We designed an exploratory survey among the population of Martignas-sur-Jalle (n=7400). A task force with the University of Bordeaux, Charles Perrens Hospital Addiction Clinic, Martignas-sur-Jalle city council, and population representatives was established to carry out and supervise the survey conducted from January 4, 2016, to February 25, 2016.

The study targeted all adults and adolescents from middle-school age (ie, from 11 to 12 years of age) with no upper age limitation. The task force agreed on this minimal age to assess screen users and ensure understanding of the questions. There were no exclusion criteria.

Participants received the questionnaire from distribution points (all city services and schools) and returned them directly through ballot boxes or mail. Of the 1200 questionnaires distributed, 401 were returned. The response rate was 33.4% and the sample represented 6.6% of the target population of the city. After a quality check, 101 questionnaires were excluded (53 with no information, 7 without age, and 41 with ScUD questions not completed). The remaining 300 questionnaires were used for the database (Figure 1).

The survey questionnaire was a 10-15-minute 2-part self-administrated questionnaire including 49 closed-ended questions designed by the task force. The first part (37 questions) explored the main types of screens used over the past month (eg, TV, computers, smartphones, tablets, and handheld consoles) and for which activities (communication, social media, work, searching information on internet, other documentation, shopping, gaming, gambling, and others). The second part assessed each ScUD diagnostic criteria in the past 12 months (9 questions) and which screens and activities were considered problematic, that is, when at least one ScUD criterion was endorsed (1 question for screens and 1 for activities). We used the previously published French translation of the 9 IGD criteria [18] and adapted them to screen use (the term “videogames” was replaced by “screens”). The original French version and the translated version of the questionnaire are provided in Multimedia Appendix 1.

Our variables of interest were sociodemographic data (age, sex), screen use (converted into hours per day), activities, main screen used, main activities, prevalence of each diagnostic criteria, and ScUD. Activities were quantified by the number of days in the past 30 days (participants had to choose out of the following 4 options: every day or almost every day; more than 1 day out of 2; less than 1 day out of 2; and never or almost never). The main screen used was defined as the prevalence of participants for whom this screen was the most used (frequency over the past 30 days multiplied by the time per day). The main activity was defined for each activity as the prevalence of participants endorsing the activity on the main screen. ScUD was defined when 5 or more criteria were reported on the scale adapted from IGD (Table 1).

The survey was anonymous and confidential, and met French regulation ethics standards for noninterventional research after institutional review board (Sanpsy/University of Bordeaux) review [43]. Participation was voluntary with no financial compensation. The questionnaire was distributed with an information note presenting the investigation, consent collection, confidentiality, and legal issues.

Of the 300 participants, 171 were women (57.0%), mean age was 27 years (SD 18.9 years), and 160 were under 18 years (53.3%). The youngest participant was 11 years and the oldest was 84 years. Almost all participants (n=297, 99.0%) reported daily screen use (Table 2).

Most of the sample reported no criteria (n=166, 55.3%), 134 participants (44.7%) reported at least one criterion (potential problem users), and 5 participants (1.7%) reported 5 criteria or more and qualified for a potential ScUD (Table 2). Adolescents (defined as 11-17 years; mean age 12.92 years, SD 1.50 years) were significantly more likely to endorse at least one ScUD criteria than adults (defined as being aged above 18 years, mean age 43.2 years, SD 16.5 years; 97/300, 32.3% vs 37/300, 12.3%; P<.001).

The prevalence of each criterion is reported in Table 3. The most endorsed were loss of control (60/300, 20.0%) and preoccupation (52/300, 17.3%). The less endorsed were losing an opportunity (6/300, 2.0%) and tolerance (7/300, 2.3%).

Unidimensionality was confirmed by all fit indices (comparative fit index 1.000; Tucker-Lewis Index 1.026; root mean square error of approximation ≤0.0001; and factor loading ≥0.4 for each criterion). Local independence was confirmed by the absence of residual correlation between the items (minimum and maximum standardized z-scores for the different combinations of item responses were equal to –1.042 and 1.129, respectively; maximal chi-square value was 2.008). All criteria had relatively high factor loading except tolerance (0.493), withdrawal (0.457), and loss of control (0.477), but these also remained above the cut-offs sanctioning unidimensionality. Factor loading for loss of interest (0.779) was higher than for any other diagnostic criterion, followed by preoccupation (0.726). The criterion preoccupation (1.279) had the lowest difficulty to be endorsed, followed by loss of control and loss of interest. Inversely, the withdrawal and tolerance criteria showed the highest difficulty. Discrimination parameters ranged from 0.884 to 2.027, indicating a good ability to delineate individuals who were higher versus lower to the latent trait (ICC; Figure 2). Both Loss of interest (2.027) and preoccupation (1.279) criteria showed a higher discrimination, while loss of control showed a lower discrimination (0.884) compared with other criteria (Table 3).

