The Problematic Internet Entertainment Use Scale for Adolescents: Prevalence of Problem Internet Use in Spanish High School Students

Many researchers and professionals have reported nonsubstance addiction to online entertainments in adolescents. However, very few scales have been designed to assess problem Internet use in this population, in spite of their high exposure and obvious vulnerability. The aim of this study was to review the currently available scales for assessing problematic Internet use and to validate a new scale of this kind for use, specifically in this age group, the Problematic Internet Entertainment Use Scale for Adolescents. The research was carried out in Spain in a gender-balanced sample of 1131 high school students aged between 12 and 18 years. Psychometric analyses showed the scale to be unidimensional, with excellent internal consistency (Cronbach's alpha of 0.92), good construct validity, and positive associations with alternative measures of maladaptive Internet use. This self-administered scale can rapidly measure the presence of symptoms of behavioral addiction to online videogames and social networking sites, as well as their degree of severity. The results estimate the prevalence of this problematic behavior in Spanish adolescents to be around 5 percent.


Introduction
I nternet addiction 1 (IA) has been frequently studied since its entry into the clinical lexicon in 1995, although it has yet to be officially recognized as a mental disorder by international organisms. Among the other terms used to refer to this condition are pathological 1,2 or problematic Internet use (PIU), 3,4 Internet dependency, 5 excessive Internet use, 6 and compulsive Internet use. 7 IA nevertheless remained the most popular term in publications 8,9,10 until the inclusion of Internet Use Disorder (IUD) in the appendix 10,11 to the fifth edition of the Diagnostic and Statistical Manual (DSM-5) of Psychiatric Disorders. The American Psychiatric Association 11 has proposed it as a possible nonsubstance addiction within the new DSM-5 category Substance Use and Addictive Disorders. This has raised the clinical legitimacy of the problem and has highlighted the need for further scientific research, even though recent advances embrace a variety of perspectives, 12,13 including more controlled designs, 14 reviews, 15 bibliometric studies, 12 meta-analyses, 16 and metasyntheses. 8,17 IUD is now considered as a behavioral addiction, a new, broad clinical entity that refers to repetitive impulsive behaviors that have negative effects on the lives of users and their relatives, and which is associated with mood, obsessive-compulsive, and, above all, substance use disorders. 18 Specifically, IUD is considered as a complex psychological construct that can be defined as a technological (behavioral) addiction related to nonessential, personal Internet activities (i.e., leisure, pleasure, or recreational) 3,5,19,20 that increase the time spent online and that cause marked disturbances in the subject's life. The symptoms include preoccupation, salience, tolerance, withdrawal, unsuccessful attempts to cut back on use, continued excessive use despite the negative consequences, engaging in Internet activities to alter moods (to escape or relieve them), relapse, craving, and conflict with others or with oneself (causing functional impairment that affects eating, sleeping, and physical activity 21 ). According to Griffith, 6 the Internet seems to provide a medium for this kind of behavioral addiction that develops certain teenagers as a result of online activities to counteract other psychological or physical deficiencies, or as a pseudocoping strategy.
Since 2001, a small number of IUD scales have been designed and validated for use with adolescents to measure their generalized PIU. 22 To date, seven scales have been developed for adolescents (see Table 1), almost all of them in Asia, and based on self-reports of high school students.
Epidemiological studies in this age group have been carried out above all in several Asian and European countries to estimate the prevalence of IUD (see Table 2). Observed prevalence rates have ranged widely, between 0 percent and 36.7 percent, due mainly to the use of scales designed for adults with different criteria and cut-off points. The most frequently used instruments are the Young Diagnostic Questionnaire (YDQ) 1 and the Internet Addiction Test (IAT), 21 although these scales have nonetheless attracted some criticism. 23 Furthermore, a few of these scales have been validated in clinical samples, which would be necessary to ensure they measure clinically relevant aspects of IUD; the Chen Internet Addiction Scale for Adolescents (CIAS) 24 is the first to draw-up clinically validated criteria for adolescents, which showed high diagnostic accuracy, specificity, and acceptable sensitivity. As Beard 24 notes, this is a good start, but further studies are needed. At present, there are at least five ways of establishing user categories according to the level of severity on IUD scales. The first is through the score on a specific scale: for example, it is diagnosed in subjects with five of the eight diagnostic criteria on the YDQ 34,40,44,47,52,64,65 ; or on the CIAS, it is diagnosed with a minimum score of 63/64. 35,37,41,48,59 A second approach would be to extract categories using percentiles (P), such as the 75 P 24 or 95 P 26 , similar to what is done in gambling research. Third, the problem user could be categorized by combining two common types of user that cause concern, namely the at-risk and addicted populations. 34,44,57,62 Fourth, one could select dependent and nondependent individuals using the mean (M) -½ standard deviation (SD). 32 Finally, the categorization could be obtained through multivariate statistical analyses, such as cluster analysis. 28,53,61 In light of the above, the present study has three objectives: 1) to design and validate a new scale for adolescents, the Problematic Internet Entertainment Use Scale for Adolescence (PIEUSA), which centers on major Internet entertainments such as online video games (OVG) and social networking sites (SNS); 2) to examine the associations between patterns of Internet use identified with the PIEUSA and both gender and age groups; and 3) to estimate PIU prevalence using the categories from the classical pathological gambling literature to assess levels of severity.

