Psychometric Validation of a German Version of the Compulsive Internet Use Scale

Abstract

Excessive Internet use and Internet addiction are currently increasing in many industrial nations. Verified and validated measuring instruments could contribute to a better understanding of this still quite recent development. The aim of this survey was to investigate the psychometric properties of a German version of the Compulsive Internet Use Scale (CIUS). We surveyed a representative German quota sample of 1,723 adolescents aged between 14 and 17 years, as well as one parent each, with standardized questionnaires. In addition, adolescents and parents were asked whether the media use by the youth was considered problematic or excessive, and whether it led to arguments at home. We conducted confirmatory factor analyses (CFA) with maximum likelihood estimation to examine the factorial validity of the German CIUS, as well as reliability and correlation analyses. The results of the CFA indicate good psychometric properties for the German version of the CIUS. They appear in line with the findings for the original version of the questionnaire. Furthermore, the CIUS showed high internal consistency, and we found significant correlations between the “CIUS-summary score” and different ratings of problems with the media usage by the youth and the parents. The German version of the CIUS seems to be a valid and suitable diagnostic tool for measuring problematic to pathological Internet use.

Introduction

Diagnostic investigation of Internet addiction

Excessive Internet use and its severe form, Internet addiction, are currently increasing in many industrial nations in Asia, North America, and Europe.¹ Verified and validated measuring instruments could contribute to a better understanding of this development.² Several assessment tools for the diagnostic investigation of pathological Internet use are currently available, but most of them lack psychometric validation. The best known and most widespread questionnaires are the Internet Addiction Test (IAT),³ the Chen Internet Addiction Scale (CIAS),⁴ and the Compulsive Internet Use Scale (CIUS).⁵

To date, the IAT is the most frequently used and psychometrically validated instrument, consisting of 20 items. Widyanto and McMurran⁶ examined the IAT and extracted six factors (with internal consistency, i.e., Cronbach's α between 0.54 and 0.82). Ferraro et al.⁷ also reported a six factor solution for the Italian version of the IAT. Chang and Law² investigated a Chinese version of the IAT and extracted three factors, whereas Ngai⁸ described a four factor structure for the Chinese IAT. Chong Guan et al.⁹ identified a five factor solution in a validation study for the Malay Version of the IAT.

In contrast to these findings, Khazaal et al.¹⁰ reported a single factor solution for the French version of the IAT (Cronbach's α=0.93), as did Korkeila et al.¹¹ for the Finnish version of the IAT (Cronbach's α=0.92). In a current study, Widyanto et al.¹²^(p143) described a three factor structure for the original version of the IAT. In another recent survey, Jelenchick et al.¹³ also used the original version of the IAT and extracted two factors (Cronbach's α 0.83 and 0.91). At the moment, the empirical findings for the IAT are heterogeneous, varying between different translations as well as between various samples investigated with the same version of the IAT.

The CIAS was developed in China and is a questionnaire consisting of 26 items. The five scales of the CIAS showed reliability coefficients (Cronbach's α) between 0.79 and 0.93.⁴ In two studies, appropriate cut-off values for the CIAS were identified, first for adolescents and afterwards for college students.^14,15 Kesici and Sahin¹⁶ described a validation of the Turkish adaptation of the questionnaire of Chen et al.,⁴ and confirmed a five factor solution. However, the reported results of the confirmatory factor analysis (CFA) indicate a rather poor model fit (root mean square error of approximation (RMSEA)=0.11, standardized root mean square residual (SRMR)=0.08, goodness of fit index (GFI)=0.66).¹⁶ The briefest of the three mentioned questionnaires is the CIUS, consisting of 14 items and “… providing a dimensional score (i.e., severity) of problematic Internet use.”^p398

