Screen time patterns and cognitive screening outcomes (MoCA-Ina) in adolescents: a cross-sectional study

Abstract

Background

Adolescent screen exposure is increasing, yet clinically interpretable thresholds for cognitive risk are unclear. This study examined associations between daily screen time and cognitive screening performance and derived a screen-time cutoff associated with cognitive impairment.

Methods

We conducted an observational cross-sectional study (March–April 2022) at a private junior high school in Indonesia during online learning. Students completed digital questionnaires reporting educational and recreational screen time and a directly reported overall estimate; a computed overall (educational + recreational) was generated to assess reporting consistency. Cognitive function was assessed using the MoCA-Ina, with < 24 as the primary impairment threshold based on recent psychometric evidence favoring lower cutoffs for improved classification accuracy.

Results

Sixty-seven adolescents were included (34 girls, 50.7%; 33 boys, 49.3%), with median age 13.0 years (12.0–16.0) and median MoCA-Ina 25.0 (19.0–31.0). MoCA-Ina did not differ by sex (girls 25.0 [19.0–31.0] vs. boys 26.0 [20.0–30.0]; p = 0.244). Recreational screen time correlated inversely with MoCA-Ina (ρ = −0.446, p < 0.001), as did overall screen time (ρ = −0.360, p = 0.003), whereas educational screen time was not associated (ρ = −0.061, p = 0.624). In adjusted regression, overall screen time remained negatively associated with MoCA-Ina (β = −0.24 per hour/day; 95% CI − 0.41 to − 0.07; p = 0.007), while age was positively associated (β = 0.96; 95% CI 0.07 to 1.85; p = 0.034). All variance inflation factors were below 2.5, indicating no substantial multicollinearity. ROC analysis showed fair discrimination (AUC 0.66; optimism-corrected AUC after bootstrap internal validation [1,000 resamples]: 0.63) with an optimal cutoff > 8.97 h/day (sensitivity 83.3%, specificity 48.8%, PPV 47.6%, NPV 84.0%); risk of impairment was higher above the cutoff (RR 2.98; 95% CI 1.15–7.72; p = 0.010; OR 4.77; 95% CI 1.40–16.31).

Conclusions

High daily screen exposure was associated with poorer cognitive screening performance. The > 8.97-hour/day threshold represents a preliminary, hypothesis-generating cutoff that may help identify adolescents at elevated likelihood of cognitive impairment, pending external validation in larger, more diverse samples.

Trial registration

071/K-LKJ/ETIK/II/2022.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12883-026-05128-5.

Background

Digital media has become a near-continuous feature of adolescent life, with population surveillance and nationally representative surveys showing that many youths spend substantial portions of the day on screens outside of schoolwork [1]. In Indonesia, internet access is now widespread; the Indonesian Internet Service Providers Association (APJII) reported national internet penetration of 79.5% in 2024 (221.6 million users) [2]. Among children and adolescents, UNICEF Online Knowledge and Practice baseline study (2023) found that the vast majority of children use the internet daily for an average of 5.4 h per day [3].

From a neurodevelopmental standpoint, adolescence is a period of heightened sensitivity because large-scale remodeling of cortical and subcortical circuitry is still underway, including synaptic pruning and progressive myelination that refine efficiency in prefrontal and association networks supporting cognitive control [4]. Several mechanistic pathways plausibly link heavier screen exposure to cognitive outcomes: (i) sleep and circadian disruption, which is consistently associated with screen use in the pediatric sleep literature; (ii) attentional fragmentation and reduced practice of sustained, effortful cognition; and (iii) displacement of developmentally protective behaviors such as physical activity, face-to-face social interaction, and cognitively enriching leisure [5, 6]. Accordingly, this study aimed to quantify the association between adolescent screen time (recreational and educational) and cognitive function measured by MoCA-Ina, and to derive a preliminary screen-time threshold associated with increased likelihood of cognitive impairment.