Participants
Students were recruited from public (state) and private schools in the city of Barcelona (Spain). The sample comprised 1131 adolescents aged 12 to 18 years (M = 14.55, SD = 1.816), and it was balanced in terms of both gender (53.5 percent male participants, 46.5 percent female participants) and age group (50.1 percent younger adolescents aged 12-14 and 49.9 percent aged 15-18).

Materials
This cross-sectional study applied an ad hoc paper and pencil questionnaire organized into three sections: 1) sociodemographic data (gender, age, family, substance consumption, and other hobbies); 2) patterns of Internet use (computer owner, user, age at first use, preferred online entertainments, self-rating of expertise from 1 inexpert to 5 very expert, frequency [days per week], duration [average length of a regular session], longest session [maximum time connected in minutes], and perceived effect, which was assessed by asking adolescents [using yes-no questions, plus three dichotomous items 67 ] whether they considered that Internet might affect them in any way); and 3) the PIEUSA, which was constructed using three sources: a text revision of the fourth edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV-TR) criteria for substance dependence and pathological gambling disorders in adults; the diagnostic criteria proposed for IUD in adults and adolescents, 1,4,24,35,68,69 and the literature on adolescent IUD prevalence. The scale contains 30 items rated on a seven-point Likert scale, from 1, strongly disagree, to 7, strongly agree. The total score therefore ranges between 30 and 210, with the highest score representing the maximum presence of the construct under study over the last 12 months (see instructions in Table 3).

Procedure
The instruments were administered by researchers during a regular classroom hour, having previously obtained permission from the head teachers of each school. To increase the validity of the students' responses, teachers were asked to be absent during the administration. The researchers instructed students, asked them to answer honestly, and guaranteed the confidentiality of their responses; all the students gave voluntary consent to participate.

Sample characteristics
Students were living with an M of four family members (M = 4.02, SD = 1.050). Their parents had at least secondary or sometimes higher education (82.15 percent), and both were employed (88.65 percent). There was a significant relationship between parents' educational level and the adolescents' total score on the PIEUSA (fathers: U = 18246.500, z = 2.426, p < 0.05, r = 0.12; mothers: U = 22547.500, z = -2.011, p < 0.05, r = 0.09), specifically in terms of the difference between parents with only primary education [median (Mdn) in relation to fathers in this group was 91; for mothers: Mdn = 92.5] and those with higher education (fathers: Mdn = 80; mothers: Mdn = 84). Almost a quarter of the students (23.2 percent) reported consumption of alcohol and/or tobacco, but this was not related to PIEUSA scores. Seventy-nine percent reported having other hobbies that did not involve technologies (mainly sports and arts), and these respondents obtained lower scores on the PIEUSA (Mdn = 84) than did those whose hobbies were all technology based (Mdn = 101) (U = 49573.00, z = -4.828, p < 0.001, r = 0.16).
Psychometric properties of the PIEUSA Factor validity. An EFA using the principal components technique was conducted on the 30 items of the PIEUSA. The value of the Kaiser-Meyer-Olkin index (KMO = 0.931) verified the sampling adequacy (N = 909), while Bartlett's test of sphericity (v 2 (435) = 9975.033; p < 0.001) indicated sufficient correlations between items. Application of the Kaiser criterion (kq1) indicated that a combination of five components explained 50.6 percent of the variance, while the screen plot showed two inflections after the first and fourth components. The largest fall was after the first factor, which explained 31.28 percent of the variance and had the highest eigenvalue. The scale could therefore be considered unidimensional. 70 A factor loading of 0.30 was used as a cut-off for items. 71

