The Digital Addiction Scale for Children: Development and Validation
Researchers worldwide have developed and validated several scales to assess various forms of adults’ digital addiction. The urge for some of these scales found support in World Health Organization’s inclusion of gaming disorder as a mental health condition in its eleventh revision of the International Classification of Diseases in June 2018. Additionally, several studies have shown that children are starting to use digital devices (DDs) (e.g., tablets and smartphones) at a very young age, including playing video games and engaging in social media. Consequently, the need for early detection of the risk of digital addiction among children is becoming more of a necessity. In the present study, the Digital Addiction Scale for Children (DASC)—a 25-item self-report instrument—was developed and validated to assess the behavior of children 9 to 12 years old in association with DD usage, including video gaming, social media, and texting. The sample comprised 822 participants (54.2 percent males), from grade 4 to grade 7. The DASC showed excellent internal consistency reliability (α = 0.936) and adequate concurrent and criterion-related validities. The results of the confirmatory factor analysis showed that the DASC fitted the data very well. The DASC paves the way to (a) help in early identification of children at risk of problematic use of DDs and/or becoming addicted to DDs and (b) stimulate further research concerning children from different cultural and contextual settings.
Scientific research investigating digital addiction among children is still in its infancy. Digital devices (DDs; tablets, smartphones, and game consoles) have become an integral part of households and their use can begin at very young ages.1,2 They provide entertainment, convenience, affordability, and portability.3 Children use these devices to play video games, watch videos (e.g., YouTube and Twitch), communicate, and interact with social media (e.g., Instagram and Snapchat). Although most device use is positive, their excessive use/misuse among a minority may become problematic and negatively affect educational, psychological, social, and/or physical well-being.4–8 Recently, the World Health Organization included gaming disorder as a mental health condition in the ICD-11.9 The ICD-11 states that the condition results in gamers having little control over gaming, gaming takes precedence over other life interests, and gaming being continued despite negative consequences.9 This was preceded by the inclusion of Internet gaming disorder (IGD) in the DSM-5 as a condition warranting further study.10
In modern society, playing video games and social media use can reinforce and complement each other. Currently, Fortnite is one of the most popular video games with 200 million users worldwide.11,12 In addition to playing video games, up to 80 percent of players watch professional gamers playing through live streaming12 and participate in related social media platforms/channels. Consequently, players can constantly switch back and forth between DDs to both play and watch video games.
An American Common Sense Media survey reported that three-quarters of teenagers own a smartphone, and 24 percent describe themselves as constantly connected online.13 The Centre for Addiction and Mental Health14 surveyed 11,438 Canadian students (grades 7–12), reporting that (a) 20 percent spent 5 hours or more on social media daily (16 percent in 2015; 11 percent in 2013), (b) 23 percent played video games daily/almost daily, (c) 9 percent played video games for 5 hours+ per day, (d) 30 percent spent 5 hours+ per day on DDs recreationally, and (e) 5 percent reported symptoms suggesting serious problematic technology use. In addition, research shows that some preteens engage in excessive DD use and a small minority experience behavioral addiction symptoms.15
The potentially problematic use of technology necessitates the development of a scale assessing children’s DD use. While several scales related to different types of digital addiction have been developed, they mainly target populations over 12 years of age and specialize in one aspect such as IGD,16 social media addiction,17 and smartphone addiction.18 To the authors’ knowledge, only one scale has been developed for children under 12 years of age and that specifically addressed video game addiction.15
The aim of this study was to develop and validate an instrument to assess children’s overall addiction to DDs. As several studies have associated digital addiction with attention deficit hyperactivity disorder and obsessive-compulsive disorder8,19; stress20–22; anxiety, depression, and narcissism7,23; low self-esteem7,24–26; and poor academic performance,4,27–29 a scale for assessing DD addiction will help in the assessment, diagnosis, and treatment of different symptoms among children aged 9–12 years.
In the present study, students from grades 4 to 7 constituted the target population. Fifteen schools were selected by a simple random sampling technique from regions geographically distributed across Lebanon. The sample comprised 822 participants (54.2 percent males) aged 9–12 years (Mage = 10.99 years, standard deviation [SD] = 1.26 years).