Empirical findings on the CIUS

For the original Dutch version of the CIUS, Meerkerk et al.¹⁷ conducted a validation study with 16,925 respondents aged 11 to 80 years (with 20% younger than 18 years). Meerkerk et al.¹⁷ reported a single factor solution with good fit indices (RMSEA=0.054, comparative fit index (CFI)=0.986) and high internal consistency (Cronbach's α=0.90). Khazaal et al.¹⁸ examined 155 Internet users (most of them adolescents, average age: 17.6 years) with an Arabic version of the CIUS. They reported that the “one-factor model of the CIUS has good psychometric properties and fits the data well” (CFA statistics: RMSEA=0.06, SRMR=0.06, CFI=0.90). They also reported a “satisfactory” internal consistency (Cronbach's α=0.78).¹⁸^(p3) Another validation study with a French version of the CIUS was conducted by Khazaal et al.¹ with 126 persons (average age: 23.3 years). Khazaal et al.¹ summarized their results saying that “the CFA indicates that a one-factor model of the CIUS has good psychometric properties and fits the data well.”^p402 They also indicated RMSEA=0.08 and CFI=0.92 as fit indices, plus a Cronbach's alpha of 0.91.

Aim and research questions of the study

The aim of this explorative study was to investigate the psychometric properties of a German translation of the CIUS. We explored the following research questions:

RQ1: Does a confirmatory factor analysis of the German version of the CIUS show results comparable to those reported for the original Dutch version of the CIUS?

RQ2: What reliability (measured by Cronbach's α) does the German version of the CIUS show?

RQ3: What are the correlations between the “CIUS-summary score” and external criteria (e.g., adolescents' and parents' perspective of excessive or problematic media use)?

Methods

Data collection

We surveyed a representative German quota sample of 1,744 adolescents (age range: 14 to 17 years) together with one parent each. Data collection was carried out by an experienced market research institute. In advance, our research group defined fixed targets concerning age and gender of the adolescents. It was defined that the target sample would consist of 14- to 17-year-old adolescents, with 50% males and 25% adolescents in every age group (14, 15, 16, and 17 years). The market research institute determined target values for all other important sample quotation features (type of school, single parents, number of inhabitants in a town, and federal state) to assure that the sample is in fact representative of Germany. In the data collection, the target achievement was very close to the target values fixed in advance. Interviewers in all 16 German federal states collected adolescent data in face-to-face interviews.

Measures

The German version of the CIUS consists of 14 items with a 5-level response format (0=“never,” 1=“seldom,” 2=“sometimes,” 3=“often,” 4=“very often”). The studies published to date consistently indicate a one-dimensional factor structure of the CIUS. The items of the CIUS are based on “… the seven criteria for substance dependence and the 10 criteria for pathological gambling in the fourth edition of the Diagnostic and Statistical Manual of Mental Disorders, as well as on the Griffiths formulated criteria for behavioral addictions.”¹^(p398),¹⁹ Meerkerk et al.¹⁷ described the core symptoms required for Internet problem behavior: “loss of control (items 1, 2, 5, and 9); preoccupation (including mental and behavioral preoccupation) (items 4, 6, and 7); withdrawal symptoms (item 14); coping or mood modification (items 12 and 13); and conflict (including inter- and intrapersonal conflict) (items 3, 8, 10, and 11).”^(p2) By summing up the values of all 14 items of the questionnaire, a “CIUS-summary score” was determined and a high sum value indicates problematic or pathological Internet use.

In addition, adolescents and parents were both examined on whether the “use of computers, of the Internet and of gaming consoles” by the youth was considered problematic (4-level response format: 1=“unproblematic,” 2=“low problem level,” 3=“medium problem level,” 4=“high problem level”) and how often this media use seemed to be “excessive” (1=“never,” 2=“seldom,” 3=“sometimes,” 4=“often,” 5=“very often”) or led to an “argument at home” (1=“never,” 2=“seldom,” 3=“sometimes,” 4=“often,” 5=” very often”) between parents and the youth. Furthermore, the adolescents were asked to rate their general life satisfaction (“How satisfied are you at the moment with your life as a whole?”).

