Academic concerns are the most important sources of chronic and sporadic stress for young people in both Western and Asian countries (41). Academic stress is defined as a student's psychological state resulting from continuous social and self-imposed pressure in a school environment that depletes the student's psychological reserves (42, 43). Students experience academic stress throughout their secondary school (41), high school (44), and university (45, 46), educational careers. Studies have shown that academic stress has been positively associated with depression (41), PSQ (24, 47), and MPA (48) among students. Jayanthi observed that, compared to adolescents who do not experience academic stress, adolescents who experienced academic stress were 2.4 times more likely to have depressive symptoms (41). Other studies have found that there is a relationship between high academic stress and PSQ (47, 49). However, scholars have not adequately addressed the adverse consequences (e.g., depression, PSQ, and MPA) of Chinese students' academic stress. Hence, we propose the following hypotheses:

H1: PAS is positively associated with depression.

H2: PAS is positively associated with MPA.

H3: PAS is positively associated with PSQ.

MPA is one of the most common behavioral (i.e., non-drug) addictions (48), and is accompanied by negative effects, such as PSQ (50), depression (35), and impaired academic performance (51). The positive relationship between MPA and PSQ has been proved in previous studies, including a longitudinal study conducted among Korean adolescents (52) and a one-year prospective study among Chinese college students (50). Zhang found that among Chinese university students, there is a significant positive relationship between smartphone addiction and bedtime procrastination, which is one of the indicators of PSQ (53). Hence, we propose the following hypothesis:

H4: MPA is positively associated with PSQ.

Similarly, the positive relationship between MPA and depression has been proved in previous studies, including a cross-sectional study conducted among Saudi university students (35), a cross-sectional study among Ukrainian college students (36), and a systematic review of relations between problematic smartphone use, anxiety and depression psychopathology (54). Furthermore, another study based on three cohorts of Korean children and adolescents confirmed the bidirectional relationship between MPA and depression (55). Hence, we propose the following hypothesis:

H5: MPA is positively associated with depression.

Researchers have documented that stress is associated with MPA, and that MPA is associated with depression. For example, according to Wan et al., smartphone addictions are significantly positively associated with both depression and stress among Malaysian public university students (56). However, it is unclear if MPA mediates the relationship between PAS and depression. Hence, we propose the following hypothesis:

H6: MPA mediates the relationship between PAS and depression.

Sleep disturbance has complex associations with depression (major depressive disorder) (31), and is a common physical symptom of depression. Numerous studies have confirmed the remarkable association between PSQ and depression (29, 57, 58). For example, Okun et al. found that PSQ is positively related to depression symptoms in postpartum women (29). Hence, we propose the following hypothesis:

H7: PSQ is positively associated with depression.

Scholars have also demonstrated that there are relationships between stress, PSQ, and depression. A prospective birth cohort study showed that PSQ is associated with stress and depression symptoms among Chinese pregnant women (58). Zhang et al. found that perceived stress is associated with sleep quality and depressive symptoms among Chinese nursing students (59). However, it has not been documented if sleep quality mediates the relationship between PAS and depression among Chinese students. Hence, we propose the following hypothesis:

H8: Sleep quality mediates the relationship between PAS and depression.

Scholars have posited that there are significant associations between MPA, depression levels, and sleep quality. Demirci found that there were positive correlations between MPA, depression levels, and sleep quality (60). The results of Kaya's multivariate regression analysis showed a relationship between smartphone usage, PSQ, and depression in university students (57). A recent meta-analysis also found that there are positive correlations between MPA, depression, and sleep quality (61). Another literature review and case study found that depressive symptoms are associated with screen time-induced poor sleep, digital device night use, and mobile phone dependency (62). Although these studies explored the correlations between MPA, sleep quality, and depression among students, several scholars have added academic stress into the relationship—for example, a review found that sleep disturbance, anxiety, stress, and depression have been associated with problematic mobile phone use (63). There still exist gaps in the literature on how PAS influences depression. First, few scholars have focused on PAS, MPA, sleep quality, and depression among Chinese students. Second, the underlying mediating mechanisms that account for this association have been disregarded partly. Based on H6 and H8, it remains unclear if MPA and sleep quality serially mediate the relationship between PAS and depression. Therefore, we propose the following hypothesis:

H9: MPA and sleep quality serially mediate the relationship between PAS and depression.