Methods

Study design and participants

This observational cross-sectional study was carried out from March 2022 to April 2022 at a private junior high school while online learning was being implemented. Ethical clearance was obtained from the Medical Research Ethical Committee (071/K-LKJ/ETIK/II/2022). This is in accordance with the Declaration of Helsinki. The population comprised junior high school students, with the exposure defined as differing amounts of daily screen use, contrasted with lower screen exposure, and the outcome defined as cognitive function. Students were eligible if they were actively enrolled in the online learning program and provided informed consent. Participants were selected through purposive sampling. After consent, students completed digital questionnaires that gathered demographic information and detailed patterns of daily screen use.

Research variables and data collection

The primary dependent variable was cognitive function, measured using the Indonesian version of the Montreal Cognitive Assessment (MoCA-Ina) [7]. To standardize administration and maintain inter-rater consistency, the assessment was conducted face-to-face by a single trained assessor. The MoCA is a commonly used screening tool for mild cognitive dysfunction that evaluates multiple domains, including attention, concentration, executive function, memory, language, visuospatial skills, abstraction, calculation, and orientation [8].

The MoCA-Ina was selected for this study because, at the time of data collection, no validated Indonesian-language cognitive screening instrument specific to the adolescent age group was available. The MoCA-Ina had undergone cross-cultural adaptation and psychometric evaluation in Indonesian populations (Fitri et al., 2022), making it the most accessible standardized tool. Although the MoCA was originally developed for adults aged ≥ 55 years, it has been applied in younger populations; the most direct evidence comes from a validation study in adolescents and young adults aged 14–21 years with congenital heart disease (Pike et al., 2017), in which MoCA scores correlated strongly with an established memory measure [9]. However, that evidence derives from a clinical sample rather than typically developing adolescents, and the instrument has not been formally validated in healthy adolescents or against age-specific normative data [9, 10]. The absence of age-appropriate normative data for adolescents is therefore acknowledged as a key limitation, and the methodological implications of applying an adult-normed instrument in this age group—including possible ceiling effects, uncertain diagnostic sensitivity, and untested construct validity—are examined in detail in the Limitations [7, 8].

For classification in diagnostic analyses, two MoCA-Ina thresholds were considered: <24 to indicate possible cognitive impairment and < 26 as the conventional “normal” cutoff. The < 24 threshold was selected as the primary definition of impairment because recent psychometric evidence has demonstrated that the conventional < 26 cutoff produces high false-positive rates, and that lower thresholds improve classification accuracy in screening contexts (Ratcliffe et al., 2023) [11]. The < 26 cutoff was treated as a normative comparator to contextualize classification stringency across thresholds. To evaluate the consistency of using the impairment definition (< 24) relative to the normal reference cutoff (26), agreement between these cutoff-based classifications was assessed using Bland–Altman methodology.

The main independent variable was duration of screen time. Students self-reported daily screen exposure and separated it into educational screen time (school-related activities and homework) and recreational screen time (gaming, social media, and entertainment). To assess internal consistency of reporting, an overall screen-time estimate was additionally computed by summing educational and recreational durations and was compared with the “overall” screen time value reported directly by the students.

Potential confounders were prespecified and accounted for in multivariable analyses, including child age, sex, breakfast habits, maternal age, and maternal education. Age and sex were included because adolescent neurodevelopment, such as trajectories of gray matter change and executive-function maturation, varies across age and differs by biological sex [12]. Maternal education was recorded as a proxy indicator of socioeconomic context, which is known to influence neurocognitive development through environmental pathways [13]. Maternal education categories were defined according to the Indonesian education system: “Primary” encompasses completion of elementary school (SD) and junior high school (SMP), representing 9 years of basic compulsory education; “Secondary lower” refers to senior high school (SMA/SMK); “Secondary higher” refers to diploma-level education; and “Tertiary” refers to university-level education. Breakfast habits were assessed given evidence linking morning food intake with short-term cognitive performance in school-aged children, particularly attention and memory outcomes [14].

