Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Factors associated with problematic internet use among University of Gondar undergraduate students, Northwest Ethiopia: Structural equation modeling

  • Werkneh Melkie Tilahun ,

    Roles Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing

    werkneh7wmt@gmail.com

    Affiliation Department of Public Health, College of Medicine and Health Science, Debre Markos University, Debre Markos, Ethiopia

  • Asefa Adimasu Tadesse,

    Roles Data curation, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Visualization, Writing – review & editing

    Affiliation Department of Epidemiology and Biostatistics, College of Medicine and Health Sciences, Institute of Public Health, University of Gondar, Gondar, Ethiopia

  • Haileab Fekadu Wolde,

    Roles Data curation, Investigation, Methodology, Project administration, Resources, Supervision, Visualization, Writing – review & editing

    Affiliation Department of Epidemiology and Biostatistics, College of Medicine and Health Sciences, Institute of Public Health, University of Gondar, Gondar, Ethiopia

  • Zenebe Abebe Gebreegziabher,

    Roles Data curation, Formal analysis, Investigation, Methodology, Project administration, Resources, Software, Supervision, Validation, Visualization, Writing – review & editing

    Affiliation Department of Epidemiology and Biostatistics, School of Public Health, Debre Birhan University, Debre Birhan, Ethiopia

  • Wondwosen Abey Abebaw,

    Roles Data curation, Investigation, Methodology, Resources, Supervision, Validation, Visualization, Writing – review & editing

    Affiliation Department of Epidemiology and Biostatistics, School of Public Health, College of Medicine and Health Sciences, Woldia University, Woldia, Ethiopia

  • Mulat Belay Simegn,

    Roles Data curation, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Visualization, Writing – review & editing

    Affiliation Department of Public Health, College of Medicine and Health Science, Debre Markos University, Debre Markos, Ethiopia

  • Lamrot Yohannes Abay,

    Roles Data curation, Investigation, Methodology, Project administration, Resources, Supervision, Validation, Visualization, Writing – review & editing

    Affiliation Department of Environmental and Occupational Health and Safety, College of Medicine and Health Sciences, Institute of Public Health, University of Gondar, Gondar, Ethiopia

  • Tigabu Kidie Tesfie

    Roles Data curation, Investigation, Project administration, Resources, Supervision, Validation, Visualization, Writing – review & editing

    Affiliation Department of Epidemiology and Biostatistics, College of Medicine and Health Sciences, Institute of Public Health, University of Gondar, Gondar, Ethiopia

Abstract

Background

For young adults and adolescents, excessive internet use has become a serious public health concern due to its negative impact on their health. It has been associated with detrimental effects on both physical and mental health. Negative academic outcomes were observed in the students, including missing classes, lower grades, and academic dismissal. Therefore, the purpose of the current study was to identify factors associated with PIU among undergraduate students at the University of Gondar.

Method

A cross-sectional study was conducted at the University of Gondar among 1514 undergraduate students from June 1–20, 2022. The study participants were selected using a stratified simple random selection procedure. Using structural equation modeling, the degree of relationship was ascertained. A p-value of less than 0.05 and an adjusted regression coefficient with a 95% confidence interval (CI) were used to interpret the data.

Results

In our study, being from non-health departments [β = 0.11, 95% CI: 0.037, 0.181], current alcohol use [β = 0.12, 95% CI: 0.061, 0.187], depressive symptoms [β = 0.23, 95% CI: 0.175, 0.291], insomnia symptoms [β = 0.12, 95% CI: 0.060, 0.196], and ADHD symptoms [β = 0.11, 95% CI: 0.049, 0.166] had a significant positive effect on PIU, while having a history of head injury had a significant negative effect [β = -0.12, 95% CI: -0.226, -0.021] on PIU.

Conclusion and recommendation

Factors such as current alcohol use, non-health department type, depressive symptoms, insomnia, and ADHD symptoms were positively associated with PIU. However, a history of head injuries was negatively associated with PIU. Therefore, strategies aimed at the early identification of PIU may lead to an improvement in the psychosocial health of university students.

Introduction

Even if the use of the internet has clear and tremendous benefits for the users, it has increased dramatically and shown cases of excessive use, which often has negative health consequences [1]. And currently, it has reached the magnitude of a significant public health concern [1]. The term "problematic internet use (PIU) refers to any use of the internet that causes a person to experience difficulties in their social, academic, professional, or psychological spheres [2]. This problem can occur at any age, in any social, educational, or financial background [3]. This problem also becomes an evident public health problem among university students in Ethiopia [4, 5].

Present-day young adults’ lives revolve around the internet, and excessive use of it has given rise to a new and developing health concern for their health [6, 7]. And it has been linked to detrimental effects on mental and physical health [6, 7]. Moreover, it has also been associated with poor academic outcomes among students, including lower academic grades, missed assignments, and academic disqualification [8]. In general, students with PIU reported poor health behaviors and can lead an unhealthy lifestyle [811]. Studies revealed that PIU significantly increased social anxiety [12], eating disorders [9, 11], problematic thoughts [13], suicidal ideation and attempt, social isolation, subjective distress [9, 10, 14, 15], and less health-protective behaviors [9, 11].

Moreover, factors such as age [10, 1618], sex [5, 9, 1921], poor sleep quality [22], less frequent exercise [23], year of study [18], marital status [10], residence [24], alcohol consumption [4, 21, 24], perceived social support [24, 25], psychological distress [4, 13, 17, 24], smoking [21], ADHD symptoms [26], depressive symptoms [5, 9, 13, 17, 23, 25, 27], chat chewing, and caffeinated drinks [4, 5] were indicated as significant factors associated with PIU.

University students should pay close attention to problematic internet use, as it is a significant public health issue [15]. Especially in underdeveloped countries, it deserves more exploitation [25]. Studies have shown that additional studies are necessary to fully comprehend the connection between internet use and academic characteristics, as well as physical and mental health [23]. College students’ abilities to evaluate and carry out PIU-related behavior changes vary widely [28]. Understanding the potential public health consequences of PIU among young adults pursuing higher education and conducting PIU screenings on university students are therefore clinically significant [29]. Thus, the aim of this study was to identify the sociodemographic, academic, behavioral, and health-related variables associated with PIU among University of Gondar undergraduate students.

Methods and materials

Data sources, study setting, design, and period

This study was conducted based on data that had been collected for another purpose. This cross-sectional study was employed from June 1–20, 2022, at the University of Gondar. All five campuses (Science Amba, Atse Tewodros, Fasiledus, Maraki, and Tseda) of the University of Gondar were involved. A total of six colleges, two institutes, two faculties, and one school are found in all campuses, offering 56 undergraduate and 64 postgraduate programs [30].

