This is an open-access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Mental Health, is properly cited. The complete bibliographic information, a link to the original publication on http://mental.jmir.org/, as well as this copyright and license information must be included.
Adolescents and young adults are among the most frequent Internet users, and accumulating evidence suggests that their Internet behaviors might affect their mental health. Internet use may impact mental health because certain Web-based content could be distressing. It is also possible that excessive use, regardless of content, produces negative consequences, such as neglect of protective offline activities.
The objective of this study was to assess how mental health is associated with (1) the time spent on the Internet, (2) the time spent on different Web-based activities (social media use, gaming, gambling, pornography use, school work, newsreading, and targeted information searches), and (3) the perceived consequences of engaging in those activities.
A random sample of 2286 adolescents was recruited from state schools in Estonia, Hungary, Italy, Lithuania, Spain, Sweden, and the United Kingdom. Questionnaire data comprising Internet behaviors and mental health variables were collected and analyzed cross-sectionally and were followed up after 4 months.
Cross-sectionally, both the time spent on the Internet and the relative time spent on various activities predicted mental health (
The magnitude of Internet use is negatively associated with mental health in general, but specific Web-based activities differ in how consistently, how much, and in what direction they affect mental health. Consequences of Internet use (especially sleep loss and withdrawal when Internet cannot be accessed) seem to predict mental health outcomes to a greater extent than the specific activities themselves. Interventions aimed at reducing the negative mental health effects of Internet use could target its negative consequences instead of the Internet use itself.
International Standard Randomized Controlled Trial Number (ISRCTN): 65120704; http://www.isrctn.com/ISRCTN65120704?q=&filters=recruitmentCountry:Lithuania&sort=&offset= 5&totalResults=32&page=1&pageSize=10&searchType=basic-search (Archived by WebCite at http://www.webcitation/abcdefg)
Depression and anxiety are two of the most prevalent psychiatric disorders among adolescents [
A major line of research has linked mental health problems to what has been termed problematic Internet use (or pathological or compulsive Internet use), which is often conceptualized as an impulse control disorder similar to gambling addiction and other behavioral addictions. The most used and validated measure of problematic Internet use, the Internet Addiction Test (IAT) [
Studies show that the problematic aspects of certain individuals’ Internet use are restricted to one or a few specific Web-based activities (eg, gaming or social media use), whereas other activities are nonproblematic [
It is thus important to differentiate between activities when investigating the mental health effects of Internet use. In some cases, it could be important because the activity in question is prone to becoming addictive, such as Web-based gambling (eg, Web-based poker, sports betting, casino spins) [
This study aimed to investigate how adolescents’ mental health is predicted by time spent on the Internet and their level of engagement in 7 types of Internet activities: social media use, gaming, gambling, pornography viewing, newsreading or watching, activities related to school or work, and targeted information searches that are not related to school or work. Second, the study also tested whether these effects would be sustained or accounted for by perceived consequences of using those Web-based activities. We investigated the impact of both negative consequences (eg, withdrawal, sleep loss) and positive consequences (eg, enjoyment, finding new friends). In addition to performing these analyses on cross-sectional data, we also tested whether these effects would predict changes in mental health over a period of 4 months.