Statistical analysis

All analyses were conducted in RStudio (Version 2026.04.1 + 583, Posit Software, PBC) using the tidyverse, pROC, and BlandAltmanLeh packages. Distributional assumptions were evaluated using the Shapiro–Wilk test together with histogram inspection. Demographic and study variables were summarized using descriptive statistics. Associations between continuous predictors and MoCA-Ina scores were explored with correlation analysis, applying Pearson correlation for normally distributed variables (mother’s age) and Spearman rank correlation for non-normally distributed variables (child age, screen-time measures, and MoCA-Ina scores).

Agreement between directly reported overall screen time and the computed total (educational plus recreational screen time) was examined using Bland–Altman methods. A difference plot was produced and limits of agreement were calculated as the mean difference ± 1.96 standard deviations. Multivariable linear regression was then performed to identify independent predictors of MoCA-Ina scores while adjusting for the prespecified confounders. Variance inflation factors (VIFs) were computed for all predictors in the multivariable model to assess multicollinearity. As a sensitivity analysis, a reduced model retaining only variables significant at p < 0.20 in univariable analysis was also fitted to evaluate the stability of coefficient estimates given the limited sample size (67 observations with eight predictors in the full model, yielding fewer than 8.4 subjects per predictor).

For diagnostic evaluation, the performance of overall screen time in identifying possible cognitive impairment was assessed using receiver operating characteristic (ROC) curve analysis, with impairment defined primarily as MoCA-Ina < 24. The area under the curve (AUC) was reported with 95% confidence intervals. Bootstrap internal validation (1,000 resamples with replacement) was performed to estimate the optimism-corrected AUC. In addition, Bland–Altman analysis was used to assess the consistency of classification when applying the < 24 impairment cutoff compared with the conventional < 26 “normal” cutoff, thereby quantifying agreement and potential systematic differences between these two threshold-based definitions. All tests were two-sided, and statistical significance was defined as p < 0.05.

Results

Participant characteristics and descriptive statistics

A total of 67 children were included in the study, consisting of 34 girls (50.7%) and 33 boys (49.3%). The baseline characteristics of the study population are summarized in (Table 1). The median age of the children was 13.00 years (IQR 12.00–16.00), and the mean mother’s age was 41.93 ± 5.20 years.

Table 1.

Participant characteristics by sex

Variable	Overall (n = 67)	Girls (n = 34)	Boys (n = 33)	P value
Mother age (years)	41.93 ± 5.20	43.29 ± 4.90	40.52 ± 5.18	0.028#
Child age (years)	13.00 (12.00–16.00)	14.00 (12.00–15.00)	13.00 (12.00–16.00)	0.171$
Recreational screen time	5.80 (0.61–16.68)	6.13 (0.61–16.68)	5.48 (1.72–15.86)	0.526$
Educational screen time	3.86 (0.71–12.00)	4.00 (0.71–6.29)	3.46 (0.75–12.00)	0.390$
Overall screen time	9.99 (2.48–21.15)	10.20 (3.77–21.15)	9.57 (2.48–21.00)	0.510$
MoCA-Ina score	25.00 (19.00–31.00)	25.00 (19.00–31.00)	26.00 (20.00–30.00)	0.244$
Breakfast habits, n (%)				1.00*
No	20 (29.9%)	10 (29.4%)	10 (30.3%)
Yes	47 (70.1%)	24 (70.6%)	23 (69.7%)
Mother’s education, n (%)				0.191*
Primary	44 (65.7%)	24 (70.6%)	20 (60.6%)
Secondary higher	2 (3.0%)	0 (0.0%)	2 (6.1%)
Secondary lower	19 (28.4%)	8 (23.5%)	11 (33.3%)
Tertiary	2 (3.0%)	2 (5.9%)	0 (0.0%)

Values are mean ± SD (approximately normal) or median (min–max) (non-normal). Categorical variables are n (%). P values compare Girls vs. Boys

^#Independent samples t-test

^$Mann–Whitney U test

*Chi-square test or Fisher’s exact test (used when expected cell counts were small)

†Maternal education categories follow the Indonesian education system: “Primary” = completion of elementary school (SD) and/or junior high school (SMP), representing up to 9 years of basic compulsory education; “Secondary lower” = senior high school (SMA/SMK); “Secondary higher” = diploma-level education; “Tertiary” = university-level education

There were no statistically significant differences between girls and boys regarding child age (p = 0.171), recreational screen time (p = 0.526), educational screen time (p = 0.390), or overall screen time (p = 0.510). Similarly, cognitive performance, as measured by the MoCA-Ina score, did not differ significantly by sex (Median: 25.00 vs. 26.00, p = 0.244; Fig. 1). Most participants (70.1%) reported eating breakfast, and most mothers (65.7%) had a primary education level.

