INTRODUCTION
As of 2023, the Koreans’ smartphone ownership rate was 96.5%, the average daily smartphone usage time among Koreans are 2 hours and 23 minutes [1]. Although these devices have many convenient uses in daily life and smartphone play an important role in modern life, problematic smartphone use (PSU) has many side effects [2-4]. When reliance on smartphones is high, serious interpersonal conflicts [5,6], daily role problems [7], and mental health problems [8,9] can be caused by a loss of control and problematic smartphone use. The proportion of the Korean population considered “at risk” for PSU is 20%; one out of five smartphone users have reported an overdependence. Smartphone overuse is reportedly increasing every year, and Korean perception of the severity of PSU has continued to rise over the past 3 years, with 78.8% of the respondents of one study indicating dependence on smartphones was serious [10].
Several previous studies examining the association between PSU and psychological factors—such as anxiety, depression, selfcontrol, and attention-deficit/hyperactivity disorder (ADHD)— have been conducted [11-14]. It reported that PSU is observed loss of control, preoccupation, tolerance, and withdrawal symptoms observed in behavioral addictions, with ADHD, depression, anxiety, reduced self-control, and impulsivity [11,15-17]. Poor self-control also leads to PSU [16,18,19], and self-control is reported to be a major risk factor in psychiatric symptoms.
Thus, it is essential to identify the psychological factors that can be utilized to develop technology, such as apps, to overcome PSU. Accordingly, we sought to clearly identify and understand the psychological risk factors associated with PSU, and their relationships, to help people overcome it.
Moreover, most PSU research targets adolescents and college students or young adults (20s to 30s) [20-22]. However, PSU is not limited to these individuals, and research across a wider range of age groups is needed [23-25]. Therefore, we targeted adults in their 20s to 50s, who are the most common smartphone users in Korea.
Hypotheses and research model
We developed a research model to identify psychological factors that influence PSU. The research model includes four factors: anxiety symptoms, depression symptoms, self-control, and ADHD. PSU was the dependent variable. Based on the variables, we hypothesized the following associations regarding psychological factors and PSU. The relationships between the variables that we expected to find are illustrated in Figure 1. There are two research models in this study.
Firstly, we aimed to elucidate how depression symptoms and anxiety symptoms affect PSU using a structural equation model (SEM). Anxiety and depression can be viewed as related concepts, they are also conceptualized in two ways [26]. Anxiety and depression symptoms can be viewed as related concepts and are conceptualized in two ways in the literature. One perspective examines individuals experiencing both anxiety and depression symptoms simultaneously. The second perspective focuses on individuals with depression symptoms accompanied by varying levels of anxiety symptoms [26]. Previous studies defined high baseline anxiety, or “anxious depression,” as a baseline Generalized Anxiety Disorder-7 (GAD-7) score of 10 or greater. Non anxious depression was defined as a GAD-7 score of less than 10.27 If anxiety symptoms and depression symptoms are intertwined, it was necessary to find out what pathway’s anxiety symptoms and depression symptoms lead to PSU. Based on this literature, we attempted to investigate the pathways through which depression symptoms and anxiety symptoms progress in relation to PSU. Based on this theoretical background, we developed Hypothesis 1 and 2.
H1: Depression symptoms are positively associated with anxiety symptoms.
H2: Anxiety symptoms are positively associated with depression symptoms.
Secondly, previous research has found that both depression and anxiety were significant PSU predictors [8,14,28-32]. Studies across a wide age range (from 17 to 67 years old) also report that compulsive smartphone use is associated with social interaction anxiety [24]. In this study context, anxiety symptoms and depression symptoms are positively associated with PSU. Based on this research, this study developed Hypotheses 3 and 4.
H3: Depression symptoms are positively associated with PSU.
H4: Anxiety symptoms are positively associated with PSU.
Several studies targeting different age groups have reported that people with lower self-control are more likely to be PSU [33-36]. Thus, in this research context, it assumed that the lower the selfcontrol, the higher the PSU. Based on this research, this study developed Hypotheses 5.
H5: Self-control is positively associated with PSU.