Data were collected as a part of the Suicide Prevention through Internet and Media Based Mental Health Promotion (SUPREME) trial (Current Controlled Trials ISRCTN65120704). The study was carried out by collaborating mental health research centers in Estonia, Hungary, Italy, Lithuania, Spain, Sweden, and the United Kingdom. As part of this project, a randomized controlled longitudinal study was carried out in 2012-2013 to evaluate a Web-based mental health intervention website, which was tested in a randomly selected sample of adolescents in a selected area of these countries. Inclusion criteria of the schools were: (1) the school authority agrees to participate; (2) the school is a state school (ie, not private); (3) the school contains at least 100 pupils within the age range of 14-16; (4) the school has more than 2 teachers for pupils aged 15 years; (5) no more than 60% of pupils are of either gender. Participants were cluster randomized, based on school affiliation, into either a full-intervention condition (with access to the intervention website) or a minimal-intervention control group (without access to the intervention website), and were administered an evaluation questionnaire at baseline and at 2 and 4 months of follow-up. The questionnaire included questions about their Internet habits, mental health and suicidal behaviors, and other variables relevant to the evaluation. This study did
Subjects were registered pupils of state schools randomly selected from a predefined area in each country: West Viru County (Estonia), Budapest (Hungary), Molise (Italy), Vilnius city (Lithuania), Barcelona city (Spain), Stockholm County (Sweden), and eastern England (the United Kingdom). Eligible state schools in these areas were randomly arranged into a contact order, the order in which schools were contacted and asked to participate. If a school declined, the next school on the list was contacted. If a school accepted participation, a team of researchers went to the school and presented the background, aims, goals, and procedures of the study to the pupils verbally and through consent forms. As the study procedure included screening for suicidal adolescents, participation was not completely anonymous, but participants’ identities were encrypted in the questionnaire. Written consent was obtained from all pupils who agreed to participate (as well as from one or both parents according to ethical regulations in the region). The study was approved by ethics committees in all participating countries.
The sampling procedure resulted in a total number of 2286 adolescents participating at baseline (Estonia=3 schools, 416 participants; Hungary=6 schools, 413 participants; Italy=3 schools, 311 participants; Lithuania=3 schools, 240 participants; Spain=3 schools, 182 participants; Sweden=9 schools, 337 participants; the United Kingdom=3 schools, 387 participants). Of the participants, 1571 (68.72%) were randomized to the full-intervention group and 715 (31.27%) to the minimal-intervention group. There was a notable dropout rate in the study. In the total sample, the number of subjects that discontinued participation comprised 467 pupils (20.42%) between T1 and T2 and 244 pupils (13.41%) between T2 and T3. Subjects were included in the longitudinal analyses if they had participated at least at T1 and T3, but participation at T2 was not necessary. This resulted in a longitudinal sample of 1544 subjects, with 56% women and a mean age of 15.8 years (standard deviation, SD=0.91 years).
Measures of Internet behaviors and uses were constructed specifically for this study. This included items that measured the regularity of Internet use (eg, using the Internet once a month vs using it once a week) and the number of hours spent on the Internet on a typical week. Participants were also asked to rate how much time they spend on 7 different activities when using the Internet (socializing, gaming, school- or work-related activities, gambling, newsreading or watching, pornography, and targeted searches that are not related to school or work). Participants rated these activities on a 7-point scale (1=I spend very little or no time doing this; 7=I spend very much time doing this). The last set of items asked participants to rate the self-perceived consequences of engaging in said activities. Participants were asked to rate the extent to which various consequences apply to them, but
For the sake of clarity, we refer to some of these consequences as “positive” (finding new friends; having fun; learning interesting things) because they are outcomes of Internet use that do not necessarily imply addictive behavior and can be expected to lead to better mental health (if at all). We refer to other consequences as “negative” (staying on the Internet longer than intended; choosing Web-based activities instead of offline social activities; staying up and losing sleep; feeling moody when Web-based activities cannot be accessed) because they suggest symptoms of problematic Internet use and can therefore be expected to lead to poor mental health. For example, these negative consequences resemble those included in the IAT [
Participants’ levels of depression, anxiety, and stress were assessed by means of the 3 subscales constituting the 42-item version of the
All study procedures took place at the respective schools in classrooms or computer rooms. The questionnaires were administered either in paper and pencil format or using a Web-based survey tool, if the school was able to provide computers for all pupils at time of data collection. The questionnaire contained items used to screen for suicidal adolescents (The Paykel Suicide Scale [
Two main analyses were performed in this study: 1 cross-sectional hierarchical multiple regression analysis and 1 longitudinal analysis. The measure of frequency of Internet use was omitted from analysis owing to a ceiling effect (90% of participants reported using the Internet at least once per day). The remaining predictor variables were thus the self-reported number of weekly hours online, the ratings of the 7 activities, and the ratings of the 9 consequences of Internet use. The composite DASS score was the dependent variable in these analyses (tests of statistical assumptions are described in
DASS-42 scores could be calculated for 2220 participants. Total DASS scores ranged between 0-3 points, where higher scores indicate more mental health problems. The mean baseline scores for males, females and the total sample are presented in
Descriptive results (means and standard deviations) for mental health and Internet use measures at baseline.
