2.1. Guam adolescents and depressive symptoms

We focus on Guam as recent research indicates that residents on the island are prone to experiencing high levels of psychological distress [16]. In particular, data from the Guam State Epidemiological Profile [17] indicate the percentage of high school students who reported they felt sad or hopeless almost daily “for 2 weeks in a row, within the past 12 months” (p. 106), which are symptoms of depression, had increased substantially between the years 2015–2019. According to 2019 Youth Risk Behavior Surveillance System (YRBSS) data, about 47% of high school students in Guam reported feeling sad or hopeless in the past 12 months compared to about 37% of them in the United States as a whole, who felt the same [18]. Around 11% of high school students in Guam reported that they did not go to school because they felt it was unsafe, compared to about 9% of them in the United States [18]. These relatively higher rates of depression and feeling unsafe at school among Guam youth should be considered mental health risks that could interfere with their healthy socialization with their parents and peers. In the more relationship-oriented cultural context of Guam, a mixed-methods study found that relational aggression and peer victimization (i.e., interpersonal harm and manipulation) and feelings of insecurity in relationships played a role in the development of depressive symptoms among young adults [19]. Given these previous findings, it is important to investigate factors that could serve to either protect or put Guam adolescents at risk for developing depression.

2.2. Parental support, depression and social media addiction

First, it is possible that for adolescents in Guam, greater levels of PS would be negatively associated with DS. There is robust evidence that PS is associated with lower levels of depression and DS among adolescents [10,11,20–23]. The general benefits model suggests that it is through having access to parental support that adolescents are more equipped and prepared to cope with adversity and stress. This, in turn, could serve to protect adolescents from psychological distress such as depression. A meta-analysis of 341 studies conducted by Rueger et al. [20] found support for the general benefits model, suggesting that PS could reduce depression. Research conducted with rural youths in the U.S. indicates that having positive relationships with both parents and peers is associated with a lesser likelihood of reporting depressive symptoms [4]. To our knowledge, there is currently no prior research on the association between PS and DS among Guam adolescents, and this was confirmed by a thorough literature search. However, research conducted in Hawaii by Wills and colleagues suggests that PS has an overall effect in buffering the positive relationship between experiencing negative life events and substance use among Hawaii adolescents [24]. We should note that the study found that this buffering effect for PS was observed only among Filipino and other Asian adolescents, but not for Caucasian or Native Hawaiian or other Pacific Islander (NHOPI) adolescents. Wills and colleagues explained that for NHOPI youths in Hawaii, particularly, a buffering effect may also come from support from a network of extended family, cousins, and siblings. Such youths belong to communities characterized by interconnected extended family networks (e.g., uncles, aunts, cousins). Additionally, a comprehensive review of research on suicide among U.S. Pacific Islanders living in Hawai’i, American Samoa, and the Micronesia region (located in the Western Pacific) found that having low family support serves as a risk factor for suicide attempts by youths [25].

Second, it is possible that for adolescents in Guam, greater levels of PS will be negatively associated with SMA. As noted above, SMA refers to the compulsive or problematic use of social media [7]. SMA is part of the overarching construct of internet addiction. Two recent studies suggest that parental support is negatively associated with adolescents’ SMA [26,27]. Hwang and Toma [27] found that higher levels of perceived PS are associated with a lesser likelihood of problematic internet use. In a similar vein, Ren and Zhu [26] found that adolescents who perceived their parents to be supportive were also less likely to both spend time online and use the Internet for leisure activities. Ren and Zhu also revealed that adolescents who view their parents as being supportive, also report that their parents play an active role in mediating their online behaviors. These include explaining why some online sites are good or bad, or talking about contents that are and are not appropriate to share online. Furthermore, active parental mediation was inversely associated with the amount of time adolescents spent online [26]. In sum, the general benefits model of social support [20] and the research reviewed above provide evidence that PS could serve as a protective factor against SMA.

2.3. Social media addiction and depression

As mentioned, extant research conducted among US adults provides evidence that excessive and problematic social media use is positively related to depression [28, 29, 30]. Although they did not examine social media use specifically, Fang et al.’s [31] recent study summarized theories and empirical research that suggest three possibilities for the direction of the relationship between the problematic use of the internet (PIU) and depression.

First, it is plausible that SMA predicts DS. This possibility is at the heart of displacement theory [31]. According to this theory, as individuals spend more time online, this displaces the time they are able to spend offline with others, including their peers and family members. Recently, however, Hall and Liu [32] have held that there is limited evidence that time spent on social media displaces time spent on face-to-face interactions. Instead, the use of social media might displace other types of media activities. Yet there is evidence that dysfunctional and excessive use of the Internet could lead to Internet addiction, which subsequently, could result in psycho-social health problems [33]. Fan, for instance, found that internet addiction positively predicts psychological distress [34]. Lam and Peng have shown that internet addiction could predict depression [35]. One mechanism that could explain why SMA is positively associated with DS among adolescents is found in the tendency toward social comparison. Specifically, when individuals use social networking sites to compare themselves to others whom they perceive to be more superior—a phenomenon referred to as upward social comparison—this puts them at a greater risk for developing depression [36,37]. Research shows that excessive use of social media often results in using social media platforms for social comparison [38]. As Pantic has held, the proclivity to use social media to compare oneself to others “may lead to incorrect conclusions regarding physical appearance, educational level, intelligence, moral integrity, as well as many other characteristics of online friends” [39]. Similarly, Kim et al. [38] have shown that addiction to social networking sites could increase upward social comparison, which subsequently could result in lower self-esteem.

Second, according to the cognitive-behavioral model of PIU, psychopathologies such as depression are predictive of problematic uses of the Internet and its symptoms. Davis’ conceptualization of the cognitive-behavioral model specifically assumes that psychopathologies (e.g., depression) could result in maladaptive cognitions such as self-doubt and holding a negative view of oneself [40], which could then lead individuals to turn to the Internet to cope. Individuals may come to believe that the Internet, as opposed to offline interaction, provides a non-threatening outlet where they feel accepted and respected. This could then lead to the eventual overuse of the Internet. Based on the cognitive-behavioral model, we can expect that psychopathologies such as depression could result in maladaptive cognitions (e.g., having a feeling of low self-worth), which could then lead individuals to turn to social media to cope. Individuals may perceive social media to be less threatening than offline interactions, which may then lead to the excessive and compulsive use of social media.

Yet recent studies have also found evidence that the association between problematic internet use and depression could also be bidirectional [41,42]. Thus, it is also plausible that changes in depression predict changes in SMA, and vice versa. Adolescents who display depression might be prone to hold maladaptive cognitions (e.g., feeling low self-worth). As a result, they may cope with feelings of depression by using social media, which could then result in the excessive and addictive use of social media. In turn, as adolescents excessively use social media, they may also be led to compare themselves to their online friends. Comparisons to online friends perceived to be more superior, could lead to a sense that oneself lacks in physical appearance, intelligence, and other characteristics [38]. In this case, SMA could then be associated positively with DS. There would then be a continuous cycle between SMA and depression and vice versa.

In sum, the above literature on the protective role of PS suggests that it would be negatively related to both SMA and DS. Also, the literature suggests SMA is positively associated with DS, however, the nature of this relationship could be either unidirectional or bidirectional. Yet given the limited research in this area conducted in Guam, we propose and investigate the following research question for Guam adolescents:

RQ1. How are PS, SMA, and DS related?

2.4. Potential role of ethnicity

As noted previously, our study also explores the potential that the associations between PS, SMA, and DS would differ based on ethnicity. According to most recent population statistics [43], CHamorus, who are the native/indigenous people of Guam, and Filipinos, respectively make up around 33% and 29% of the island’s population. As the two largest ethnic groups in Guam, CHamorus and Filipinos in Guam share certain cultural norms and values, which can be attributed to their parallel historical experiences, namely centuries of Spanish and U.S. colonialism [44]. Non-CHamoru Micronesians in Guam, who originate from the Freely Associated States (FAS), are the third largest ethnic group in the island, and make up more than 11% of the population. The FAS group includes indigenous peoples originating from the Federated States of Micronesia (FSM; including Chuukese, Kosraeans, Pohnpeians, and Yapese), the Republic of Palau (Palauans), and the Republic of the Marshall Islands (Marshallese). FAS migrants are allowed to enter Guam or the United States without visa restrictions under the Compacts of Free Association with the United States, which were signed in 1986 for the FSM and the Republic of the Marshall Islands and in 1994 for the Republic of Palau.

While there is scant research relevant to PS and SMA among youths representing these three ethnic groups living in Guam, a study conducted by Bosqui et al. suggested that adult CHamorus and other Pacific Islanders in Guam (namely, those from the FAS) are at a greater risk for experiencing psychological distress than Filipinos [16]. The study’s results also suggested that CHamorus have the greatest risk for psychological distress among the three groups. The researchers partly attributed these findings to the higher rates of poverty and levels of disadvantages experienced by CHamorus and other Pacific Islanders in Guam, compared to Filipinos. In sum, given these potential differences in psychological distress between the three largest ethnic groups in Guam, we investigate the following research question:

RQ2: Will the relationships between PS, SMA, and DS differ based on ethnicity?

3.1. