German 
Report to the Federal Ministry of Health, Germany
Project duration: November 15, 2010 to February 14, 2011
Funding code: Chapter 15 02 Title 684 69
Funding amount: € 14,580
Hans-Jürgen Rumpf, Christian Meyer, Anja Kreuzer & Ulrich John
Funding amount: € 14,580
Ad Vermulst (Department of Developmental Psychopathology, Behavioural Science Institute, Radboud University Nijmegen, Niederlande)
Gert-Jan Merkeerk (IVO Addiction Research Institute, Rotterdam, Niederlande)
Management and contact address: PD Dr. Hans-Jürgen Rumpf
University of Lübeck, Clinic for Psychiatry and Psychotherapy, Ratzeburger Allee 160, 23538
Lübeck, Tel. 0451/5002871, email: hans-juergen.rumpf@psychiatrie.uk-sh.de
Greifswald and Lübeck, 31.05.2011
Table of contents
- Conclusion
- Introduction
1.1 Starting point of the project
1.2 State of research
1.3 Aim of the PINTA study
1.4 Project structure, structures and responsibilities - Survey and evaluation methodology
2.1 sample
2.1.1 Landline sample
2.1.2 Mobile-only sample
2.2 Survey procedure
2.3 Statistical Analysis
2.3.1 Procedure for estimating prevalence
2.3.2 Weighting of the telephone sample and consideration of the sample design
in data evaluation - Implementation, work plan and schedule
- Results
4.1 Prevalence estimate based on the cut-off
4.2 Prevalence Estimation Based on Latent Class Analysis (LCA) - Discussion of the results
- Gender Mainstreaming Aspects
- Overall assessment
- Dissemination and publicity of the project results
- Utilization of the project results (sustainability / transfer potential)
- List of publications
- Literature
0. Summary
Background: The data on the prevalence of internet addiction are deficient. The available findings for Germany show methodological deficiencies; in particular, they are not based on representative samples. The present analysis can fall back on a large and representative sample that was recruited as part of the study Pathological Gambling and Epidemiology (PAGE). In this project, Internet addiction was also recorded as a comorbidity using the Compulsive Internet Use Scale (CIUS).
Method: The sample consisted of 15,024 people aged 14–64 who were interviewed by telephone and of whom 1,001 could only be reached via mobile phone and not via landline. In addition to a proportional random sampling approach, the representativeness was ensured through extensive weightings. The prevalence was estimated in the PINTA study via the CIUS, in which 1. used a cut-off from another study and 2. A latent class analysis was calculated on the basis of the CIUS items.
Results: Based on the cut-off of 28, the estimated prevalence for internet addiction is 1.5% (women 1.3%, men 1.7%). When using the LCA, the rates are slightly lower at 1% (women 0.8%, men 1.2%). In the 14–24 age group, the prevalence increases to 2.4% (women 2.5%, men 2.5%). If only 14–16 year olds are considered, 4.0% are internet addicts (women 4.9%, men 3.1%). The high proportion among young girls can be found in both methodological approaches. The girls and women (14–24 years old) who were conspicuous mainly use social networks on the Internet (77.1% of addicts according to LCA) and rarely use online games (7.2%). The young men also use social networks, albeit to a lesser extent (64.8%), but more often online games (33.6%). With the help of the LCA, a further group with problematic internet use can be identified in addition to the presumably addicts, which affects a total of 4.6% of the respondents (women 4.4%, men 4.9%). Here, too, there are high rates among young cohorts, and particularly among women.
Conclusion: Compared to an earlier estimate of 3.2% based on another study, PINTA has lower, yet significant rates. The accuracy of the estimate is significantly improved here on the basis of the representative sample. The high prevalence among girls and young women is striking. Further studies with in-depth analyzes are necessary.
1 Introduction
1.1 Starting point of the project
Internet addiction is a still little researched form of non-substance addiction. Much attention is currently being paid to it, including: because it could be a problem of increasing importance. So far it has not been clarified whether (1) addiction problems with Internet use represent a significant disorder with clinical relevance and (2) whether their prevalence in the population is of a magnitude that requires federal political action. However, due to the lack of sufficiently valid data, there have been no meaningful studies of the problem so far.