Population

All regular undergraduate students at the University of Gondar were the source population. Those students who had registered for the 2022 academic year and available during the data collection period were included in the study population. However, students with thyroid disease and below the age of 18 were excluded.

Sample size determination

The sample size required for structural equation modeling (SEM) is dependent on the complexity of the model. As a general rule of thumb, the minimum sample size should be no less than 5–20 times the number of parameters to be estimated [31]. Even if the data were collected for another purpose, we checked whether our sample was sufficient to estimate the parameters specified in the hypothesized model. Accordingly, there were 34 observed endogenous variables (6, 9, 9, 7, and 3 for ADHD, depression, problematic internet use (PIU), insomnia, and social support respectively). Therefore, 29 path coefficients were needed since five of them were fixed to 1, in order to give the latent variable measurement scale. One disturbance term for PIU and 34 error terms for those observed endogenous variables. There were 30 exogenous variables (26 observed and four latent) in the specified model. Therefore, the total number of free parameters to be estimated is equal to

  1. 34*2–5 = 63 path coefficients and error terms for latent variables
  2. 30*2 = 60 path coefficients and variance for exogenous variables and
  3. 1 disturbance term for PIU, making a total of 124 free parameters to be estimated. Therefore, the N:q ratio becomes approximately 12:1. Thus, we can conclude that the sample size is sufficient to estimate the parameters specified in the hypothesized model

Sampling procedure

A stratified simple random sampling technique was applied to select the study participants. A detailed description was written and published elsewhere [32].

Variables of the study

Outcome variables.

Problematic internet use (latent endogenous variable).

Independent variables.

Latent exogenous variables (insomnia, ADHD, depression and social support)

Observed exogenous variables such as;.

Socio demographic factors: sex, age, marital status, monthly allowance, maternal education, father education, prior residence, self-rated family economic level and presence of partner.

Health related factors: history of chronic disease, history of head injury, history of parental psychiatric illness, birth order and history of stressful life events.

Behavioral factors: history of alcohol use, current alcohol use, cigarette smoking, chat chewing, cannabis use, and physical exercise

Academic related factors: year of study, number of studying hours per day, number of sleeping hours per day, history of academic failure, worry about academic performance and department type.

Measurement and data collection tool

In addition to questions about PIU, ADHD, insomnia, social support and depression, questions about behavioral, clinical, academic, and sociodemographic aspects were also included.

Problematic internet use.

A simplified version of the long PIU-18 item, the Problematic Internet Use Questionnaire-9 (PIUQ-9), was used to measure it. Nine questions total, with five alternative answers: 1 for "never," 2 for "rarely," 3 for "sometimes," 4 for "often," and 5 for "always/almost always" that can be found elsewhere [33, 34]. Higher scores indicate a serious internet use issue.

ADHD symptom.

The six-item Adult Self-Report Scale-V1.1 (ASRS-V1.1), a simplified screener from the World Health Organization Composite International Diagnostic Interview, was used to measure it. There are five options for each question: never, rarely, sometimes, often, and very often. An increased risk of ADHD symptoms is indicated by higher scores [35].

Depressive symptom.

It was determined by the Patient Health Questionnaire-9 (PHQ-9) tool. It has nine questions that ask about the frequency of the occurrence of symptoms. All can be scored from zero to three as not at all, several days, more than half the days, and nearly every day, respectively. There can be a minimum score of 0 and a maximum of 27. Higher scores indicate a high risk of depressive symptoms [36].

Insomnia.

The measurement was made using the insomnia severity index, which has seven questions that ask about the last two weeks of insomnia problems. It has 5 alternatives, which can be scored from 0 to 4, and higher sores indicate a higher risk of insomnia [37].

Social support.

It was evaluated with the Oslo Social Support Scale (OSSS-3), which consists of items. Four options, numbered 1–4, are provided for the first question; five options, numbered 1–5, are provided for the remaining questions. Larger scores on the OSSS-3 suggest stronger social support [38].

Ever-smoking was assessed by asking a single question: “Have you ever smoked cigarettes in your lifetime?” and the responses were “yes” or “no” [39].

Physical exercise.

Exercising or engaging in any sport, such as walking for at least 20 minutes per day, to which the answers were either "yes" or "no" [40].

History of alcohol use was assessed by asking a single question: “Have you ever drunk at least one of the alcoholic beverages (beer, wine, whiskey, Areki, Tela, Tej, etc.) for nonmedical purposes?” and possible responses were “yes” or “no” [41].

Current alcohol use was assessed by asking a single question: “Have you ever drunk at least one of the alcoholic beverages (beer, wine, whiskey, Areki, Tela, Tej, etc.) for nonmedical purposes in the last three months?” and possible responses were “yes” or “no” [41, 42].

Chat chewing was assessed by asking a single question: “Have you ever chewed chat during your lifetime?” and possible responses were “yes” or “no” [41].

Cannabis use was assessed by asking a single question “Have you ever used cannabis during your lifetime?” and possible responses were “yes” or “no” [43].

The history of parental psychiatric illness was assessed using one question: “Do your parents have a history of any medically confirmed psychiatric disorders? (It can be any type). Possible responses were “no”, “father”, “mother”, and “both” [32].

Data collection procedure

A standardized, self-administered questionnaire was used to collect data. The data was collected by four trained data collectors. Following the selection of an identification number for a student using Excel, data collectors located the relevant participants. The questionnaire was given out after the purpose, advantages, and disadvantages of the study were discussed and students gave their consent to participate. The data collector and the principal investigators promptly verified the completeness and consistency of the responses. The principal investigators had direct oversight of every process.

Data quality assurance and management

Before actual data collection, a pilot study was carried out, and senior and practicing psychiatrists verified the instrument’s face validity. A widely spoken Amharic language, which is also Ethiopia’s official working language, was used to translate the original English questionnaire, and then another person translated it again to English to ensure proper translation of terminologies. Two days of training were given to data collectors regarding the subject matter, confidentiality, ethical behavior, and data collection methods. The researchers kept in regular, timely communication with the collectors throughout the data collection process to address any doubts or concerns regarding the procedures and offer assistance as needed. Early corrective action was taken once the collected data were promptly examined for consistency and completeness. Consistency, any entry errors, missing values, and outliers were carefully reviewed during the data entry process by referring back to the questionnaire.

Reliability and validity of tools.

