Research questions and hypotheses

Study design and setting

A cross-sectional analytical study was conducted at the outpatient department (OPD) at Abbasi Shaheed Hospital, Karachi, Pakistan. Ethical approval for the study was obtained prior to data collection from the Ethics Review Committee of Center for Integrated Health Research and Innovation (CIHRI) (Reference No: MS/ERC/029/2025; Date: 01 March 2025). The hospital provides multidisciplinary outpatient services, including neurology, general medicine, and psychiatry clinics, making it an appropriate setting for recruiting participants with varied digital exposure patterns and cognitive profiles.

The study was carried out over a five-month period from 14 March 2025 to 14 August 2025, and was designed in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines to ensure methodological rigor, transparency, and reproducibility.

Study population

The study population included adults aged 18–65 years who were able to communicate effectively, provide informed consent, and complete all study assessments, including questionnaires and cognitive testing. Participants were recruited from OPDs, accompanying attendants, and community visitors attending the hospital during the study period. Individuals with previously diagnosed major neurocognitive disorders, dementia, severe psychiatric illness requiring hospitalization, significant neurological deficits affecting cognitive assessment, acute medical illness at the time of enrollment, severe sensory impairment, or inability to complete study procedures were excluded.

Sample size and sampling technique

Sample size was determined using a standard single-proportion cross-sectional formula, assuming a 50% prevalence of altered cognitive performance associated with behavioral risk exposure, a 95% confidence interval, and a 5% margin of error. The minimum required sample size was calculated to be 384 participants. To enhance statistical robustness and compensate for potential incomplete responses, the final sample size was increased to 400 participants. A non-probability consecutive sampling technique was employed, whereby all eligible individuals presenting during the data collection period were approached sequentially until the target sample size was achieved after obtaining informed consent. A total of 427 individuals were initially approached. Twenty-seven individuals were excluded prior to final analysis. Of these, 11 did not meet eligibility criteria, while 16 had incomplete responses. An incomplete response was defined as failure to complete one or more primary study measures, including the DNS assessment items, Montreal Cognitive Assessment (MoCA) evaluation, Connor-Davidson Resilience Scale (CD-RISC) questionnaire, Patient Health Questionnaire-9 (PHQ-9) questionnaire, or key sociodemographic variables required for multivariable analysis. Only participants with complete datasets were included in the final analysis.

Study variables and operational definitions

The primary exposure variable in this study was the DNS, conceptualized as a study-specific composite representation of daily behavioral patterns related to digital technology use and lifestyle routines [17].

DNS was developed as an exploratory composite behavioral index intended to capture multiple dimensions of digital lifestyle behavior within a single summary measure. The four selected domains (screen exposure, sleep behavior, physical activity, and digital engagement type) were chosen based on previous literature reporting associations between these behavioral domains and cognitive or psychological outcomes [18–20]. Because no established composite instrument incorporating all four domains currently exists, equal weighting was applied to each component to ensure methodological simplicity and transparency and to avoid introducing arbitrary differential weighting in the absence of empirical evidence supporting the relative importance of individual domains. The scoring framework was designed to represent cumulative behavioral risk rather than to serve as a validated diagnostic or predictive instrument. To address reproducibility and methodological transparency, this construct was operationalized as a composite DNS score derived from four predefined behavioral domains: screen exposure duration, sleep timing and regularity, physical activity patterns, and type of digital engagement. Each domain was quantified using a standardized three-point ordinal scoring system (0–2), where higher scores indicated more maladaptive behavior.

Screen exposure duration was assessed based on self-reported average daily screen time (including smartphones, computers, and television) and categorized as ≤ 2 hours/day (0 points), > 2 to ≤ 5 hours/day (1 point), and > 5 hours/day (2 points). Sleep timing and regularity were evaluated using self-reported sleep duration and consistency of sleep-wake cycles and classified as regular sleep (7–9 hours/night with consistent timing = 0 points), mildly irregular sleep (> 6 to < 7 hours/night or variable timing = 1 point), and poor sleep (≤ 6 hours/night, irregular timing, or frequent disturbances = 2 points). Physical activity patterns were measured based on weekly duration of moderate-to-vigorous activity and categorized as ≥ 150 minutes/week (0 points), 60–149 minutes/week (1 point), and < 60 minutes/week (2 points). Digital engagement type was classified based on predominant usage behavior into active/goal-directed use (e.g., educational or occupational activities = 0 points), mixed use (1 point), and passive consumption (e.g., social media scrolling or streaming = 2 points).