The survey included demographic questions concerning age, gender, most recent school grade average, and DD usage habits, including types of DDs used, purpose of using DDs, and average number of hours of DD usage. Also included was the 30-item version of the Revised Child Anxiety and Depression Scale (RCADS-30),30 which was used to validate the Digital Addiction Scale for Children (DASC) criterion validity.
Digital Addiction Scale for Children
The DASC is a 25-item self-report instrument that was developed based on nine diagnostic DSM-5 IGD criteria and also mapped onto Griffiths31 six core addiction criteria (preoccupation, tolerance, withdrawal, mood modification, conflict, and relapse). Added to those were three additional criteria (i.e., problems, deception, and displacement). The problems criterion refers to life necessities that could become uncontrollable due to digital addiction such as sleep, discord with parents, money management, and academic achievement. Deception refers to how children lie to their parents about the amount of time and what they do on their DDs. Displacement refers to parents feeling disconnected from their children, which results in the compromising of the family unit.
Discriminant validity was established because none of the nine criteria were highly correlated with each other (Table 1). All the correlations were <0.7 (0.389–0.696). Therefore, discriminant validity existed among all criteria because they assessed different aspects of the overall construct (addiction).
Items were created based on the theoretical definition of each addiction criterion, assuming addiction is a human trait not confined to a specific age group. During piloting, an initial set of items for each addiction criterion were created guided by the criterion’s theoretical definitions and a thorough review of relevant literature. The initial pool included a substantial number of items with the aim of sampling all possible and known alternatives, language simplicity, items’ straightforwardness, and appropriateness to children’s reading level. This initial item pool was reviewed by knowledgeable field experts with special focus on content validity. Experts provided suggestions for adding, removing, and amending items. Additionally, items were subjected to content analysis by educators associated with the target population. Experts provided suggestions for adding, removing, and amending items.
Two major factors led to deciding on the number of items per addiction criterion. Some criteria encompassed a smaller set of items to start with because by definition, they represented narrower content and consequently resulted in a fewer number of items. The second major factor was given that there are no known specific rules about the number of items to retain for each criterion, the authors developed their own strategy to guide the decision on retention of a necessary and sufficient number of items per addiction criterion. To start with, many more items were generated than would be needed for the final scale. Following this, the authors
(a) retained an adequate parsimonious number of items that had satisfactory conceptual consistency with each other on the addiction criterion being assessed;
(b) deleted items that had the lowest criterion loadings and highest cross-loadings; and
(c) retained items that used simple language appropriate for children’s reading level.
Once developed, the scale was pretested for content adequacy of the items. The item selection followed an iterative process involving several periods of item writing and rewriting, implementing the conceptual and psychometric analysis at each iteration to ensure that each addiction criterion-specific set of items was relevant and representative to the criterion under focus. Two English language teachers, scale development professionals, and a child psychologist reviewed the questions and made sure they were appropriate for the target population. Sample questions include “When I am not at school, I spend a lot of time using my device” (preoccupation) and “I have spent more and more time on my device” (tolerance). All items are rated on a five-point Likert scale: 1 (never), 2 (rarely), 3 (sometimes), 4 (often), and 5 (always). Scores range from 25 to 125, with higher scores indicating higher digital dependency.
Selected schools were contacted for securing prior permission from their administrations for data collection. The survey was emailed along with an explanation of the nature and purpose of the study and how the survey would be administered. All schools opted to obtain parental consent for their child’s participation. Upon parent/guardian consent form approval, dates and times were scheduled. Before its distribution, a trained researcher explained the survey procedure to all participants and assured them of confidentiality and anonymity.
Given the DASC’s 25 items, the study required at least 500 participants for robust analytical purposes, and the present study’s sample size exceeded this (N = 822). The Statistical Package for the Social Sciences, version 24.0 for Windows (SPSS, Inc., Chicago, IL), and AMOS were used to conduct statistical analysis. The analysis provided (a) descriptive data, (b) internal consistency (Cronbach’s α), (c) dimensionality and factorial validity, (d) criterion-related validity, and (e) confirmatory factor analysis (CFA). The CFA utilized structural equation modeling with IBM SPSS Amos Graphics 24.0 to test the structure underlying the 25 items of the DASC. The latent construct was digital addiction, which was the endogenous variable. The 25 items were the exogenous variables used to assess participants’ digital dependency level.