Sample

The survey included 1,744 adolescents, with 50% males and 25% in every age group (14, 15, 16, and 17 years) each. On average, the juveniles were 15.50 (SD=1.12) years old. In total, 39% of the sample attended the equivalent to a grammar school, 23% attended the junior high school, and another 19% a secondary modern school. In our sample, 14% of the adolescents went to a comprehensive school and 3% to a special school. Almost 2% had started their vocational training. A total of 27% of the sample lived in large cities (with more than 100,000 inhabitants), 29% lived in medium-sized towns with between 20,000 to 100,000 inhabitants, and 44% in small towns (20,000 inhabitants maximum).

Statistical analyses

All statistical analyses described below were performed with the Predictive Analysis SoftWare v18 (PASW) and Analysis of MOment Structures v18 (AMOS) (SPSS Inc., Chicago, IL). All data sets with valid values in the CIUS summary score (n=1,723) were included in the different analyses. One percent of the sample or 21 cases had missing CIUS values and was therefore excluded. We conducted confirmatory factor analyses (CFA) with maximum likelihood estimation to examine the factorial validity of the German translation of the CIUS. Analogous to the available findings,^1,17,18 we conducted a model specification with 14 measured or observed variables (all items of the CIUS) and one latent factor. In the first specified model, correlations between the residual error terms were not permitted, while in our second model, correlations of the error variances (analogous to the other published CIUS-validation studies) were approved. In selecting which residual error terms were correlated, we followed the proposals of Meerkerk et al.,¹⁷ and otherwise based them on the results for the modification indices of the first model. The RMSEA and the SRMR were calculated as measures of overall model fit. We computed the CFI and the Tucker–Lewis Index (TLI) as established measures based on model comparisons. The results were in accordance to the cutoff criteria for fit indexes in covariance structure analysis.”²⁰ In addition to this, we assessed the reliability of the German version of the CIUS. We also calculated Pearson's product–moment correlations between the CIUS summary score and external criteria.

Results

Results of the CFA

For the first model (without correlations of the error variances), all goodness-of-fit measures (except for the SRMR) failed the recommended cutoff values (see Table 1).

Table 1.

Factor Loadings, Goodness-of-Fit Measures, and Reliability Coefficient

Item	English	German	First Model	Second Model
Item 1	How often do you find it difficult to stop using the Internet when you are online?	Wie häufig finden Sie es schwierig, mit dem Internetgebrauch aufzuhören, wenn Sie online sind?	0.75	0.75
Item 2	How often do you continue to use the Internet despite your intention to stop?	Wie häufig setzen Sie Ihren Internetgebrauch fort, obwohl Sie eigentlich aufhören wollten?	0.74	0.73
Item 3	How often do others (e.g., partner, children, parents, friends) say you should use the Internet less?	Wie häufig sagen Ihnen andere Menschen (z.B. Partner, Kinder, Eltern, Freunde), dass Sie das Internet weniger nutzen sollten?	0.75	0.76
Item 4	How often do you prefer to use the Internet instead of spending time with others (e.g., partner, children, parents, friends)?	Wie häufig bevorzugen Sie das Internet, statt Zeit mit anderen zu verbringen (z.B. Partner, Kinder, Eltern, Freunde)?	0.72	0.74
Item 5	How often are you short of sleep because of the Internet?	Wie häufig schlafen Sie zu wenig wegen des Internets?	0.62	0.62
Item 6	How often do you think about the Internet, even when not online?	Wie häufig denken Sie an das Internet, auch wenn Sie gerade nicht online sind?	0.69	0.68
Item 7	How often do you look forward to your next Internet session?	Wie oft freuen Sie sich bereits auf Ihre nächste Internetsitzung?	0.67	0.65
Item 8	How often do you think you should use the Internet less often?	Wie häufig denken Sie darüber nach, dass Sie weniger Zeit im Internet verbringen sollten?	0.45	0.45
Item 9	How often have you unsuccessfully tried to spend less time on the Internet?	Wie häufig haben Sie erfolglos versucht, weniger Zeit im Internet zu verbringen?	0.50	0.49
Item 10	How often do you rush through your (home) work in order to go on the Internet?	Wie häufig erledigen Sie Ihre Aufgaben zuhause hastig, damit Sie früher ins Internet können?	0.76	0.75
Item 11	How often do you neglect your daily obligations (work, school, or family life) because you prefer to go on the Internet?	Wie häufig vernachlässigen Sie Ihre Alltagsverpflichtungen (Arbeit, Schule, Familienleben), weil Sie lieber ins Internet gehen?	0.71	0.71
Item 12	How often do you go on the Internet when you are feeling down?	Wie häufig gehen Sie ins Internet, wenn Sie sich niedergeschlagen fühlen?	0.71	0.68
Item 13	How often do you use the Internet to escape from your sorrows or get relief from negative feelings?	Wie häufig nutzen Sie das Internet, um Ihren Sorgen zu entkommen oder um sich von einer negativen Stimmung zu entlasten?	0.71	0.68
Item 14	How often do you feel restless, frustrated, or irritated when you cannot use the Internet?	Wie häufig fühlen Sie sich unruhig, frustriert oder gereizt, wenn Sie das Internet nicht nutzen können?	0.69	0.70
Fit indices:
χ²			2045.97	516.78
df			77	72
RMSEA			0.122	0.060
SRMR			0.058	0.032
CFI			0.856	0.968
TLI			0.830	0.959
Cronbach's α			0.929	0.929

