Longitudinal influences of depressive moods on problematic mobile phone use and negative school outcomes among Korean adolescents

Abstract

The study aimed to examine the longitudinal causal relationships of depressive moods, problematic mobile phone use, and negative school outcomes based on the cognitive-behavioral model among Korean adolescents. The changes within each construct over time were also explored. A total of 1,610 valid responses from three-year longitudinal data from the Korean Children and Youth Panel Survey were analysed and multivariate latent growth modeling was used. Depressive moods, problematic mobile phone use, and negative school outcomes at earlier ages each increased in severity across the three years. Initial levels of depressive moods increased initial levels of problematic mobile phone use and negative school outcomes, including changed rates of negative school outcomes. Additionally, changed rates of depressive moods positively predicted changed rates of problematic mobile phone use and negative school outcomes. Lastly, initial levels and changed rates of problematic mobile phone use predicted initial levels and changed rates of negative school outcomes respectively. Implications of the findings in the context of adolescents' psychological problems and problematic mobile phone use are discussed.

Keywords

cognitive-behavioral model depressive moods Korean adolescents longitudinal data multivariate latent growth modelling negative school outcomes problematic mobile phone use

Problematic mobile phone use (PMU) has become common among adolescents. A Korean national survey in 2016 showed that 30.6% of teenagers were dependent users of mobile phones and had experienced adverse consequences due to excessive use of and loss of control over cell phones. The ratio of dependent users among teenagers, 30.6%, was much higher than that of adults: 19% of people in their 20s, 13.3% of those in their 30s, and 11.8% of those in their 40s (Ministry of Science, ICT and Future Planning of Korea & National Information Society Agency of Korea, 2017). The problem is that PMU is related to many adverse effects in their lives including health problems, delinquencies, and psychological and behavioral problems (Hardell, 2017; Livingstone & Smith, 2014; Vernon, Modecki, & Barber, 2018).

In particular, depressive moods are salient predictors of PMU (Bickham, Hswen, & Rich, 2015; Jun, 2016), which ultimately results in adverse school outcomes (Sánchez-Martínez & Otero, 2009). Prior studies have shown that depressive symptoms of teenage students increase PMU (Augner & Hacker, 2012; Ge, 2014; Kim, Seo, & David, 2015) and maladaptive school outcomes (Fröjd, Nissinen, Pelkonen, Marttunen, Koivisto, & Kaltiala-Heino, 2008; McCarthy, Young, Benas, & Gallop, 2018). PMU increases negative school outcomes as well (Egger, Costello, & Angold, 2003; Owens, Stevenson, Hadwin, & Norgate, 2012). More generally, depressive feelings are often precursors to unhealthy and unsuccessful development among adolescents (Webb, Panico, Bécares, McMunn, Kelly, & Sacker, 2017), and adverse school outcomes are typical problems among teenage students (Werner-Seidler, Perry, Calear, Newby, & Christensen, 2017).

On the basis of previous findings, this study proposed causal relationships among depressive feelings, PMU, and negative school outcomes. However, few studies have attempted to reveal the unified relationships among the constructs based on a theoretical model. Therefore, it is necessary to use a theoretical model that allows understanding of the causes and consequences of PMU. Thus, we applied the cognitive-behavioral model of problematic Internet use (PIU) (Caplan, 2002; Davis, 2001) to PMU. According to Davis's early work (2001), the cognitive-behavioral model of PIU suggests that individuals with psychopathological problems are likely to experience cognitive and behavioral symptoms of PIU, leading to adverse outcomes. Subsequent studies from Caplan (2003, 2005) demonstrated that social skill deficit and preference for online social interactions, rather than face-to-face interactions, increases PIU, which results in troubles at work, school or in the family. Caplan (2010) demonstrated that a preference for online social interaction and mood regulation predicts the increase of deficient self-regulation, which leads to adverse outcomes. For example, individuals with severe psychological distress including depression or anxiety are likely to experience symptoms of PIU (Caplan & High, 2011), resulting in difficulties in individuals' management of their own lives (Kim, LaRose, & Peng, 2009).

Concerning teenagers, Gámez-Guadix, Villa-Georg, and Calvete (2012) confirmed the cognitive-behavioral model of PIU among Mexican adolescents and indicated that the preference for online social interaction caused by depression and anxiety increased PIU, which, consequentially, was significantly influenced adverse life outcomes. A longitudinal study among Spanish teenagers demonstrated that deficient self-regulation at T1 positively affected negative consequences of the Internet and problematic alcohol use at T2, six months apart (Gámez-Guadix, Calvete, & Hayas, 2015).

PMU, similar to PIU, is a type of technology addiction that is representative of modern society. Griffiths (1999) introduced technology addiction as a concept of behavioral addiction rather than the concept of pathological addiction including alcoholism and drug addiction. Subsequent studies have defined PMU as an obsession with mobile phones, including withdrawal symptoms and a decreased tolerance for daily life upon separation, due to mobile phone overuse and dependence (Bianchi & Phillips, 2005; Jun, 2014). Thus, like PIU, PMU involves cognitive and behavioral disorders, and by testing a mediating effect of PMU on depressive moods and negative school outcomes based on the cognitive-behavioral model, the associations among etiology, development, and consequences of PMU are expected to be revealed.