Table 2 shows the means, SDs, corrected item–total correlation, and loadings for each DASC item. Most items were skewed toward the less frequent tail of the distribution and demonstrated adequate variability. More than 25 percent of the items had averages ranging between never (value = 1) and rarely (value = 2). The item with the highest mean was “When I am not at school, I spend a lot of time using my device” (M = 3.36). The item with the highest rating for always was “Using my device helps me to forget my problems” (22.3 percent).
|Preoccupation (α = 0.61) 0.65|
|1. When I am not at school, I spend a lot of time using my device||3.36||1.19||0.556||0.595|
|11. When I do not have my device, I think about what I do on it (video games, social media, and texting, etc.)||2.57||1.35||0.601||0.639|
|14. Using my device is the most important thing in my life||1.71||1.14||0.616||0.661|
|Tolerance (α = 0.67) 0.71|
|2. I feel the need to spend more time using my device||2.46||1.24||0.596||0.642|
|7. I have spent more and more time on my device||2.59||1.25||0.706||0.747|
|Withdrawal (α = 0.87)|
|3. I feel upset when I am not able to use my device||2.74||1.40||0.658||0.705|
|8. I feel upset when I am asked to stop using my device||2.66||1.36||0.694||0.740|
|12. I feel frustrated when I cannot use my device||2.37||1.32||0.731||0.770|
|21. I feel frustrated when I am asked to stop using my device||2.30||1.33||0.731||0.777|
|Problems (α = 0.65) 0.69|
|10. I am sleeping less because I am using my device||2.20||1.37||0.609||0.652|
|13. I have problems with my parents about the amount of time I spend using my device||2.04||1.20||0.629||0.676|
|23. I spend too much money on things for my device||1.86||1.22||0.469||0.505|
|25. I continue using my device despite the fact that my grades at school are getting lower and lower||1.79||1.10||0.435||0.477|
|Conflict (α = 0.59) 0.63|
|9. My parents try to stop or limit me using my device, but they fail||2.22||1.36||0.581||0.626|
|22. I argue with my parents when they ask me to stop using my device||2.05||1.24||0.611||0.659|
|Deception (α = 0.56) 0.60|
|4. I lie to my parents about the amount of time I spend using my device||1.49||0.95||0.500||0.542|
|16. I lie to my parents about what I do on my device||1.47||0.94||0.415||0.457|
|Displacement (α = 0.62) 0.66|
|6. I do not spend time with my family members because I prefer using my device||1.60||0.96||0.592||0.636|
|18. I have lost interest in hobbies or other activities because I prefer using my device||1.59||1.03||0.544||0.594|
|20. I check my device when I am doing homework or other important things||2.26||1.37||0.544||0.592|
|Relapse (α = 0.59) 0.63|
|17. I am not able to control using my device||2.05||1.23||0.602||0.647|
|19. When I stop using my device, it is not long before I start using it again||2.71||1.29||0.658||0.702|
|Mood modification (α = 0.68) 0.72|
|5. Using my device helps me to forget my problems||2.99||1.43||0.460||0.497|
|15. Using my device is more enjoyable than doing other things||2.30||1.23||0.638||0.679|
|24. Using my device makes me feel better when I feel bad||3.21||1.26||0.537||0.572|
Less than half of the participants (45.4 percent; 46.1 percent males) endorsed none of the DSM-5 nine criteria (Table 3). A total of 42.2 percent (59.9 percent males) endorsed one to four criteria. Participants who endorsed four or less criteria (87.6 percent) were classified as nonaddicts. The remaining 12.4 percent participants (61.8 percent males) endorsed five or more criteria and were classified as addicts. The prevalence of criterion fulfillment among participants was as follows: mood modification (31.6 percent), preoccupation (28.7 percent), withdrawal (26.9 percent), problems (18.5 percent), relapse (16.2 percent), conflict (15.1 percent), displacement (14.1 percent), tolerance (13.7 percent), and deception (4.4 percent).