1.2 State of research
Internationally, prevalence rates between 1 and 14% can be found (Christakis, 2010). The data on the frequency of internet addiction internationally and for Germany have been viewed and summarized in a project funded by the Federal Ministry of Health (BMG) (Petersen et al., 2010). The authors conclude that there are a number of methodological problems so that only preliminary estimates are possible. The main problems are that in many cases the samples are casual and cannot claim to be representative, and that survey methods have been used that have not been validated. In addition, there is currently no uniform definition of internet addiction (Byun et al., 2009).
The only previous study for Germany that also covered the area of adults comes from Hahn and Jerusalem (2001). On the basis of an online sample of more than 7,000 people, the prevalence is 3.2% across all age groups; the proportion rises among younger subjects. Boys under the age of 18 were about twice as likely to be affected compared to female participants, who were, however, overall underrepresented in the study. However, a large proportion of the studies on internet addiction only refer to adolescents. Based on a survey in schools, Meixner (2010) reports a frequency of 1.4% among 12 to 25-year-olds. Further studies are limited to computer game behavior. A representative student survey found a rate of 1.7% for this (Rehbein, Kleimann & Mossle, 2010).
Since the detailed review by Petersen and Thomasius (2010) on pathological Internet use in Germany, another survey has been carried out at the University of Mainz on 2,512 people including adulthood with the aim of estimating the prevalence. However, the data have not yet been published.
In summary, the data situation with regard to the frequency of problematic Internet use or Internet dependency in Germany is currently incomplete. The previous studies have examined subpopulations or did not show a representative basis.
1.3 Aim of the PINTA study
The aim of PINTA was to provide the most reliable figures possible for the extent of internet addiction by overcoming two methodological weaknesses of previous studies: a) by the inclusion of both adolescents and adults and b) by ensuring representativeness. For the assessment of Internet addiction in Germany, the analysis of a data set from the study “Pathological Gambling and Epidemiology (PAGE)” was carried out. In this study, internet addiction was recorded as a comorbidity of pathological gambling. Since PAGE offers a representative and large sample in the age group 14–64, a more precise estimate of the prevalence was possible.
1.4 Project structure, structures and responsibilities
The project was led by the University of Lübeck. There was close cooperation with the Institute for Epidemiology and Social Medicine of the University Medicine in Greifswald. This was ensured structurally through telephone conferences and regular bilateral telephone contacts.
In order to obtain additional expertise in the field of internet addiction, Dr. Gert-Jan Meerkerk (IVO Addiction Research Institute, Rotterdam, Netherlands). For statistical analysis, the project was continued by Dr. Ad A. Vermulst (Department of Developmental Psychopathology, Behavioral Science Institute, Radboud University Nijmegen, Netherlands), who is an expert in the calculation of latent class analyzes. Two scientific employees were hired who had already worked on the PAGE study, so that synergies could be created here.
2. Survey and evaluation methodology
An analysis of the data from the PAGE project was carried out. PAGE was funded by the federal states as part of the State Treaty on Gambling and carried out from December 1st, 2009 to February 28th, 2011. The methodology and the first results are available in the form of a final report (Meyer et al., 2011). PAGE’s multimodal recruiting approach included, among other things, a nationwide telephone survey. This forms the basis for the data analysis carried out in PINTA.
2.1 Sample
Two samples were taken for the telephone survey. In addition to a random selection of landline numbers, people who can only be reached via mobile phones were also recruited. This strand is of particular importance because the proportion of people who can only be reached this way is significant and continues to rise. Evidence emerges that this is a population with specific characteristics. For example, the prevalence of pathological gambling is increased in this group (Meyer et al., 2011). A computer-aided telephone interview (CATI) of 15 minutes on average was carried out with all participants by trained interviewers.
2.1.1 Fixed line sample
The sampling was carried out in a multi-stage procedure by infas. A detailed description can be found in the corresponding method report (Hess, Steinwede, Gilberg & Kleudgen, 2011), which is available on request from the authors of this report. In the first stage, primary sampling units (PSU) were drawn. The probability of selection of the municipalities (sample points) was proportional to the resident population in the target group. Insgesamt wurden 53 Sample Points in 52 Gemeinden bestimmt (Berlin war mit zwei Sample Points vertreten). For the random selection of the PSUs, stratifications were made according to the federal states, administrative districts and counties as well as the slot machine density. The selected municipalities are shown in Figure 1.
In a second step, the households were determined on the basis of telephone numbers. These Secondary Sampling Units (SSUs) were assigned to the municipalities via the area codes. In each municipality, 5,800 telephone numbers were drawn for the gross sample. In the third step, the target persons were determined (Third Sampling Unit, TSU). If there was more than one person belonging to the target group (ages 14 to 64), the person who was last born was chosen.
