Digital media exposure and cognitive functioning in European children and adolescents of the I.Family study

The digital environment can pose health risks through exposure to unhealthy content. Yet, little is known about its relation to children’s cognitive functioning. This study investigates the association between digital media (DM) exposure and children’s cognitive functioning. This cross-sectional study is based on examinations of children aged 8–18 years (N = 8673) of the I.Family cohort (2013–2014). Exposure to television, computer, smartphone and internet was self-reported (hours/day). Media multitasking (MMT) was defined as simultaneous use of computers with other digital or non-screen-based activities. Standard instruments were used to assess cognitive inflexibility (score: 0–39), decision-making ability (− 100 to + 100) and impulsivity (12–48). Adjusted regression coefficients and 99.9%CIs were calculated by generalized linear mixed-effects models. In total, 3261 participants provided data for impulsivity, 3441 for cognitive inflexibility and 4046 for decision-making. Exposure to smartphones and media multitasking were positively associated with impulsivity (βsmartphone = 0.74; 99.9%CI = 0.42–1.07; βMMT = 0.73; 99.9%CI = 0.35–1.12) and cognitive inflexibility (βsmartphone = 0.32; 99.9%CI = -0.02–0.66; βMMT = 0.39; 99.9%CI = 0.01–0.77) while being inversely associated with decision-making ability. Extensive smartphone/internet exposure combined with low computer/medium TV exposure was associated with higher impulsivity and cognitive inflexibility scores, especially in girls. DM exposure is adversely associated with cognitive functioning in children and adolescents. Children require protection against the likely adverse impact of digital environment.


Data collection
Core questionnaire and assessment of media use Age, sex, and country of residence were self-reported by adolescents (≥ 12 years) and proxy-reported by parents of younger children (< 12 years).Information on DM use duration, MMT, and confounding variables were measured using standardized questionnaires, previously tested for relative validity and reproducibility 30 .Participants reported the DM use duration separately for weekdays and weekend days, including TV/DVD/video, computer/ game console (PC), and internet use (Supplementary Methods).In order to prevent a potential overlap with internet use, for PC use participants were asked "How long do you usually sit at a computer/game console per day?(Please disregard the time spent on internet use.)", which enabled the assessment of off-line use of PC and game consoles.Total duration of TV, PC and internet use was respectively calculated as the sum of the weighted durations during weekdays and weekend days (hours/week), and quantified as hours/day.We measured smartphone use asking: "Thinking only about yesterday, about how much time did you spend watching TV shows, movies or music videos on a cell phone?".On a 5-point Likert-scale, answers ranged from 0 (not at all) to 5 (> 3 h/ day).An attributed time was assigned to each category to calculate the duration (hours/day) of smartphone use.Moreover, we measured MMT asking whether children engaged in other activities while using PCs, including TV, sending text messages, playing video-games, listening to music and reading.Based on dichotomized answers ("yes" or "no), a composite score of media multitasking behaviour ranging from 0 to 5 was calculated.

Emotion-driven impulsiveness
To assess emotion-driven impulsiveness (EDI), 3400 children aged ≥ 8 years self-completed the 12-item negative urgency subscale from the Urgency, Premeditation, Perseverance, Sensation seeking, and Positive urgency (UPPS-P) questionnaire 31 .Participants rated items on a 4-point Likert scale ranging from 1 (agree strongly) to 4 (disagree strongly).All items were recoded except for one item, to make sure that all items ran in the same direction.For participants who completed all items (those with incomplete items were excluded), a sum score for EDI was calculated ranging from 12 to 48 31 .A higher score indicated higher impulsivity.After all exclusion criteria (incomplete items of the subscale in addition to primary exclusion criteria, Supplementary Fig. S1) were applied, 3261 children aged 9.9-17.9years remained for the final analyses on impulsivity.