|Criteria fulfilled||N (%)|
|0||373 (45.4)||172 (38.8)||200 (53.5)|
|1||134 (16.3)||76 (17.2)||57 (15.2)|
|2||98 (11.9)||58 (13.1)||40 (10.7)|
|3||64 (7.8)||42 (9.5)||21 (5.6)|
|4||51 (6.2)||32 (7.2)||18 (4.8)|
|5||26 (3.2)||15 (3.4)||11 (2.9)|
|6||26 (3.2)||14 (3.2)||11 (2.9)|
|7||30 (3.6)||21 (4.7)||9 (2.4)|
|8||11 (1.3)||6 (1.4)||5 (1.3)|
|9||9 (1.1)||7 (1.6)||2 (0.5)|
Participants spent a mean daily average of 2.14 hours on DDs during weekdays (SD = 2.41 hours) and 5.87 hours at weekends (SD = 5.44 hours). Approximately 14.4 percent (65.2 percent males) reported not using DDs during weekdays, compared with 0.9 percent (71.4 percent males) at weekends. Two-thirds used DDs for 2 hours or less on weekdays (69.8 percent; 53.1 percent males), compared with 28.3 percent (42.8 percent males) at weekends. Table 4 indicates that more than two-thirds of 9-year-old children reported using mobile phones (67.7 percent) rising to 90 percent among 12-year-olds.
|Age (years)||I use a mobile phone|
|% within: Age||66.7||33.3||100.0|
|% within: I use a mobile phone||13.5||24.0||15.8|
|% within: Age||66.3||33.7||100.0|
|% within: I use a mobile phone||18.5||33.5||21.8|
|% within: Age||78.3||21.7||100.0|
|% within: I use a mobile phone||23.2||22.9||23.2|
|% within: Age||89.9||10.1||100.0|
|% within: I use a mobile phone||32.2||12.8||27.9|
|% within: Age||87.0||13.0||100.0|
|% within: I use a mobile phone||12.6||6.7||11.3|
|% within: Age||78.1||21.9||100.0|
|% within: I use a mobile phone||100.0||100.0||100.0|
An independent-samples t test showed a significant difference in DASC scores for males (M = 59.1, SD = 19.6) and females [M = 53.5, SD = 18.9; t(815) = 4.1, p < 0.005]. The magnitude of differences in mean scores (mean difference = 5.7, confidence interval [95% CI] 2.9–8.2) was small (η2 = 0.02). An independent-samples t test showed no significant difference in DASC scores for addicted males (M = 84.3, SD = 12.1) and addicted females [M = 85.6, SD = 11.8; t(189) = −0.7, p = 0.475]. The magnitude of differences in the means (mean difference = −1.3, 95% CI [−4.9 to 2.3]) was minimal (η2 = 0.003).
Gender-based correlations were calculated between DASC scores and other variables. Weekday and weekend gameplay time among males and females had significant positive medium correlations with DASC score (Table 6). Within the addict cohort, weekend gameplay time among males (10.4 hours) and females (8.1 hours) did not correlate with DASC (Table 5). Table 6 shows that female addicts had higher correlation on DASC scores (r = 0.454, p < 0.01) with device usage on weekdays compared with male addicts (r = 0.191, p < 0.05). The DASC score among addicted males and females did not correlate with age (Table 7).
|Males||Females||t test for equality of means||η2||Effect|
|Hours weekdays||2.3||2.8||2.0||1.8||1.7||729.7||0.3||−0.0 to 0.6||0.004||Very small|
|DASC||84.3||12.1||85.6||11.8||−0.7||189||−1.3||−4.9 to 2.3||0.003||Very small|
|Hours weekdays||3.5||4.0||3.0||1.9||1.2||182.7||0.5||−0.3 to 1.4||0.008||Very small|
|Hours weekends||10.4||7.2||8.1||5.5||2.3*||152.7||2.2||0.3–4.2||0.002||Very small|
|Hours weekdays||1.8||2.1||1.8||1.6||0.0*||608||0.0||−0.3 to 0.3||0.003||Very small|
|Hours weekends||5.3||4.7||4.2||4.2||3.0**||598.3||1.1||0.4–1.8||0.003||Very small|
The DASC produced excellent internal consistency reliability with excellent Cronbach’s alpha (0.936). The computed α showed that the index of measurement error in the DASC was very small (0.116).