Item information curves (Figure 3) showed that most discriminating criteria were, in order, loss of interests, preoccupation, deceive/cover up, and risk/lose relationship/opportunities, which also provided the most information on the measurement of the latent trait. Loss of interests and preoccupation criteria also provided the greatest amount of information and high precision across the latent trait severity continuum of ScUD. Loss of control criterion was identified as the least discriminating and the least informative.

Total information curves (Figure 3) showed an increased information across the severity spectrum for the 9 IGD criteria group. Removing the loss of control criterion did not seem to affect the ability of the test to capture the disorder phenomenon. However, removing the loss of interests criterion changed the amount of severity information provided by the test. The 3 models brought roughly the same range of severity.

In univariate analysis, participants with no ScUD criterion were more likely to report television (P<.001) as the most used screen (Table 4). Participants with at least one ScUD criterion were more likely to have smartphone (P=.04) and computer (P=.04) as the most used screens, which they also reported as the most problematic screens: smartphone (69/269, 25.7%) and computer (61/269, 22.7%). For tablets and handled console there was no difference between participants with and without at least one ScUD criteria.

Compared with participants with no ScUD criteria, participants with at least one ScUD criteria reported more video gaming (P<.001) and communication/social network (P<.001), which they also reported as the most problematic activity: 36.6% (71/194) and 31.4% (61/194), respectively.

In multivariate analysis, when controlled on age and gender, participants with at least one ScUD criterion were more likely to have a computer (P=.004) as the most used screen type. For activities, they reported more videogaming (P=.002) and communication/social network (P=.03) compared with participants with no ScUD criteria. Besides, a new association was found between having at least one ScUD criterion and watching news and research of information (P=.002) that was not observed in the univariate analysis.

This is the first study to combine description of screen use and exploration of the dimensionality and psychometric validity of the 9 IGD DSM-5 criteria adapted to a potential “ScUD” among a general population sample. Prevalence of ScUD was 1.7% (5/300) in our sample. Our results confirm the initial hypothesis of unidimensionality of the 9 IGD DSM-5 criteria adapted to ScUD.

Almost all participants (297/300, 99.0%) of this survey used screens daily, reflecting a high level of equipment use in daily life. ScUD criteria were characterized by the DSM-5 IGD criteria adapted to screen use. The majority of our participants (166/300, 55.3%) self-reported none of the criteria in the past 12 months. However, a notable proportion (134/300, 44.7%) self-reported at least one criterion and a screen type or screen activity as problematic in the past 12 months. This can be interpreted as a need for support and advice for better use of screens in that population. “Screen addiction” prevalence (≥5 criteria endorsed) was 1.7% (5/300), which is in range with the prevalence of IGD (2.0%) in population-based studies [23,44,45]. Two recent meta-analysis on gaming disorder prevalence, a “screen-related addiction,” showed prevalence in the same order of magnitude, 3.0% and 3.3%, respectively [44,45]. Our results are interesting in that they go against lay beliefs of a very high prevalence of “screen addiction.” For those people satisfying 5 or more ScUD criteria, a persistent and recurrent use of screens leading to clinically significant impairment or distress could be assumed, similar to IGD in the DSM-5 [17,18]. In this particular situation, it should be assumed that advice on screen use would be insufficient, and that an addiction-oriented intervention would be useful. There were more adolescents than adults with at least one ScUD criterion (97/300, 32.3% vs 37/300, 12.3%). As many as 2 adults and 3 teenagers met the threshold of 5 criteria for ScUD.

Screens most used differed between participants with at least one ScUD criteria or no criteria. When controlled for age and gender, participants with at least one ScUD criterion were significantly more likely to use computers as the main screen. This may be explained by the activities performed on computers. These participants reported more video gaming, communication/social network, and watching news and research of information, all of which are commonly done on computers. Screens and activities reported as problematic by participants with at least one ScUD criterion were similar to the screens used (eg, smartphone, computers) and activities (eg, video gaming, communication/social network) performed the most, a result that may be of interest for prevention. There was a group of screen users that reported some problem with use and as such is likely to be responsive to interventions focused on related support.