In addition, the current study explored the longitudinal causal associations of the constructs among Korean adolescents to reveal teenagers' growth over time. A few previous studies identified that each occurrence of depressive moods, PMU, and maladaptive school outcomes increase the seriousness of future conditions over time (Bowen, Jenson, & Clark, 2004; Jun, 2014, 2016; Rushton, Forcier, & Schectman, 2002). However, to our knowledge, there are no longitudinal studies that analyse the relationships among depressive moods, PMU, and negative school outcomes based on a theoretical model.

This study aimed to test the longitudinal effects of depressive moods on PMU and negative school outcomes among Korean adolescents based on the cognitive-behavioral model. For example, we proposed causal associations between initial levels and changed rates of depressive feelings and those of PMU over time. We also added a novel path, from depressive moods to negative school outcomes, to the cognitive-behavioral model, to test if PMU mediates depressive feelings and adverse school outcomes partially or fully. Prior to testing longitudinal relationships among the constructs, the study explored the changes within each construct, namely depressive feelings, PMU, and negative school outcomes over time. In this context, Gámez-Guadix and his colleagues (2015) identified the temporal and reciprocal relations between PIU and negative consequences among Spanish teenagers and Jun (2015) showed that PMU and depressive symptoms of Korean adolescents became continuously worsened over time and they had bidirectional relationships across the three years. Based on previous studies, this study developed two hypotheses: (a) depressive moods, PMU, and negative school outcomes at an earlier time positively predicts an increase in the respective construct at a later time; and (b) depressive moods increase PMU and negative school outcomes, and PMU increases negative school outcomes over time among adolescents.

Methods

Participants and procedure

The current study used three-year longitudinal data from the Korean Children and Youth Panel Survey (KCYPS). The KCYPS was established in 2010 and is conducted every year by the National Youth Policy Institute (NYPI) in Korea. The NYPI is the only national youth research institute in Korea and provides primary statistical data for youth research and policies through comprehensive panel surveys. The KCYPS used stratified multistage cluster sampling based on 2009 school statistics collected by the Ministry of Education in Korea. Participants responded to surveys conducted by experienced interviewers at their homes in the context of parental prior written consent. The initial number of participants in 2010 was 2,378, and the surveys were conducted annually with the same sample without sample substitution. We used data from the third (2012; T1), fourth (2013; T2), and fifth (2014; T3) years of the KCYPS and the number of participants per year was 2,219 in 2012, 2,092 in 2013, and 2,070 in 2014. Only participants who had used mobile phones successively during the three years, from 2012–2014, were included in the current study and there were no other inclusion or exclusion criteria. Finally, data from a total of 1,610 adolescents (790 girls and 820 boys) were used. The participants were sixth graders in 2012, and their average age was 11.94 years (SD = 0.25).

Measures

In this study, the scales used in KCYPS were used at all-time points. Participants responded to questionnaires through face-to-face interview surveys. All questionnaire items of the primary variables used in this study are as follows.

Depressive moods

Depressive moods were measured by coding the three items established by Kim, Lim, Kim, Park, Yoo, Choi, and Lee (2006), as used in the KCYPS: ‘I am not happy to be alive’, ‘I have many worries’, and ‘I think my life is depressing’. Each item was rated on a scale from 1 (never true) to 4 (always true). The Cronbach's alpha coefficients were 0.860 at T1, 0.839 at T2, and 0.813 at T3.

Problematic mobile phone use

PMU was measured using the seven items established by Lee, Kim, Nah, Lee, Kim, and Bae (2002), as used in the KCYPS: ‘I am worried when I do not hear from anybody on my mobile phone’, ‘I cannot stand being bored without a mobile phone when I am alone’, ‘I am using my mobile phone more often’, ‘I feel nervous without a mobile phone when I go out’, ‘I feel isolated when I do not have a mobile phone’, ‘I do not notice how time flies when I am using a mobile phone’, and ‘I cannot live without a mobile phone’. Each item was rated on a scale from 1 (never true) to 4 (always true). The Cronbach's alpha coefficients were 0.890 at T1, 0.893 at T2, and 0.881 at T3.

Negative school outcomes

This study measured negative school outcomes by coding the four items established by Min (1991) with reference to the school adjustment scales of Yoon and Moon (1977) and Lee (1990), as used in the KCYPS: ‘I do not like school classes’, ‘I do not complete my homework’, ‘I do not understand what I learned in classes’, and ‘I do not ask teachers or friends when I do not understand’ These items were rated on a scale of 1 (never true) to 4 (always true). The Cronbach's alpha coefficients for these items were 0.745 at T1, 0.799 at T2, and 0.793 at T3.

