the absence of diagnosis, perceive that they would develop IA in the future, without knowing that they would have been classified as IA according to an external instrument. Furthermore, although health care workers worry about IA among adolescents, adolescents themselves might not see IA as problematic, as seen from our data. That may also explain why perceived severity was not negatively associated with IA, as some of those with IA might even see heavy use of the Internet as beneficial rather than harmful. Although some theories, such as those related to fear appeal [
77], have indicated that perceived susceptibility and severity would result in perceived threat which may deter adoption of a risk behavior [
77], our results suggest that IA prevention programs emphasizing the threat (perceived susceptibility and severity) associated with IA are unlikely to be effective. Therefore, fear appeal approaches [
78], which have commonly been used for prevention of other risk behaviors (e.g., substance use), may not be applicable for prevention of IA among students.
The adolescents’ reliance of the Internet may have become a barrier against reducing its use, as over half of them agreed with at least one type of perceived barrier for reducing Internet use (i.e., reduced communication with friends, feeling lost, bored and old fashion). Internet use may have become an “indispensable” means of social interaction and even an internalized identity of the student [
79]. In addition, perceived barriers were significantly and strongly associated with IA. Future intervention programs should hence consider how to remove such barriers. For example, school-based campaigns which encourage controlled Internet use and organize regular offline social activities for students may lower their perceived barriers for reducing Internet use and thus risk for developing IA [
80].
Consistently, strong social benefits of Internet use were perceived by the sampled students, as about 60 % of them reported at least one type of perceived social benefit. They may see Internet use as a key means of developing and maintaining friendships and obtaining instrumental and emotional support. The Internet has hence become an important resource, supporting the formation of new social networks and establishing social capital among students [
4]. According to social marketing principles, a behavior will be initiated if the perceived benefit exceeds the perceived cost [
81], which is likely to be true for the case of Internet use among students. Health workers hence cannot underestimate difficulties for preventing or treating IA problems among students.
Given that Internet use may provide strong perceived and/or actual benefits, students should be trained to control their Internet use. As expected, low perceived self-efficacy to reduce Internet use was significantly associated with IA. It is hence important to increase students’ perceived self-efficacy to regulate their own Internet use [
20,
25,
82]. School-based training like impulse-control techniques, time management and interpersonal communication skills have shown to be effective in improving perceived control over time and preventing Internet overuse [
83].
Contrary to our hypothesis based on the HBM, the Cue to Action Scale (i.e., reminders from parents and teachers/social workers) to reduce Internet use was positively associated with risk of IA. It is likely that those with IA were often reminded by their parents, teachers and social workers to reduce Internet use. Hence, IA might be a cause of parental reminders instead of the reverse that parental reminders are a cause of IA. Again, similar reversed causality problems have been reported when applying the HBM to other cross-sectional studies [
36]. However, it is important to point out that about one third of the IA cases reported that they had not been reminded by their parents to reduce Internet use. Previous studies have shown that parental control is protective against IA [
84]. Parents should hence be trained on how to recognize their child’s IA symptoms, communicate with their child on IA issues, and exercise appropriate control on his/her Internet use.
We further observe that the presence of peers with IA was very common. Males, as compared to female students, perceived more peers with IA. The sex difference is consistent with the observation that males have higher IA prevalence than females. However, the relatively small sex difference in IA (<3 %) could at most provide a partial explanation of the higher PNPIA among males. It is also possible that male students with very high intensity of Internet use might be more likely than females to be clustered together to play online games or perform other online activities, or be more aware of one’s peers’ Internet use. Such speculations need to be confirmed by future studies. We did not define IA when asking the question about PNPIA as the definition is a complicated one and the question was asked prior to asking the items of CIAS. It is true that the participants may have different definitions about IA, but as the term is used colloquially in Hong Kong, it is likely that our participants were referring to peers that were likely to have apparent problematic Internet use. Therefore, the variable of PNPIA may be assessing such a perception, which is a potential source of peer influence on students’ IA status. Peer influence is one of the strongest predictors of risk behaviors among adolescents [
41‐
45].