To validate the tool, 156 students from Debre Markos University participated in an external pilot study. To further verify the factor loadings, internal consistency, construct, and statistical validity, confirmatory factor analysis was performed. Composite reliability (CR) was used to verify internal consistency. As a result, all constructs’ composite reliability falls between 0.64 and 0.9. As a result, every construct satisfied the required threshold of 0.60 [44]. Average variance extracted (AVE) and CR were used to verify construct validity, also known as convergent validity. Consequently, the AVE falls between 0.38 and 0.49. Despite AVE being less than 0.5, the tool’s convergent validity held true. We can determine that the construct’s convergent validity is sufficient based only on CR, as AVE represents a more conservative estimation of the measurement model’s validity [44, 45]. A thorough explanation was composed and released in another location. A detailed description was written and published elsewhere [32].

Data processing and analysis

Epi-Data Version 4.61 was used to enter the data. For analysis, Amos SPSS version 21 and STATA version 16 were utilized. We verified a number of assumptions, including multivariate normality, sphericity, sample adequacy, missing data, and outliers. A descriptive analysis and then SEM were employed to test and estimate complex relationships between variables. The adjusted regression coefficients with a 95% confidence interval (CI) and a corresponding p-value < 0.05 were used to interpret the degree of association. Model specification, identification, parameter estimation, model evaluation, and modification were the five logical steps in the model-building process [46].

1. Model specification.

In order to ascertain each relationship and parameter in the model that is relevant to the researcher, the model specification is the initial phase in which the researcher identifies the concept and describes the proposed relationships among the variables [47]. In our case, we hypothesized that all the independent variables mentioned in the variable section were related to PIU (Fig 1).

thumbnail
Fig 1. Hypothesized structural model, UoG, Northwest Ethiopia, 2022.

Circles indicate latent variables or error terms or disturbances, rectangles indicate observed variables, single arrows indicate factor loadings or regression coefficients, and double arrows indicate the covariance between latent variables; ADHD = Adult Attention Deficit Hyperactivity Disorder, d1-d9 –depression items, AD1-AD6 = ADHD items, S1-S3 = items for Social Support, I_1 –I_7- Insomnia Items, exercise = Physical exercise, year = year of study, cann = cannabis, current_alco = current alcohol, alco = alcohol use, failure = history of academic failure, fam_pych = history of family psychiatric illness, allow- Monthly Allowance, worry = worry about academic performance. Event = history of stressful life event, FEL = family economic level, read and sleep = reading and sleeping hours per day, FE = father education level, ME = mother education level, BO = birth order, PIU = Problematic Internet Use, int1-9 = items for Problematic Internet Use, DPTT = department type and CD = Chronic Disease.

https://doi.org/10.1371/journal.pone.0302033.g001

2. Model identification.

In order to obtain a solution, the number of free parameters to be estimated, Q, must be equal to or less than the number of non-redundant elements in the sample covariance matrix (Q ≤ P*), where P* = K(K + 1)/2 and K is the number of measured variables. There are three types of model identification. Under-identified: If the degree of freedom (df) is less than zero, just-identified: if the degree of freedom (df) is zero, and over-identified: the degree of freedom will be positive (> 0), and all Q parameters can be estimated, df = (P*—Q) [48]. Model coefficients can only be estimated in the just-identified or over-identified model [47]. In our study, there were 50 observed endogenous and exogenous variables: (50(50+1))/2 = 2550/2 = 1275, non-redundant elements in the sample covariance, and Q = 124. Thus, df = 1275–124 = 1151, which indicates that the model is over-identified.

3. Model estimation.

It is a process of finding numerical values for unknown (free) parameters [49]. In this study, the data didn’t satisfy the multivariate normality assumption, so the maximum likelihood estimation approach with 3500 bootstrap samples was used.

4. Evaluation of model fit (model testing).

The fit indices in SEM are used to evaluate each path coefficient. A few indices that are less affected by sample size are the root mean square error of approximation (RMSEA) and comparative fit index (CFI) [47, 50, 51], and the Tucker-Lewis index (TLI) [52]. Models with the least amount of overfitting can be chosen using information criteria (AIC and BIC); however, they are not helpful for testing null hypotheses, even if BIC is another measure of a model’s parsimony among potential models [53]. Therefore, RMSEA, CFI, TLI, AIC, and BIC were used. For each fit index, there are recommended values that should be followed, although none are absolutes [53]. CFI and TLI (NNFI) > 0.90 [47, 50, 51], and RMSEA < 0.05 [51] suggest improved model fitness.

5. Model modification.

It entails either fixing or freeing parameters that were free in order to modify a stated and estimated model. To help choose which parameters to add to a model to improve the model fit, the modification index (MI) and the standardized expected parameter change (SEPC) were used. If fixed parameters were linked to a high MI value (> 4) with one degree of freedom and a 0.05 alpha level, they could be included in the model and freely estimated.

Ethical consideration

The UoG institutional review board granted ethical clearance under reference number Ref No. /IPH/2131/2014. A list of student ID numbers was gathered with permission from the main registrar’s office of the University of Gondar. To collect student data at any moment, a permission letter was requested from every college, institute, faculty, school, or department. Before the questionnaires were distributed, each participant provided written informed consent. During the data gathering process, no personal identifiers were recorded, and the remaining acquired information was stored in a manner that guaranteed data confidentiality and anonymity. The participants were made aware of their right to withdraw from the study at any moment and that their participation was entirely voluntary.

Result

Socio-demographic factors

A total of 1514 students participated in this study. Of the participants, males represented almost a third (65%). The participants’ median age was 21 years old (IQR = 2).The majority of respondents (96.9%) were single in their marital status. Roughly a third (66.8%) were urban residents before. More than three-fourths (79.7%) of respondents said they have a stable, intimate partner. About 30.6% and 36.8% of the participant’s mother and father were unable to read and write and attended higher education, respectively. Regarding health and behavioral-related factors, a month before the data was collected, over half (58.5%) of the study participants said they had gone through stressful situations. According to the majority (91.7% and 92.3%) of participants, none of their parents had ever experienced mental health issues or chronic diseases, respectively. The majority of participants (93.4% and 89.8%) indicated that they had never chewed chat or smoked cigarettes, respectively. The median birth order was 2 (IQR = 3). The majority (90.4%) of the participants had not experienced a previous head injury. About two-thirds (66.5%) and half (50.8%) of the participants revealed that they had a history of previous alcohol use and were currently using alcohol, respectively. Almost all (96.9%) and about three-fourths (74.4%) had never used cannabis or regular physical exercise, respectively. The mean score of the PIU and social support scores was 23.2 and 10.3 with a standard deviation of 8.4 and 2.2, respectively. The median scores for ADHD symptoms, depressive symptoms, and insomnia were 12, 8, and 7 with an IQR of 7, 8, and 7, respectively. Regarding factors related with education, more than half (58.7%) of respondents said they were concerned about their academic performance. One-fourth of the responders were from the College of Medicine and Health Science, while the College of Social Science and Humanities accounted for roughly one-fifth. The median hours of study and sleep were likewise recorded at 5 and 8, respectively. For more information, you can refer to the previous work [32].