Figure 1: Sample points (represented by blue markings; pink markings represent therapy facilities for pathological gambling)
Between June 7 and October 22, 2010, a total of 14,022 telephone interviews from the fixed network sample were carried out. Of 26,736 households in which a target person aged between 14 and 64 lived, an interview could be carried out with 52.4% of the target persons after using the last birthday question, 38.9% refused to take part in the survey. 8.7% of the target persons did not take part because the contact person refused access (4.6%), because they were too seriously ill for a survey (1.4%) or because they could not be reached (2.7%; Hess & Steinwede, 2011).
2.1.2 Mobile-only sample
The random sample of people who cannot be reached via the fixed network, but only via mobile phone, was drawn nationwide because of the unrealizable regional allocation. The target group were again people between the ages of 14 and 64 years. From the randomly drawn phone numbers, only those people who can only be reached via a mobile phone were selected via a screening.
A total of 1,001 telephone interviews were carried out between November 22, 2010 and February 1, 2011 in the mobile-only sample. To this end, 13,273 people had to be screened to ensure that they could only be reached by mobile phone. Of the 1,767 potential target persons identified, 747 persons (42.3%) refused to take part. 7 people (0.4%) were unable to take part in the telephone interview due to illness or disability. 12 people (0.7%) dropped out due to insufficient knowledge of German (Hess & Steinwede, 2011).
2.2 Survey procedure
Im Mittelpunkt der Prävalenzschätzung für pathologischen Internetgebrauch stand die Compulsive Internet Use Scale (CIUS; Meerkerk, Van Den Eijnden, Vermulst & Garretsen, 2009), ein Fragebogenverfahren zur Erfassung von Merkmalen der Internetabhängigkeit. Your 14 items have a five-level answer format (Figure 2), with between 0 and 56 points being achieved. The method was developed in several subsamples and shows a one-factorial structure throughout. There are also data from the general population available, which speaks in favor of choosing this method for use in epidemiological surveys. The Cronbach’s alpha as a measure of internal consistency was .89 and indicates good reliability. A convergent validity with similar procedures was shown. There is currently no recommended cut-off based on a broad database. Initial indications suggest a threshold value of 28 (see 3.3.1; Van Rooij, Schoenmakers, Vermulst, Van Den Eijnden & Van De Mheen, 2011).
Figure 2: Items of the CIUS (answer categories: never, rarely, sometimes, often, very often)
- How often do you find it difficult to stop using the internet while online?
- How often do you continue to use the internet when you wanted to stop?
- How often do other people, e.g. your partner, children, parents or friends tell you that you
- should use the internet less?
- How often do you prefer to use the internet instead of spending time with others, e.g. with your partner,
- Children, parents, friends?
- How often do you sleep too little because of the internet?
- How often do you think of the internet when you are not online?
- How often do you look forward to your next internet session?
- How often do you think about spending less time online?
- How many times have you tried unsuccessfully to spend less time online?
- How often do you rush to do your chores at home so you can get on the internet sooner?
- How often do you neglect your everyday responsibilities (work, school, family life) because
- you prefer to go online?
- How often do you go online when you feel down?
- How often do you use the internet to escape your worries or
- relieve a negative mood?
- How often do you feel restless, frustrated, or irritable when you cannot use the internet?
All persons who stated that they had used the Internet for private purposes either for at least one hour on a weekday or one day on the weekend were asked the questions of the CIUS. The procedure was used in conjunction with other survey instruments. In terms of content, the focus was on the recording of gambling behavior and pathological gambling. In the order in which they were presented, a procedure for recording social capital was first used, followed by CIUS, the interview on gambling and socio-demographic questions. Social capital was recorded using 12 questions about participation in social events in the last 12 months (cinema, sporting event, art exhibition, further education etc .; Hanson, Östergren, Elmstahl, Isaacsson & Ranstam, 1997). In addition to this information on the “Social Participation” construct, the procedure also includes a question as to the extent to which there is a general feeling of being able to trust other people (“Trust”). Pathological gambling was recorded using the gambling section of the Composite International Diagnostic Interview (CIDI) (WHO, 2009). It was also asked what activities more than 50% of the time is spent on the Internet. The free text information has been combined into main categories. In the case of multiple responses, the first response was evaluated.
2.3 Statistical Analysis
2.3.1 Procedure for estimating prevalence
The frequency of internet addiction was estimated in two ways:
- A cut-off was used for the screening questionnaire CIUS, which was taken from a different sample (Van Rooij et al., 2011). This limit comes from two random samples of 13 to 16 year old students (n = 1,572 / 1,476). The aim of the study was to identify a suspicious group with addiction to online video games. Such a group was identified by means of a latent class analysis (LCA) based on the CIUS. A cut-off of 28 or more points was found to be favorable. With the help of this threshold value, a rough estimate of Internet dependency could be made in PAGE.
- With the help of a modern statistical method, the latent class analysis (LCA), which is particularly informative for our purpose, a group was identified which, based on its response pattern, can be viewed as likely dependent. LCA is a method for identifying significant groups of people, whose response behavior is similar. The calculation was done with Mplus 5.1. (Muthén & Muthén, 1998). Several models were calculated to determine a subgroup showing characteristics of Internet addiction and a number of goodness-of-fit measures were used to select a model. These measures included: Bayesian Information Criterion (BIC) ‑value (low values indicate better fit), Entropy measure (higher values indicate better fit), Vuong-Lo-Mendell-Rubin likelihood ratio test, and adjusted Lo-Mendell ‑Rubin Likelihood Ratio Test (a p‑value <.05 indicates that the model is better than the previous one). The bootstrapped likelihood ratio test (BLRT) was not applicable because it cannot be calculated with weighted data. Another important criterion is the usefulness of the model based on theoretical or practical considerations. The characteristics of the classes are checked by means of inferential statistical comparisons with the other classes. The LCA calculations were done by Dr. Ad A. Vermulst, Department of Developmental Psychopathology, Behavioral Science Institute, Radboud University Nijmegen, Holland.
2.3.2 Weighting of the telephone sample and consideration of the sample design in the data evaluation
All data were analyzed using sample weights. Since estimates based on the German population should be made on the basis of the telephone sample, special attention was paid to the development of adequate weighting variables. The weights were developed separately for the landline and mobile network samples. First, so-called design weights were determined to compensate for different selection probabilities through the sample design. The distorting effects of the multi-stage drawing process were compensated for. For example, households that can be reached via several telephone connections are more likely to be selected than households with only one connection. By contrast, people who live in multi-person households have a lower selection probability than people in single households due to the design, as only one person in the specified age range per household was surveyed. Finally, the mobile phone and landline phone samples had to be merged. The weighting had to be used to reproduce the proportion of people in the population with a landline connection or who can only be reached via a mobile phone connection (“Mobile-Onlys”) in the sample. Based on current findings, a proportion of “mobile-onlys” in the 14 to 64 year old population of 14% (infas social research; unpublished data) is assumed.
A second weighting step consisted of balancing the different willingness to participate in different socio-demographically defined population groups on the basis of existing marginal distributions from official statistics (redressment). Due to the significant association of gambling problems with various social indicators according to previous findings, in order to avoid distorted prevalence estimates it was of great importance, in addition to age and gender distributions, to adapt the characteristics of schooling, unemployment and migration background to the population. Basically, the expansion of the characteristics for the redressment is accompanied by a reduction in the effective number of cases in the sample and thus an increase in the sampling error. Thus, when selecting features to be considered, a trade-off must be made between possible distortion and the accuracy of the point estimates. Against this background, the following analyzes were based on a weighting that includes the social indicators mentioned, but not the characteristic of the political size class of the community of residence. A distortion of the results is hardly to be expected as the political size class of the residential community was already taken into account as a stratification criterion in the community selection and no significant connection with the main investigation criterion gambling problems can be assumed.
For the inferential statistical validation of the findings presented below, the sample design was taken into account when estimating the sampling errors, as far as methodologically possible, since an analysis using standard methods, which assume a simple random sample, would lead to significant distortions.
3. Implementation, work plan and schedule
Three months were planned for the implementation of the project. The work to be performed included the review of the literature, the preparation of the data sets, the data analysis, publication and reporting. The work was delayed for two reasons: 1. The complex statistical procedure based on the LCA in cooperation with the researchers from the Netherlands required a greater amount of time, especially for the respective agreements and content-related discussions of the procedure and the results. 2. In the PAGE study, as part of the telephone survey, the inclusion of people who can only be reached via mobile phone was made possible. The corresponding surveys were not completed until February 1, 2011. However, it was decided to use this sub-sample and also to include it in the analyzes because it represents a significant improvement in the methodology. In addition to the planned work, we have used the opportunity to provide services beyond the scope of the resources in this project. This expansion of our services has significantly improved the quality of the results. After the end of the funding period, the work was continued by the project management using its own resources, so that it was completed by the time the report was drawn up. A publication of the results in a specialist journal is still pending due to the delays.
4. Results
Of the 15,023 people surveyed, 8,130 (54.1%) stated that they had used the Internet for private purposes either for at least one hour on a weekday or one day on the weekend and received the questions from the CIUS. All of the following analyzes have been carried out on the basis of weighted data, unless otherwise noted.
4.1 Prevalence estimate based on the cut-off
Based on the CIUS cut-off of 28, as described in the methodology, the estimated prevalence of probable internet addiction for the total sample of 14- to 64-year-olds is 1.5% based on all participants (subjects, who were not asked about internet usage due to the filter are considered inconspicuous). Findings by gender and the respective confidence intervals can be found in Table 1.
Table 1: Prevalence estimate based on cut-off 28 of the CIUS, ages 14–64 (n = 15,023)
Prevalence (%) | Confidence interval (%) | |
Total | 1,5 | 1,3–1,7 |
women | 1,3 | 1,0–1,7 |
men | 1,7 | 1,3–2,1 |
If the younger age group is considered separately, the prevalence figures are higher and there is a shift within the sexes to a higher proportion among the female participants (Tables 2 and 3).
Table 2: Prevalence estimate based on cut-off 28 of the CIUS, ages 14–24 (n = 2,937)
Prevalence (%) | Confidence interval (%) | |
Total | 3,8 | 3,0–4,6 |
women | 4,5 | 3,3–6,0 |
men | 3,0 | 2,3–4,3 |
Table 3: Prevalence estimate based on cut-off 28 of the CIUS, ages 14–16 (n = 693)
Prevalence (%) | Confidence interval (%) | |
Total | 6,3 | 4,6–8,4 |
women | 8,6 | 5,5–13,0 |
men | 4,1 | 2,6–6,3 |
If one considers the group of 14 to 24 year-olds separated by sex, one can see for the suspect in the CIUS (presumed internet addicts) that female participants mainly stated social networks as the first mentioned main activity on the internet (81.4%; Table 4) . This also applies to a large extent to the male participants (61.4%), who, however, in contrast to the girls and women, often name online games (28.9%). Overall, the preferences differ significantly from each other (<0.001).
Table 4: First mention of the main activities on the Internet among 14–24 year olds with a conspicuous CIUS result (28 or more points) by gender
Activities online | Frequency (%) | Confidence interval | |
Feminine | Social networks | 81,4 | 64,4–91,4 |
E‑Mail | 12,7 | 4,7–30,2 | |
Online games | 3,8 | 3,3–17,7 | |
Entertainment (music, films, etc.) | 2,1 | 0,3–14,3 | |
Masculine | Social networks | 61,4 | 43,5–76,7 |
Online games | 28,9 | 15,7–47,1 | |
Information | 3,5 | 0,4–23,0 | |
E‑Mail | 2,5 | 0,4–15,6 | |
Shopping / Selling | 2,4 | 0,3–16,8 | |
Internet telephony | 1,2 | 0,1–9,0 |
4.2 Prevalence estimate based on the latent class analysis (LCA)
Latent class models with 2 to 7 classes were calculated. The models with 5 and 6 classes showed an identically large group that showed extreme values in the CIUS. Compared to the 5‑class solution, the 6‑class solution showed the better model adaptation (BIC: 22417 vs. 22490; Entropy 0.769 vs. 0.762). The 6 groups do not overlap, as can be seen in Figure 2.
Figure 2: Box plot of the CIUS total values for the 6 classes
A further analysis of the 6‑class solution revealed features that speak for the existence of a group that can be viewed as dependent (class 6). A second group (class 5) probably shows an increased risk in terms of problematic Internet use. The corresponding findings are named below: This shows that class 6 has higher CIUS values. This can also be observed to a lesser extent for class 5 (Table 5). The same applies to the number of hours that are spent on the Internet during the week. Overall, grade 6 shows a lower level of social activities and social trust. Grade 5 is the youngest of the groups, followed by Grade 6.
Table 5: Characteristics of the 6 classes: mean values (standard error)
Class | CIUS- Average | Hours in the internet /Week | Social participation | Social Trust | Age |
1 | 15,4 (0,03) | 8,7 (0,19) | 5,7 (0,07) | 2,6 (0,02) | 40,4 (0,34) |
2 | 19,4 (0,03) | 11,4 (0,34) | 5,8 (0,74) | 2,7 (0,02) | 36,8 (0,44) |
3 | 23,7 (0,03) | 13,3 (0,33) | 5,7 (0,73) | 2,7 (0,02) | 33,8 (0,41) |
4 | 29, 6 (0,06) | 16,0 (0,42) | 5,5 (0,08) | 2,7 (0,02) | 31,1 (0,51) |
5 | 37,3 (0,13) | 22,5 (0,69) | 5,3 (0,13) | 2,7 (0,03) | 27,6 (0,47) |
6 | 48,7 (0,53) | 29,2 (1,64) | 5,0 (0,21) | 2,5 (0,63) | 30,0 (1,08) |
significance ℗* | <0,001 | <0,001 | <0,001 | <0,001 | <0,001 |
* ANOVA (CIUS and age) or Kruskal-Wallis‑H test based on unweighted data
In the following, the occurrence of class 6 in the total population and in partial samples is used to estimate the prevalence.
This results in an estimated prevalence for probable internet addiction of 1.0% for all participants in the total sample of 14 to 64 year olds. Findings by gender and the respective confidence intervals can be found in Table 6.
Table 6: Prevalence estimate of internet addiction based on the LCA (frequency of class 6), ages 14–64 (n = 15,023)
Prevalence (%) | Confidence interval (%) | |
Total | 1,0 | 0,9–1,2 |
women | 0,8 | 0,6–1,1 |
men | 1,2 | 1,0–1,6 |
If the younger age group is considered separately, this approach also results in higher prevalence rates and the prevalence levels are initially equal with regard to the sexes. Among 14 to 16 year olds, the prevalence is higher in female participants (Tables 7 and 8).
Table 7: Prevalence estimate of internet addiction based on the LCA (frequency of class 6), ages 14–24 (n = 2,937)
Prevalence (%) | Confidence interval (%) | |
Total | 2,4 | 1,9–3,1 |
women | 2,4 | 1,6–3,5 |
men | 2,5 | 1,7–3,5 |
Table 8: Prevalence estimate of internet addiction based on the LCA (frequency of class 6), ages 14–16 (n = 693)
Prevalence (%) | Confidence interval (%) | |
Total | 4,0 | 2,7–5,7 |
women | 4,9 | 2,8–8,5 |
men | 3,1 | 1,8–5,3 |
When looking at the first-mentioned main activities on the Internet, it again becomes apparent that social networks are in the foreground for both sexes, but that these are mentioned even more frequently for girls and women (Table 9). In contrast, boys and men play online games much more often. Overall, the preferences also differ significantly from each other here (<0.001). Some of the activities of the male participants listed in Table 4 no longer appear here. Die Schätzung auf Basis der LCA führt somit zu einer stärkeren Fokussierung auf die Hauptaktivitäten Soziale Netzwerke und Onlinespielen, was zusätzlich für die Validiät des LCA-Ansatzes sprechen mag.
Table 9: First mention of the main activities on the Internet of 14–24 year-olds in class 6 of the LCA by gender
Activities online | Frequency (%) | Confidence interval | |
Feminine | Social networks | 77,1 | 52,8–91,0 |
E‑Mail | 11,7 | 2,7–39,3 | |
Online games | 7,2 | 1,5–28,3 | |
Entertainment (music, films, etc.) | 4,0 | 0,5–24,5 | |
Masculine | Social networks | 64,8 | 9,1–27,4 |
Online games | 33,6 | 2,3–16,7 | |
Internet telephony | 1,5 | 0,5–1,3 |
In addition to class 6, which can be viewed as dependent, prevalence rates can be given for class 5, which presumably has problematic Internet use. The corresponding data can be found in Tables 10 to 12. Overall, the respective proportions are significantly higher than in the case of the dependency. There are again higher rates in the younger samples and the predominance of women in the younger age groups.
Table 10: Prevalence estimate of problematic internet use based on the LCA (frequency of class 5), ages 14–64 (n = 15,023)
Prevalence (%) | Confidence interval (%) | |
Total | 4,6 | 4,2–5,1 |
women | 4,4 | 3,9–5,0 |
men | 4,9 | 4,3–5,5 |
Table 11: Prevalence estimate of problematic internet use based on the LCA (frequency of class 5), ages 14–24 (n = 2,937)
Prevalence (%) | Confidence interval (%) | |
Total | 13,6 | 12,4–14,8 |
women | 14,8 | 13,0–16,8 |
men | 12,4 | 10,4–14,7 |
Table 12: Prevalence estimate of problematic internet use based on the LCA (frequency of class 5), ages 14–16 (n = 693)
Prevalence (%) | Confidence interval (%) | |
Total | 15,4 | 12,8–18,5 |
women | 17,2 | 13,2–22,2 |
men | 13,7 | 10,5–17,7 |
5. Discussion of the results
The prevalence estimates found in the PINTA study are below the previously available data from Hahn and Jerusalem (2001), who found a rate of 3.2% on the basis of an occasional sample in an online survey. The estimates of the present sample are between 1% and 1.5%. The higher value was found using a cut-off from another study (Van Rooij et al., 2011). This estimate has the following sources of error: The comparative study is a sample that is limited to 13- to 16-year-old students. Furthermore, it was about the recording of online game addiction. The transferability is therefore limited. In addition, a prevalence estimate based on a screening process is always associated with a high error rate. Significant overestimations can be made, particularly with low prevalence and low specificity (Gambino, 1997). Corrections using formulas that take sensitivity and specificity into account, as originally planned, were not possible in this case. The reason is that these two validity measures are based on response probabilities of the LCA in the study by Van Rooij et al. could have been calculated, but it would have been a circular argument, since the conspicuous class was also calculated from the same analysis. An external criterion is missing.
The second approach of the present study was able to identify a class based on the LCA, which is very likely to represent a group of internet addicts. This is supported by a number of findings that distinguish it from the other classes. This group had the highest values in the CIUS, spent most of the time on the Internet,
showed less social activity, felt less social trust and was more likely to be young. A second group also stood out and can be viewed as problematic with regard to Internet behavior. This procedure has the following sources of error: 1. The group formation is based solely on the CIUS. Features that are not included here are not taken into account. 2. Even if a number of criteria have been used to find the most suitable model, there is always some room for interpretation. Taken together, there is no external validation here either.
Overall, it is assumed that the estimate based on the LCA is closer to the true prevalence, since the sources of error in the other estimate must generally be assessed as significantly larger. For the group as a whole, the estimates are also far apart.
In relation to the project objectives, it has clearly been possible to make a more precise estimate of the prevalence possible. The clear advantage is the basis of a large and representative sample which, in addition to the fixed network sample, also includes people who can only be reached via mobile phones.
When looking at the age groups and the distribution within the sexes, it is noticeable that in the young age groups the prevalence rates of girls exceed those of boys. Compared with earlier findings (Hahn & Jerusalem, 2001; Petersen et al., 2010) this was not to be expected. The finding is all the more striking because this trend was found in both estimates using the different methodological approaches. This can also be found in the LCA for the second conspicuous class, whose Internet use can be viewed as problematic. If one looks at the respective conspicuous in the group of 14 to 24 year olds, one finds differences in the preferences of the Internet activities. It is true that both groups most frequently state that social networks are used, but this is particularly pronounced among women, who, on the other hand, rarely use online games. Overall, and especially for these unexpected findings in the young female test persons, future studies will have to clarify whether the abnormalities found are actually to be understood as a disorder for which help is needed. For this it is necessary to conduct in-depth interviews that capture the clinical significance on the level of symptoms and criteria as well as the associated impairments.
6. Gender mainstreaming aspects
Due to the representativeness of the sample and the separate analyzes for women and men, aspects of gender mainstreaming could be fully taken into account.
7. Overall assessment
The project objectives were fully achieved. In particular, the statistically complex LCA could be used, which methodologically enabled a more precise estimate of the prevalence. The delays in the project process due to this analysis and the addition of the sample, which can only be reached via mobile phone, can be clearly justified by the respective methodological gain.
8. Dissemination and publicity of the project results
Due to the very short duration of the project, no broad publication activity has been possible so far. This is planned for the coming period. There are already plans for presentations at the Addiction Congress in Frankfurt (September 28th-October 1st, 2011) and at the Scientific Discussion of the German Society for Addiction Research and Addiction Therapy (DG-Sucht) in Lübeck (December 2nd-4th, 2011). A presentation at the Criminological Research Institute in Hanover has already taken place. Publications in specialist journals are to follow.