A composite DNS score was calculated by summing all domain scores (range: 0–8), with higher scores reflecting more maladaptive behavioral patterns. Classification thresholds were established a priori for descriptive and analytical purposes, whereby scores of 0–2 represented adaptive behavioral profiles, scores of 3–5 represented intermediate profiles, and scores of 6–8 represented maladaptive profiles. These thresholds were selected to facilitate risk stratification across increasing levels of behavioral burden and were not derived from external validation studies. Consequently, the DNS categories should be interpreted as exploratory groupings developed for the present investigation rather than clinically validated classifications.

Since the DNS represents a study-specific composite behavioral construct, formal psychometric validation, predictive performance assessment, calibration analysis, and external validation were beyond the scope of the present study and warrant investigation in future research.

The primary outcome variable was cognitive function, assessed using the MoCA, a widely validated screening instrument for global cognitive performance [21]. The MoCA evaluates multiple cognitive domains, including attention, executive function, memory, language, visuospatial ability, and orientation, with higher scores reflecting better cognitive functioning.

Psychological resilience was considered a key secondary outcome and potential mediator, conceptualized as the ability to maintain or regain psychological well-being in the face of stressors. It was assessed using the CD-RISC [22], which provides a standardized measure of resilience capacity across emotional, cognitive, and behavioral domains.

Depressive symptoms were measured as a key covariate using the PHQ-9 [23], given its known influence on both cognitive performance and behavioral patterns.

All psychometric instruments (MoCA, CD-RISC, and PHQ-9) were administered in their validated versions, and standardized scoring procedures were followed according to established guidelines. Moreover, sociodemographic and clinical variables, including age, gender, education level, occupation, socioeconomic status, and chronic comorbid conditions such as diabetes and hypertension, were recorded to control for potential confounding effects in the final analysis.

Data collection procedure

Data collection was conducted through a structured, interviewer-administered questionnaire to ensure clarity, reduce missing data, and accommodate participants with varying levels of literacy. Trained research assistants approached eligible individuals in outpatient waiting areas and explained the study objectives in detail before obtaining informed consent. Interviews were conducted in a private and quiet setting within the hospital to ensure confidentiality and minimize response bias. Each participant underwent a standardized assessment sequence, beginning with sociodemographic data collection, followed by evaluation of digital neurobehavioral patterns, cognitive assessment using MoCA, and psychological assessment using validated scales. The average duration of each interview was approximately 20 to 25 minutes. To ensure data quality and consistency, all research assistants received prior training, and periodic supervision was conducted. Completed questionnaires were reviewed daily for completeness and accuracy, and double data entry was performed to minimize entry errors.

Statistical analysis

Data were entered and analyzed using SPSS version 26. Continuous variables were summarized as means and standard deviations, while categorical variables were presented as frequencies and percentages. Initial comparisons between groups were performed using independent sample t-tests and one-way analysis of variance (ANOVA), as appropriate, to explore differences in cognitive function and psychological resilience across digital behavioral categories. When the ANOVA was statistically significant, Bonferroni-adjusted post-hoc pairwise comparisons were performed to identify specific between-group differences. Pearson correlation analysis was used to examine bivariate relationships among continuous variables. For these analyses, screen time was entered as self-reported average daily screen exposure (hours/day) and sleep quality as the self-reported sleep duration and sleep-wake regularity measure described above, both treated as continuous variables. Normality of continuous variables was assessed using the Shapiro-Wilk test and visual inspection of histograms. Appropriate parametric or non-parametric tests were applied accordingly.

To determine independent associations, multivariable linear regression models were constructed with cognitive function as the dependent variable. DNS score was entered as the primary independent variable, while age, gender, education level, comorbidities, and depressive symptoms were included as covariates to adjust for confounding. Multicollinearity was assessed using the variance inflation factor (VIF), and model assumptions (linearity, homoscedasticity, and normality of residuals) were verified prior to interpretation. VIF values below 5 were considered indicative of acceptable multicollinearity. In the present analysis, VIF values ranged from 1.08 to 2.34, indicating no evidence of problematic multicollinearity among the predictors. Model assumptions were assessed through examination of standardized residual plots, normal probability (Q-Q) plots, residual-versus-fitted value plots, and tests of homoscedasticity, which demonstrated acceptable linearity, homoscedasticity, and approximate normal distribution of residuals. Residual independence was evaluated using the Durbin-Watson statistic (1.92), indicating acceptable independence of residuals. Regression results are reported using standardized beta coefficients (β) to facilitate comparison of the relative strength of associations across predictors measured on different scales. Standard errors (SE), t-values, and p-values correspond to the unstandardized regression coefficients (B). Psychological resilience was subsequently evaluated as a potential mediating variable using bootstrapped mediation analysis with 5,000 resamples, allowing estimation of indirect effects with greater statistical reliability. Because all study variables were measured at a single time point, the mediation analysis was conducted to explore potential indirect associations rather than establish causal pathways. The temporal ordering assumed in the mediation model was based on theoretical considerations and prior literature; however, alternative causal directions remain possible, and the findings should therefore be interpreted as exploratory. A p-value of less than 0.05 was considered statistically significant.