Assumptions for structural equation modeling

Omitted variable bias.

When a certain regression model leaves out a third variable that has the potential to influence both the independent and dependent variables in the causal pathway, it can lead to a condition known as "omitted variable bias" (OVB) [54]. As a result, we made sure that no independent PIU-affecting variable was overlooked that wasn’t included in the model. In order to make sure that the omitted variables aren’t leading to model misspecification, we performed a regression specification-error test (RESET). The results showed that there were no issues with the omitted variables in our study because the P-value was 0.39 > 0.05, indicating that the model has no OVB.

Sample size adequacy and sphericity.

We evaluated the results of Bartlett’s test for sphericity and the Kaiser-Meyer-Olkin (KMO) measure of sample adequacy. Excellent partial correlation, as indicated by the overall KMO (0.90), and a significant Bartlett’s test of sphericity. A detailed description was written and published elsewhere [32].

Multivariate normality, missing and outliers.

It was evaluated using the Henze-Zirkler, Doornik-Hansen, and Mardia’s skewness and kurtosis tests, and the results were unsatisfactory. Using the Mahalanobis distance (probability < 0.001 [55]), outliers were examined, and 77 observations show that there is an outlier. Errors in data entry and measurement were verified. All the measurements, though, made sense, and we thought they were all good outliers. Ten observations had values that were missing. Because they made up less than 5% of the sample as a whole and were deemed to be absent entirely at random, a list-wise deletion was carried out.

Independence of observation.

When clustering at the departmental and college levels was examined, the ICC was significantly below the necessary threshold (0.1) (S1 Table).

Common method bias (CMB).

An ordinal scale was used in our study to measure the outcome variables (PIU) and additional exogenous latent factors (social support, insomnia, ADHD, and depression). The Harman single-factor test for CMB was performed. The results showed that the variance explained by a single factor was 19%. Therefore, we can conclude that there was no CMB problem in our study.

Confirmatory factor analysis (CFA).

First, CFA was used to assess measurement models for proposed and empirically confirmed factor structures in our SEM (Fig 2). If it is thought that items have common sources other than the latent factors, residuals and errors have been permitted to correlate. We then merged the structural path models and CFA models to create our generic SEM framework.

thumbnail
Fig 2. Initial measurement model with standardized estimates displayed, UoG, Northwest Ethiopia, 2022.

Circles indicate latent variables or error terms, rectangles indicate observed variables, single arrows indicate factor loadings or regression coefficients, and double arrows indicate the covariance between latent variables.

https://doi.org/10.1371/journal.pone.0302033.g002

The measurement model’s model fit indices (TLI = 0.87, CFI = 0.88) fell below the needed cutoff point of 0.9, despite other goodness of fit metrics (CMIN/DF = 4.99, RMSEA (P-value) = 0.05 (0.10)) meeting the necessary threshold. Additionally, the BIC was 2741.96 and the AIC was 2738.24. This suggests that the sample variance-covariance data does not support the analysis and that the CFA of the original models is inadequate. As a result, we kept adding potential adjustments depending on MI and SEPC. In construct indicators, covariance between error terms was included (S2 Table and S1 Fig). All of the model fit metrics (CMIN/DF = 2.97, TLI = 0.93, CFI = 0.94, RMSEA (P-value) = 0.04 (1.00), AIC = 1678.48, and BIC = 1682.73) thus become acceptable. But the model wasn’t parsimonious. Consequently, a type of parceling called average partial factorial parceling was used.

Item parceling. Three constructs, ADHD, insomnia, and depression, were subjected to average partial factorial parceling. The standard recommendation is three parcels [56]. Hence, every construct had three parcels (S3 Table). In terms of model fitness, CMIN/DF = 7.24, TLI = 0.91, CFI = 0.92, RMSEA (P-value) = 0.06 (0.00), AIC = 1400.3, and BIC = 1676.75 were the majority of model fit indices at the necessary level. But in order to raise the model fit indices, we added more potential adjustments based on MI and SEPC (S2 Fig). In conclusion, every goodness of fit index (CMIN/DF = 3.47, TLI = 0.96, CFI = 0.97, RMSEA (P-value) = 0.04 (1.00), AIC = 711) fell within an acceptable range. BIC, a measure of model parsimony, dropped dramatically from 1676.8 to 713.8. As a result, we chose to move on with the adjusted measurement model that included parceled indicators.

Structural model

An analysis of a set of relations between variables can be done using a statistical technique called a structural model. Variables can be both dependent and independent variables (Fig 3).

thumbnail
Fig 3. Full structural model after parceling, UoG, Northwest Ethiopia, 2022.

AF1-3- represents parcel 1–3 for ADHD, IF1, 1F25, IF34- represents parcel 1–3 for insomnia and d614, d927 and d835 represents parcel 1–3 for depression respectively; exercise = Physical exercise, year = year of study, cann = cannabis, current_alco = current alcohol, alco = alcohol use, failure = history of academic failure, worry = worry about academic performance. Event = history of stressful life event, read and sleep = reading and sleeping hours per day, ME = mother education level, BO = birth order, PIU = Problematic Internet Use, int1-9 = items for Problematic Internet Use, DPTT = department type and CD = Chronic Disease.

https://doi.org/10.1371/journal.pone.0302033.g003

Model selection.

Prior to allowing correlation between independent predictors, insignificant variables were eliminated from the model. After that, error terms for the indicators within the same construct were added. Fit indices and information criteria were used to compare models (Table 1).

thumbnail
Table 1. Model selection after parceling, UoG, Northwest Ethiopia, 2022.

https://doi.org/10.1371/journal.pone.0302033.t001

Factors associated with PIU

The model was modified to exclude variables such as sex, age, residence, average monthly allowance, marital status, year of study, cannabis use, smoking history, family history of mental health issues, physical activity history, academic failure history, and self-reported household income level, worry about academic performance, presence of a stable partner, history of stressful events, presence of medically confirmed chronic disease, history of smoking and chat chewing, sleeping and reading time per day, social support, previous history of alcohol use, father and mother education, and birth order by considering their statistically insignificant effect on PIU. Once the path coefficients have been adjusted and all the remaining variables have been controlled, the model is summarized (Fig 4).

thumbnail
Fig 4. Final SEM showing factors associated with PIU, UoG, Northwest Ethiopia, 2022.