Variablea | Total | Women | Men | Gender differenceb | ||
M (SD) | M (SD) | M (SD) | ||||
Depression | 0.52 (0.59) | 0.62 (0.64) | 0.40 (0.49) | 9.15 | <.001 | 0.40 |
Anxiety | 0.48 (0.49) | 0.54 (0.51) | 0.40 (0.45) | 7.02 | <.001 | 0.30 |
Stress | 0.72 (0.60) | 0.83 (0.63) | 0.57 (0.53) | 10.39 | <.001 | 0.37 |
DASS (total) | 0.57 (0.52) | 0.67 (0.54) | 0.46 (0.45) | 9.71 | <.001 | 0.42 |
Time spent on the Internet | 17.12 (17.72) | 16.43 (17.04) | 17.96 (18.50) | −1.99 | .046 | −0.09 |
Socializing | 4.94 (1.73) | 5.29 (1.62) | 4.51 (1.77) | 10.80 | <.001 | 0.46 |
Gaming | 3.05 (2.04) | 2.03 (1.42) | 4.33 (1.98) | −31.95 | <.001 | −1.33 |
School or work | 3.71 (1.54) | 4.01 (1.49) | 3.34 (1.52) | 10.44 | <.001 | 0.45 |
Gambling | 1.31 (0.98) | 1.09 (0.51) | 1.58 (1.30) | −12.06 | <.001 | −0.50 |
News | 2.96 (1.66) | 2.93 (1.63) | 2.99 (1.69) | −0.83 | .41 | NS |
Pornography | 1.73 (1.48) | 1.10 (0.55) | 2.53 (1.86) | −25.42 | <.001 | −1.04 |
Targeted searches | 3.28 (1.68) | 3.28 (1.68) | 3.29 (1.68) | −0.19 | .85 | NS |
Finding friends | 3.42 (1.79) | 3.40 (1.81) | 3.44 (1.76) | −0.45 | .65 | NS |
Learning | 4.07 (1.64) | 4.02 (1.60) | 4.12 (1.68) | −1.35 | .18 | NS |
Having fun | 4.66 (1.77) | 4.49 (1.73) | 4.88 (1.80) | −5.08 | <.001 | −0.22 |
Meaningfulness | 4.12 (1.22) | 4.10 (1.15) | 4.14 (1.30) | −0.69 | .49 | NS |
Impact on grades | 3.95 (1.24) | 3.95 (1.24) | 3.93 (1.24) | 0.37 | .71 | NS |
Staying on the Internet longer | 4.03 (1.86) | 4.22 (1.84) | 3.79 (1.86) | 5.34 | <.001 | 0.23 |
Prefers Web-based relations | 2.14 (1.44) | 1.99 (1.38) | 2.31 (1.49) | −5.16 | <.001 | −0.22 |
Sleep loss | 3.07 (1.97) | 3.05 (1.98) | 3.09 (1.95) | −0.51 | .61 | NS |
Withdrawal (negative mood when inaccessible) | 2.25 (1.52) | 2.24 (1.54) | 2.26 (1.49) | −0.22 | .83 | NS |
aMental health scores (depression, anxiety, stress, DASS total) range between 0 and 3. Time spent on the Internet is measured in hours. All other Internet-related measures range between 1 and 7.
bGender differences were determined through independent samples t-tests;
The cross-sectional hierarchical multiple regression analysis was used to predict DASS scores at T1 by means of Internet use at T1. The first model comprising the control variables (gender, age, experimental condition) was highly significant (
Results from the cross-sectional hierarchical multiple regression analysis. Statistics are presented for each predictor variable in each model.