Cognitive inflexibility
To measure cognitive inflexibility (CIF), 4034 children and adolescents performed a computerised version of the Berg Card Sorting test 32,33 .Four cards of different colours and shapes, and a deck consisting of 64 stimulus cards were shown to the participant.Participants had to sort cards according to a particular rule (by symbol, number or colour) that was unknown to them, by choosing one of the key cards (e.g., if 'by colour' is the correct rule, the colour of symbols on the stimulus card should match the colour of symbols on the key card).A feedback message ('correct' or 'incorrect') was provided to the participants after sorting each card.The rule was changed without notice after 10 consecutive correct trials, and the participant had to find out the new rule.The number of perseverative errors after the rule had changed, i.e., the number of cards sorted according to the previous rule, was used as the measure of CIF.A higher number of errors indicates higher CIF.After all exclusion criteria were applied (Supplementary Fig. S1), 3441 children aged 8-17.9 years remained in the final analysis group.

Decision-making ability
Decision-making ability (DMA) was measured in 4169 children aged 8-18 years, using a computerised version of the Hungry Donkey Test 34 , the child-friendly version of the Iowa Gambling Task 35 , consisting of 100 trials.In each trial, participants should help a hungry donkey to collect apples by choosing one of the four doors presented on the screen.Each choice resulted in reward (apples) or in punishment (loss of apples).Doors 1&2 were disadvantageous doors because they yielded larger immediate reward but led to losing more apples in the long-term, resulting in net loss.Doors 3&4 were advantageous doors because they yielded smaller immediate rewards but led to winning more apples in the long-term, resulting in net gain 34 .DMA was calculated by subtracting the number of advantageous choices (doors 3&4) from the number of disadvantageous choices (doors 1&2), resulting in a score ranging from − 100 to + 100.A higher DMA is characterised by more advantageous choices than disadvantageous ones.After all exclusion criteria were applied (Supplementary Fig. S1), 4046 children aged 8-17.9 years remained for the final analyses on DMA.

Potential confounders
A vast array of confounders was self-or proxy-reported via questionnaires.Besides age, sex and country of residence, we also included information on parental highest education attainment 36 , the number of media rules at home 37 , pubertal status 38,39 , weight status (objectively-measured) 40 , total daily sleep duration and psychosocial well-being 41 .Moreover, via a kinship and household interview 42 , parents reported on family structure, including whether the participating child was an only child in the household and whether he/she lived in a one-parent or a two-parent family.Detailed explanation on measurements, operationalization and the rationale of including the confounding variables is provided in Supplementary Methods.

Statistical analyses
Response proportions differed for the cognitive tests and UPPS-P questionnaire resulting in three overlapping analysis groups (Supplementary Fig. S1), thus descriptive analyses were conducted separately for each cognitive outcome.Characteristics of age, sex, parental education attainment, weight status, pubertal status, family structure, country, MMT and media rules at home were tabulated for each analysis group.Children's performance on the cognitive tests was also calculated.To account for missing values, standard fully conditional specification multiple imputation (MI) was performed with 10 replications.This procedure has demonstrated unbiased handling of missing values, and enables the inclusion of both continuous and categorical variables in the imputation model 43 .All exposures and covariates used in the analyses were included in the MI, except the outcomes due to high percentage of missing values (> 50%).The relative efficiency of imputation (i.e., how well the true population parameters are estimated) for all variables was ≥ 98%, indicating good imputation quality.The proportion of missing information ranged from 0.5% for weight status to 26% for puberty based on Tanner stages.The characteristics of imputed and non-imputed analysis groups are shown in Supplementary Table S1.To examine the role of DM exposure on cognitive outcomes, a two-step analysis approach was conducted: Step 1: Association of single DM exposures with cognitive outcomes The associations between duration of using single DM and single outcomes were examined using generalized linear mixed regressions, adjusting for confounders.To assess potential multicollinearity of DM variables, we included TV, PC, internet, smartphone, MMT and media rules in the same regression model (step two) and calculated the tolerance and variance inflation factor (our limit: < 10) 44 , which indicated lack of multicollinearity (data not shown).
Step 2: Latent profiles of DM use in association with cognitive outcomes To identify underlying latent profiles of DM use, latent class analyses (LCA) were conducted 45 with three categories of duration for each DM variable (low duration: < 1 h/day; medium duration: 1-2 h/day; high duration: > 2 h/day).LCA was performed using two to six latent profiles of four variables (TV, PC, smartphone and internet), clustered by country, as we previously observed country-differences on DM use 46 .Models were then compared based on the Bayesian Information Criterium (BIC) and a clear distinction of latent profiles in terms of conditional probabilities (Supplementary Table S2).The chosen profiles (Supplementary Table S3) were then used in generalized linear mixed-effect models as predictors for each outcome, adjusting for covariates, including MMT (categorized as "no MMT", "1-2 MMT", " > 2 MMT").
To adjust for multiple testing (56 regressions in total: crude and adjusted models for three outcomes and five exposures; the latent class analyses for each outcome, and the stratified analyses), the statistical significance level was set at α = 0.001, using the Sidak method 47 .Non-standardized regression coefficients (β) and 99.9% confidence intervals (99.9%CI) were estimated and then combined for the multiple imputed datasets.All analyses were performed with the statistical software SAS version 9.4 (Statistical Analyses System, SAS Institute Inc., Cary, NC, USA).