Statistical analysis

To examine the changes over time in depressive moods, PMU, and maladaptive school outcomes as well as the longitudinal associations among the three constructs, multivariate latent growth modeling was performed in this study. Latent growth modeling (LGM) is an advanced statistical method conceptualized by the structural equation modeling framework, and analyses univariate trajectories of longitudinal variables, which means the changes within individuals over time, and associations of changes in single or multiple longitudinal variables (Duncan & Duncan, 2009; Muniz-Terrera, Robitaille, Kelly, Johansson, Hofer, & Piccinin, 2017). LGM estimates means of latent variables consisting of an intercept and a slope and path coefficients between constructs: An intercept describes the mean of a construct at the initial time, and a slope illustrates the changed rate of the mean of a construct over time (Duncan, Duncan, Strycker, Li, & Alport, 1999). Multivariate LGM can test how a change in one longitudinal variable influences the change in other longitudinal variables and provides a dynamic view of the associations between changes in multiple processes (Duncan, Duncan, & Strycker, 2001). In multivariate LGM, modeling of intercept-intercept, intercept-slope, and slope-slope associations between constructs is examined (Kim, Kim, & Hong, 2007). Overall, the path-diagram of our multivariate LGM is represented in Figure 1. AMOS ver. 20.0 was used to perform multivariate LGM.

Figure 1.

Multivariate latent growth model.

Results

The validity of all main variables was ascertained. Confirmatory factor analysis was conducted on the three key variables at each time point. Model fits were acceptable: χ² = 708.190, df = 74, TLI = 0.912, CFI = 0.938, RMSEA = 0.073 at T1; χ2 = 800.859, df = 74, TLI = 0.918, CFI = 0.933, RMSEA = 0.078 at T2; and χ2 = 675.201, df = 74, TLI = 0.923, CFI = 0.938, RMSEA = 0.071 at T3. We applied three fit indexes to evaluate the model fit: The values of CFI and TLI over 0.9 and close to 0.95, the values of RMSEA under 0.08 and below 0.05 were interpreted as indicating acceptable fit (Browne & Cudeck, 1993; Hu & Bentler, 1998). However, these cut-off values should be applied with caution because they are often not met in more complicated studies (Kim, Hong, & Choo, 2007; Marsh, Hau, & Wen, 2004). All items of the constructs had 0.5 or higher factor loadings in every time: Depressive moods (0.686 ∼ 0.915), PMU (0.584 ∼ 0.840), and negative school outcomes (549 ∼ 0.835). AVE values were more than 0.5 and C.R values were more than 0.8 at T1 ∼ T3: depressive moods (AVE = 0.763 ∼ 0.800, C.R = 0.901 ∼ 922), PMU (AVE = 0.564 ∼ .591, C.R =0.893 ∼ 0.909), and negative school outcomes (AVE = 0.573 ∼ 0.660, C.R =0.841 ∼ 0.885), which represented that the convergent validity of the construct was ascertained. Next, a multi-time invariance test was conducted to ensure the validity of longitudinal data. The multi-time invariance represents the equivalence of the measurement parameters over time and consists of three invariances: (a) weak factorial invariance with the equivalence of factor loadings over time (i.e., metric invariance), (b) strong factorial invariance with the equality of intercepts across time in addition to the same factor loadings (i.e., scalar invariance), and (c) strict factorial invariance adding the equality of error variances over time to the invariant factor loadings and intercepts (i.e., error variance invariance). As presented in Table 1, full metric invariance was established for the three key variables: Depressive moods (Δχ2 = 5.699, Δdf = 4, p > 0.05, ΔTLI = 0.001, ΔCFI = 0.000, ΔRMSEA = -0.003), problematic mobile phone use (Δχ2 = 92.599, Δdf = 12, p < 0.001, ΔTLI = 0.003, ΔCFI = −0.005, ΔRMSEA = −l0.001), and negative school outcomes (Δχ2 = 5.868, Δdf = 6, p > 0.05, ΔTLI = 0.001, ΔCFI = 0.000, ΔRMSEA = −l0.002). Full scalar invariance of depressive moods (Δχ2 = 148.883, Δdf = 6, p < 0.001, ΔTLI = −l0.030, ΔCFI = −0.018, ΔRMSEA = 0.034) and partial scalar invariance of the other variables were established: Problematic mobile phone use (Δχ2 = 353.994, Δdf = 6, p < 0.001, ΔTLI = −l0.009, ΔCFI = −l0.008, ΔRMSEA = 0.006) and negative school outcomes (Δχ2 = 10.443, p < 0.05, Δdf = 4, ΔTLI = −0.001, ΔCFI = −0.001, ΔRMSEA = 0.002). Error variance invariance for the three main variables was achieved: Depressive moods (Δχ2 =4.596, Δdf = 6, p > 0.05, ΔTLI = 0.008, ΔCFI = 0.000, ΔRMSEA = −0.007), problematic mobile phone use (Δχ2 = 7.091, Δdf = 12, p > 0.05, ΔTLI = 0.002, ΔCFI = 0.002, ΔRMSEA = −0.011), and negative school outcomes (Δχ2 = 82.36, Δdf = 8, p < 0.001, ΔTLI = −0.015, ΔCFI = −0.012, ΔRMSEA = .014).

Table 1.

Model fit indices of measurement invariance models.