Standardized estimates are displayed.

https://doi.org/10.1371/journal.pone.0302033.g004

Our final model indicated that department type, current alcohol use, history of head injury, depressive symptoms, insomnia symptoms, and ADHD were significant factors. Thus, being from non-health departments [adjusted β = 0.11, 95% CI: 0.037, 0.181], current alcohol use [adjusted β = 0.12, 95% CI: 0.061, 0.187], depressive symptoms [adjusted β = 0.23, 95% CI: 0.175, 0.291], insomnia symptoms [adjusted β = 0.12, 95% CI: 0.060, 0.196], and ADHD symptoms [adjusted β = 0.11, 95% CI: 0.049, 0.166] had a significant positive effect on PIU. Having a history of head injury had a significant negative effect [adjusted β = -0.12, 95% CI: -0.226, -0.021] on PIU (Table 2).

thumbnail
Table 2. Factors associated with PIU among University of Gondar undergraduate students, northwest Ethiopia, 2022.

Unstandardized estimates.

https://doi.org/10.1371/journal.pone.0302033.t002

Discussion

This study applied SEM to investigate factors associated with PIU among University of Gondar undergraduate students in Northwest Ethiopia. And our final reduced structural equation model revealed that depressive symptoms had a positive effect [β = 0.23, 95% CI: 0.175, 0.291] on PIU. Which indicates students with high depressive symptom scores had a high PIU score as compared to their counterparts, keeping other variables in the model controlled. This finding was supported by previous reports [5, 13, 17, 23, 25, 27, 5762]. This might be due to the fact that students with depressive symptoms may use the internet and other social media to relieve their loneliness, emotional distress, and unhealthy lifestyle [9, 63].

Moreover, insomnia symptoms were also positively associated with a higher PIU score [β = 0.12, 95% CI: 0.060, 0.196] in our study. Which indicates students with high insomnia symptom scores had a high PIU score as compared to their counterparts. In addition, a report from China and Palestine [35, 64] also revealed a strong positive association between PIU and insomnia. This could be the result of spending more time online, as using a computer at night can raise arousal levels and disrupt the restorative processes needed to fall asleep [6568]. This interaction and relationship between these disorders may indicate a bidirectional relationship between them [69].

Being a non-health department student had a positive association [β = 0.11, 95% CI: 0.037, 0.181] with PIU. This indicates that being from a non-health department increases the level of PIU by 0.11 compared to their counterparts. This might be due to the fact that students from non-health departments have more free time to spend on social media, which can increase the likelihood of developing PIU [9]. Clinical rotations, classes, and exam preparation are all common parts of medical students’ hectic schedules. Particularly on weekends and during the pauses between semesters, they have some spare time [70]. Another study conducted in Argentina also revealed that the majority of medical students’ free time was spent on social and cultural activities, sports activities, sharing time with friends, and being with family [71].

Moreover, symptoms of ADHD have a significant positive effect on PIU [β = 0.11, 95% CI: 0.049, 0.166]. This indicates students with higher ADHD symptoms scores had a higher PIU level compared to their counterparts, keeping other variables in the model controlled. This finding is supported by a systematic review [7274], a study conducted in Britain [75], New York [61], Japan [76], Turkey [77, 78], and Malaysia [79]. This can be due to the fact that individuals with ADHD are more sensitive to rewards. People with ADHD can greatly benefit from the motivating payoff that comes from feeling in control and having the opportunity to express themselves online [80]. In addition, compulsive and risky habits, such as problematic internet use, have a high tendency to cause comparable dopamine release and euphoric sensations in people with ADHD [81].

Current alcohol use was significantly associated with the PIU score [β = 0.12, 95% CI: 0.061, 0.187]. This indicates current alcohol user students had an increased PIU score compared to their counterparts. Previous findings also indicated a close relationship between alcohol use and PIU [4, 8285]. Furthermore, our research demonstrated that insomnia significantly increased PIU. This could be because alcohol drinkers may have a stronger dependence on the internet, and insomnia may be a symptom of internet addiction. There is also a complex link between alcohol consumption, insomnia, and PIU. Extensive research on adolescents in South Korea and Japan has also shown that adolescents who drink alcohol have a greater likelihood of sleep disturbance [82, 86, 87].

Our study also revealed a negative relationship between history of head injury and PIU [β = -0.12, 95% CI: -0.226, -0.021], where students with a history of injury had a decreased PIU compared to their counterparts, keeping other variables in the model controlled. This may be as a result of the long-term cognitive issues that can arise from head traumas, such as issues with memory, concentration, and problem-solving skills, which can then lead to emotional or learning challenges [32, 88, 89]. Thus, students with these problems may be forced to spend their free time on their education to cope with the highest burden in higher education institutions rather than spending their time on the internet. In contrast to this finding, a study conducted in China revealed a positive association between childhood trauma and PIU [90]. This difference might be due to the fact that the previous study included other types of trauma, like emotional and psychological trauma, in addition to physical trauma, which the current study didn’t address.

Our findings imply that public health policy planners may benefit from taking into account university students’ internet usage and how it relates to their mental health state when creating context-based treatments. Additionally, it encourages university administration to step in when inappropriate internet use unfolds.

Strength and limitation of the study

Our findings indicated the relationship between different mental health problems and PIU and highlighted the mechanisms of the problem in order to understand and inform public health policymakers and planners on internet use. However, this study does not show any cause-and-effect relationship because of the cross-sectional nature of the study design. Moreover, all our reports were based on self-reported data, which may result in socially desirable bias.

Conclusion

Our final SEM model revealed that factors such as current alcohol use, non-health department type, depressive symptoms, insomnia symptoms, ADHD symptoms, and history of head injury were significantly associated with PIU. Therefore, it is better to support future studies with clinical diagnoses to investigate the relationship between PIU and other behavioral and mental health disorders. Moreover, strategies aimed at the early identification of PIU may lead to an improvement in the psychosocial health of university students.