Sensitivity analyses
In post-hoc analyses, the associations in step two were stratified by sex, family structure (one-vs.two-parent) and parental education attainment, adjusted for remaining covariates, in order to explore underlying differences.To account for self-and proxy-reported data by adolescents and parents, we stratified the impulsivity analysis group by age group (< 12 years vs. ≥ 12 years) and further adjusted for continuous age.

Informed consent
All procedures followed were in accordance with the ethical standards of the Helsinki Declaration of 1975, and its later amendments.Ethical approval was obtained from local institutional review boards at each study centre.Informed consent was obtained from all participants included in the study.

Latent profiles of DM use in association with cognitive functioning
The LCA of four latent DM profiles showed the lowest BIC and a clear interpretable distinction of conditional probabilities for the respective variables (Supplementary Table S3).Profile names were chosen based on the highest conditional probabilities.The majority of participants (57%) had low usage of all media types, i.e., < 1 h/ day for each media.This was considered the most favorable DM profile and was thus used as reference category in regression models of step two.Circa 13% of participants had high DM use except smartphone, while 10% had The adjusted regression models in step two showed that participants with "high DM use, except smartphone", had an almost 2-point higher impulsivity score (β, 1.81; 99.9%CI 0.67-2.96)compared to those with low use of all media, independent of MMT (Table 3).This association remained significant among girls (β, 2.32; 99.9%CI 0.66-3.99),adolescents (β, 1.80; 99.9%CI 0.65-2.95) and those living in two-parent family (β, 1.79; 99.9%CI 0.54-3.04).The stratified analyses by parental education level (Supplementary Table S6) showed positive associations across all strata.Statistically significant associations were observed in participants from families with a medium educational background (β, 2.21; 99.9%CI 0.41-4.01).Children and adolescents with "high smartphone/ internet, medium TV/low PC use" also showed higher impulsivity scores (β, 1.67, 99.9%CI 0.47-2.87),especially Table 1.Characteristics of European children and adolescents who took part in the cognitive tests for assessing impulsivity, cognitive inflexibility and decision-making ability.a Results are based on imputed samples (10 replications).Due to rounding of decimals, percentages may not add up to 100%.b Polish children and adolescents reported information only on impulsivity, as the computerized tests were not performed in this sample.c Abbreviations: IQR-interquartile range.d For DMA and CIF, the median and IQR are reported because the scores were skewed.However, the values of skewness (0.8 and -0.14 respectively) did not exceed the threshold (-1, + 1), hence transformation was not necessary 75 .3).Table 4 shows the association between latent profiles of DM use and cognitive inflexibility in children and adolescents.Although not statistically significant, the results indicate a negative association between the profile "high DM use, except smartphone" and CIF.In contrast, a positive association between "high smartphone/ internet, medium TV/low PC use" profile and CIF was observed in the overall sample and across strata, except for boys (Table 4) and for participants from families with low educational level (Supplementary Table S7).