Dimensionality and factorial validity analysis
Exploratory principal component analysis to examine the DASC’s dimensionality was performed using SPSS, version 24. The interitem correlation matrix contained no negative values, indicating that items assessed the same characteristic. The corrected item–total correlation values were all positively associated with the total score and ranged from 0.415, item 16, to 0.731, items 12 and 21 (Table 2), indicating that scale items were assessing the same construct. No loadings (Table 2) were considered poor. Additionally, no values in the “Alpha if Item Deleted” column were higher than 0.936, indicating DASC’s homogeneity and suggesting that no DASC item should be removed.
Next, DASC items were subjected to principal component analysis (PCA) extraction. The rotation method was Oblimin with Kaiser Normalization. Before performing PCA, data suitability factor analysis was assessed. Inspection of the correlation matrix revealed the presence of many coefficients of 0.3 and above (77.05 percent). The Kaiser–Meyer–Olkin measure of sampling adequacy was 0.960, exceeding the recommended value (0.600), and Bartlett’s test of sphericity reached statistical significance (chi-squared = 8732.443, df = 300, p < 0.0005), supporting the factorability of the correlation matrix. In fact, PCA demonstrated the presence of two components with eigenvalues exceeding 1, explaining 40.63 percent and 5.69 percent of the variance, respectively. An inspection of the scree plot showed a clear break after the first component. Hence, based on Cattell’s scree test, one component was retained. However, the results of the parallel analysis showed two components with an eigenvalue (10.159 and 1.421, respectively) exceeding the corresponding criterion value (1.331 ± 0.029 and 1.279 ± 0.024, respectively) for a randomly generated data matrix of the same size (25 variables × 828 participants). This was further supported by Velicer’s minimum average partial (MAP) test, implemented on the correlation matrix using PCA extraction. The MAP test suggested a two-factor solution. There were slight variations in the individual factor loadings between PCA and MAP test.
Furthermore, the component correlation matrix indicated that the relationship between the two factors was strong (−0.559). Consequently, it was appropriate to use the Oblimin rotation solution, which is an oblique rotation used when factors are assumed to be correlated.32 The Oblimin rotation provided the pattern matrix (Table 8). Thirteen items loaded on component 1, including the complete item sets of conflict, displacement, and problems criteria. Items pertaining to component 1, especially those with the highest loadings and respective criteria, represented interpersonal factors. Most items pertaining to this factor expressed consequences relating to relationships/communication between individuals. Twelve items loaded on component 2, including the complete item sets of mood modification, withdrawal, and tolerance criteria. Items of component 2, especially the ones with the highest loadings and the respective criteria, represented intrapersonal factors. Most items pertaining to this factor expressed consequences taking place or existing in the mind. The preoccupation and relapse criteria had their items split between the two components. The pattern matrix was supported by the structure matrix (Table 9) showing the correlations between the 25 scale items and the interpersonal and intrapersonal factors.
|Mood modification||Item 5||−0.785|
Confirmatory factor analysis
The χ2-to-df ratio was 2.434 (p < 0.00005) (Table 10), indicating that the model was an adequate fit.33 The root-mean-squared error of approximation (RMSEA) was 0.0418, indicating a good model fit (i.e., <0.05). Since the computed PCLOSE (0.994)—testing the null hypothesis that RMSEA is no greater than 0.05—was significantly greater than 0.05, there was no evidence to reject the null hypothesis. Additionally, the normed fit index, comparative fit index, and Tucker-Lewis coefficient were 0.933, 0.959, and 0.951, respectively, suggesting that the model fitted very well. The SRMR (the standardized RMR, root mean square residual) was 0.0337, indicating a very good fit (i.e., <0.05).