Supporting information

S1 Table. ICC result for outcome variables (PIU), UoG, Northwest Ethiopia, 2022.

https://doi.org/10.1371/journal.pone.0302033.s001

(DOCX)

S2 Table. Modification indices among the error terms of the constructs, as covariance between each pair of items during CFA, UoG, Northwest Ethiopia, 2022.

https://doi.org/10.1371/journal.pone.0302033.s002

(DOCX)

S3 Table. Parceling applied for three construct variables, UoG, Northwest Ethiopia, 2022.

https://doi.org/10.1371/journal.pone.0302033.s003

(DOCX)

S4 Table. Participant’s responses on each items of PIU, UoG, Northwest Ethiopia, 2022.

https://doi.org/10.1371/journal.pone.0302033.s004

(DOCX)

S1 Fig. Modified measurement model with standardized estimates displayed, UoG, Northwest Ethiopia, 2022.

https://doi.org/10.1371/journal.pone.0302033.s005

(TIFF)

S2 Fig. Modified measurement model after parceling with standardized estimates displayed, UoG, Northwest Ethiopia, 2022.

https://doi.org/10.1371/journal.pone.0302033.s006

(TIFF)

Acknowledgments

We are thankful to all the data collectors, all study participants, and the University of Gondar.

References

  1. 1. Organization, W.H., Public health implications of excessive use of the internet, computers, smartphones and similar electronic devices: Meeting report, Main Meeting Hall, Foundation for Promotion of Cancer Research, National Cancer Research Centre, Tokyo, Japan, 27–29 August 2014. 2015: World Health Organization.
  2. 2. Beard K.W. and Wolf E.M., Modification in the proposed diagnostic criteria for Internet addiction. Cyberpsychology & behavior, 2001. 4(3): p. 377–383. pmid:11710263
  3. 3. Young K.S., Caught in the net: How to recognize the signs of internet addiction—and a winning strategy for recovery. 1998: John Wiley & Sons.
  4. 4. Zenebe Y., et al., Prevalence and associated factors of internet addiction among undergraduate university students in Ethiopia: a community university-based cross-sectional study. BMC psychology, 2021. 9: p. 1–10.
  5. 5. Mengistu N., et al., Prevalence and Factors Associated with Problematic Internet Use among Ethiopian Undergraduate University Students in 2019. J Addict, 2021. 2021: p. 6041607. pmid:34925933
  6. 6. Moreno M.A., Jelenchick L.A., and Christakis D.A., Problematic internet use among older adolescents: A conceptual framework. Computers in Human Behavior, 2013. 29(4): p. 1879–1887.
  7. 7. Rumpf H.-J., Effertz T., and Montag C., The cost burden of problematic internet usage. Current Opinion in Behavioral Sciences, 2022. 44: p. 101107.
  8. 8. Chen S.-Y. and Tzeng J.-Y., College female and male heavy internet users’ profiles of practices and their academic grades and psychosocial adjustment. Cyberpsychology, Behavior, and Social Networking, 2010. 13(3): p. 257–262. pmid:20557244
  9. 9. Kożybska M., et al., Problematic Internet Use, health behaviors, depression and eating disorders: a cross-sectional study among Polish medical school students. Annals of General Psychiatry, 2022. 21(1): p. 5. pmid:35148793
  10. 10. Poorolajal J., et al., Prevalence of problematic internet use disorder and associated risk factors and complications among Iranian university students: a national survey. Health Promot Perspect, 2019. 9(3): p. 207–213. pmid:31508341
  11. 11. Mahmid F., Bdier D., and Chou P., The association between problematic Internet use, eating disorder behaviors, and well-being among Palestinian university students. Psicol Reflex Crit, 2021. 34(1): p. 32. pmid:34674078
  12. 12. Jaiswal A., et al., Burden of internet addiction, social anxiety and social phobia among University students, India. J Family Med Prim Care, 2020. 9(7): p. 3607–3612. pmid:33102337
  13. 13. Kakul F. and Javed S., Internet gaming disorder: an interplay of cognitive psychopathology. Asian Journal of Social Health and Behavior, 2023. 6(1): p. 36–45.
  14. 14. Herruzo C., et al., Suicidal Behavior and Problematic Internet Use in College Students. Psicothema, 2023. 35(1): p. 77–86. pmid:36695853
  15. 15. Pino Osuna M.J., et al., Problematic Internet Use and Psychological Problems among University Students with Disabilities. Adicciones, 2023. 35(2): p. 177–184.
  16. 16. Romero-López M., et al., Problematic Internet Use among University Students and Its Relationship with Social Skills. Brain Sci, 2021. 11(10). pmid:34679366
  17. 17. Fernández-Villa T., et al., Problematic Internet Use in University Students: associated factors and differences of gender. Adicciones, 2015. 27(4): p. 265–75. pmid:26706809
  18. 18. Stevens C., et al., Problematic internet use/computer gaming among US college students: Prevalence and correlates with mental health symptoms. Depress Anxiety, 2020. 37(11): p. 1127–1136. pmid:32939888
  19. 19. Ali R., Mohammed N., and Aly H., Internet addiction among medical students of Sohag University, Egypt. Journal of Egyptian Public Health Association, 2017. 92(2): p. 86–95. pmid:30184405
  20. 20. Kamal N.N. and Kamal N.N., Determinants of Problematic Internet use and its Association with Disordered Eating Attitudes among Minia University Students. Int J Prev Med, 2018. 9: p. 35. pmid:29721236
  21. 21. Frangos C.C., Frangos C.C., and Sotiropoulos I., Problematic Internet Use among Greek university students: an ordinal logistic regression with risk factors of negative psychological beliefs, pornographic sites, and online games. Cyberpsychol Behav Soc Netw, 2011. 14(1–2): p. 51–8. pmid:21329443
  22. 22. Wang Q., et al., Problematic internet use and subjective sleep quality among college students in China: Results from a pilot study. J Am Coll Health, 2022. 70(2): p. 552–560. pmid:32407209
  23. 23. Derbyshire K.L., et al., Problematic Internet use and associated risks in a college sample. Compr Psychiatry, 2013. 54(5): p. 415–22. pmid:23312879
  24. 24. Ramón-Arbués E., et al., Prevalence and Factors Associated with Problematic Internet Use in a Population of Spanish University Students. Int J Environ Res Public Health, 2021. 18(14).
  25. 25. Stubbs M., Bateman C.J., and Hull D.M., Problematic Internet Use Among University Students in Jamaica. Int J Ment Health Addict, 2022: p. 1–12. pmid:35571574
