Addiction is a manifestation of attachment disorder, and there is an inverse relationship between addiction and healthy interpersonal attachment (29). Attachment styles have frequently been referred to in studies of addictive behavior, and a growing number of studies (30) have examined the influence of attachment styles on technology addiction (20).

Attachment theory was proposed by the psychologist John Bowlby in the 1960s (31). The theory conceptualizes the tendency of humans to establish strong emotional ties with intimate people. According to the theory, individuals who experience unwilling separation and loss are at risk of various forms of emotional distress and personality disorder, including anxiety, anger, depression, and emotional detachment (32). Attachment styles can be divided into secure attachment and insecure attachment, and the latter can be divided into anxious attachment and avoidant attachment (33). People with an anxious attachment style have a hyperactive attachment system and tend to seek constant support and comfort. People with an avoidant attachment style, in contrast, have a deactivated attachment system that leads to the inhibition of psychological and social relationship needs, which means that they deliberately maintain a clear distance from others (34). Unlike attachment avoidance, people with an anxious attachment style are more impacted during the confinement, because they are more invested in the relationship out of fear of being abandoned and alone (33).

Attachment anxiety has been reported to cause intensive and dysfunctional Internet and social media use, and is closely related to smartphone addiction (17, 35). People with attachment anxiety are afraid of being abandoned and are likely to engage in intimacy-seeking, relationship-related disclosure behaviors (36). If their relationships with family members and peers (30) do not satisfy their relationship needs, anxiety about being abandoned or rejected by others may drive them to use smartphones for comfort (10), which eventually increases the risk of addiction (19). Therefore, we theorized that students who are unable to return to campus or meet their teachers and peers at school during home quarantine find it hard to satisfy their attachment needs, which increases the probability of smartphone addiction. More specifically, we hypothesized that attachment anxiety directly affects smartphone addiction during the confinement, leading to our first hypothesis.

Numerous studies have examined how interpersonal relationships may contribute to smartphone addiction (37, 38). Individuals with high attachment anxiety are reported to get less social support and be less satisfied with the interpersonal relationships they build. They frequently worry that their friends will abandon them, so they try to be as emotionally close to others as possible. They could turn to smartphones for safety as a result of unhappy relationships (39). For example, individuals with high attachment anxiety frequently use their cell phones to contact others in order to maintain stable relationships (40). They are more likely to be on social media and to be concerned about how others perceive them (41). Student-student and teacher–student relationships are important types of interpersonal relationships among university students (42). When emerging adults leave home for university, there is a gradual shift from parental attachment to peer attachment (43), which means that students are more likely to turn to their peers and teachers to satisfy their attachment needs (27). Although teachers' work is mainly focused on education, providing care and emotional support is also an important part of their role (44). Therefore, although a teacher–student relationship is not a complete attachment bond, it can create a certain sense of security for the student (45). In terms of student–student relationships, an insecure attachment style is negatively related to perceived peer support (46). Attachment anxiety can predict the quality of university students' peer relationships, which is manifested through friendship quality and the level of loneliness (27).

Meanwhile, the quality of interpersonal relationships is closely related to smartphone addiction. Studies have demonstrated both direct (47) and indirect (48) effects of teacher–student relationships on smartphone addiction. Specifically, perceived support from teachers can reduce problematic smartphone use by promoting students' self-esteem (47). Deviant peer affiliation, representing poor student–student relationships, is positively related to Internet game addiction (49). For the same reason, healthy student–student relationships are negatively associated with smartphone addiction (50). Based on our review of the relevant literature, we hypothesized that attachment anxiety would negatively predict a good interpersonal relationship and a tendency toward smartphone addiction:

School connectedness is a composite variable describing students' psychological sense of safety and membership to their school environment (51). The importance of school connectedness has been suggested because students need meaningful and supportive relationships outside of their family, and such bonds usually occur in the school environment (52). Attachment anxiety directly affects students' connectedness and life satisfaction, and individuals with insecure attachment styles are likely to experience difficulties in fulfilling their need to belong (53). Students with positive school connectedness enjoy their school life and have better sense of belonging in school (54). School connectedness has been shown to be a salient predictor of students' psychological wellbeing and prosocial behavior (55).