Model		χ²	df	TLI	CFI	RMSEA
DM	Configural invariance	44.474	15	0.989	0.996	0.035
	Full metric invariance	50.173	19	0.990	0.996	0.032
	Full scalar invariance	199.056	25	0.960	0.978	0.066
	Full error variance invariance	203.652	31	0.968	0.978	0.059
PMU	Configural invariance	1617.614	165	0.910	0.935	0.059
	Full metric invariance	1710.213	177	0.913	0.930	0.058
	Full scalar invariance	2856.421	197	0.857	0.882	0.075
	Partial scalar invariance	2064.207	183	0.904	0.922	0.064
	Full error variance invariance	2071.298	195	0.906	0.924	0.053
NSO	Configural invariance	61.241	39	0.993	0.997	0.019
	Full metric invariance	67.109	45	0.994	0.997	0.017
	Full scalar invariance	597.111	56	0.884	0.917	0.077
	Partial scalar invariance	77.552	49	0.993	0.996	0.019
	Full error variance invariance	159.912	57	0.978	0.984	0.033

Note: DM represents depressive moods; PMU represents problematic mobile phone use; NSO represents negative school outcomes. All χ² values were significant at p < .001.

The mean values of all main variables at T1, T2, and T3 showed values of around 2 points of 4 points: Depressive moods (M = 1.78, 1.86, 1.96; SD = 0.67, 0.63, 0.59), PMU (M = 2.06, 2.26, 2.28; SD = 0.71, 0.70, 0.66), and negative school outcomes (M = 2.01, 2.07, 2.10; SD = 0.56, 0.60, 0.56).

In advance of fitting a multivariate LGM, we fit the single-construct LGMs for depressive moods, PMU, and negative school outcomes respectively, and we examined linear changes of the constructs over time. First, the single-construct LGM for depressive moods fit the data well: χ2 = 10.769 (df = 3, p < .05), CFI = 0.991, TLI = 0.991, RMSEA = 0.040. The means of intercept and slope were positively significant (Mi = 1.773, SEi = 0.016, p < 0.001; Ms = 0.093, SEs = 0.009, p < 0.001). It showed the initial levels of respondents' depressive moods were between low and moderate, and there was an increase in depressive mood levels over time. Additionally, there were significant variances in both intercept and slope (Vi = 0.246, SEi = 0.016, p < 0.001; Vs = 0.023, SEs = 0.006, p < 0.001), revealing individual differences in initial levels as well as changes in depressive moods across time.

Next, the single-construct LGM for PMU had an acceptable model fit, χ2 = 63.426 (df = 3, p < 0.001), CFI = 0.920, TLI = 0.920, RMSEA = 0.082. The average intercept and slope for PMU were positively significant (Mi = 2.094, SEi = 0.017, p < 0.001; Ms = 0.104, SEs = 0.010, p < 0.001). The findings indicated that the initial levels of participants' PMU were between moderate and high, and the levels of PMU have increased over time. The variance of intercept was significant (Vi = 0.227, SEi = 0.018, p < 0.001), but the variance of slope was not significant (Vs = 0.012, SEs = 0.007, p > 0.05). These findings showed that there were individual differences in initial levels of PMU, but not in changes in PMU across time. Lastly, the single-construct LGM for negative school outcomes fit the data well, χ2 = 27.184 (df = 3, p < 0.001), CFI = 0.970, TLI = 0.970, RMSEA = 0.061. We found positively significant means of intercept and slope (Mi = 2.019, SEi = 0.014, p < 0.001; Ms = 0.044, SEs = 0.008, p < 0.001). These results demonstrated that the levels of participants' negative school outcomes were between moderate and high in the beginning, and their levels of negative school outcomes increased across time. The variance of intercept was significant (Vi = 0.227, SEi = 0.018, p < 0.001), but the variance of slope was not significant (Vs = 0.012, SEs = 0.007, p > 0.05). The findings showed that there were individual differences in initial levels of negative school outcomes, but no individual differences in the change in negative school outcomes over time.

Based on the cognitive-behavioral model, this study fit a model including depressive moods, PMU, and negative school outcomes and modeled paths between the intercepts and slopes of the constructs as shown in Figure 1. The study demonstrated standardized coefficients as well as bootstrapped 95% confidence intervals based on 2,000 iterations. To decide a final approved model of this study, we compared the fits of two models: Model 1, with a direct path from depressive moods to negative school outcomes, and Model 2, without the direct path between them. We found that Model 1 had better fit indices (χ2 = 225.707, df = 23, TLI = 0.931, CFI = 0.937, RMSEA = 0.074) than Model 2 (χ2 = 421.941, df = 26, NFI = 0.871, TLI = 0.830, CFI = 0.878, RMSEA = 0.097) and the difference of χ2 was statistically significant (Δχ2 = 196.234, Δdf = 3. P < 0.001). Thus, the study selected Model 1 as the final approved model of the study shown in Figure 1.