Fig. 1.

Bar chart showing the distribution of MoCA-Ina by sex and breakfast habits

Correlates and predictors of cognitive performance

In bivariate analyses, higher recreational screen time showed a moderate inverse correlation with MoCA-Ina scores (ρ = −0.446, p < 0.001), and overall screen time was also negatively correlated with MoCA-Ina (ρ = −0.360, p = 0.003) (Fig. 1; Table 2). In contrast, educational screen time (ρ = −0.061, p = 0.624), child age (ρ = 0.191, p = 0.122), and mother’s age (r = − 0.076, p = 0.543) were not significantly associated with MoCA-Ina in unadjusted analyses (Table 2). After adjustment in the multivariable linear regression model, overall screen time remained independently associated with lower cognitive scores (β = −0.24 per additional hour/day; 95% CI − 0.41 to − 0.07; p = 0.007), while child age was positively associated with MoCA-Ina (β = 0.96; 95% CI 0.07 to 1.85; p = 0.034) (Table 3). All variance inflation factors were below 2.5, indicating no substantial multicollinearity (Supplementary Table S1). A sensitivity analysis using a reduced model (retaining recreational screen time, child age, and overall screen time) yielded consistent results: overall screen time remained negatively associated with MoCA-Ina (β = −0.22; 95% CI − 0.38 to − 0.06; p = 0.009), confirming the robustness of the primary finding.

Table 2.

Correlation between predictors and outcome

Predictor	Correlation coefficient	P value
Mother age	r = − 0.076¹	0.543
Child age	ρ = 0.191²	0.122
Recreational screen time	ρ = −0.446²	< 0.001
Educational screen time	ρ = −0.061²	0.624
Overall screen time	ρ = −0.360²	0.003

¹Pearson correlation coefficient (r); ²Spearman rank correlation coefficient (ρ)

Table 3.

Multivariable linear regression model for outcome (adjusted associations)

Variable	β (95% CI)	P value
(Intercept)	17.34 (5.08 to 29.60)	0.006*
Sex: Boys (ref: Girls)	1.13 (− 0.58 to 2.84)	0.191
Child age	0.96 (0.07 to 1.85)	0.034*
Mother age	−0.06 (− 0.23 to 0.11)	0.479
Breakfast: Yes (ref: No)	−0.64 (− 2.53 to 1.24)	0.498
Overall screen time	−0.24 (− 0.41 to − 0.07)	0.007*
Mother’s education: Secondary higher (ref: Primary)	−1.01 (− 6.19 to 4.16)	0.696
Mother’s education: Secondary lower (ref: Primary)	−0.53 (− 2.47 to 1.40)	0.583
Mother’s education: Tertiary (ref: Primary)	0.90 (− 3.83 to 5.63)	0.704