Covariables
The adjusted associations between latent DM profiles and DMA, depicted in Table 5 and Supplementary Table S8, showed a positive, but not statistically significant association for the profile of "high DM use, except smartphone" and DMA.Children with high smartphone/internet, but medium TV/low PC use" showed lower DMA scores across all strata.Although not statistically significant, participants living in one-parent households (Table 5) showed more than 4-point lower DMA score when exposed to smartphones/internet for > 2 h/day (β, − 4.36; 99.9%CI, -16.1-7.43)compared to low use of all media (< 1 h/day).
Table 2.The association of digital media exposure with impulsivity, cognitive inflexibility and decisionmaking ability in European children and adolescents.a Models are based on regressing single DM exposures on each of the outcomes.b All models are adjusted for basic confounders including sex, continuous age, parental education level (low, medium, high), country of residence, total sleep duration (continuous), pubertal status (pre-or early pubertal vs. pubertal), well-being score (continuous), in addition to media rules at home (continuous), being an only child (yes vs. no) and family structure (one-parent vs. two-parent family).In all models, a random effect for family-id was included, to consider family influences, thus to partially account for genetic factors.c Due to missing value for the family id, one participant was not included in the analysis.d   Table 3. Association of latent profiles of digital media use with impulsivity in European children and adolescents.a Models are based on regressing the latent profiles of DM exposure on impulsivity, adjusting for basic confounders, including sex (not in the models stratified by sex), continuous age, parental education level (low, medium vs. high), country of residence, total sleep duration (continuous), pubertal status (pre-or early pubertal vs. pubertal), well-being score (continuous), in addition to media rules at home (continuous), being an only child (yes vs. no), family structure (one-vs.two-parent family; not in the models stratified by family structure) and media multitasking (no MMT, 1-2 MMT vs. > 2 MMT).In all models, a random effect for family id was included, to consider family influences and to partially account for genetic factors influencing the cognitive functioning.b Due to missing value for the family id, one participant was not included in the analyses.c Bold numbers indicate statistical significance based on 99.9% confidence intervals.