In this study, our primary aim was to investigate the prevalence of depression and sleep disturbance among Chinese students. Our secondary aim was to test if there were relationships between PAS, MPA, sleep quality, and depression. First, we tested if there was a relationship between PAS and depression among Chinese students (H1: PAS is positively associated with depression). Second, we tested if MPA was a mediator of the relationship between PAS and depression (H2: PAS is positively associated with MPA, H5: MPA is positively associated with depression, and H6: MPA mediates the relationship between PAS and depression). Third, we tested if sleep quality was a mediator of the relationship between PAS and MPA (H3: PAS is positively associated with PSQ, H7: PSQ is positively associated with depression, and H8: Sleep quality mediates the relationship between PAS and depression). Finally, we also tested if MPA and sleep quality together played a serial mediating role in the influence of PAS on depression (H4: MPA is positively associated with PSQ and H9: MPA and sleep quality serially mediate the relationship between PAS and depression).

Data were collected from a cross-sectional questionnaire survey that was conducted from September to December 2018 in Heilongjiang Province, China, by the Heilongjiang Center for Disease Control and Prevention. A multistage cluster sampling method was used. In the first stage, three cities of Heilongjiang province were randomly selected by economic characteristics. In the second stage, one urban district and one rural township were chosen at random. In the third stage, two middle schools were randomly selected in each urban district and rural township; Since nine-year compulsory education was implemented in China, high school education is not included in the nine-year compulsory education, high schools are more in urban districts than in rural townships, two high schools and one high school were randomly selected in urban district and rural township, respectively; Since vocational high schools and universities are scarce in rural townships, one vocational high school and one college were randomly selected from the urban district. In the fourth stage, two classes were randomly selected from each grade of middle school, high school, vocational high school, and from grades 1, 2, and 3 in college. Since senior students may have been looking for a job or working as an intern, some of them were not on campus, they were not been investigated. Finally, four middle schools, three high schools, one vocational high school, and one college were randomly selected within each city (Harbin, Jiamusi, and Jixi) of Heilongjiang Province. Data were collected through a self-administered questionnaire distributed in class. Students completed the survey within 1 h, while a well-trained member of the research group supervised. All the students were informed of the purpose of the study and assured that their identities would remain confidential. Students and their parents provided written informed assent to participate in the study.

Finally, we recruited 6,480 students in our investigation; 6,430 (99.23%) valid questionnaires were analyzed after excluding those with incomplete information. Participants were included in the sample if they had one constant internet-accessible mobile phone, which is similar with previous studies (64–68). A total of 5,109 (79.46%) participants reported having one constant internet-accessible mobile phone at the time of the survey. The final sample comprised 1,904 middle school students from grades 1, 2, and 3, respectively; 1,859 high school students from grades 1, 2, and 3, respectively; 660 vocational high school students from grades 1, 2, and 3, respectively; and 686 college students from grades 1, 2 and 3, respectively. Of these participants, there were 2,422 (47.41%) boys and 2,687 (52.59%) girls; on average, the mean age of participants was 15.53 years, with a standard deviation of 2.22, ranging from 11 to 25 years. Approval was obtained from the Medical Research Ethics Committee of Harbin Medical University and the principals of the participating schools.

The Pittsburgh Sleep Quality Index (PSQI) was used in our study (87). PSQI scale contains 19 items covering seven components: subjective sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbances, use of sleep medication, and daytime dysfunction. Each component was scored from 0 (no difficulty) to 3 (severe difficulty). The total score was calculated from the seven component scores, ranging from 0 to 21. A score of more than 5 implied poor sleep (87). The CFA on the seven-factor model showed a good model fit, with χ² = 79.49, df = 11, P < 0.000, RMSEA = 0.04, SRMR = 0.02, CFI = 0.99, TLI = 0.98. Additionally, Cronbach's alpha coefficient was 0.624 for the PSQI scale in our study.

SPSS version 19.0. and Mplus 7.0 were used to analyse data in our study. Descriptive analyses were first conducted of participants' characteristics, participants' mobile phone use and the prevalence of sleep disturbance, MPA, and depression. We tested the reliability and validity of the MPAI scale, PSQI scale and CES-D scale by examining their Cronbach's alpha coefficient and performing a CFA. Spearman's correlation analysis was performed to examine the general relationships among the four variables—PAS, MPA, sleep quality, and depression. A structural equation model (SEM) was built to examine hypotheses. We tested the mediating role of MPA and sleep quality; the constructed serial mediation model included three latent variables (MPA, sleep quality and depression) and one manifest variable (PAS), PAS was the independent variable, depression was the dependent variable, and MPA and sleep quality were the mediating variables (88). The bootstrapping analyses used 5,000 samples at the 95% confidence interval (CI) to indicate significance.