Also, students' feelings about school act as prominent predictors of smartphone addiction. For example, problematic Internet use is lower among students who show appreciation for their school than among those who do not (56); lower school connectedness is related to increased Internet game addiction (57); and maladjustment in school is associated with smartphone addiction (58). Given that school connectedness is likely to be weaker when students are far away from campus, we propose the following hypotheses in the context of the confinement. We hypothesized that school connectedness acts as a mediator in the relationship between anxiety attachment and smartphone addiction. This is based on the significant relationship between anxiety attachment and school connectedness, school connectedness and smartphone addiction, and anxiety attachment and smartphone addiction.

The hypothetical model is shown in Figure 1.

In the worst year of the epidemic in China, most universities took quarantine measures. Due to the confinement policy, a two-stage sampling survey method was used, combining probability and non-probability sampling. Based on geographic position and socioeconomic development, China is divided into four major regions: Northeastern region, Eastern region, Central region, and Western region. Then, five typical administrative cities were selected from each region. The participants were recruited online via messages sent to undergraduate students through WeChat in 2020. To ensure the validity of the research, screening was performed to ensure the participants (1) were Chinese university students, (2) were studying at a university that conducted confinement policy requiring home study, and (3) possessed a smartphone. After screening, 1,052 eligible participants remained, but data from 53 of them were excluded because their response times were too short or their responses to different items were too similar, according to the standard practice for questionnaire-based research. Eventually, 999 participants (384 male and 615 female, Mage = 20.05, SDage = 1.175, age range 17–29) provided valid data. We obtained informed consent forms from all the participants and one from a 17-year-old participant's parents. This study was approved by the Research Ethics Committee of the corresponding's author's university.

The participants completed a questionnaire designed to collect information on demographic variables, attachment anxiety, teacher–student relationships, student–student relationships, school connectedness, and smartphone addiction. After completing the questionnaire, the participants received modest monetary compensation in return for their contribution to the study. 52.7% of the participants were junior and 26.2% were sophomore. More than half of the participants studied in Northeastern region and Eastern region. 74.6% of the participants experienced home learning exceeding 4 months. The detailed demographic information is shown in Table 1.

We conducted descriptive statistical analysis using SPSS version 19 to examine the characteristics of the participants and the measured constructs. Structural equation modeling (SEM) was performed with Mplus version 7 to test the relationships among the latent constructs.

Following a two-stage procedure (69), we first performed confirmatory factor analysis (CFA) to evaluate the measurement model, which verified the associations between the latent constructs and the observable variables by assessing the reliability and validity of the measurement tools.

Second, the structural equation model tested the relationships among the constructs by examining the standardized path coefficients. We performed a bootstrap procedure to test the mediation effects of our hypothetical model, as this has been shown to be a valid and powerful method of verifying mediation effects (70).

We also evaluated the fitness of the model using the chi-square test, degree of freedom (df), standardized root mean square residual (SRMR), Tucker–Lewis index (TLI), comparative fit index (CFI), and root mean square error of approximation (RMSEA). TLI and CFI values > 0.90 and RMSEA values < 0.08 indicate an acceptable model fit (71).

To evaluate the measurement model, we used Cronbach's alpha and composite reliability (CR) values to examine the reliability of the constructs. According to Fornell and Larcker (72), CR values above 0.700 represent good composite reliability. In the current study, Cronbach's alpha ranged from 0.682 to 0.925, and the CR values ranged from 0.784 to 0.927, suggesting that the reliability of the latent constructs was good or acceptable (see Table 2).

To test the validity of the measurement model, convergent and discriminant validity were evaluated. Convergent validity was evaluated by the average variance extracted (AVE) of each construct. The AVE values ranged from 0.548 to 0.762 (see Table 2), and all were above the recommended value of 0.500 (72), thus indicating good convergent validity. To evaluate discriminant validity, we compared the square root of the AVE with the absolute values of the correlation coefficients between each construct and the other constructs. According to Table 3, the square roots of the AVE of each construct were larger than the absolute values of the correlation coefficients between constructs. Therefore, discriminant validity was also acceptable.