Entered in model no | Predictor variable | Model noa | Standardized Beta | 95% CI | Tolerance | ||
1 | (Constant) | 1 | 1.07 | .29 | |||
2 | 0.41 | .68 | |||||
3 | 0.03 | .97 | |||||
4 | −0.30 | .77 | |||||
1 | Exp. Conditionb | 1 | .00 | −0.05 to 0.04 | −0.12 | .90 | 1.00 |
2 | −.01 | −0.05 to 0.04 | −0.28 | .78 | 1.00 | ||
3 | .00 | −0.05 to 0.05 | 0.02 | .98 | 0.99 | ||
4 | .02 | −.03 to 0.06 | 0.80 | .42 | 0.98 | ||
1 | Genderc | 1 | −.21 | −0.26 to −0.16 | −8.80 | <.001 | 1.00 |
2 | −.22 | -0.26 to −0.17 | −9.10 | <.001 | 1.00 | ||
3 | −.26 | −0.32 to −0.20 | −8.03 | <.001 | 0.52 | ||
4 | −.22 | −0.28 to −0.16 | −7.10 | <.001 | 0.51 | ||
1 | Age | 1 | −.03 | −0.07 to 0.02 | −1.04 | .30 | 1.00 |
2 | −.01 | −0.06 to 0.04 | −0.39 | .69 | 0.98 | ||
3 | .00 | −0.05 to 0.05 | −0.02 | .99 | 0.97 | ||
4 | .01 | −0.04 to 0.05 | 0.31 | .76 | 0.92 | ||
2 | Time spent on the Internet | 2 | .12 | 0.08-0.17 | 5.10 | <.001 | 0.98 |
3 | .10 | 0.05-0.14 | 3.88 | <.001 | 0.91 | ||
4 | .02 | −0.03 to 0.07 | 0.93 | .35 | 0.84 | ||
3 | Socializing | 3 | .05 | 0.00-0.10 | 2.06 | .04 | 0.90 |
4 | −.01 | −0.06 to 0.03 | −0.57 | .57 | 0.78 | ||
3 | Gaming | 3 | −.02 | −0.07 to 0.04 | −0.54 | .59 | 0.64 |
4 | −.06 | −0.12 to −0.01 | −2.17 | .03 | 0.57 | ||
3 | School or work | 3 | −.05 | −0.10 to 0.00 | −2.02 | .04 | 0.83 |
4 | −.03 | −0.08 to 0.02 | −1.22 | .22 | 0.78 | ||
3 | Gambling | 3 | .08 | 0.03-0.13 | 3.11 | .002 | 0.89 |
4 | .05 | 0.01-0.10 | 2.30 | .02 | 0.87 | ||
3 | News | 3 | .01 | −.04 to 0.06 | 0.50 | .62 | 0.85 |
4 | .03 | −0.02 to 0.07 | 1.08 | .28 | 0.82 | ||
3 | Pornography | 3 | .07 | 0.01-0.12 | 2.46 | .01 | 0.72 |
4 | .02 | −0.03 to 0.07 | 0.78 | .44 | 0.71 | ||
3 | Targeted searches | 3 | .13 | 0.08-0.18 | 4.94 | <.001 | 0.84 |
4 | .09 | 0.04-0.14 | 3.56 | <.001 | 0.75 | ||
4 | Finding friends | 4 | .03 | −0.01 to 0.08 | 1.38 | .17 | 0.79 |
4 | Learning | 4 | .01 | −0.04 to 0.06 | 0.34 | .73 | 0.67 |
4 | Having fun | 4 | −.05 | −0.10 to 0.00 | −1.80 | .07 | 0.71 |
4 | Meaningfulness | 4 | −.05 | −0.10 to −0.01 | −2.22 | .03 | 0.90 |
4 | Impact on grades | 4 | −.07 | −0.11 to −0.02 | −2.78 | .005 | 0.88 |
4 | Staying on the Internet longer | 4 | .01 | −0.04 to 0.07 | 0.53 | .60 | 0.66 |
4 | Prefers Web-based relations | 4 | .12 | 0.07–0.17 | 4.74 | <.001 | 0.79 |
4 | Sleep loss | 4 | .13 | 0.08-0.19 | 4.95 | <.001 | 0.65 |
4 | Withdrawal (neg. mood when inaccessible) | 4 | 0.22 | 0.17-0.27 | 8.80 | <.001 | 0.76 |
aThe model numbers designate which values were obtained when (1) only control variables were analyzed, (2) when time spent over the Internet was added to the model, (3) when Web-based activities were added to the model, and (4) when consequences of Internet use were added to the model.
bFor experimental condition, the minimal-intervention condition constitutes the reference group.
cFor gender, females constitute the reference group.
The longitudinal hierarchical multiple regression analysis was used to predict change in overall psychopathology (the score difference between T1 and T3) by means of change in Internet use. There was no indication of problematic levels of collinearity in the model, as all variables had a tolerance value above 0.7. The first model comprising the control variables (gender, age, experimental condition) was not significant (
Thus, Internet use that was reported to result in staying up late and losing sleep (“sleep loss”) and to produce negative mood when it could not be accessed (“withdrawal”) were the only variables that consistently predicted longitudinal change in mental health. To further investigate these negative consequences, 2 standard multiple regressions were calculated to predict longitudinal changes in each of these variables by means of changes in time spent on the Internet and the different Web-based activities. The regression model that predicted sleep loss was significant (
Results from the multiple regression analysis predicting changes in “sleep loss” by means of change in Internet use.
Predictor variable | Standardized beta | 95% CI | ||
Constant | 0.82 | .42 | ||
Time spent on the Internet | .07 | 0.01-0.13 | 2.25 | .03 |
Socializing | .06 | 0.00-0.11 | 1.89 | .06 |
Gaming | .08 | 0.02-0.14 | 2.59 | .01 |
School or work | −.10 | −0.16 to −0.04 | −3.16 | .002 |
Gambling | .01 | −0.05 to 0.07 | 0.36 | .72 |
News | .04 | −0.02 to 0.10 | 1.20 | .23 |
Pornography | .06 | 0.01-0.12 | 2.14 | .03 |
Targeted search | .08 | 0.02-0.14 | 2.56 | .01 |
Results from the multiple regression analysis predicting changes in “withdrawal” by means of change in Internet use.
Predictor variable | Standardized beta | 95% CI | ||
Constant | 3.47 | .001 | ||
Time online | .12 | 0.06-0.17 | 3.93 | <.001 |
Socializing | .03 | −0.03 to 0.09 | 1.03 | .31 |
Gaming | .08 | 0.02-0.13 | 2.56 | .01 |
School or work | .00 | −0.06 to 0.06 | −0.03 | .97 |
Gambling | .14 | 0.08-0.20 | 4.75 | <.001 |
News | .04 | −0.02 to 0.10 | 1.27 | .20 |
Pornography | .10 | 0.04-0.16 | 3.38 | .001 |
Targeted search | .02 | −0.04 to 0.08 | 0.57 | .57 |
The purpose of this study was to identify Internet-related risk and protective factors for mental health problems and to test if the effects of time spent on the Internet and on various Web-based activities could be accounted for by a number of perceived consequences of those activities. This was investigated by examining the association between adolescents’ general mental health (combined levels of depression, anxiety, and stress or tension) and those Internet-related behaviors, both cross-sectionally and longitudinally over a 4-month period.
The cross-sectional results showed that mental health was predicted by Internet-related behaviors at baseline (15.3% explained variance after adjusting for the number of predictors in the model). Individual effect sizes were rather small (standardized ß=.05-.22). Time spent on the Internet had a larger effect than most individual activities, but consequences of Internet use explained the largest variance in DASS scores (11.1%). Of these, 3 of the 4 negative consequences were the most important predictors (preference for Web-based activities over offline social activities, sleep loss, and withdrawal), whereas the positive consequences were nonsignificant. Internet use that was perceived to increase life meaning or improve school grades or work performance was associated with better mental health, but the effects were smaller than for the negative consequences.