To determine whether the model fits the data well, multiple indices were tested, including (1) the model χ² and its p value, in which non-significance is desirable for good fit. With increasing sample size and a fixed degree of freedom, the χ² value increases. It is difficult to get a nonsignificant chi-square (indicative of good fit) when sample sizes are over 200 (89). This can lead to a problem where plausible models might be rejected. Because this statistic is sensitive to the sample size, inspection of the other fit indices is recommended (90). (2) The root mean square error of approximation (RMSEA) in which values ranging from 0.05 to 0.08 represent adequate fit, and values <0.05 indicate good fit. (3) The standardized root mean square residual (SRMR) in which values are ≤0.08 indicate good fit. (4) The comparative fit index (CFI), in which values range from 0.90 to 0.95 indicate an adequate fit and values ≥0.95 indicate a good fit, and (5) the Tacker-Lewis index (TLI) in which values >0.90 indicate a good fit.

The mean scores of PAS were 2.61 ± 1.03, 2.68 ± 1.06, 2.13 ± 0.98, and 2.29 ±0.96 for middle school students, high school students, vocational high school students, and college school students, respectively.

Among the participants, 45.55% used their mobile phone more than 2 h daily; 39.5% of the participants had a mobile phone for more than 3 years; 53.24% of the participants spent more than or equal to 30 yuan on mobile phone charges every month (Table 1). The mean MPAI score of all the participants was 30.62 ± 11.92.

The prevalence of depressive symptoms was 28.69% (n = 1,466) with a mean CES-D score of 12.52 ± 8.86. Prevalence of depression at a mild level (CES-D ≥ 16 and CES-D < 21), moderate level (CES-D ≥ 21 and CES-D < 25), and severe level (CES-D ≥ 25) was 12.62, 6.95, and 9.12%, respectively. The prevalence of depressive symptoms among high school students (33.14%) was the highest, while the prevalence of depressive symptoms among college students (21.43%) was the lowest (Table 2).

The prevalence of sleep disturbance was 27.95% (n = 1,428) with a mean global PSQI score of 4.29 ± 2.59. The prevalence of sleep disturbance among high school students (36.47%) was the highest, while the prevalence of sleep disturbance among middle school students (20.75%) was the lowest. The average sleep time and sleep latency were 7.40 ± 1.28 h and 15.81 ± 12.48 min, respectively. Among the participants, 14.50% reported that they had bad or very bad sleep quality; 36.29% reported that their sleep latency was more than 15 min; 50.89% reported that they slept ≤7 h a day; 12.62% reported that their sleep efficiency was ≤85%; 67.59% reported that they experienced sleep disturbances; 2.90% of them reported that they used sleep medication; and 78.80% reported that they had daytime dysfunction (Table 3).

Means, Standard Deviation (SD) and correlations of the main variables in the mediation model are shown in Table 4. The results, indicating that the variables were significantly and positively correlated, provide initial support for the hypotheses of this study, and act as a foundation of the serial mediation model.

SEM was used to provide the fit indexes of the serial mediation model. A model was constructed with MPA (M1) as a mediator and sleep quality (M2) as another mediator. In this model, PAS was set as the predictor (X) and depression as the outcome (Y). Results of the serial mediation model indicated that the constructed model exhibited a satisfactory fit with the data: χ² = 1,196.50, df = 95, P < 0.000, SRMR = 0.04, RMSEA = 0.05, CFI = 0.95, and TLI = 0.94.

First, PAS was positively associated with depression (B = 0.10, SE = 0.02, 95% CI = 0.06–0.13). Higher levels of PAS were related to higher levels of depression, and thus H1 was supported. Second, PAS positively predicted MPA (B = 0.18, SE = 0.02, 95% CI = 0.15–0.21). Higher levels of PAS were related to higher levels of MPA, and thus H2 was supported. Third, PAS was positively associated with PSQ (B = 0.23, SE = 0.02, 95% CI = 0.19–0.26). Higher levels of PAS were related to poorer sleep quality, and thus H3 was supported. Fourth, MPA was positively associated with PSQ (B = 0.51, SE = 0.02, 95% CI = 0.47–0.54). Higher levels of MPA were related to poorer sleep quality, and thus H4 was supported. Fifth, MPA was positively associated with depression (B = 0.17, SE = 0.02, 95% CI = 0.13–0.22). Higher levels of MPA were related to higher levels of depression, and thus H5 was supported. Last, PSQ was positively associated with depression (B = 0.44, SE = 0.03, 95% CI = 0.39–0.49). PSQ was related to higher levels of depression, and thus H7 was supported.

Table 5 shows all possible indirect effects of the mediation model. First, the indirect effect of PAS on depression through MPA was significant (B = 0.08, 95% boot CI = 0.06–0.11), and thus H6 was supported. Second, the indirect effect of PAS on depression through sleep quality was significant (B = 0.27, 95% boot CI = 0.22–0.33), and thus H8 was supported. Third, the indirect effect of PAS on depression through MPA and sleep quality was also significant (B = 0.11, 95% boot CI = 0.08–0.14), and thus H9 was also supported. The total indirect effect was B = 0.46, 95% boot CI = 0.40–0.53, and the mediating effect of MPA and sleep quality were significant (P < 0.001), accounting for 64.01% (total indirect effect/total effect) of the total effect. The indirect effect related to sleep quality accounted for 82.61% of the total indirect effect, that is, (indirect effect 2 + indirect effect 3)/total indirect effect (Table 5).

According to studies that have focused on depression among Chinese students, the prevalence of depression varies from 22.0 to 68.5% (5, 91–95). In our study, the prevalence of depressive symptoms was 28.69%. The differences across these studies may have resulted from temporal or regional disparities or variations in depression definitions and assessment methods. Depressive symptoms are related to many negative consequences, such as increased suicide risk among students (96) and increased college withdrawal rates (97). Controlling depressive symptoms among students can both protect human capital value from the societal perspective and maintain students' physical and mental health from the individual perspective. In our study, the most stressed, depressed, and sleep-deprived students were high school students. Thus, Chinese high school students' physical and mental health requires attention. In China, high school students are admitted to colleges and universities based on gaokao, the standardized National College Entrance Examination (98). These admission decisions are extremely important, as they impact high school students' future educational opportunities, career paths, and life experiences. Our research results prove that Chinese students experience the most stressful and competitive academic environment of their academic careers when they are in high school.

Chinese students spend considerable time on mobile phones−45.55% of the participants spent more than 2 h daily on their mobile devices. 39.5% of participants had had a mobile phone for more than 3 years, while the mean age of participants was 15.53 years. Using the mediation model, we illustrated the mediating role of MPA in line with our hypothesis. As H6 predicted, MPA played a role in the path from PAS to depression. MPA could partially explain the association between PAS and depression among Chinese students—hence, MPA was not only an outcome of PAS, but also a catalyst of depression. First, we found that high levels of PAS were associated with high levels of MPA. This finding is consistent with previous research results (48) and suggests that PAS may be a significant trigger for students' negative behaviors—such as MPA. Scholars have posited that young people's digital distraction activities—including playing computer games and online surfing—may be interpreted as a way to avoid problems, reality, and stress (99, 100). High levels of PAS were associated with high levels of MPA, which may be due to students' use of mobile phones to escape from academic stress. Second, we found that high levels of MPA were associated with high levels of depression, which is in line with existing research results (35, 36, 101). Students who experience MPA may neglect real-world social engagement (102) resulting in academic underperformance (103), clinical health symptoms (68), which are related to negative emotions—such as depression. Our findings add to the existing research that suggests that when students are facing academic stress, they may be addicted to their mobile phones to escape from academic stress, and thus the negative consequences of MPA may lead to depression in students.

As H8 predicted, sleep quality is not only an outcome of academic pressure—it is also a catalyst of depression. Moreover, the indirect effect related to sleep quality accounted for 82.61% of the total indirect effect. Thus, compared to MPA, sleep quality played a more important role in the path from PAS to depression. We found that higher levels of PAS were associated with poorer sleep quality. This finding is consistent with previous research findings (24, 47). For example, Waqas et al. demonstrated that perceived stress is a significant predictor of PSQ (47). In China, students exist in a prolonged competitive learning environment and experience unrelenting academic stress. To achieve better academic performance and meet the extraordinarily high expectations of parents and educators, Chinese students have heavy homework burdens and learning burdens, resulting in sleep deprivation. Furthermore, academic stress decreases sleep quality. According to Almojali et al., students who are not suffering from academic stress are less likely to experience PSQ (104). Previous studies have proved that sleep deficiency and sleep health problems are common among Chinese students (105). Our research results may explain why higher levels of PAS were related to poorer sleep quality.