The longitudinal associations among the three constructs are reported in Table 2. With regard to the paths among the intercepts, the intercept of depressive moods positively affected the intercept of PMU (β = 0.297, p < 0.001) and the intercept of PMU positively affected the intercept of negative school outcomes (β = 0.179, p < 0.001). These findings demonstrated that the initial levels of PMU of respondents who were originally depressed were high, and their levels of negative school outcomes were also high. The result was consistent with the cognitive-behavioral model. Additionally, the intercept of depressive moods positively affected the intercept of negative school outcomes (β = 0.397, p < 0.001). Then, PMU partially mediated between initial levels of depressive moods and negative school outcomes.

Table 2.

Structural path coefficients in the final approved model.

Parameter	B	SE	CR	β
DM intercept → PMU intercept	0.297	0.039	7.567***	0.332
DM intercept → PMU slope	0.087	0.049	1.754	0.125
DM slope → PMU slope	0.349	0.064	5.460***	0.450
PMU intercept → NSO intercept	0.179	0.037	4.819***	0.250
PMU intercept → NSO slope	−0.058	0.050	−1.164	−0.079
PMU slope → NSO slope	0.253	0.112	2.256*	0.265
DM intercept → NSO intercept	0.397	0.033	11.865***	0.620
DM intercept → NSO slope	0.126	0.049	2.591*	0.191
DM slope → NSO slope	0.502	0.082	6.144***	0.676

*p < 0.05; ***p < 0.001.

Regarding the paths between intercepts and slopes, the intercept of depressive moods positively influenced the slope of negative school outcomes, β = 0.126, p < 0.05. Thus, negative school outcomes of participants with high levels of depressive moods initially increased further over time. The paths that were depressive moods intercept → PMU slope, and PMU intercept → negative school outcomes slope, were not statistically significant.

Concerning the paths among slopes, the slope of depressive moods positively influenced the slope of PMU (β = 0.349, p < 0.001), and PMU positively influenced the slope of negative school outcomes (β = 0.253, p < 0.05). The findings showed that participants who experienced increased depressive moods over time increased their PMU more over time to deepen negative school outcomes more over time, supporting the cognitive-behavioral model from a longitudinal perspective. Additionally, the slope of depressive moods positively affected the slope of negative school outcomes (β = 0.502, p < 0.001), indicating that PMU partially mediated the change in depressive moods and negative school outcomes.

In addition, the study found that the indirect effect of depressive moods on negative school outcomes through PMU was smaller than the direct effect of depressive moods on negative school outcomes, as shown in Table 3. Furthermore, the indirect effect of the intercept of depressive moods on the slope of negative school outcomes was not significant compared to the direct effect of the associations, which was significant.

Table 3.

Standardized estimates of the direct, indirect, and total effects of depressive moods on negative school outcomes.

Parameter (SE)	Direct effect	Indirect effect	Total effect
DM intercept → NSO intercept	0.620 (0.052)*	0.083 (0.025)***	0.703 (0.034)**
DM intercept → NSO slope	0.191 (0.052)*	0.007 (0.034)	0.198 (0.072)**
DM slope → NSO slope	0.676 (0.052)**	0.119 (0.042)**	0.795 (0.048)**

*p < 0.05; **p < 0.01; ***p < 0.001.

Discussion

The present study examined the longitudinal causal relationships among depressive moods, PMU, and negative school outcomes based on the cognitive-behavioral model. We adopted the longitudinal data of 1,610 Korean students over three years and used multivariate latent growth modeling. The study indicated linear changes in depressive moods, PMU, and negative school outcomes, finding associations between the initial levels and changed rates of the constructs across the three years.

The findings showed that depressive feelings, PMU, and negative school outcomes increased persistently across the three years. These results support the previous studies showing that depressive moods (Rushton et al., 2002), PMU (Jun, 2014, 2016), and negative school outcomes (Bowen et al., 2004) worsen over time. In particular, PMU increased more in middle school students (seventh and eighth graders) than in elementary school students (sixth graders) due to mental vulnerabilities and excessive academic stress in Korea (Kim, 2012). The initial levels of PMU and negative school outcomes were between moderate and high, and there were no individual differences in changes in PMU and negative school outcomes over time. This means that most students are likely to experience PMU and negative school outcomes, and the levels of PMU and negative school outcomes, regardless of their initial levels, increase after entering middle school. Therefore, PMU and negative school outcomes should be regarded as prevalent problems for teenage students. Supervision of cell phone use and school lives is needed in families and schools for all adolescents, not just for some, to protect them from increases in PMU and negative school outcomes. Additionally, individual differences in the initial levels of depressive moods and changes over time were found in the respondents whose depressive mood levels were between low and moderate. These results suggest that some teenage students are more vulnerable to depressive feelings than others. Thus, it is necessary to identify the characteristics of the vulnerable teenagers and why they feel more depressive moods than others. Also, there is a need for particularly intensive attention and care for them, beyond what is offered to the youth in general.