9. Utilization of the project results (sustainability / transfer potential)
The results indicate high rates of problematic or addictive internet use in young age groups, especially among women. In order to be able to assess whether there is a particular need for prevention or treatment offers, further clarification of these initial findings in a detailed follow-up study is urgently required.
10. List of publications lectures
Rumpf, H. J., Meyer, C. & John, U. (2011). Prevalence of internet addiction (PINTA), Criminological Research Institute Lower Saxony. Hannover, 09.05.2011.Rumpf, H. J., Meyer, C. & John, U. (2011). Prevalence of Internet Addiction (PINTA): Results and Outlook, Federal Ministry of Health. Berlin, 07.04.2011.
11. Literature
Byun, S., Ruffini, C., Mills, J. E., Douglas, A. C., Niang, M., Stepchenkova, S., Lee, S. K., Loutfi, J., Lee, J. K., Atallah, M. & Blanton, M. (2009). Internet addiction: metasynthesis of 1996–2006 quantitative research. Cyberpsychol Behav, 12, 203–207.
Christakis, D. A. (2010). Internet addiction: a 21(st) century epidemic? Bmc Medicine, 8, 3. Gambino, B. (1997). The correction for bias in prevalence estimation with screening tests.
Journal of Gambling Studies, 13, 343–351.
Hahn, A. & Jerusalem, M. (2001). Internetsucht: Jugendliche gefangen im Netz. In J. Raithel (Ed.), Risikoverhalten Jugendlicher: Erklärungen, Formen und Prävention (pp. 279- 293). Berlin: Leske + Budrich.
Hanson, B. S., Östergren, P. O., Elmstahl, S., Isaacsson, S. O. & Ranstam, J. (1997). Reliability and validity assessments of measures od social networks, social support and control — results from the Malmö Shoulder and Neck Study. Scandinavian Journal of Public Health, 25, 249–257.
Hess, D., Steinwede, A., Gilberg, R. & Kleudgen, M. (2011). Methodenbericht PAGE — Pathologisches Glücksspielen und Epidemiologie. Bonn: infas Institut für angewandte Sozialwissenschaften GmbH.
Meerkerk, G. J., Van Den Eijnden, R., Vermulst, A. A. & Garretsen, H. F. L. (2009). The Compulsive Internet Use Scale (CIUS): Some Psychometric Properties. Cyberpsychology & Behavior, 12, 1–6.
Meixner, S. (2010). Exzessive Internetnutzung im Jugendalter. Kinder und Jugendschutz in Wissenschaft und Praxis, 55, 3–7.
Meyer, C., Rumpf, H.-J., Kreuzer, A.-., de Brito, S., Glorius, S., Jeske, C., Kastirke, N., Porz,
S., Schön, D., Westram, A., Klinger, D., Goeze, D., Bischof, G. & John, U. (2011). Pathologisches Glücksspielen und Epidemiologie (PAGE): Entstehung, Komorbidität, Remission und Behandlung. Endbericht an das Hessische Ministerium des Innern und für Sport. Universitäten Greifswald und Lübeck. Retrieved, from the World Wide Web:
Muthén, L. K. & Muthén, B. O. (1998). Mplus User’s guide, 5th edn. Los Angeles: Muthén & Muthén.
Petersen, K. U., Thomasius, R., Schelb, Y., Spieles, H., Trautmann, S., Thiel, R. & Wey- mann, N. (2010). Beratungs- und Behandlungsangebote zum pathologischen Internetgebrauch in Deutschland. Endbericht an das Bundesministerium für Gesundheit (BMG). Hamburg: Universitätsklinikum Hamburg-Eppendorf, Deutsches Zentrum für Suchtfragen des Kindes- und Jugendalters (DZSKJ).
Rehbein, F., Kleimann, M. & Mossle, T. (2010). Prevalence and risk factors of video game dependency in adolescence: results of a German nationwide survey. Cyberpsychol Behav Soc Netw, 13, 269–277.
Van Rooij, A. J., Schoenmakers, T. M., Vermulst, A. A., Van Den Eijnden, R. J. & Van De Mheen, D. (2011). Online video game addiction: identification of addicted adolescent gamers. Addiction, 106, 205–212.