  26. 26. Zhao Y., et al., Association of Symptoms of Attention Deficit and Hyperactivity with Problematic Internet Use among University Students in Wuhan, China During the COVID-19 Pandemic. J Affect Disord, 2021. 286: p. 220–227. pmid:33740639
  27. 27. Phetphum C., Keeratisiroj O., and Prajongjeep A., The association between mobile game addiction and mental health problems and learning outcomes among Thai youths classified by gender and education levels. 2023.
  28. 28. Moreno M.A., et al., College Students and Problematic Internet Use: A Pilot Study Assessing Self-Appraisal and Independent Behavior Change. J Adolesc Health, 2019. 64(1): p. 131–133. pmid:30254007
  29. 29. Pal Singh Balhara Y., et al., Correlates of Problematic Internet Use among college and university students in eight countries: An international cross-sectional study. Asian J Psychiatr, 2019. 45: p. 113–120. pmid:31563832
  30. 30. University Of Gondar, University Of Gondar History. 2021, Gondar, ETHIOPIA: UoG.
  31. 31. Lei P.-W. and Wu Q., Introduction to Structural Equation Modeling: Issues and Practical Considerations. Educational Measurement: Issues and Practice, 2007. 26(3): p. 33–43.
  32. 32. Tilahun W.M., et al., Magnitude, relationship and determinants of attention deficit hyperactivity disorder and depression among University of Gondar undergraduate students, Northwest Ethiopia, 2022: Non-recursive structural equation modeling. PLoS One, 2023. 18(10): p. e0291137. pmid:37796847
  33. 33. Laconi S., et al., Psychometric Evaluation of the Nine-Item Problematic Internet Use Questionnaire (PIUQ-9) in Nine European Samples of Internet Users. Frontiers in Psychiatry, 2019. 10. pmid:30984037
  34. 34. Koronczai B., et al., Confirmation of the three-factor model of problematic internet use on off-line adolescent and adult samples. Cyberpsychology, Behavior, and Social Networking, 2011. 14(11): p. 657–664. pmid:21711129
  35. 35. Adler L., Kessler R.C., and Spencer T., Adult ADHD self-report scale-v1. 1 (ASRS-v1. 1) symptom checklist. New York, NY: World Health Organization, 2003.
  36. 36. Kroenke K., Spitzer R.L., and Williams J.B., The PHQ‐9: validity of a brief depression severity measure. Journal of general internal medicine, 2001. 16(9): p. 606–613. pmid:11556941
  37. 37. Shahid A., et al., Insomnia severity index (ISI), in STOP, THAT and one hundred other sleep scales. 2011, Springer. p. 191–193.
  38. 38. Bøen H., Dalgard O.S., and Bjertness E., The importance of social support in the associations between psychological distress and somatic health problems and socio-economic factors among older adults living at home: a cross sectional study. BMC geriatrics, 2012. 12(1): p. 1–12. pmid:22682023
  39. 39. Telayneh A.T., et al., Cigarette smoking prevalence and associated factors among college students, Amhara, Ethiopia. Pan Afr Med J, 2021. 40: p. 170. pmid:34970412
  40. 40. WHO, WHO Global Recommendations on Physical Activity for Health. 2011.
  41. 41. Tessema Z.T. and Zeleke T.A., Prevalence and predictors of alcohol use among adult males in Ethiopia: multilevel analysis of Ethiopian Demographic and Health Survey 2016. Tropical medicine and health, 2020. 48(1): p. 100–100. pmid:33353567
  42. 42. Gebeyehu E.T. and Srahbzu Biresaw M., Alcohol Use and Its Associated Factors among Adolescents Aged 15–19 Years at Governmental High Schools of Aksum Town, Tigray, Ethiopia, 2019: A Cross-Sectional Study. Journal of addiction, 2021. 2021: p. 5518946–5518946. pmid:33824774
  43. 43. Ayele M. and Mengistu A., Psychosocial Problems of Jimma University Students, Southwest Ethiopia. Ethiopian Journal of Health Sciences, 2004. 14(1).
  44. 44. Fornell C. and Larcker D.F., Structural equation models with unobservable variables and measurement error: Algebra and statistics. 1981, Sage Publications Sage CA: Los Angeles, CA.
  45. 45. Lam L.W., Impact of competitiveness on salespeople’s commitment and performance. Journal of Business Research, 2012. 65(9): p. 1328–1334.
  46. 46. Byrne B.M., Structural equation modeling with Mplus: Basic concepts, applications, and programming. 2013: routledge.
  47. 47. Fan Y., et al., Applications of structural equation modeling (SEM) in ecological studies: an updated review. Ecological Processes, 2016. 5(1): p. 19.
  48. 48. Kline R.B, Principles and Practice of Structural Equation Modeling. Third ed. Methodology in the Social Sciences, ed. Little T.D.. 2011, New york, USA: Guilford Press. 11–12.
  49. 49. Suhr D., The Basics of Structural Equation Modeling 2006: p. 3.
  50. 50. Hu L.t. and Bentler P.M., Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural equation modeling: a multidisciplinary journal, 1999. 6(1): p. 1–55.
  51. 51. Smith T.D. and McMillan B.F., A Primer of Model Fit Indices in Structural Equation Modeling. 2001.
  52. 52. Bentler P.M., Comparative fit indexes in structural models. Psychological bulletin, 1990. 107(2): p. 238. pmid:2320703
  53. 53. Hoyle R.K., Structural equation modeling for social and personality psychology. Structural Equation Modeling for Social and Personality Psychology, 2011: p. 1–120.
  54. 54. Wilms R., et al., Omitted variable bias: A threat to estimating causal relationships. Methods in Psychology, 2021. 5: p. 100075.
  55. 55. Tabachnick B.G., Fidell L.S., and Ullman J.B., Using multivariate statistics. Vol. 5. 2007: pearson Boston, MA.
  56. 56. Matsunaga M., Item parceling in structural equation modeling: A primer. Communication methods and measures, 2008. 2(4): p. 260–293.
  57. 57. Boumosleh J. and Jaalouk D., Depression, anxiety, and smartphone addiction in university students–A cross-sectional study. 2017. PLoS One. Aug. 4(12): p. 8.
  58. 58. Obeid S., et al., Internet addiction among Lebanese adolescents: the role of self-esteem, anger, depression, anxiety, social anxiety and fear, impulsivity, and aggression—a cross-sectional study. The Journal of nervous and mental disease, 2019. 207(10): p. 838–846. pmid:31503174
  59. 59. Wang P.-W., et al., Association between problematic cellular phone use and suicide: The moderating effect of family function and depression. Comprehensive psychiatry, 2014. 55(2): p. 342–348. pmid:24262117
  60. 60. Morita M., et al., Bidirectional relationship of problematic Internet use with hyperactivity/inattention and depressive symptoms in adolescents: a population-based cohort study. European Child & Adolescent Psychiatry, 2022. 31(10): p. 1601–1609. pmid:34021782