As Hu and Bentler²⁰^(p27) state, “… a cutoff value close to .95 …” for the CFI and the TLI, “… a cutoff value close to .08 for SRMR …,” and “… a cutoff value close to .06” for the RMSEA seem to result in lower type II error rates (with acceptable costs of type I error rates).” In the first model; the RMSEA was 0.122 and the SRMR was 0.058. The values for the CFI (0.856) and TLI (0.830) were also below the recommended cutoff value of 0.95.²⁰ The calculated CFA with one latent factor did not show an optimal fit. Therefore, the first model was rejected.

In the next step, a new model was specified, including the correlations of residual error terms analogous to the study of Meerkerk et al.¹⁷ In general, correlations of the error variances should meet two criteria. On the one hand, Pearson's correlation coefficient should be greater than 0.3, and on the other hand, connecting two residual error terms always has to be theoretically grounded.²¹ Along the lines of Meerkerk et al.,¹⁷ we correlated the error variances of the item pairs 1 and 2 (r=0.69), 6 and 7 (r=0.62), 8 and 9 (r=0.62), 10 and 11 (r=0.67), as well as 12 and 13 (r=0.78). All the Pearson correlations are reported in parentheses and were greater than 0.3. Meerkerk et al.¹⁷^(p3) justified the way they proceeded: “Correlating the error variances of these items is justifiable because the items show overlap in content.” Following Meerkerk et al.,¹⁷^(p3) the first two items describe “problems with stopping use of the Internet,” items 6 and 7 “thinking about or looking forward to using the Internet,” items 8 and 9 “using the Internet less often or spending less time on the Internet,” items 10 and 11 “neglecting daily obligations,” and items 12 and 13 “going on the Internet when feeling down or to escape from negative feelings.”

For the second model, with the described correlated error variances, substantially improved goodness-of-fit measures were obtained (see Table 1). The values for the RMSEA (0.060), the SRMR (0.032), the CFI (0.968), and TLI (0.959) reached the recommended cutoff values of Hu and Bentler.²⁰ The observed factor loadings were between 0.45 (item 8) and 0.76 (item 3; see Table 1). Regarding all parameters, the second model shows an acceptable model fit and was chosen as a suitable statistical solution.