It was shown that depressive moods positively predicted PMU, and PMU positively predicted negative school outcomes. Additionally, the study verified that the changed rates of depressive moods increased those of PMU, which led to an increase in the changed rates of negative school outcomes. Thus, it is important to focus on teenagers whose levels of depressive moods and PMU are rapidly increasing, even if their current levels are not high. These results resemble the prior studies illustrating causes, depressive symptoms (Bickham et al., 2015; Jun, 2016; Webb et al., 2017) and consequences, including maladaptive school outcomes (Sánchez-Martínez & Otero, 2009), of PMU. Also, these findings were in accordance with the cognitive-behavioral model among teenagers, which identified the positive relations among psychological distress, PIU, and troubles at school and family (Gámez-Guadix et al., 2012; Gámez-Guadix et al., 2015). According to the results of the study and the cognitive-behavioral theory, the more teenagers feel depressed, the more they interact with others in the virtual world using the Internet and mobile phones, which disturbs their daily lives.

This study revealed that depressive feelings directly increased adverse school outcomes and PMU partially mediated between depressive moods and negative school outcomes, which have not been verified in previous studies on the cognitive-behavioral model (Davis, 2001; Caplan, 2002, 2003, 2005, 2010). In order to reduce negative school outcomes, parents and teachers should make efforts to relieve not only PMU but also depressive feelings. Particularly, the direct paths from depressive feelings to negative school outcomes, in the initial levels and the changed rates, were significant, as mentioned above, and the direct effect of depressive moods on adverse school outcomes was bigger than the indirect effect through PMU. Thus, it is most urgent and essential for students to alleviate depressive feelings.

The study has several limitations. First, this study used longitudinal data for early teenagers aged between 12- and 14-years-old and found stable changes in depressive moods, PMU, and negative school outcomes. However, we did not explore data for adolescents of other ages such as high school students. Because the Korean national survey showed that the levels of PMU of high school students were different from those of middle school students (Ministry of Science, ICT, and Future Planning of Korea & National Information Society Agency of Korea, 2017), we suggest that follow-up studies should use data for the entire adolescent population. Second, we used the national secondary data consisting of the measures, which were (a) developed for Korean teenagers, and (b) not well-known and widely used internationally. Thus, future studies which verify the cognitive-behavioral model for PMU among adolescents in other countries are needed, using internationally validated measures in psychological well-being, PMU, and negative outcomes. Third, the dependent variables in the study are negative school outcomes and we explored the effects of depressive moods and PMU on negative school outcomes. However, in addition to school, friends and family are important variables in teenagers' daily lives. In the cognitive-behavioral theory, troubles at work, school, and family were dependent variables. Therefore, follow-up studies should identify the effects of depressive moods and PMU on negative consequences of family and friends among adolescents.

Despite these limitations, this study contributes to enhancing our understanding of the cognitive and behavioral process of the way in which depressive moods predict negative school outcomes through PMU over time. Specifically, this pioneering study is noteworthy in applying a cognitive-behavioral model to PMU theoretically: First, while PIU was limited to social skill deficit, PMU added (a) entertainment functions, including mobile games and video watching, and (b) business functions, namely email and Internet lectures. Therefore, the current study extended the scope of cognitive and behavioral disorders. Second, unlike the existing PIU studies, this study focused on maladjustment to school life which was most important for adolescents and provided more useful and practical implications for them. Third, we identified longitudinal associations among the constructs and revealed a new path from depressive moods to negative school outcomes, which was not tested in the cognitive-behavioral model.

Internationally, depressive symptoms and PMU are significant causes of difficulty in adolescents' daily lives including their school lives (Webb et al., 2007), and many younger teenagers have started using mobile phones (Seo & Choi, 2018). Therefore, the findings of the present study can be a source for education and public policies to decrease such negative impacts of PMU from a global perspective. Depressive feelings are prevalent, and their alleviation is the fundamental factor in preventing PMU and adverse school outcomes during adolescence.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Sangmin Jun is an Assistant Professor of the Department of Consumer Science at Chungbuk National University in Chungju, Republic of Korea. She completed her PhD degree at Seoul National University. Her current research focuses on problematic mobile phone use.

References

Augner

Hacker

G. W.

(2012) Associations between problematic mobile phone use and psychological parameters in young adults. International Journal of Public Health 57: 437–441. doi:10.1007/s00038-011-0234-z.

Bianchi

Phillips

J. G.

(2005) Psychological predictors of problem mobile phone use. CyberPsychology Behavior 8: 39–51. doi:10.1089/cpb.2005.8.39.

Bickham

D. S.

Hswen

Rich

(2015) Media use and depression: Exposure, household rules, and symptoms among young adolescents in the USA. International Journal of Public Health 60: 147–155. doi:10.1007/s00038-014-0647-6.

Bowen

Jenson

W. R.

Clark

(2004) School-based interventions for students with behavior, NY: Kluwer Academic/Plenum Publishers, doi:10.1007/978-1-4419-9102-7.

Browne

M. W.

Cudeck

(1993) Alternative ways of assessing model fit. In: Bollen

K. A.

Long

J. S.

(eds) Testing structural equation models, Newbury Park, CA: Sage.

Caplan

S. E.

(2002) Problematic Internet use and psychosocial well-being: Development of a theory-based cognitive-behavioral measurement instrument. Computers in Human Behavior 18(5): 553–575. doi:10.1016/S0747-5632(02)00004-3.