The objective of the present study was to examine cross-sectional associations between digital behavioral patterns, psychological resilience, and cognitive function rather than to develop, validate, or evaluate the predictive performance of a clinical prediction model. Therefore, assessments of model discrimination, calibration, predictive accuracy, or external validation were not performed. Future longitudinal studies involving independent populations are warranted to evaluate the psychometric properties, predictive validity, and generalizability of the DNS framework.

Sociodemographic and clinical characteristics

The study included 400 participants with a mean age of 39.6 ± 12.41 years, representing a relatively young to middle-aged adult population. Males constituted 53.50% (n = 214) of the sample, while females accounted for 46.50% (n = 186). As shown in Table 1, educational attainment was relatively balanced, with 40.00% of participants having graduate-level education or higher, indicating variation in educational attainment within the study population. Regarding clinical comorbidities, 28.00% had hypertension and 24.00% had diabetes mellitus, reflecting the distribution of chronic diseases in the study population. Overall, the sociodemographic distribution indicated a heterogeneous sample suitable for evaluating behavioral and cognitive associations across diverse participant characteristics.

Distribution of DNS

The classification of participants into digital neurobehavioral profiles was based on the composite DNS score (range: 0–8), derived from four domains: screen exposure duration, sleep timing and regularity, physical activity patterns, and digital engagement type, as described in the Methods section. Participants were categorized into adaptive (score 0–2), intermediate (score 3–5), and maladaptive (score 6–8) profiles using predefined thresholds. The distribution of digital neurobehavioral profiles revealed that 35.50% (n = 142) of participants were classified as adaptive, 42.00% (n = 168) as intermediate, and 22.50% (n = 90) as maladaptive. Participants in the maladaptive category had higher cumulative DNS scores, reflecting a greater burden of adverse behavioral characteristics across the assessed domains, including prolonged screen exposure, irregular or insufficient sleep, low physical activity, and predominantly passive digital engagement. In contrast, the adaptive group demonstrated lower DNS scores, characterized by balanced digital usage, regular sleep patterns, and adequate physical activity levels. This distribution indicated that nearly two-thirds of participants were classified within the intermediate or maladaptive categories according to the exploratory DNS scoring framework, reflecting a greater cumulative burden of the behavioral characteristics incorporated into the composite index (Table 2).

Cognitive and psychological outcomes across digital profiles

As shown in Table 3, there was a clear graded decline in cognitive performance across digital neurobehavioral categories. Participants with adaptive digital behavior had the highest mean MoCA scores (27.2 ± 1.8), while those with maladaptive profiles demonstrated significantly lower scores (23.4 ± 2.9), with a statistically significant difference across groups (p < 0.001). Bonferroni-adjusted post-hoc pairwise comparisons were subsequently performed to identify differences between individual digital behavior profiles. A similar trend was observed for psychological resilience, which progressively decreased from adaptive to maladaptive groups. Conversely, depressive symptoms measured by PHQ-9 increased substantially across worsening digital behavior profiles. These findings indicated a graded cross-sectional association between digital behavioral patterns and cognitive and psychological outcomes. Although DNS categories demonstrated significant associations with cognitive performance and psychological resilience, these findings represented observed cross-sectional relationships and should not be interpreted as evidence of predictive validity or clinical utility.

Correlation between study variables

The correlation analysis presented in Table 4 demonstrated significant interrelationships among digital behavior, cognition, and psychological variables. Cognitive function (MoCA scores) showed moderate positive correlations with psychological resilience (r = 0.42, p < 0.01) and sleep quality (r = 0.44, p < 0.01), while demonstrating significant negative correlations with depressive symptoms (PHQ-9; r = −0.46, p < 0.05) and screen time (r = −0.39, p < 0.01). Similarly, psychological resilience was positively associated with sleep quality (r = 0.40, p < 0.01) and inversely correlated with depressive symptoms (r = −0.51, p < 0.01) and screen time (r = −0.33, p < 0.01). Depressive symptoms were positively related to screen time (r = 0.41, p < 0.01) and negatively associated with sleep quality (r = −0.48, p < 0.01). Finally, screen time was inversely correlated with sleep quality (r = −0.34, p < 0.001). Taken together, these findings support the presence of a tightly interconnected pattern of associations, wherein maladaptive digital behaviors and less favorable sleep behavior are linked to diminished resilience and cognitive outcomes, while higher resilience was associated with lower depressive symptoms and better cognitive performance.