ADHD reported to be significantly associated with PSU [20-22,37]. The subjects of the study are diverse, including preschoolers, adolescents, children, college students, and young adults (20s and 30s). Based on this literature review, Hypothesis 6 was developed.
H6: ADHD is positively associated with PSU.
METHODS
Participants
We attempted to identify psychological factors that influence PSU with data from an online survey. Thus, this study collected smartphone users’ data from an anonymous web-based survey. The online survey was executed by an established polling company (dataSpring, Inc., https://www.d8aspring.com/). The survey was conducted from December 8, 2019, through January 17, 2020.
The company sent emails to an online panel; the emails included a survey link and an informed consent link. If panelists formally agreed to participate in the survey, they could complete an anonymous questionnaire. The inclusive criteria were that participants had to be adults between the ages of 20 and 59 years and be smartphone users. Responses from participants who repeated the same answer for all the survey items were deleted. In addition, we calculated the response time for the entire survey. The minimum time required was estimated to be approximately 3 minutes. We also excluded the responses of those who responded faster than the minimum time, as they were deemed “meaningless responses.” Finally, we used 691 smartphone users’ response in the analysis.
Measures
A total of five variables were used to develop the research model: PSU, anxiety symptoms, depression symptoms, selfcontrol, and ADHD (Table 1). We used PSU as the dependent variable. The four independent variables included 1) anxiety symptoms, 2) depression symptoms, 3) self-control, and 4) ADHD. After reviewing the literature on psychological factors, four variables were finally selected.
First, we used the Korean Smartphone Addiction Proneness Scale for Adults (S-Scale) to evaluate PSU [38,39]. The S-Scale has 15 questions rated on a 4-point Likert scale (Supplementary Table 1). The higher the total score, the higher the PSU. The S-Scale has three risk group based on cutoff criteria: no-risk (total score 15-39 points), potential-risk (40-43), and high risk (over 44). Second, this study employed the GAD-7 scale to evaluate anxiety symptoms [40]. The GAD-7 has seven questions rated on a 0- to 3-point Likert scale. The higher the total score, the higher the level of anxiety (Supplementary Table 2). The GAD-7 has four groups based on cutoff criteria: no anxiety (total score 0-4), mild anxiety (5-9), moderate anxiety (10-14), and severe anxiety (15-21). Third, for depression symptoms, we used the Patient Health Questionnaire-9 (PHQ-9) [41]. The PHQ-9 has nine questions rated on a 0- to 3-point Likert scale (Supplementary Table 3). The higher the total score, the higher the degree of depression. The PHQ-9 has five group based on cutoff criteria: none (total score 0-4), mild (5- 9), moderate (10-14), moderately severe (15-19), and severe (20-27). Fourth, this study measured the Self-control using the Brief Self-Control Scale (BSCS) [42]. The BSCS includes 13 questions rated on a 5-point Likert scale (Supplementary Table 4). The higher the total score, the less self-control. There is no cutoff criterion for BSCS. Lastly, we used the Adult ADHD Self-Report Scale Symptom Checklist (ASRS-v.1.1) part A to measure ADHD (Supplementary Table 5). The ASRS-v.1.1 consists of 18 criteria described in the Diagnostic and Statistical Manual of Mental Disorders (DSM)-IV-TR. Part A of the ASRS-v.1.1 consists of 6 items on a 5-point Likert scale. According to the scoring instructions, four or more marks in the darkly shaded boxes indicate that the respondent has symptoms “highly consistent with ADHD in adults” (Supplementary Table 5) [43]. Thus, we used those 6 items to assess ADHD in this study.
Statistical analysis
First, this study performed the exploratory factor analysis (EFA) to determine the discernible construct structures. We performed a principal component analysis with varimax rotation to test construct validity. Second, this study calculated Cronbach’s alpha to assess the reliability of all constructs. Third, confirmatory factor analysis (CFA) was conducted to check construct validity: construct reliability (CR) score, average variance extracted (AVE), and discriminant validity. Next, we tested the hypotheses to assess the research model.