Furthermore, the results showed that time spent on the Internet, social media use, pornography viewing, and school or work activities were only significant predictors when perceived consequences were not accounted for, which suggests that the mental health effects of these activities were explained by the consequences. Web-based gaming, gambling, and targeted searches, on the other hand, were significant predictors of mental health even when controlling for perceived consequences, suggesting that the content of these activities was relatively important in comparison with perceived consequences, with regard to mental health. Together, these results indicate that all Web-based activities measured in this study are predictive of mental health, but only some of them seem to have content-based effects large enough to be detected in a fully adjusted model. The other activities seemed to only affect mental health by means of their perceived consequences, mainly the preference for Web-based interactions, sleep loss, and withdrawal. As these negative consequences are indicative of problematic Internet use [
Previous studies have linked sleep loss and withdrawal symptoms to mental health problems and problematic Internet use [
The results of this study confirm that problematic (or unhealthy) Internet use cannot simply be equated to high-intensity or frequent Internet use. First, although time spent on the Internet was found to be negatively associated with mental health, some activities, such as school work, were positively associated. Second, time spent on the Internet was not an independent risk factor for mental health after accounting for the perceived consequences of Internet use, underlining that Internet use is not intrinsically harmful. Even when it comes to specific activities, for example, gaming, the relationship could be complex. Previous studies have established that gaming has a negative effect on mental health (eg, [
The causal link between general Internet use and mental health also seems complex. Previous authors have acknowledged the possibility that the risk associated with Internet use could reflect an already present disorder, which may have an effect on how the Internet is used [
Furthermore, although Internet-related sleep loss was found to be a longitudinal predictor of mental health, there is an established bidirectional link between sleeping problems and depression [
In this study, we found no effect of perceived positive consequences of Internet use on mental health, and it is possible that this is because they are actually rather motives for using the Internet. In other words, participants may have reported consequences they hoped for rather than what actually happened. Sagioglou and Greitemeyer [
This study is limited by the nature of the measurements used to estimate the participant’s Internet use. One issue of validity concerns the consequences of Internet use, which cannot be assumed to perfectly reflect the real outcomes. In addition to the difficulty of observing the impact of daily activities on one’s own health and behaviors, this measure might also be particularly vulnerable to recall biases and expectancy effects. Hence, this study only intended to measure the perceived consequences. It is also difficult to know whether the perceived consequences are produced by the Internet behaviors or some third factor, such as comorbid disorders. Another limitation of this study is that we did not make in-depth measures of the Web-based content that participants use. Therefore, one should take caution when applying these results to uses of more specific content; for example, different types of games and social networking activities may have different effects on both perceived consequences and mental health. Furthermore, our measurements did not include any problematic Internet use diagnostic tool. It is possible that if we had included more negative consequences of Internet use, or specific problematic Internet use criteria, this would have explained a larger proportion of the effects of the Web-based activities. Finally, there was a notable dropout rate between baseline and follow-up measurements (34%), which reduced the statistical power in the longitudinal analyses compared with the cross-sectional analyses. Also, participation in this study was not completely anonymous, and participants with high suicidal risk were excluded from the data analysis, which could mean that some of the adolescents with the most severe psychopathology were not represented in the analyses.
Different Web-based activities or content can have specific effects on mental health, even when used in moderate levels and when adjusting for the number of hours spent on the Internet. Web-based activities differ in how consistently, how much, and in what direction they affect mental health. Activities also differ regarding which negative consequences they produce, and those consequences (especially sleep loss and withdrawal) seem to predict mental health outcomes to a greater extent than the activities themselves. Therefore, it seems that time spent on the Internet and Web-based content are predictive of mental health mainly because they predict such negative consequences. These results underscore the importance of differentiating between generalized and specific forms of problematic Internet use. It also confirms that Internet use is not intrinsically harmful, but it depends on the activity that one engages in, and how it affects the individual. Change in mental health over time appears to be best predicted by changes in Internet-related sleep loss and withdrawal, and interventions to reduce harmful Internet use should therefore target such consequences. Positive consequences of Internet use may not predict mental health directly but might predict the propensity to engage in certain Web-based activities excessively or problematically. However, the causality between Internet use and mental health morbidity is complex and likely to be reciprocal, which means interventions or treatments of problematic Internet use might have to be multifaceted to be effective.
Depression Anxiety Stress Scale
Diagnostic and Statistical Manual of Mental Disorders
Internet Addiction Test
Suicide prevention through Internet and media based mental health promotion
All authors except J Westerlund were involved in the planning or execution stages of the SUPREME project, including the Randomized Controlled Trial, where V Carli was the principal investigator. J Balasz, A Germanavicius
None declared.