Results of the reliability and correlation analyses

Measured by Cronbach's coefficient, the internal consistency of the German version of the CIUS was high (Cronbach's α=0.929). The Pearson's product–moment correlations between the “CIUS-summary score” and different ratings of the parents were: r=0.57 (p<0.001 for “problematic use of computers, of the Internet and of gaming consoles”), r=0.63 (p<0.001 for “frequency of excessive media use”), and r=0.60 (p<0.001 for “frequency of controversies at home because of media use”). Similar values were calculated for the correlations between the “CIUS-summary score” and the self-ratings of the adolescents: r=0.50 (p<0.001 for “problematic media use”), r=0.60 (p<0.001 for “frequency of excessive media use”) and r=0.64 (p<0.001 for “frequency of controversies at home because of media use”). In addition to this, we found a negative significant correlation between the CIUS summary score and the satisfaction with life (r=−0.36, p<0.001).

Discussion

The results of this study indicate good psychometric properties for the German version of the CIUS in a representative quota sample of adolescents (n=1,723). Our findings for the German version confirm the results Meerkerk et al.¹⁷ reported for the original Dutch version of the CIUS, and Khazaal et al.¹⁸ described for the Arabic version of the questionnaire or for the French version of the CIUS respectively.¹ We obtained good values for the goodness-of-fit measures analogous to the findings of Meerkerk et al.¹⁷ For a large population sample (n=16,925), Meerkerk et al.¹⁷ reported a value of 0.054 for the RMSEA (measure of overall model fit), whereas in our study, the RMSEA was 0.060 and the SRMR (alternative measure of overall model fit) was 0.032. For the CFI, Meerkerk et al.¹⁷ reported a value of 0.986, and we received comparable results for the CFI of 0.968 and for the TLI of 0.959. As a specific value for the reliability, Meerkerk et al.¹⁷ reported a Cronbach's alpha of 0.90, and we obtained a Cronbach's alpha of 0.929 as the result of our reliability analysis. In summary, it can be stated that the results for the German version of the CIUS in our survey seem to be comparable to the results of the original version of the CIUS described by Meerkerk et al.¹⁷

The CIUS instrument appears to have several advantages. The briefness of this standardized questionnaire (14 items) makes it suitable for applications in research as well as in the clinical area. From an economic point of view, the use of the CIUS in online surveys or more generally in epidemiological studies seems reasonable.¹ The results of the other published studies and of our investigation consistently confirm a one factor model of the CIUS,^1,17,18 whereas the current empirical findings for the factor structure of the IAT are heterogeneous.^2,6–13 Furthermore, the CIUS showed high reliability (internal consistency), ranging from values of 0.90¹⁷ to 0.91¹ to 0.93 in the present study. Only the internal consistency for the Arabic version of the CIUS (Cronbach's α=0.78) was slightly lower than in the other surveys.¹⁸

The age structure of our sample surely is a limitation of this study. Due to systematic sampling, our results are transferable to the population of all adolescents in this age group in Germany, but only the children and not the parents were required to complete the CIUS questionnaire in our survey. In contrast, Meerkerk et al.¹⁷ examined participants aged between 11 and 80 years, whereas Khazaal et al.¹⁸ also investigated many adolescents in their validation study for the Arabic version of the CIUS (average age of the sample: 17.6 years).

Future validation studies should combine different questionnaires measuring Internet addiction (e.g., IAT, CIAS, and CIUS) and describe the interdependencies and the differences between the results. Chong Guan et al.⁹ conducted a study with the Malay version of the IAT and the CIUS, and reported a highly significant correlation of 0.838. The parallel use of different diagnostic methods would make it easier to compare the established questionnaires. If the findings were similar, the diagnostic tool could be chosen depending on the research question. If a differentiated measurement of Internet addiction is needed, then a multidimensional instrument like the IAT or the CIAS could be used. If, however, an investigation should be as economical as possible, the CIUS should be the instrument of choice.

Improving diagnostics for the growing area Internet addiction is absolutely necessary.¹ For a better understanding of this phenomenon, psychometric validations of the measuring instruments (e.g., concerning ecological validity, accordance with standardized interviews) are essential.² With appropriate screening and diagnostic tools, a more reliable identification of affected persons is possible. This in turn facilitates a more effective treatment as well as the better implementation of secondary preventive approaches.