Discussion
This cross-sectional study using data from European children and adolescents shows that the duration of exposure to contemporary DM, including smartphones and internet as well as media multitasking, are positively associated with emotion-driven impulsiveness and cognitive inflexibility, independent of well-being, sleep duration and weight status.The strength of these associations differed by sex and family structure.Longer exposure to smartphones, internet and MMT was associated with a higher impulsivity score among children and adolescents in the present study, while no association was observed for non-internet-based media including TV and PC use.These findings suggest that the perpetual flow of information and input received simultaneously from the online environment such as emails, notifications and SM posts, may act as stressors.It is hypothesized that these exposures exceed the cognitive capacity of youth to handle and process that information, thus leading to "digital stress" 48 .Children and adolescents may be particularly vulnerable to digital stress because the neuronal myelination and synaptic pruning within the parietal and prefrontal cortex (areas related to attention control and delayed reinforcement) are not fully developed, leading to compromised emotionalregulation 49 and reduced control of impulses 50 .These findings are worrisome considering the obesogenic (digital) food environment.The impact of smartphones and internet on EDI might lie in the pathway of mindless eating in front of screens, especially in reward-seeking contexts.Moreover, the prolific content accessible via internetbased DM (smartphones, SM platforms) which provides short and continuous gratifications that may activate the reward system (caudate, insula) and subsequent emotional and behavioural responses, such as snacking 24 .
Table 4. Association of latent profiles of digital media use with cognitive inflexibility in European children and adolescents.a Models are based on regressing the latent profiles of DM exposure on cognitive inflexibility on the same model, adjusting for basic confounders, including sex (not in the models stratified by sex), continuous age, parental education level (low, medium vs. high), country of residence, total sleep duration (hours/day), pubertal status (pre-or early pubertal vs. pubertal status), well-being score (continuous), in addition to media rules at home (continuous), being an only child (yes vs. no), family structure (one-vs.two parent family; not in the models stratified by family structure) and media multitasking (no MMT, 1-2 MMT, > 2 MMT).In all models, a random effect for family id was included, to consider family influences and to partially account for genetic factors influencing the cognitive functioning.None of the associations were statistically significant after adjustment for multiple testing.Table 5. Association of latent profiles of digital media use with decision-making ability in European children and adolescents.a Models are based on regressing the latent profiles of DM exposure on the decision-making ability on the same model, adjusting for covariates including sex (not in the models stratified by sex), continuous age, parental education level (low, medium vs. high), country of residence, total sleep duration (hours/day), pubertal status (pre-or early pubertal vs. pubertal status), well-being score (continuous), in addition to media rules at home (continuous), being an only child (yes vs. no), family structure (one-vs.two parent family; not in the models stratified by family structure) and media multitasking (no MMT, 1-2 MMT, > 2 MMT).In all models, a random effect for family id was included, to consider family influences and to partially account for genetic factors influencing the cognitive function.None of the associations were statistically significant after adjustment for multiple testing.Another potential explanation may lie in the fact that DM displaces (real-life) social interactions such as parent-child, sibling-or peer relationships, often known as technoference.Social interactions are crucial for a healthy development because they built the foundation of processes related with personality and cognitive development, such as emotion regulation 51 .The interference of DM with parent-child interactions may compete with children's ability to concentrate and regulate their emotions, leading to internalizing and externalizing problems like reduced ability to control impulses 52 .
Children and adolescents who used all DM except smartphone for > 2 h/day, had almost 2-point higher impulsivity score compared to children with low use of all media.Although the sole exposure to TV and PC was not associated with impulsivity, using them for prolonged duration in addition to constantly checking internet content seems to have a higher negative impact on children's capabilities to regulate emotions.The association between "high smartphone/internet use, medium TV/low PC" and impulsivity was stronger among girls compared to boys, potentially because girls use smartphones and internet mainly for socializing and navigating SM, while boys mainly use them to play games 2 .Previous evidence shows that SM exposure impacts girls' and adolescents' psycho-emotional well-being 53 and body-image 54 via social comparisons over images posted on these platforms.This may lead to emotional overeating or restrained eating, as maladaptive coping strategies for relieving negative emotions 55 .Remarkably, neuro-developmental differences between children and adolescents might explain the stronger association between DM exposure and impulsivity in adolescents.The limbic subcortical system (affective/hot system) matures early on and the control system (cold) matures later in adolescence 56 , hence adolescents are more prone to engage in risky habits, also under digital stress.
Our results show that smartphone exposure and MMT were associated with higher cognitive inflexibility, suggesting that the digital environment may adversely impact youth's ability to smoothly shift between tasks.Smartphones and MMT encourage high levels of flicking between information sources at the expense of brain circuitry needed to sustain concentration 57 .This may explain why smartphone use and MMT lead to poor academic performance in youth 27,28 .Additionally, repeated exposure to fast-paced content, like short-edited video segments in SM (e.g., Instagram reels) or online game applications might trigger individuals into seeking higher arousal levels, which in turn hamper engagement in activities that require sustained attention (e.g., homework) 58 .The frequency of checking smartphones and internet might also lie in the pathway of the aforementioned associations.One longitudinal study conducted among Japanese children of a similar age range observed that increasing internet use frequency was associated with reduced increases of the grey and white brain matter volume, which are responsible for attention control and executive functioning 16 .Although not significant, the negative association between prolonged exposure to all DM, except smartphone, and CIF suggests that children may be using those media for educational purposes and this could positively influence their mental multitasking abilities.Prolonged exposure to smartphones/internet but medium TV/low PC use was associated with higher CIF, indicating that smartphones particularly may disrupt children's cognitive multitasking compared to other DM, as they are mostly used for entertainment purposes 59,60 , rather than for education 12 .
The adjusted associations between DM exposure and decision-making ability in a reward-related context were not statistically significant, but suggested a negative association of smartphone exposure and MMT with DMA.This aligns with a previous study where excessive smartphone use was related to reduced connectivity in the orbitofrontal cortex, a brain region related to DMA in reward-seeking behaviours 61 .One potential explanation could be that DM may interfere with children's capacities to weigh short-term rewards against long-term negative outcomes, especially in a highly rewarding (digital) food environment (e.g., by consuming energy-dense foods).This is supported by previous research, which showed that children exposed to multiple DM tend to make unhealthy food choices 62 .Nevertheless, to our best knowledge, no other study has investigated the role of DM exposure on decision-making ability using the Hungry Donkey Test.Given the limited research in this area, our non-significant results should be interpreted with caution.Studies conducted in children of a similar age range using both the Hungry Donkey Test 63 and Iowa Gambling Test 64 have shown that DMA varies with age in a U-shaped curve.Younger children perform better in the task compared to early-adolescents, with performance becoming again better in late adolescents.This indicates that although DM use increases with age 46,65 , other mechanisms impact the development of DMA.The ventromedial prefrontal cortex, for instance, which is specialized in decision-making, functionally matures during adolescence and continues until young adulthood 66 .In our study, we measured the self-reported DM use duration, while previous studies that found significant association between DM exposure and activation of decision-making related brain areas, used the Smartphone 61 and Internet Addiction Scale 67,68 .Thus, more longitudinal research is warranted to understand the extent and the underlying mechanisms (e.g., structural changes in the brain) via which DM exposure may impact DMA, using more detailed measures of DM, including SM exposure.