Our study showed unidimensionality of the 9 IGD DSM-5 criteria adapted to ScUD. The model showed adequate fit and the criteria reflected 1 underlying latent trait (ScUD). Moreover, we found no residual correlation between the items, and thus confirmed local independence, a fundamental assumption in IRT models. This means that the items were correlated only through the latent trait that the test is measuring [46]. Some criteria had specific psychometric characteristics. Loss of interest (losing interest or reducing participation in other recreational activities) and preoccupation (being absorbed by screen use and thinking about it) loaded more strongly than other diagnostic criteria, indicating that they fit well with the 1-factor model, similar to results from a parent-reported survey of screen media “addiction” in children [14]. These criteria were among the more frequently endorsed, and had higher discrimination than others. Thus, both loss of interest and preoccupation criteria seem to capture the less severe end of the diagnostic spectrum, and the criteria well differentiated between respondents with high and low screen use severity. By identifying participants with less severe ScUD, these items are potentially useful as early indicators of ScUD [47]. It would be interesting to assess, within a prospective cohort of adolescents, whether the occurrence of these criteria predicts a subsequent ScUD.

Withdrawal and tolerance criteria had the lowest factor loading and showed the highest difficulty and moderate discrimination power, similar to results in a general population study of children [14]. Our results suggest that these criteria may not be relevant to define ScUD. By contrast, in some IGD surveys including population of video gamers with significant gaming time, withdrawal and tolerance had higher factor loadings and seemed discriminating [19,20], suggesting that very high and regular level of gaming practice may promote tolerance and withdrawal symptoms. Recently, the World Health Organization (WHO) specified its own gaming disorder criteria in the 11th revision of the International Classification of Diseases (ICD-11) [48]. Tolerance and withdrawal criteria were removed, as well as preoccupation, deceive/cover up, and escape adverse mood. Additional studies among the general population are thus needed to determine to what extent withdrawal and tolerance are related to the intensity of screen use and characterize potential ScUD.

The loss of control criterion (feeling that you should use less screens, but being unable to cut back on the amount of time spent watching it) had a lower factor loading, a lower discrimination power, and was among the less difficult (more frequent) criteria. This suggests that this criterion is frequent in a population without ScUD, perhaps due to high overall screen use exposure [1,3]. Including a criterion with poor discrimination may increase the risk for false-positive diagnosis, especially at the lower range of difficulty (high frequency) [47]. In previous studies on IGD [19,20] this criterion had low standard in terms of factor loading, discrimination, and difficulty. However, this result is questionable because this criterion is reported to be a central criterion of addiction [17,49]. By contrast, in another study about screen addiction, loss of control showed the highest factor loading in children [14], possibly because reports were from children’s caregivers, and cessation of use is a source of conflict between parents and children. More studies are therefore needed to evaluate the potential importance of this criterion in ScUD.

Study limitations are to be noted. This was a convenience sample with a somewhat low response rate. Survey respondents represented 6.60% (401/6075) of the target population (men and women above 11 years from Martignas-sur-Jalle). Compared with the target population, our final sample was younger (24 years vs 40.5 years), mainly due to a higher proportion of 12-18 year olds (160/300, 53.3%, in our sample vs 784/6075, 12.9%, in the target population). Gender ratio was comparable (3159/6075, 52.0%, women in target population vs 171/300, 57.0%, in our sample). As our questionnaire was self-administered, risk of participant misinterpretation should be considered. However, we used the operationalized formulations for IGD assessment [18]. As a result of missing information, 101 questionnaires (responses) were excluded and there was a higher rate of adolescents among our sample. This could suggest that the questionnaire may have been of little interest to some participants, or might have been difficult to understand, or that adolescents might feel more concerned by this survey. An important element for the validity of the ScUD diagnostic criteria is to determine whether criteria or criteria sets function differently across population subgroups, such as age or sex. As our sample is composed of adolescents and adults, it would be interesting to see whether criteria behave differently according to age. However, in this study, the prevalence of some criteria was too small and thus such an analysis could not be performed here. Additional data in samples more likely to endorse ScUD criteria should be collected and analyzed for differential item functioning. Further studies should investigate the relationships between ScUD items to determine whether there is some local dependence, indicating a possible redundancy. Finally, because craving was not part of IGD criteria [18], no craving criterion was assessed. Some studies suggest that craving should be included [50,51], as it has a high prevalence in samples of those with IGD [52,53], and is the most specific criterion for many substance use disorders [37]. Additional studies should thus be carried out by including craving.

We described screen use in a French community sample and have shown that the adaptation of the DSM-5 IGD criteria to “ScUD” has good psychometric validity. Endorsement of diagnostic criteria in the past 12 months could be interpreted as current complaints and impairment of the users, strengthening the possibility for ScUD to qualify as a disorder. Further studies are needed to confirm the validity of ScUD diagnosis and its negative consequences. We suggest that there may be similarities between different screen-related addictions, thus allowing for a broader tool to encompass the screen activities. Future studies will have to determine whether the type of screen/screen activity is related to the likelihood of ScUD diagnosis, the validity of a craving criterion, if all criteria are needed or if some should be removed or replaced, and if the diagnosis threshold of 5 is appropriate. Screen use and its consequences represent an important emerging field for addiction research.