70/100
Item analysis and internal consistency. Table 3 shows the items that obtained the highest scores (M > 4, SD < 2.31; items 2, 10, 15, 17, 18, and 19) and the lowest scores (M < 2, SD < 1.6; items 20, 24, and 25). According to homogeneity indices, only one item (24: ''I have spent money on OVG or SNS entertainment'') showed a low correlation with the corrected total score. However, its correlation was near the cut-off point, and omitting it did not improve the value of Cronbach's alpha. We therefore decided to maintain this item. The internal consistency of the scale was excellent (a = 0.923).   Patterns of Internet usage in relation to gender and age group. In this sample, 91.7 percent owned a computer with a home Internet connection, and 82.9 percent reported using it regularly for entertainment (at least once a week) during the last year. Only a slight difference was found between age groups, with younger adolescents using the Internet more than older ones (v 2 (2) = 20.375; p < 0.001; V = 0.135). In the selfassessment of Internet expertise, male participants were significantly more likely to rank themselves as highly expert (v 2 (4) = 89.626; p < 0.001; V = 0.282), as were younger subjects (v 2 (4) = 32.813; p < 0.001; V = 0.171). Regular Internet use was initiated around the age of 10 (Mdn = 10, M = 9.93, SD = 2.698). Male participants started at a younger age (U = 88709.500, z = 2.798; p < 0.01, r = 0.09), and the members of the younger age group had also started at earlier ages (U = 53907.500, z = 12.007; p < 0.001, r = 0.4). Female participants reported a preference for only SNS or both forms of entertainment, while male participants preferred only OVG or both (v 2 (2) = 30.592; p < 0.001; V = 0.184). Younger subjects were more likely to use OVG (v 2 (2) = 20.064; p < 0.001; V = 0.149). With regard to the frequency of use, 71 percent reported using the Internet around five days per week (Mdn = 5, M = 4.75, SD = 2.212). The usual time per regular session ranged from one hour (in minutes: Mdn = 60, M = 78.68, SD = 74.191) to two hours (in minutes: Mdn = 120, M = 154.05, SD = 156.522). Overall, 31.6 percent of adolescents reported that the time spent affected them in some way (i.e., with regard to homework, friends, and sleep hours).
PIEUSA total score in relation to gender and age group. Inferential bivariate analyses showed that M scores on the PIEUSA (see Table 4) were higher in male participants (F (1,903) = 10.251; p < 0.01; r = 0.12) and in the younger age group (F (1,903) = 9.030; p < 0.01; r = 0.1). However, no interaction effect was found, and nor were there other statistical differences among the variables related to the adolescents' characteristics and their Internet use. A weak inverse correlation was detected between the PIEUSA total score and age (r = -0.113; p < 0.01; r 2 = 0.01), as well as in relation to the age when Internet use was initiated (r = -0.139; p < 0.01; r 2 = 0.02).
Prevalence for the use of Internet entertainment. Following standard practice in gambling studies, we applied rigorous statistical criteria based on the 15, 80, and 95 percentiles, 26 for which the corresponding PIEUSA scores : v 2 = 654.643, p < .001), as well as between each pair of them. When combining at-risk and problem users, the potential PIU prevalence increased to 20.5 percent.
The group of problematic users (N = 45) did not differ from the other groups in terms of sociodemographic characteristics, but significant differences were found in relation to their Internet use (Table 5). Problematic users started using the Internet as entertainment earlier (F (3,739) = 5.182; p < 0.01; r = 0.15), did so with almost daily frequency (H (3) : v 2 = 51.886; p < 0.001), spent more than twice as long as occasional users (H (3) : v 2 = 60.743; p < 0.001), and were the group most likely to report that Internet entertainments were affecting them in some way (v 2 (3) = 13.001; p < 0.01; V = 0.136). Of this group, 62.2 percent were male participants, and their average age was 14 (SD = 1.413); furthermore, 30.2 percent consumed tobacco or alcohol; 28.9 percent had technologies as their main hobby; and 68.2 percent considered themselves to be highly expert. Finally, the median score of these problem users was above 5 on almost all the scale items, which means that they presented all the symptoms of IUD. The only items on which the median scores were five or lower were items 24 and 25, which correspond to an indicator of possible gambling and to the symptom of conflict, respectively.