We also found that high levels of PSQ were associated with high levels of depression, which is consistent with prior research findings (31, 50, 57). Scholars have proved that PSQ is related to multiple negative consequences that may lead to depression—including daytime dysfunction, poor academic performance, and fatigue (106, 107). Our findings add to the existing research that suggests that sleep quality is a mediator between PAS and depression among students, which means that higher levels of PAS were related to poorer sleep quality—such as sleep deficiency and daytime dysfunction—which was related to higher levels of depression.

As per H9, MPA and sleep quality together play a serial mediating role in the influence of PAS on depression. The results of our study showed that higher levels of PAS were related to higher levels of MPA, which was associated with poorer sleep quality, which was associated with higher levels of depression. Numerous studies have documented that there is a positive relationship between MPA and PSQ (50, 52). For example, Kang et al. found that there were bidirectional longitudinal relationships between MPA and PSQ (50). Scholars have posited that the more screen time young people use, the less sleep time they have (108). Moreover, young people often use their mobile phone in the bedroom—bedtime mobile phone use is related to higher insomnia scores and increased fatigue (109), and both insomnia and fatigue are related to depression (110, 111). This may explain why MPA and PSQ together play a serial mediating role in the influence of PAS on depression. Our findings suggests that Chinese students are likely to distract themselves from PAS by using their mobile phones, and thus shortening their sleep duration, decreasing their sleep quality, leading to PSQ, and resulting in depressive symptoms.

To reduce depressive symptoms among students, their PAS should be managed. Given the multiple, negative consequences (MPA, PSQ, and depression) of PAS, stakeholders—family members, educators (including teachers, school administrators, and school health professionals), and policy makers—should take preventative measures to help students manage and relieve academic stress, such as provide counseling services (112), foster their psychological resilience (113), and increase social support (19) to improve their overall well-being. Second, students' sleep quality should be ensured to reduce depressive symptoms. Stakeholders should actively promote counseling and intervention for students experiencing sleep disturbances. Third, given that higher levels of MPA are associated with poorer sleep quality and higher levels of depression, stakeholders should develop mitigating strategies to manage mobile phone use to ensure students' sleep quality and to relieve their depressive symptoms. Rational and normative mobile phone use should be advocated and classroom management strategies enforced to ensure that students use their mobile phones at restricted times and places for positive purposes, such as online learning. Fourth, regular psychological assessment of depression, MPA, and PSQ will help stakeholders detect and manage students' health problems. Last, parents and family members, educators, and policy makers should encourage students to exercise more to alleviate MPA (114), improve sleep quality (115), and reduce depression (116).

This study has several limitations. First, although we conducted our research based on previous studies, due to the cross-sectional design of our study, we could not confirm causal relationships among the study variables. Second, the study period was September to December 2018, which was more than 2 years ago. However, we believe that the results of our study are valuable for understanding the mechanisms of how PAS influences students' depression through MPA and sleep quality, and our study can provide a basis for future research. Third, the study measured the participants' perceived academic stress using a single item, which may not have captured various other relevant stressors, such as parental learning expectations. Future studies should use a multiple-item scale to assess the participants' perceived academic stress. Forth, this study was limited to middle school, high school, vocational high school, and college students. Future research on Chinese students at all education levels from primary school to postgraduate levels is necessary. Fifth, perceived academic stress can increase during stressful conditions (117), such as during exams or major change in life (e.g., from high-school student to freshman). While our study was conducted in 27 schools in three cities, it is impossible that we conducted the survey when the participants had no examinations or changes. Future studies can control for stressful academic conditions in the analyses to enhance their accuracy. Last, gender, age, and other factors are important influencing factors of PAS, MPA, sleep quality, and depression. Since the main aim of this study was to test if there was a relationship between PAS and depression and if MPA and sleep quality together play a serial mediating role in the influence of PAS on depression, the aforementioned factors were not considered in this study. Future studies should consider these factors and test the relationships between PAS, MPA, sleep quality, depression, and other health indicators.