The structural model was analyzed to test the relationships between the constructs. The hypothetical model and path coefficients (standard errors) are shown in Figure 2. According to the path coefficients, attachment anxiety was positively related to smartphone addiction (β = 0.29, p < 0.001), thus confirming H1. The paths between attachment anxiety and teacher–student relationships (H2a, β = −0.10, p = 0.014), attachment anxiety and student–student relationships (H2b, β = −0.23, p < 0.001), and attachment anxiety and school connectedness (H2c, β = −0.38, p < 0.001) were all statistically significant, thus supporting H2a, H2b, and H2c. Teacher–student relationships (H3a, β = −0.13, p = 0.001) and school connectedness (H3c, β = −0.41, p < 0.001) were both negatively associated with smartphone addiction, thus confirming H3a and H3c. However, the path between student–student relationships and smartphone addiction (H3b, β = 0.04, p = 0.370) was not significant, thus H3b was not confirmed. Only one of the seven paths in the model was not significant. The hypothetical model provided a good fit to the data (χ² = 1011.215; df = 196; SRMR = 0.081; TLI = 0.914; CFI = 0.925; RMSEA = 0.065).

The bootstrap method was used to test the indirect effects of attachment anxiety on smartphone addiction. The indirect effects of the three mediators (teacher–student relationships, student–student relationships, and school connectedness) in the hypothetical model were estimated with 10,000 randomly assigned samples from the original data (Table 4). The mediation effects of teacher–student relationships and school connectedness were both significant because the 95% confidence intervals did not include zero, thus confirming H4a and H4c. However, the mediation effect of student–student relationships was not significant as the 95% confidence interval included zero (p = 0.383); thus, H4b was not confirmed. Comparison of the mediation effects among the three constructs showed that the contrast between the mediation effects of teacher–student relationships and school connectedness was significant (p < 0.001), implying the existence of different mediation effects on the relationship between attachment anxiety and smartphone addiction. Table 5 shows that most of our hypotheses were verified.

We found that attachment anxiety was negatively associated with good teacher–student relationships. Meanwhile, teacher–student relationships had a negative influence on smartphone addiction, which is in line with the study by Peng et al. (47), which found that teachers' support for students' autonomy reduced students' problematic smartphone use. Teacher–student relationships were a significant mediator in our study. These results suggest that during the COVID-19 pandemic, university students with high attachment anxiety tend to have poor relationships with their teachers, which increases the potential for problematic smartphone use.

The mediation effect of teacher–student relationships implies that unlike face-to-face regular periods, students with a high level of attachment anxiety tend to lack interactions and bidirectional emotional exchanges with their teachers during home quarantine. They try to compensate for their sense of emptiness and insecurity through the use of smartphones, and excessive use can easily lead to smartphone addiction. The mediating effect demonstrates that teacher–student relationships have an important role in preventing smartphone addiction. The findings also imply that if teachers provide care and support to enhance their relationships with students, students will be less likely to develop smartphone addiction during the confinement.

As expected, attachment anxiety was negatively associated with student–student relationships. This suggests that in the confinement students with high attachment anxiety are likely to have poor relationships with their school peers. Previous studies (74) have found that student–student relationships are negatively correlated with addictive behavior, whereas healthy student–student relationships can alleviate addiction. However, similar to Ihm's (75) study, we found no significant association between student–student relationships and smartphone addiction; therefore, the mediating effect was also non-significant. This suggests that even though university students with attachment anxiety tend to have problematic student–student relationships, they are unlikely to compensate for the lack of companionship by excessively using their smartphones.