Footnotes

Acknowledgment

The EXIF-study (directed by Professor Rudolf Kammerl) was funded by the Federal Ministry of Family Affairs, Senior Citizens, Women and Youth (BMFSFJ).

Author Disclosure Statement

No competing financial interests exist.

References

Khazaal

, Chatton

, Horn

, et al. French validation of the Compulsive Internet Use Scale (CIUS). Psychiatric Quarterly, 2012; 83:397–405.

Chang

, Law

SPM

. Factor structure for Young's Internet Addiction Test: a confirmatory study. Computers in Human Behavior, 2008; 24:2597–2619.

Young

. (1998) Caught in the net. New York: John Wiley.

Chen

, Weng

, Su

, et al. Development of Chinese Internet Addiction Scale and its psychometric study. Chinese Journal of Psychology, 2003; 45:279–294.

Meerkerk

, Van Den Eijnden

RJJM

, Garretsen

HFL

. Predicting Compulsive Internet Use: it's all about sex!. CyberPsychology & Behavior, 2006; 9:95–103.

Widyanto

, McMurran

. The psychometric properties of the Internet addiction test. CyberPsychology & Behavior, 2004; 7:443–450.

Ferraro

, Caci

, D'Amico

, et al. Internet addiction disorder: an Italian study. CyberPsychology & Behavior, 2007; 10:170–175.

Ngai

. Exploring the validity of the Internet Addiction Test for students in grades 5–9 in Hong Kong. International Journal of Adolescence & Youth, 2007; 13:221–237.

Chong Guan

, Isa

, Hashim

, et al. Validity of the Malay version of the Internet Addiction Test: a study on a group of medical students in Malaysia. Asia-Pacific Journal of Public Health, 2012; doi:10.1177/1010539512447808

10.

Khazaal

, Billieux

, Thorens

, et al. French validation of the Internet Addiction Test. CyberPsychology & Behavior, 2008; 11:703–706.

11.

Korkeila

, Kaarlas

, Jaaskelainen

, et al. Attached to the web—harmful use of the Internet and its correlates. European Psychiatry, 2010; 25:236–241.

12.

Widyanto

, Griffiths

, Brunsden

. A psychometric comparison of the Internet Addiction Test, the Internet-related Problem Scale, and self-diagnosis. CyberPsychology, Behavior & Social Networking, 2011; 14:141–149.

13.

Jelenchick

, Becker

, Moreno

. Assessing the psychometric properties of the Internet Addiction Test (IAT) in US college students. Psychiatry Research, 2012; 196:296–301.

14.

, Yen

, et al. Screening for Internet addiction: an empirical study on cut-off points for the Chen Internet Addiction Scale. The Kaohsiung Journal of Medical Sciences, 2005; 21:545–551.

15.

, Yen

, Chen

, et al. Proposed diagnostic criteria and the screening and diagnosing tool of Internet addiction in college students. Comprehensive Psychiatry, 2009; 50:378–384.

16.

Kesici

, Sahin

. Turkish adaptation study of Internet Addiction Scale. CyberPsychology, Behavior, & Social Networking, 2010; 13:185–189.

17.

Meerkerk

, Van Den Eijnden

RJJM

, Vermulst

, et al. The Compulsive Internet Use Scale (CIUS): some psychometric properties. CyberPsychology & Behavior, 2009; 12:1–6.

18.

Khazaal

, Chatton

, Atwi

, et al. Arabic validation of the Compulsive Internet Use Scale (CIUS). Substance Abuse Treatment, Prevention, & Policy, 2011; 6:32.

19.

Griffiths

. Internet addiction: fact or fiction?. Psychologist, 1999; 12:246–250.

20.

, Bentler

. Cutoff criteria for fit indexes in covariance structure analysis: conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 1999; 6:1–55.

21.

Kline

. (2011) Principles and practice of structural equation modeling. 3rd ed. New York: Guilford Press.