Caplan

S. E.

(2003) Preference for online social interaction: A theory of problematic Internet use and psychosocial well-being. Communication Research 30(6): 625–648. doi: 10.1177/0093650203257842.

Caplan

S. E.

(2005) A social skill account of problematic Internet use. Journal of Communication 55(4): 721–736. doi:10.1111/j.1460-2466.2005.tb03019.x.

Caplan

S. E.

(2010) Theory and measurement of generalized problematic Internet use: A two-step approach. Computers in Human Behavior 26(5): 1089–1097. doi: 10.1016/j.chb.2010.03.012.

10.

Caplan, S. E., & High, A. C. (2011). Online social interaction, psychosocial well-being, and problematic Internet use. In K. S. Young, & C. N. de Abreu (Eds.), Internet addiction: A handbook and guide to evaluation and treatment (pp. 35–53). Hoboken, NJ: Wiley & Sons. doi: 10.1002/9781118013991.

11.

Davis

R. A.

(2001) A cognitive-behavioral model of pathological Internet use. Computers in Human Behavior 17(2): 187–195. doi:10.1016/S0747-5632(00)00041-8.

12.

Duncan

T. E.

Duncan

S. C.

(2009) The ABC's of LGM: An introductory guide to latent variable growth curve modeling. Social and Personality Psychology Compass 3(6): 979–991. doi:10.1111/j.1751-9004.2009.00224.x.

13.

Duncan, T. E., Duncan, S. C., Strycker, L. A., Li, F., & Alport, A. (1999). An introduction to latent variable growth curve modeling: Concepts, issues, and applications. Mahwah, NJ: Lawrence Erlbaum. doi:10.4324/9780203879962.

14.

Duncan

S. C.

Duncan

T. E.

Strycker

L. A.

(2001) Qualitative and quantitative shifts in adolescent problem behavior development: A cohort-sequential multivariate latent growth modeling approach. Journal of Psychopathology and Behavioral Assessment 23(1): 43–50. Retrieved from https://link-springer-com-443.web.bisu.edu.cn/article/10.1023/A:1011091523808.

15.

Egger

H. L.

Costello

E. J.

Angold

(2003) School refusal and psychiatric disorders: a community study. Journal of the American Academy of Child and Adolescent Psychiatry 42: 797–807. doi:10.1097/01.CHI.0000046865.56865.79.

16.

Fröjd

S. A.

Nissinen

E. S.

Pelkonen

M. U.

Marttunen

M. J.

Koivisto

A. M.

Kaltiala-Heino

(2008) Depression and school performance in middle adolescent boys and girls. Journal of Adolescence 31(4): 485–498. doi:10.1016/j.adolescence.2007.08.006.

17.

Gámez-Guadix

Calvete

Orue

Las Hayas

(2015) Problematic Internet use and problematic alcohol use from the cognitive-behavioral model: A longitudinal study among adolescents. Addictive Behaviors 40: 109–114. doi:10.1016/j.addbeh.2014.09.009.

18.

Gámez-Guadix

Villa-George

F. I.

Calvete

(2012) Measurement and analysis of the cognitive-behavioral model of generalized problematic Internet use among Mexican adolescents. Journal of Adolescence 35: 1581–1591. doi:10.1016/j.adolescence.2012.06.005.

19.

Ge, X. H. (2014). Adolescent attachment and the mobile phone addiction: Mediating effects of social support. In H. C. Liu, W. P. Sung, & W. Yao (Eds.), Computer, Intelligent Computing and Education Technology (pp. 41–44). London: CRC Press. doi: 10.1201/ b16698-10.

20.

Griffiths

M. D.

(1999) Internet addiction: Fact or fiction?. The Psychologist: Bulletin of the British Psychological Society 12: 246–250. Retrieved from http://psycnet.apa.org/record/1999-13829-005.

21.

Hardell

(2018) Effects of mobile phones on children's and adolescents' health: A commentary. Child Development 89(1): 137–140. doi:10.1111/cdev.12831.

22.

L. T.

Bentler

P. M.

(1998) Fit indices in covariance structure modeling: Sensitivity to underparameterized model misspecification. Psychological Methods 3: 424–253. doi:10.1037/1082-989X.3.4.424.

23.

Jun

(2014) Longitudinal relationship between overuse/addictive use of mobile phones and depression in adolescents: Adapting the autoregressive cross-lagged model and multiple group analysis across gender. Family and Environment Research 52: 301–312. doi: 10.6115/fer.2014.026.

24.

Jun

(2016) The reciprocal longitudinal relationships between mobile phone addiction and depressive symptoms among Korean adolescents. Computers in Human Behavior 58: 179–186. doi: 10.1016/j.chb.2015.12.061.

25.

Kim

(2012) The influence of adolescents' psychological variables and attachment with parents upon internet game addicts and cellular phone addicts. The Korean Journal of Human Development 19(4): 1–22. Retrieved from http://www.kjhd.org/?page=7&a=45338.

26.

Kim

Hong

(2009) Writing articles using structural equation modeling, Communication Books: Seoul.

27.