We also found that PNPIA was strongly associated with IA. Similar findings have not been reported in literature. Compared to adults, adolescents are less psychosocially mature [
85] and more prone to peer influences during the cognitive processes of decision making regarding their risk behaviors [
86]. Clustering and inducement effects may co-exist in this case [
54]. Students’ beliefs about the outcomes of reducing one’s Internet use may be directly adopted from other peers with IA, or students with IA may meet and become friends with other IA peers via the Internet. This study further contributes to the understanding of the indirect effect of PNPIA on IA development by identifying three partial mediators that were related to the HBM. As mentioned, we cannot assume adolescents with or without IA would perceive IA as harmful; they might even see it as beneficial. Unlike the case of having peers who are drug users or smokers, some adolescents and even their family members might not perceive any need to avoid peers with IA. Thus, there was no obvious obstacle filtering adolescents’ exposure to influences on Internet use exerted by peers with IA. Such influences might operate through mediators including increases in perceived social benefits of Internet use and perceived barriers for reducing Internet use, and decreases in perceived self-efficacy for reducing Internet use, resulting in an increased risk of developing IA. Interventions may consider modifying these constructs.
The students’ social context may offer some potential explanations for observed mediating effects. To the students, having peers with IA may increase perceived social benefits of Internet use as they can connect with their peers with IA in the cyber world. These heightened perceived social benefits could increase Internet use and may lead to IA. Similarly, having peers with IA may increase the perceived barriers for reducing IA, as reduction of such use may result in potential loss of approval from and connection with those peers with IA, and hence “the world”. The contention that students connect with some peers with IA via the Internet therefore explains the mediating effects of the two constructs related to perceived social benefits of Internet use and perceived barriers for reducing Internet use. Furthermore, the presence of IA peers may adversely influence students’ subjective judgment on self-competence in controlling Internet use through vicarious learning [
57].
The mediating effects have important implications for designing interventions. Ability to change perceptions on the three partial mediators of the associations between PNPIA and IA may neutralize potential influences of peers with IA on IA development among secondary school students. The findings of this study suggest the usefulness of group cognitive-behavioral interventions [
83], which can involve both IA cases and their peers with IA. Such interventions may focus on altering their cognitions of the three specific constructs of the HBM. Given the strong effect of peer influence, changing relevant cognitions related to IA through group interventions (e.g., class discussion) may be a good entry point for preventing IA among students. Peer education programs led by those peers who have recovered from IA and act as role models are also potentially useful alternatives for changing perceived benefits and barriers related to Internet use, as well as promoting perceived self-efficacy toward reducing Internet use. However, caution should be exercised when implementing group interventions involving unidentified IA cases, as such events may connect such non-IA cases with new and old peers with IA, and expose non-IA cases to negative cognitive influences from such peers with IA, hence increasing their risk of developing IA.
It is important to point out that the partial but not full mediation effect implies that other mechanisms may exist in explaining the association between students’ PNPIA and IA. We have mentioned potential clustering effects of students with IA as one of the potential mechanisms. Modeling effects and emotional factors may also be involved [
87,
88]. This is the first study attempting to discern such mediation effects. Future studies are warranted and allow us to understand better the mechanisms of peer influences on risk behaviors among students.
The study had some limitations. First, identification of IA and PNPIA relied solely on self-reported data and reporting bias may exist. In particular, the way that PNPIA was measured by one single question presents a limitation, as participants may have different understandings on the definition of IA and they might have referred to their responses as heavy Internet use or overuse, instead of IA. We were hence actually assessing perception instead of the true number of peers with IA. However, as IA is commonly used in daily language among adolescents in Hong Kong, the participants seemed to show no problem answering the question. We believe that PNPIA was a reasonable indicator of potential peer influences on Internet use that might increase students’ Internet use and hence risk of developing IA. Second, it was a cross-sectional survey, and no causal relationship could be established, as seen from the findings that some significant associations did not concur with the HBM. Third, a panel was formed to construct some scales to assess cognitions related to Internet use based on the HBM since such scales were not available. We restricted ourselves to social benefits when assessing perceived benefits of Internet use because social media is growingly influential, but other aspects of perceived benefits could have been used. Some constructs were only measured by one or two items (e.g., perceived susceptibility and perceived severity) and reliability of those indicators could not be assessed. One of the limitations is that the subscales formed had an uneven number of items, according to the panel’s suggestions. Last, we did not ask about age as we thought that age was relatively homogeneous with school grades. However, age might interact with the associations found due to maturation of the students. Future studies should investigate whether associations found in this study vary as age increases.
Authors’ contributions
YW performed the statistical analysis and participated in data interpretation and manuscript writing. AW participated in literature review, data interpretation, and manuscript writing. JL was responsible for the research conception and design, supervision, data interpretation, and finalizing of the manuscript. All authors have approved the final manuscript.