We believe that student–student relationships failed to mediate the relationship between attachment anxiety and smartphone addiction because the policy of home quarantine served as an external attribution source. In the absence of face-to-face communication, students with attachment anxiety have a reasonable explanation for the lack of concern and understanding from their peers, and the sense of loss caused by online alienation from school peers seems more acceptable than feeling isolated offline. In addition, without face-to-face communication, individuals can still perceive the presence of their school peers through multiple channels, such as “liking” and commenting on their social media posts. In contrast, during the confinement, online courses are almost the only way for students to perceive the existence of their teachers and schools, and to maintain their relationships through interaction with them.

Individuals with attachment anxiety are prone to show aggressive behavior after being treated unfairly (76). They also hold contradictory attitudes toward their attachment figures, exhibiting dependent and rejecting behaviors simultaneously (30). Therefore, if the need for security is unsatisfied, individuals with attachment anxiety tend to behave in extreme ways and try to impose their psychological pain on their attachment objects. Students who perceive neglect and threat from their school peers during home quarantine may respond in this way, which is actually facilitated by distance and online communication. However, when individuals believe that they have been unfairly treated by their schools and teachers, it is difficult for them to retaliate and they may turn to their smartphones to vent their emotions. Furthermore, this study only focused on student–student relationships in the school context. During the confinement, students can compensate for the loss of companionship at school by strengthening their relationships with their neighborhood peers. Hence, it is understandable that no significant relationship was found between student–student relationships and smartphone addiction.

Similar to teacher–student relationships, school connectedness was negatively correlated with attachment anxiety and smartphone addiction. School connectedness also significantly mediated the relationship between attachment anxiety and smartphone addiction. These findings indicate that students with high attachment anxiety tend to lack school connectedness during the confinement, which could lead to smartphone addiction. In contrast, university students with strong school connectedness are less likely to develop smartphone addiction in the confinement period.

When away from school during quarantine, students have little direct contact with their schools. The lack of interaction and participation in school life can lead to university students feeling detached from their schools. Students in Chinese universities are provided with uniform housing, and they typically socialize in dorms of four to eight people. Students who board on campus can focus more on their studies, form better relationships with others, and feel more connected to the school (77). The lack of parental love and care may also be partially made up for by the campus community. The safety and health of students with high attachment anxiety, as well as the development of wholesome living and studying practices, are also ensured by school policies, peer supervision, and teachers' guidance (78). It is critical to cultivate a student-first campus culture and a diversified boarding life, where students can express emotions, learn to collaborate, and deal with peer relationships through numerous school activities (79). Generally speaking, Chinese universities have complete, student-oriented facilities with varied activities and caring living arrangements to provide emotional support for students. However, the quarantine policies and home-based learning brought by COVID-19 forced students to leave school suddenly. If universities are not able to provide good services, varied activities, and more supervision during the confinement via the internet, there is a greater chance that students will become addicted to their smartphones. If schools could introduce more measures to increase the sense of belonging during the confinement period, smartphone addiction could be alleviated to a great extent.

A multiple mediation model was established in this study. Examination of the mediation effects revealed significant differences between teacher–student relationships and school connectedness, and student–student relationships and school connectedness. The contrasting mediation effects indicate that these two mediators relieve smartphone addiction rather differently.

The literature review suggested that teacher–student relationships often appear to be part of the school relationship. Relationships with teachers and school peers have been used to measure students' relational maladjustment in school (58). In the current study, we examined school connectedness in the same dimension as interpersonal relationships with teachers and school peers. The significant difference in mediation effects between school connectedness and teacher–student relationships confirms our expectation that teacher–student relationships can operate as an independent mediator. Teachers and schools alleviate smartphone addiction differently. Positive relationships with teachers can provide students with warmth and security, which may satisfy their psychological needs and in turn relieve their Internet addiction (48). Likewise, encouraging school appreciation (56) and strengthening school connectedness can offer students a sense of participation and security that could help to prevent smartphone addiction.

Although student–student relationships did not have a significant mediation effect, there was a significant difference in mediation effects between student–student relationships and school connectedness. This highlights the importance of the mediating effect of school connectedness in the multiple mediation model, and suggests that a strong sense of school connectedness is more helpful than student–student relationships in alleviating students' mobile phone addiction.