Kim

LaRose

Peng

(2009) Loneliness as the cause and the effect of problematic Internet Use: The relationship between Internet use and psychological well-being. CyberPsychology & Behavior 12(4): 451–455. doi: 10.1089/cpb.2008.0327.

28.

Kim

J. H.

Hong

S. H.

Choo

B. D.

(2007) Application of structural equation modeling in management studies: A critical review. Korean Management Review 36(4): 897–923. Retrieved from http://www.dbpia.co.kr/Journal/ArticleDetail/NODE06079619.

29.

Kim

S. Y.

Lim

J. Y.

Kim

S. W.

Park

S. H.

You

S. R.

Choi

J. Y.

Shin

W. Y.

(2006) Korean adolescents' development index survey I, Korea Institute for Youth Development: Seoul.

30.

Kim

J. H.

Seo

David

(2015) Alleviating depression only to become problematic mobile phone users: Can face-to-face communication be the antidote?. Computers in Human Behavior 51: 440–447. doi:10.1016/j.chb.2015.05.030.

31.

Lee

Kim

Nah

Lee

Kim

Bae

(2002) A study on the mobile phone use among adolescents and its effects, Seoul: Samsung Life Public Welfare Foundation.

32.

Lee, S. P. (1990). A study on the differences between academic performance and behavioral characteristics according to school adaptation levels. Master's Thesis, Hongik Graduation School of Education: Seoul. Retrieved from http://dl.nanet.go.kr/SearchDetailView.do?cn=KDMT1199120615_1.

33.

Livingstone

Smith

P. K.

(2014) Annual research review: Harms experienced by child users of online and mobile technologies: The nature, prevalence, and management of sexual and aggressive risks in the digital age. Journal of Child Psychology and Psychiatry 55: 635–654. doi:10.1111/jcpp.12197.

34.

Marsh

H. W.

Hau

K.-T.

Wen

(2004) In search of golden rules: Comment on hypothesis-testing approaches to setting cutoff values for fit indexes and dangers in overgeneralizing Hu and Bentler's (1999) findings. Structural Equation Modeling 11(3): 320–341. doi:10.1207/s15328007sem1103_2.

35.

McCarthy, A. E., Young, J. F., Benas, J. S., & Gallop, R. J. (2018). School-related outcomes from a randomized controlled trial of adolescent depression prevention programs. Journal of Emotional and Behavioral Disorders, 26(3), 170–181. doi: 10.1177/1063426617717739.

36.

Min, B. S. (1991). The effect of school adjustment and self-concept on academic achievements. Master's Thesis, Hongik University: Seoul. Retrieved from http://dl.nanet.go.kr/SearchDetailView.do.

37.

Ministry of Science, ICT, and Future Planning of Korea & National Information Society Agency of Korea (2017). The survey on Internet overdependence in 2016. Retrieved from http://www.korea.kr/archive/expDocView.do?docId=37489.

38.

Muniz-Terrera

Robitaille

Kelly

Johnsson

Hofer

Piccinin

(2017) Latent growth models matched to research questions to answer questions about dynamics of change in multiple processes. Journal of Clinical Epidemiology 82: 158–166. doi:10.1016/j.jclinepi.2016.09.001.

39.

Owens

Stevenson

Hadwin

J. A.

Morgate

(2012) Anxiety and depression in academic performance: An exploration of the mediating factors of worry and working memory. School Psychology International 33(4): 433–449. doi:10.1177/0143034311427433.

40.

Rushton

J. L.

Forcier

Schectman

R. M.

(2002) Epidemiology of depressive symptoms in the national longitudinal study of adolescent health. Journal of the American Academy of Child & Adolescent Psychiatry 41: 199–205. doi:10.1097/00004583-200202000-00014.

41.

Sánchez-Martínez

Otero

(2009) Factors associated with cell phone use in adolescents in the community of Madrid (Spain). CyberPsychology & Behavior 12: 131–137. doi:10.1089/cpb.2008.0164.

42.

Seo

Choi

(2018) Classes of trajectory in mobile phone dependency and the effects of negative parenting on them during early adolescence. School Psychology International 39(2): 156–169. doi:10.1177/0143034317745946.

43.

Vernon

Modecki

K. L.

Barber

B. L.

(2018) Mobile phones in the bedroom: Trajectories of sleep habits and subsequent adolescent psychosocial development. Child Development 89(1): 66–77. doi: 10.1111/cdev.12836.

44.

Webb

Panico

Bécares

McMunn

Kelly

Sacker

(2017) The inter-relationship of adolescent unhappiness and parental mental distress. Journal of Adolescent Health 60(2): 196–203. doi:10.1016/j.jadohealth.2016.10.001.

45.

Werner-Seidler

Perry

Calear

A. L.

Newby

J. M.

Christensen

(2017) School-based depression and anxiety prevention programs for young people: A systematic review and meta-analysis. Clinical Psychology Review 51: 30–47. doi:10.1016/j.cpr.2016.10.005.

46.

Yoon

K. S.

Moon

S. M.

(1977) A study on the background characteristics and school adaptation of students. The Kyungsang University Journal of Student Guidance 4: 19–28. Retrieved from http://dl.nanet.go.kr/SearchDetailView.do.