In detail, the GAD-7, PHQ-9, and S-Scale have cutoff criteria, while the BSCS and ADHD do not. For those with cutoffs, the total score of each variable was used for analysis. There-fore, the values measured by multiple indicators were used in the SEM using a single indicator (total score). Cronbach’s alpha was calculated to measure the reliability of three variables: GAD-7, PHQ-9, and S-Scale (Supplementary Table 6). Meanwhile, the BSCS and ADHD scales, which do not have cutoffs, were subjected to EFA and CFA based on multiple indicators. Descriptive statistics and the internal consistency of the data were assessed using R. SEM was applied to identify causal relationships between the model parameters using R. We used the R package lavaan for SEMs (version 0.6-5) (http://cran.r-project.org/web/packages/lavaan/) [44]. We conducted a one-way analysis of variance using PSU as the independent variable in order to investigate potential bias and respondent characteristics.
RESULTS
Participants’ characteristics
Of the 691 respondents whose data were used for analysis, sex was nearly evenly split (female 51.5%, male 48.5%) (Table 2). The age distribution ranged from 20 to 59. Nearly half of the respondents were married (49.5%), and most (82.1%) had education levels University graduation or above (including college students). The largest group by salary (35.6%) was those in the income range of 1,825.82 to 3,651.63 USD per month. Regarding the respondents’ residence areas, the largest group were those living in Seoul (32.6%). For the majority of participants (69.5%), the overall survey scores indicated normal smartphone use; 54.6% of respondents were in the “normal” group for anxiety symptoms; 41.4% indicated having no symptoms of depression.
Reliability and validity
First, these are the results of the EFA and a principal component analysis of self-control and ADHD. Two factors emerged with non-cross construct loadings greater than 0.07. Two factors explain 65.098% of the total variance. Communality ranged from 0.593 to 0.718, with all items achieving the 0.50 threshold. Thus, two constructs are distinct uni-dimensional scales (Table 3).
Second, this was the result of reliability of all constructs using Cronbach’s alpha value. The internal consistency values for all constructs were significant: 0.801 for self-control, 0.758 for ADHD, 0.909 for PSU, 0.908 for depression symptoms, 0.922 for anxiety symptoms. The internal consistency values for the two constructs were greater than 0.70.45 Thus, all variables were proven reliable.
Third, this was results of convergent validity and reliability: AVE, CR scores, and discriminant validity. The CFA were as below: chi-square/degree of freedom (χ2/df=1.582), comparative fit index (CFI=0.995), Tucker-Lewis index (TLI=0.992), and root mean square error of approximation (RMSEA=0.029). All values met the recommended threshold [46]. The convergent validity of each construct was evaluated based on AVE and CR. All variables had an AVE loading greater than 0.50 (0.503 of self-control and 0.527 of ADHD) [47]. Two variables had CR scores greater than 0.70 (0.793 of self-control and 0.758 of ADHD) [48]. Thus, we used four of the 13 items for self-control and three of 6 items for ADHD.
Finally, this was results of the discriminant validity of selfcontrol and ADHD. We can achieve discriminant validity when the square root of the AVE for each construct is greater than the correlations between each construct and the other constructs [46]. The square root of the AVE for each construct was greater than the corresponding inter-construct correlations (Supplementary Table 7).
Hypothesis testing
Firstly, this is hypothesis testing results of the proposed model-1. We assessed structural model fit using general goodnessof-fit measures: χ2/df (chi-square/degree of freedom)=2.449, CFI=0.986, TLI=0.979, and RMSEA=0.046. The value of χ2/df is acceptable when it is 3 or less. CFI and TLI are acceptable when it is 0.9 or more. RMSEA is acceptable when it is 0.08 or less. The overall values of model_1 were within acceptable range.
The structural path diagram supported the positive relationships described in four of the hypotheses (Figure 2). The structural path diagram supported the relationships postulated for H1, H3, H5, and H6 (Supplementary Table 8). As predicted, H1 was supported: depression symptoms are positively associated with anxiety (β=0.939, p<0.001). H3, depression symptoms are positively associated with PSU, was supported by a positive path coefficient (β=0.451, p<0.01). H5, the hypothesis that self-control is positively associated with PSU, was supported by a positive path coefficient (β=0.261, p<0.001). H6, the hypothesis that ADHD is positively associated with PSU, was supported by a positive path coefficient (β=0.242, p<0.001). However, H4 was rejected.
Secondly, this is hypothesis testing results of the research model_2. We assessed structural model fit using general goodness-of-fit measures: χ2/df=2.995, CFI=0.981, TLI=0.971, and RMSEA=0.054. Thus, the overall values also were within acceptable ranges. The structural path diagram supported the positive relationships described in four of the hypotheses (Figure 2). The structural path diagram supported the relationships postulated for H2, H3, H5, and H6 (Supplementary Table 8). H2 was supported: anxiety symptoms have a positive effect on depression (β=0.944, p<0.001). H3 also was supported by a positive path coefficient (β=0.520, p<0.001). H5 also was supported by a positive path coefficient (β=0.286, p<0.001). H6 also was supported by a positive path coefficient (β=0.253, p<0.001). However, H4 was also rejected in research model_2.
DISCUSSION
We attempted to find psychological risk factors of PSU among smartphone users in their 20s to 50s. We sought to identify psychological factors that could be leveraged to develop technologies to overcome PSU. Some interesting results influence PSU are as follows.
First, our results strongly suggest that PSU has the characteristics of behavioral addiction. All factors have been reported to be related to behavioral addiction as well as traditional addiction issues, such as substance use and gambling disorders [49-52]. However, there is still controversy about determining PSU as an addiction. Nevertheless, from the standpoint of addiction in terms of PSU, it would at least be practical to use IT technology to help people who use their smartphones excessively.
Second, we found that anxiety symptoms lead to PSU through depression symptoms. In other words, anxiety symptoms do not directly affect PSU but affect it through depression symptoms. According to previous research, PSU are observed loss of control, preoccupation, tolerance, and withdrawal symptoms observed in behavioral addictions, with depression, anxiety, reduced self-control, and impulsivity [17,53,54]. Our finding differs from previous studies that have reported anxiety is a risk factor for PSU [55]. Additionally, our results showed the depression symptoms is a predictor of PSU, as seen in other studies [25,56]. According to previous studies, depression and anxiety were positive predictors of PSU, with depression score being a more powerful predictor than anxiety [11]. The relationship between depression and anxiety and smartphone dependence can be approached in two ways. The first approach is to identify if problematic smartphone overuse has occurred due to smartphones being used as a means to relieve negative moods, such as depression or anxiety [55]. The second approach is to recognize that excessive use of smartphones leads to anxiety, depression, and sleep problems [43]. Our results support the first approach. In these cases, breaking the relevant mechanism is necessary. Thus, these results are very meaningful clinically. In developing technologies to prevent and manage PSU, it is important to develop content or services that can relieve depression symptoms and anxiety symptoms.
Third, another interesting result is that the link between anxiety symptoms and depression symptoms is very strong. According to previous studies, some factors—avoidance, close relationships, group relationships, sociability, and interpersonal hypersensitivity—are mediators in the relationship between anxiety and later depressive symptoms [57-59]. Additionally, avoidance and interpersonal relationships have been reported to mediate the relationship between early anxiety and depression, and this relationship applies to both symptoms and disorders [12]. In future research, it will be of great clinical significance to deal with avoidance, sociality, and interpersonal hypersensitivity, among other factors, for PSU. A therapeutic approach to social and interpersonal hypersensitivity may encourage functional and healthy smartphone use rather than PSU.
Fourth, consistent with previous studies, the lower an individual’s self-control, the more dependent he/she is on the smartphone [33,60]. Additionally, the higher the ADHD, the more dependence on the smartphone [61,62]. Thus, self-control and ADHD are important risk factors to address in IT technology to help people who use their smartphones excessively.