Prior to conducting multivariable regression analyses, model assumptions were evaluated. Multicollinearity was assessed using the VIF, and all predictor variables demonstrated acceptable VIF values (< 5), suggesting no evidence of problematic multicollinearity. Assessment of standardized residual plots and probability plots suggested acceptable linearity and approximate normality of the residuals. Residual-versus-fitted value plots did not indicate substantial heteroscedasticity, and Durbin-Watson statistics were consistent with acceptable independence of residuals. Overall, the diagnostic assessments supported the use of linear regression models for further analyses.

Factors associated with cognitive function

Multivariable regression analysis, shown in Table 5, revealed that the DNS score was independently associated with cognitive performance after adjusting for age, education, comorbidities, and depressive symptoms. Specifically, higher DNS scores were significantly associated with lower MoCA scores (β = −0.34, p < 0.001). Depressive symptoms were independently associated with lower cognitive performance, while higher educational attainment was associated with better cognitive performance. Older age was also independently associated with lower cognitive scores (β = −0.21, p < 0.001), whereas hypertension showed no significant independent association (β = −0.12, p = 0.27). The model explained 48% of the variance in cognitive function (R² = 0.48, p < 0.001), indicating substantial explanatory capacity of the included variables.

Factors associated with psychological resilience

As shown in Table 6, psychological resilience was significantly associated with both behavioral and psychological factors. Higher depressive symptom scores were strongly associated with reduced resilience (β = −0.41, p < 0.001), while higher DNS scores were also independently associated with lower resilience levels (β = −0.28, p < 0.001). Education was positively associated with resilience, although the cross-sectional design precluded conclusions regarding a protective causal effect. Older age was associated with modestly lower resilience scores (β = −0.14, p = 0.005). The overall model explained 44% of the variance in resilience scores (R² = 0.44, p < 0.001).

Mediation analysis: role of psychological resilience

Mediation analysis results, summarized in Table 7, indicated a statistically significant indirect association between DNS and cognitive function through psychological resilience. The total effect of digital behavior on cognition was significant (β = −0.34, p < 0.001), and remained significant after adjustment, although reduced in magnitude in the direct effect pathway (β = −0.25, p < 0.001). Importantly, the indirect association through psychological resilience was statistically significant (β = −0.09, p = 0.01), consistent with a statistically detectable indirect pathway in the cross-sectional data. These findings suggested that part of the observed association between maladaptive digital behavioral patterns and cognitive function may be statistically accounted for by differences in psychological resilience; however, the cross-sectional design precluded conclusions regarding temporal ordering or causal mechanisms.

Limitations

This study had several limitations. First, the cross-sectional, single-center design limited the ability to establish causal relationships between digital behavioral patterns, psychological resilience, and cognitive performance. Although mediation analysis suggested potential indirect associations, the temporal sequence of these relationships cannot be confirmed. Second, digital behavior measures were based on self-report, which may have introduced recall and reporting bias, and residual confounding from unmeasured factors could not be completely excluded. Third, cognitive function was assessed using screening-based measures rather than comprehensive neuropsychological testing, which limited detailed interpretation across specific cognitive domains. Moreover, the DNS represents an exploratory composite behavioral index developed for this study and has not undergone formal psychometric validation. Therefore, its component weighting, classification thresholds, reproducibility, and any potential predictive utility require rigorous evaluation in longitudinal studies and independent populations before clinical or prognostic applications can be considered.

Future research

Future research should focus on validating and refining the DNS framework through longitudinal, multicenter studies. Future investigations should incorporate probability-based sampling strategies, culturally diverse populations, and objective digital behavior measurements obtained through wearable devices or passive digital monitoring approaches. No clinical application should be inferred from the present findings until the construct has undergone appropriate psychometric assessment, longitudinal evaluation, and external validation in independent populations.

Conclusion

This study found that digital behavioral patterns, represented through the exploratory DNS framework, are associated with cognitive function and psychological resilience. The findings suggest that integrating multiple digital lifestyle domains, including screen exposure, sleep patterns, physical activity, and digital engagement behaviors, may provide a useful approach for investigating behavioral factors related to cognitive and psychological outcomes. However, the DNS developed in this study represents a study-specific composite behavioral construct and should not be interpreted as a validated digital biomarker or clinical risk classification tool. Furthermore, given the cross-sectional design, the observed associations and mediation findings should be interpreted cautiously and do not establish causal pathways or direct neuroprotective effects. Future longitudinal, psychometric, and external validation studies are required to determine the reproducibility and generalizability of the DNS framework and to establish whether it has any meaningful predictive utility. Until such evidence is available, the DNS should be regarded as an exploratory research construct, and no clinical or prognostic application should be inferred.