|Goodness-of-fit measure||Perfect fit||Range||Good fit||Model|
|CMIN/DF||1||<3 to 1||2.434|
|NFI (Delta1)||1||0 to 1||>0.95||0.933|
|CFI||1||0 to 1||>0.95||0.959|
|TLI||1||0 to 1||>0.90||0.951|
Concurrent validity analysis
Concurrent validity was investigated by examining the bootstrapped Pearson’s correlation coefficient with 10,000 bootstrap samples and 95% BCa CI between the total scores and the item “Choose the option that best describes your addiction to using your devices,” which yielded adequate results (r = 0.61, R2 = 0.37, p < 0.0001, 95% BCa CI [0.56–0.66]), supporting concurrent validity.
Criterion-related validity analysis
Criterion-related validity was confirmed by the statistically significant and positively medium correlation between the DASC and both the weekday gameplay time (r = 0.385, p < 0.0005, 95% CI [0.318–0.446]) and the weekend gameplay time (r = 0.454, p < 0.0005, 95% CI [0.399–0.513]). Several studies adopted the positive correlations between gameplay time as a criterion-related validity test, including the IGD-20 test,16 Video Game Dependency Scale,34 and Game Addiction Scale (GAS).35 Moreover, there was a statistically significant and positively strong correlation between the DASC and self-reported digital addiction (r = 0.612, p < 0.0005, 95% CI [0.564–0.653]) and participants’ description of parents’ digital addiction (r = 0.173, p < 0.0005, 95% CI [0.096–0.248]). Results also showed that the DASC and RCADS-30 were significantly correlated with each other (r = 0.414, p < 0.01, 95% CI [0.340–0.482]), further suggesting the validity of the DASC.
This study’s aim was to develop and validate a psychometric scale to assess digital addiction among children aged 9–12 years old. The DASC was developed using the theoretical framework built upon IGD DSM-5 criteria10 and the components model of addiction.31 Consequently, the DASC underwent rigorous psychometric examination. The results supported the internal consistency of DASC, as assessed by Cronbach’s alpha and corrected item–total correlation.
Based on a recommendation of meeting five or more DSM-5 IGD criteria, 12.4 percent were identified as at risk of addiction to DDs (14.2 percent males and 10.2 percent females), and within the addict cohort, 62.4 percent were males. Mood modification (31.6 percent) and preoccupation (28.7 percent) were the most endorsed criteria, similar to the findings of a German study.34 Results also demonstrated that female addicts had higher correlated DASC scores with weekday device usage compared with male addicts despite there being no difference between times spent on devices. This suggests that weekday device usage among females causes more conflicts compared with males perhaps due to higher social media use. The exploratory principal component analysis showed that all items assessed the same construct and the scale was homogeneous. Analysis of both concurrent and criterion validity also yielded good results, further highlighting the concurrent and criterion validity of the DASC. Additionally, the CFA provided adequate results regarding DASC construct validity.
Future studies can further expand on the findings here by using the DASC in different samples and cultural contexts. The DASC appears promising, but requires further testing using clinical samples. Although the study provided robust findings in relation to rigorous psychometric testing, it is not without limitations. Although most scales in the field use self-report questionnaires, there are associated possible biases (short-term memory recall and social desirability). In addition, criterion-related validation was carried out by correlating the DASC score with time spent using DDs. It should also be noted that comparing correlations between addicts and nonaddicts may be problematic given potential differences in variances. Finally, other criteria should be considered in future studies of the DASC, such as correlation with other digital addictions and different psychiatric disorders.
The present study investigated the psychometric properties of the DASC. Reliability analysis showed that internal consistency was excellent and the DASC is a valid, reliable, and psychometrically robust instrument for use with 9–12-year-old children. The main value of this study is that development of the DASC will (a) help clinically identify children at risk of digital addiction and (b) stimulate further research concerning children from different cultural and contextual settings.
Author Disclosure Statement
No competing financial interests exist.
This research was supported by a grant from the National Council for Scientific Research (CNRS)—Lebanon and Notre Dame University–Louaize.