Finally, we used single indicators of scales such as GAD-7 and PHQ-9 in SEM and confirmed good model fit. For psychological risk factors such as anxiety symptoms and depression symptoms, a total score is often used to assess mental symptoms and disorders. Some previous studies were conducted after carrying out validation and reliability verification with detailed items of each scale and performing SEM analysis with some remaining items [63]. However, in these cases, we can omit some items we want to measure for each factor. Additionally, if the total score is already used when diagnosing patients in psychiatric counseling, it is also possible to consider using the total score of the scale. Scales such as the GAD-7 and PHQ-9 are considered the gold standard of clinically valid scales. It is very wasteful to remove some items for SEM analysis, and using the total score will help researchers properly reflect symptoms. It is desirable to use a single indicator to include as much information as possible and examine the causal relationship; thus, we conducted SEM using the total score of the scale to evaluate each psychological factor as a single indicator [64]. In the marketing field, SEM has long been conducted using a single indicator [65]. There is still controversy over whether multiple or single indicators are best [66], but this can be adapted according to the research situation. This study can be viewed as an example of a structural model that can be used in such cases. We consider this model useful for researchers who have similar concerns.
Although there are meaningful results from this research, there are some limitations. First, we used the term “PSU,” although there are diverse related terms: “excessive smartphone usage,” “smartphone overdependence,” and “smartphone addiction [67].” The choice of term depends on the researcher or research environment. However, Panova and Carbonell [68] suggest that behavior can have similar expressions to addiction in terms of overuse, impulse control problems, and negative consequences, but that does not mean it should be considered an addiction. They also proposed using terms such as “problematic use” when conducting technical behavior research. If there is an accepted unification of terms in the future, it would be desirable to interpret problematic use according to those standardized terms. Second, we used Part A (6 items) of the ASRS-v1.1 to measure ADHD. In the case of ADHD, the score was reflected cross-sectionally, and the onset time point included in the DSM-5 diagnostic criteria for ADHD were not included in this study. An additional clinician’s evaluation is required to accurately determine adult ADHD. Third, data were collected using a self-response survey. Because the data were not collected under the evaluation of a psychiatrist, it may include errors or mistakes by the respondents. For future studies, it would be more desirable to have verified accurate data accompanied by evaluations from a psychiatrist.
Fourth, in this study, the total score was used only when the measurement items had a total score and cutoff points, and in other cases (BSCS and ASRS-v1.1 Part A), EFA and CFA were conducted and then applied to the SEM. ASRS has 6 items in Part A and Part B. In addition, Part B provides additional clues. If 4 or more items are marked in the darkly shaded boxes of the main 6 items, it can be seen as showing symptoms very similar to ADHD. However, when scoring the checked areas, there are cases where the scores for non-ADHD cases are higher than those for ADHD symptoms. Therefore, it was judged that there are limitations in converting ASRS to a total score and using it in structural equation modeling. Since the 6 questions in Part A best predict the disease and are most suitable for use as a screening tool, the reliability and validity of the 6 items in Part A were reexamined and used. In social science, even for measurements that have been previously proven, the research model is verified after conducting EFA and CFA depending on the subject. Therefore, this study also tried to analyze these items by applying them to the SEM, going beyond the limitations of the application of these items.
Fifth, in this study, anxiety symptoms are not a risk factor for PSU. These results differ from previous studies that reported anxiety is a risk factor for PSU. The discrepancy in previous research results is likely due to sample specificity. The anxiety scores of our study subjects were skewed toward the normal group. The total score on the GAD-7, which measures anxiety symptoms, ranged from 0 to 21. In our data, the mean was 5, the median was 4, and the mode was 0. As shown in Table 2, only 39 (5.6%) of the 691 subjects scored at the severe level (scores 15-21). Future studies should be conducted with a proportion of anxiety symptom groups.
This study identified psychological factors and relationships among them that can prevent overdependence on smartphones among smartphone users in their 20s to 50s: anxiety symptoms, depression symptoms, self-control, and ADHD. In particular, the link between anxiety symptoms, depression symptoms, and dependence on smartphones was revealed in detail. Key psychological factors can be utilized in developing an app service and other programs to mediate dependence on smartphones.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Supplement1
Print