Strengths and limitations
This is the first observational study to investigate the role of DM exposure on several measures of cognition in European children and adolescents.In contrast to most other studies focusing only on TV and video-gaming, we examined various DM exposures including contemporary DM like smartphone and internet, as well as MMT.Although all the single media (TV, PC, internet and smartphones) are digital media, we aimed at examining the association of each media with measures of cognitive functioning, because the patterns of use and the content children and adolescents are exposed to differs from one media to another.Using LCA to identify underlying patterns of DM exposure represents an advantage.We accounted for various important confounders of the associations between DM use and cognition, including sleep, well-being, pubertal and weight status, and family structure.All analyses were corrected for multiple testing, strengthening the reliability of the reported associations.
The cross-sectional nature of our study limits our ability to draw causal conclusions.Hence, we cannot exclude that certain psychological characteristics and personality traits predispose certain forms of DM use.Applying www.nature.com/scientificreports/ the reverse causality hypothesis, it can be assumed that prior delays in cognitive functioning may have led to a prolonged DM use among children and adolescents.Recent evidence also suggests that genetic variants and neuro-biological mechanisms commonly observed in behavioural addictions (i.e., dopamine release) are related to the excessive use of smartphones, internet, and video-games [69][70][71] .On the other side, an increasing number of studies have suggested various potential mechanisms through which DM may lead to poor cognitive functioning in children (as already discussed in this paper: digital stress, technoference, overstimulation, reduced attention control or impact on brain structures).Given that research on this topic is still at its infancy, future longitudinal studies are warranted to investigate how DM exposure over time impacts cognitive functioning while accounting for genetic and psychological characteristics, including the cognitive abilities in early childhood.Of note, when we controlled for factors like psychosocial well-being and partially accounted for family influences, the observed associations between DM exposure and cognitive functioning remained robust.When comparing the characteristics of participants completing the cognitive tests vs. those who did not, we observed no differences (expect for age and subsequently for pubertal status).Hence, it is unlikely that results are affected by selection bias, although the external validity may be limited given the non-representative sampling frame for each included country.Recall bias may have led to an underreported EDI, as the impulsivity sub-scale was self-reported.DM exposure was measured based on self-reports, thus a recall and social-desirability bias may have resulted to overand under-estimation of DM use 72 .However, previous studies have shown that self-reported DM usage (e.g., smartphones) adequately distinguished between high and low use among adolescents 73 .The recency of the data is a limitation, as the digital environment and media skills of children have dramatically changed since 2013/2014.Our assessment of smartphone duration included the exposure to content such as TV shows, music videos or movies.Smartphones can be used for various purposes both offline and online, including playing games (offline) or social interactions (e.g., video-calling, texting etc.), which we could not account for with the available data.MMT was defined as the simultaneous use of a computer with other media, without considering smartphone and SM use, which are also significant contributors to MMT behaviour 74 .Therefore, the observed associations between MMT and cognitive functioning could be much more prominent in real life.We urge future studies to consider all sources of screen time and MMT to capture the complete picture of DM exposure during childhood.Moreover, we did not distinguish between smartphone and internet use for academic and entertainment purposes, which might lead to different results, and future studies should examine this hypothesis.We also lacked information on ease of access to DM at home, although we partly accounted for this by using information on media rules at home and MMT.Finally, we could not obtain information on SM exposure (Instagram, TikTok) 60 and we urge further research to investigate the role of SM on children's cognitive functioning, by considering the patterns of use (duration or problematic/addictive use of SM) and type of content provided.

Conclusions
Smartphone, internet and media multitasking were found to be positively associated with emotion-driven impulsiveness and cognitive inflexibility, independent of psychosocial well-being and family structure.Our study provides evidence on a potential underlying mechanism by which DM exposure affects cognitive development and related health behaviours.These findings ask for parents, paediatricians and policy makers to help youth implement sound media use habits, in order to build the foundation for developing healthy psycho-physiological resilience against the likely adverse impact of digital environment.