*<0.05

Diagnostic value of screen time for cognitive impairment

ROC analysis indicated that overall screen time had fair discrimination for cognitive impairment (MoCA < 24), with an AUC of 0.66 (95% CI 0.52–0.80); bootstrap internal validation (1,000 resamples) yielded an optimism-corrected AUC of 0.63, confirming only fair discriminatory ability (Fig. 2). The optimal cutoff was > 8.97 h/day (Fig. 2), which classified 42 children as “high screen time” and 25 as “low screen time” (Table 4). Using this threshold, sensitivity was 83.3% (20/24 impaired correctly identified) and specificity was 48.8% (21/43 normals correctly identified), with a positive predictive value of 47.6% (20/42) and a negative predictive value of 84.0% (21/25) (Table 4). The relatively low specificity and modest PPV indicate that this cutoff should be interpreted as a preliminary screening signal rather than a definitive clinical diagnostic threshold. The risk of impairment was 47.6% in the high screen time group versus 16.0% in the low screen time group, corresponding to a relative risk of 2.98 (95% CI 1.15–7.72; p = 0.010) (Fig. 2; Table 4); the corresponding odds ratio from the 2 × 2 table was 4.77 (95% CI 1.40–16.31) (Table 4). The Bland–Altman plot shows no systematic bias between observed and predicted MoCA-Ina scores (mean bias = 0.00). However, the 95% limits of agreement are relatively wide (− 5.87 to + 5.87), suggesting meaningful individual-level prediction error that could affect classification for scores near a cutoff such as 26 (Fig. 3).

Table 4.

2x2 Contingency Table

	Impaired (MoCA < 24)	Normal (MoCA ≥ 24)	Total
High Screen Time (≥ 8.97 h)	20	22	42
Low Screen Time (< 8.97 h)	4	21	25
Total	24	43	67

Fig. 2.

ROC curve analysis

Fig. 3.

Bland-Altman Plot

Discussion

Principal findings and interpretations

Our findings are consistent with the broader literature suggesting that heavier recreational screen exposure is more consistently associated with weaker cognitive performance, likely because it clusters with behaviors that disrupt neurodevelopmental “inputs” such as sleep regularity, sustained attention, and cognitively enriching activities [15]. Large-scale evidence aligns with this pattern: in a prospective cohort study, higher screen time predicted later developmental outcomes with small but measurable cross-lagged effects (e.g., standardized β = −0.06 at 24 months and β = −0.08 at 36 months) [16]. Importantly, meta-analytic work emphasizes that context matters as much as duration—program viewing and background TV exposure show pooled negative correlations with cognitive outcomes (e.g., r = − 0.16 for program viewing; r = − 0.10 for background TV), while co-use with caregivers can be positively associated with cognition (e.g., r = 0.14), supporting the idea that social scaffolding and content quality modify neurocognitive risk [17]. These associations are plausible: frequent rapid-reward, high-salience media can bias dopaminergic reward learning toward immediate reinforcement, reduce tolerance for delayed reward, and fragment attentional control—striatal networks [18, 19].

The relationship between sustained sedentary behavior and cognitive outcomes is more nuanced than a simple dose–response model. A recent meta-analysis by Li et al. examined whether interrupting prolonged sitting with acute physical exercise breaks improves cognitive performance and found mixed results, suggesting that even active interruptions during sedentary episodes do not reliably benefit cognition [20]. This evidence adds important context to our findings: if physical activity breaks alone cannot consistently reverse the cognitive effects of prolonged sedentary behavior, then the case for passive, uninterrupted recreational screen exposure as a risk factor becomes more compelling and requires more careful theorizing. Specifically, the attentional and reward-processing demands inherent to recreational screen content—as distinct from the sedentary posture itself—may constitute an independent contributor to cognitive load and attentional depletion.

Diagnostic thresholds and clinical implications

While our study proposes a preliminary threshold for identifying higher-risk adolescents, the clinical message should be framed less as a single number and more as a signal of cumulative exposure plus behavioral patterning. Population data show that meeting recommended limits on recreational screen time is consistently associated with better cognition: in ABCD (n = 4,524; ages 9–10), meeting the screen-only recommendation was associated with notably higher global cognition (β = 4.25 points) compared with meeting none, and the combination of meeting sleep + screen recommendations showed similarly strong associations (β = 5.15) [21]. This “whole-day” framing is clinically useful: rather than only counseling reduction in total hours, clinicians can target the most neurobiologically sensitive windows (late evening), prioritize content and co-use, and protect sleep routines [21]. However, given the modest AUC (0.66; optimism-corrected 0.63) and low specificity (48.8%) of our proposed cutoff, the > 8.97 h/day threshold should not be used as a standalone diagnostic criterion but rather as a preliminary screening signal warranting further assessment. External validation in independent, larger samples is required before clinical application.