Discussion
This psychometric study concerns the construction and validation of a new PIU scale, which constitutes a novel instrument for the rapid and accurate measurement of maladaptive patterns in the use of online entertainments such as OVG and SNS. At the time of the study design, the PIEUSA was set to be the first scale to be developed for adolescents in a Western language, since the CIAS was in Chinese. However, it can now be considered the fourth scale created exclusively for adolescents and incorporating user categories, after the CIAS, 24 the Problem Internet-Use Screening Tool (SCREEN), 26 and the Internet Dependence Scale (IDS). 28 The findings should be regarded as preliminary as the study presents several limitations. First, the cross-sectional nature of the design means that causality cannot be inferred, and it therefore remains to be confirmed if higher PIEUSA scores or the category of problem users could predict the symptoms of IUD in adolescents. Second, the sample comprised a nonrandom selection of Barcelona high school students, and this narrows the generalizability of the findings; note however that the large number of participants and their answers, which are similar to those reported in other Spanish studies involving adolescents, 72,73 suggest that the data are representative of the wider secondary student population. Third, the PIEUSA is a self-report instrument, although it was answered voluntarily, and in the presence of the researchers in an appropriate setting, which should increase the honesty of responses. Fourth, it was not possible to validate the scale with a clinical sample, this being a limitation of all the studies reviewed, except those concerning the CIAS. 24,35 Despite the above limitations, the PIEUSA does present similar psychometric properties to other IUD scales, showing a high reliability and acceptable validity. 16,74 Furthermore, it fills a gap in the literature by addressing one of the most prominent issues in this field: assessing levels of problem use of online entertainment among adolescents. 31 Having established the statistically highest cut-off points for PIU prevalence, the rates of problematic Internet entertainment use and potential PIU in these Spanish adolescents were found to be 5 percent and 20.5 percent, respectively, which are within the prevalence range reported in adolescent epidemiological studies. 23,27,28,53 The prevalence of potential PIU was similar to that obtained with the CIAS, 24,35,37,41,48,59 which seems to be the best IUD scale for use with adolescents; perhaps it could be considered as a gold standard.
Our study agrees with other adolescent Internet use studies in identifying the following as possible risk factors to consider: being male, being a young adolescent, living with parents, 36 daily Internet use, and recognition of a behavioral problem. We also detected a cohort effect: in that, younger adolescents start to use Internet as a form of entertainment earlier and gain expertise faster, and the longer they spend on it the more likely, they are to develop problem use. However, in contrast to other studies, 24,34 the problematic user profile presented no evidence of sociodemographic differences; 48 only as regard the time of Internet use, they spent twice as long as nonproblematic users, 45 with an average of 2.25 hours per day and around 15-25 hours weekly. 26,36,45,49 To distinguish problem use from high engagement or temporary absorption, exploratory qualitative studies are Note: The first four variables are quantitative, the numbers reflects the mean, and the standard deviation between brackets, the last variable is qualitative; the frequency is shown and the percentage between brackets. now required to shed light on IUD symptoms in adolescent patients, 75 and to assess the testimony of their counselors. 76 For example, screening clinical interview protocols 30,75 could be used to obtain evidence of IUD to complement the data provided by the existing scales; likewise, confirmatory studies could help to establish new advances (for instance, in the psychological mechanisms underlying IUD). Proposals for diagnostic criteria in adolescents are also needed, 24,31 as are scales aimed specifically at this population, with clear and common cut-off points.
In summary, after seventeen years of scientific production, this young field of research currently has at least three goals 12 : (1) to reach a consensus on the IUD construct and its operationalization; (2) to develop more reliable and validated scales for adolescents, scales that not only use clear methods to categorize severity, but that are also psychometrically adapted to different languages and cultures to facilitate comparative studies; and (3) to validate scales in clinical populations, which would help to identify and diagnose it more accurately, offering both sensitivity and specificity.