  61. 61. Restrepo A., et al., Problematic internet use in children and adolescents: associations with psychiatric disorders and impairment. BMC psychiatry, 2020. 20: p. 1–11.
  62. 62. Alavi S.S., et al., Behavioral Addiction versus Substance Addiction: Correspondence of Psychiatric and Psychological Views. Int J Prev Med, 2012. 3(4): p. 290–4. pmid:22624087
  63. 63. Davis R.A., A cognitive-behavioral model of pathological Internet use. Computers in human behavior, 2001. 17(2): p. 187–195.
  64. 64. Mahamid F.A., Berte D.Z., and Bdier D., Problematic internet use and its association with sleep disturbance and life satisfaction among Palestinians during the COVID-19 pandemic. Current Psychology, 2022. 41(11): p. 8167–8174. pmid:34334988
  65. 65. Islamie Farsani S., et al., Some Facts on Problematic Internet Use and Sleep Disturbance among Adolescents. Iran J Public Health, 2016. 45(11): p. 1531–1532. pmid:28028510
  66. 66. Kokka I., et al., Exploring the Effects of Problematic Internet Use on Adolescent Sleep: A Systematic Review. Int J Environ Res Public Health, 2021. 18(2).
  67. 67. Younes F., et al., Internet addiction and relationships with insomnia, anxiety, depression, stress and self-esteem in university students: A cross-sectional designed study. PloS one, 2016. 11(9): p. e0161126. pmid:27618306
  68. 68. Mahmoud O.A.A., Hadad S., and Sayed T.A., The association between Internet addiction and sleep quality among Sohag University medical students. Middle East Current Psychiatry, 2022. 29(1): p. 23.
  69. 69. Johnson E.O., Roth T., and Breslau N., The association of insomnia with anxiety disorders and depression: exploration of the direction of risk. Journal of psychiatric research, 2006. 40(8): p. 700–708. pmid:16978649
  70. 70. Do med students have any free time? March 2, 2024]; Available from: https://www.quora.com/Do-med-students-have-any-free-time.
  71. 71. Bassan N., et al., Availability and use of leisure time in students of an Argentinean Medical School: a contribution pursuing wellbeing. Educational Research, 2012. 3: p. 468–472.
  72. 72. Augner C., Vlasak T., and Barth A., The relationship between problematic internet use and attention deficit, hyperactivity and impulsivity: A meta-analysis. Journal of Psychiatric Research, 2023. 168: p. 1–12. pmid:37866293
  73. 73. Wang B.-q., et al., The association between attention deficit/hyperactivity disorder and internet addiction: a systematic review and meta-analysis. BMC Psychiatry, 2017. 17(1): p. 260. pmid:28724403
  74. 74. Werling A.M., et al., Problematic use of digital media in children and adolescents with a diagnosis of attention-deficit/hyperactivity disorder compared to controls. A meta-analysis. Journal of Behavioral Addictions, 2022. pmid:35567763
  75. 75. Panagiotidi M. and Overton P., The relationship between internet addiction, attention deficit hyperactivity symptoms and online activities in adults. Comprehensive psychiatry, 2018. 87: p. 7–11. pmid:30176388
  76. 76. Tateno M., et al., Internet addiction and self‐evaluated attention‐deficit hyperactivity disorder traits among Japanese college students. Psychiatry and clinical neurosciences, 2016. 70(12): p. 567–572. pmid:27573254
  77. 77. Cakmak F.H. and Gul H., Factors associated with problematic internet use among children and adolescents with Attention Deficit Hyperactivity Disorder. North Clin Istanb, 2018. 5(4): p. 302–313. pmid:30859160
  78. 78. Dündar C. and Karabıçak C., Problematic Internet Use Associated with Attention Deficit Hyperactivity Disorder in Turkish College Students. Erciyes Medical Journal/Erciyes Tip Dergisi, 2022. 44(2).
  79. 79. Zakaria H., et al., Internet addiction and its relationship with attention deficit hyperactivity disorder (ADHD) symptoms, anxiety and stress among university students in Malaysia. PloS one, 2023. 18(7): p. e0283862. pmid:37506072
  80. 80. Zhou B., et al., Motivational but not executive dysfunction in attention deficit/hyperactivity disorder predicts internet addiction: Evidence from a longitudinal study. Psychiatry research, 2020. 285: p. 112814. pmid:32036155
  81. 81. Vural P., Yeşim U., and Kilic E.Z., Relationship between symptoms of disruptive behavior disorders and unsafe internet usage in early adolescence. Nöro Psikiyatri Arşivi, 2015. 52(3): p. 240. pmid:28360717
  82. 82. Morioka H., et al., The association between alcohol use and problematic internet use: A large-scale nationwide cross-sectional study of adolescents in Japan. J Epidemiol, 2017. 27(3): p. 107–111. pmid:28142042
  83. 83. Qeadan F., Egbert J., and English K., Associations between problematic internet use and substance misuse among US college students. Computers in Human Behavior, 2022. 134: p. 107327.
  84. 84. Dib J.E., et al., Factors associated with problematic internet use among a large sample of Lebanese adolescents. BMC Pediatrics, 2021. 21(1): p. 148. pmid:33781242
  85. 85. Müller M. and Montag C., The Relationship Between Internet Addiction and Alcohol Consumption Is Influenced by The Smoking Status in Male Online Video Gamers. Clinical Neuropsychiatry, 2017. 14(1).
  86. 86. Morioka H., et al., Associations between sleep disturbance and alcohol drinking: A large-scale epidemiological study of adolescents in Japan. Alcohol, 2013. 47(8): p. 619–628. pmid:24188738
  87. 87. Choi K., et al., Internet overuse and excessive daytime sleepiness in adolescents. Psychiatry and Clinical Neurosciences, 2009. 63(4): p. 455–462. pmid:19490510
  88. 88. Head, E. Head injuries in children linked to reduced brain size and learning difficulties 2022 [cited November 24, 2023; Available from: https://www.imperial.ac.uk/news/238052/head-injuries-children-linked-reduced-brain/.
  89. 89. Graham N.S. and Sharp D.J., Understanding neurodegeneration after traumatic brain injury: from mechanisms to clinical trials in dementia. Journal of Neurology, Neurosurgery & Psychiatry, 2019. 90(11): p. 1221–1233. pmid:31542723
  90. 90. Wang X., Li D., and Li S., Childhood trauma and problematic internet use: A meta-analysis based on students in mainland China. Frontiers in Psychology, 2023. 14. pmid:37123295