Sociodemographic factors and null findings

Null associations for sex, breakfast habits, and maternal education in our sample are not inconsistent with prior work showing that screen–cognition relationships are often small in magnitude and highly dependent on measurement, content type, and confounding by family routines. For example, a meta-analysis examining screen media use and academic performance found overall effects that were generally small (e.g., effect sizes around − 0.05 to − 0.07), implying that many studies will not detect subgroup differences unless samples are large and exposures are precisely characterized [22]. Additionally, “education” and “SES” effects may be partially diluted when screen exposure is pervasive across strata, and when the quality of the screen ecology (co-use, content appropriateness, bedtime access) drives outcomes more than raw duration [23].

Strengths and limitations

A key strength is modeling cognitive outcomes with adjustment for major covariates while distinguishing types of exposure, which aligns with the literature’s move away from a single-dimensional “screen time” metric [24].

However, several important limitations must be acknowledged. First, the cross-sectional design precludes causal inference; the observed associations could reflect reverse causality, whereby adolescents with lower baseline cognitive control are more susceptible to prolonged or problematic media use [25]. Second, all screen-time data were self-reported, which introduces recall bias and social desirability bias. Adolescents may underreport recreational screen time due to social desirability, which would bias associations toward the null, potentially making our estimates conservative. Differential misclassification by cognitive status cannot be excluded. No objective measures (e.g., device usage logs, passive monitoring applications) were used to verify self-report, and future studies should incorporate such measures to improve exposure accuracy.

Third, the MoCA-Ina was developed and validated as a screening tool for mild cognitive impairment in adults aged ≥ 55 years. Applying it to 12–16-year-old adolescents introduces unknown psychometric properties, and three methodological implications warrant explicit consideration. First, with respect to ceiling effects, a screening tool calibrated to detect age-related decline may compress score variance in cognitively intact younger respondents and limit discrimination across the upper range of ability; in our sample, however, the anticipated ceiling was not observed, as the median MoCA-Ina score (25.0) fell below the conventional adult-normal threshold of ≥ 26 and a substantial proportion of adolescents were consequently classified as ‘abnormal’ under adult criteria. This downward shift relative to adult norms is itself informative, suggesting that adult-derived cut-points may misclassify typically developing adolescents and that the score distribution partly reflects a mismatch between adult normative expectations and adolescent performance rather than true cognitive impairment. Second, with respect to diagnostic sensitivity, the established cut-points (< 24 and < 26) and the tool’s reported sensitivity and specificity were derived in older clinical populations; younger individuals may score above adult cut-offs even when genuine deficits are present, and age and education are known to exert a substantial influence on MoCA performance, so that fixed thresholds increase the risk of misclassification in this age group [10]. In the absence of an age-appropriate gold-standard reference, the ‘impairment’ classification applied here is operational rather than strictly diagnostic. Third, with respect to construct validity, the MoCA was designed to capture the cognitive profile of neurodegenerative and age-related decline, and whether its domain structure, item difficulty, and factor loadings behave equivalently in the developing adolescent brain is untested; the only direct adolescent validation to date was conducted in a clinical sample with congenital heart disease using a ≥ 26 cut-off, leaving validity in healthy, typically developing adolescents unestablished [9]. While no validated Indonesian-language adolescent cognitive screening instrument was available at the time of data collection, the use of an adult-normed tool in this population limits the precision and interpretability of the cognitive outcome measure.

Fourth, the ROC-derived cutoff of > 8.97 h/day was both derived and evaluated on the same dataset without external validation. Although bootstrap internal validation (1,000 resamples) yielded an optimism-corrected AUC of 0.63, this remains only fair discriminatory ability. The low specificity (48.8%) and modest PPV (47.6%) indicate that the cutoff cannot serve as a standalone diagnostic criterion and should be regarded as a preliminary, hypothesis-generating threshold requiring validation in independent samples.

Fifth, the multivariable regression model contained eight predictors fitted to only 67 observations (fewer than 8.4 subjects per predictor), which is below conventional recommendations (10–20 per variable) and raises concerns about overfitting and coefficient instability. Although the sensitivity analysis using a reduced model yielded consistent results and all VIFs were below 2.5, replication in larger samples is necessary to confirm the stability of these estimates.

Sixth, participants were selected through purposive sampling from a single private junior high school during the COVID-19 online learning period. This restricts representativeness to the broader Indonesian adolescent population and introduces the possibility that screen-time patterns during mandatory online learning were elevated and qualitatively different from habitual exposure under normal conditions. Generalizability to non-pandemic settings and to adolescents from different socioeconomic, geographic, and educational backgrounds is therefore limited.

Seventh, despite being referenced in the theoretical framework, key potential confounders and mediators—including sleep quality, physical activity, and mental health status—were not measured. The absence of these variables limits our ability to disentangle the pathways through which screen time may be associated with cognitive performance and represents a conceptual inconsistency that should be addressed in future prospective studies incorporating objective and multidimensional exposure and outcome measures.

Mechanistic interpretation should therefore lean on converging evidence: neuroimaging work has linked higher screen-based media use to differences in neurodevelopmentally relevant brain metrics [26]. Separately, screen time is associated with changes in attention-related outcomes and brain structure, including reduced cortical thickness in frontal regions that support executive control and self-regulation [27]. These convergent findings strengthen biological plausibility, even when any single observational study cannot establish causality.

Conclusions

In this cross-sectional study of 67 Indonesian adolescents during online learning, higher recreational and overall screen exposure were independently associated with lower cognitive screening scores on the MoCA-Ina after adjustment for demographic covariates. A preliminary screen-time threshold of > 8.97 h/day showed fair discriminatory ability (optimism-corrected AUC 0.63) for identifying adolescents with MoCA-Ina scores below 24, although the modest specificity and the absence of external validation preclude its use as a standalone clinical diagnostic criterion.

The practical implications of these findings suggest prioritizing reduction of passive/recreational and evening screen exposure, encouraging co-use and age-appropriate content when screens are used, and protecting sleep as a core neurocognitive safeguard. External evidence suggests that sleep disruption may mediate part of this association, although this pathway was not assessed in the current study. Future research should employ prospective designs with larger, probability-based samples, objective screen-time measurement, and concurrent assessment of sleep, physical activity, and mental health to clarify the direction, magnitude, and mechanistic pathways underlying the observed associations. .

Abbreviations

ABCD

Adolescent Brain Cognitive Development (study)

APJII

Indonesian Internet Service Providers Association (Asosiasi Penyelenggara Jasa Internet Indonesia)

AUC

Area under the curve

CI

Confidence interval

IQR

Interquartile range

MoCA

Montreal Cognitive Assessment

MoCA-Ina

Indonesian version of the Montreal Cognitive Assessment

PBC

Public Benefit Corporation

PICO

Population, Intervention/Exposure, Comparison, Outcome

pROC

An R package for ROC curve analysis

ROC

Receiver operating characteristic

SES

Socioeconomic status

TV

Television

UNICEF

United Nations Children’s Fund

VIF

Variance inflation factor

Authors' contributions

PYG: Concept., Meth., Superv., W–R&E; A: Data cur., Inv., W–OD; JHW: Formal anal., Viz., W–R&E; YEA: Concept., Meth., Inv., Superv., W–R&E; PLG: Meth., Res., W–R&E. All auth. read & appr. the final MS.

Funding

This research received no external funding.

Data availability

The datasets generated and/or analyzed during the current study are available from the corresponding author on reasonable request.

Code Availability

Not applicable. No custom code or software was developed for this study.

Declarations

Ethics approval and consent to participate

This observational cross-sectional study was reviewed and approved by the institutional ethics committee of the Faculty of Medicine, Pelita Harapan University (Approval No.: 071/K-LKJ/ETIK/II/2022). The study was conducted in accordance with the Declaration of Helsinki and applicable local regulations. Written informed consent to participate was obtained from all participants. For any participant under the age of 16 years, written informed consent was obtained from their parent or legal guardian prior to participation.

Informed consent was obtained from all individual participants included in the study.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.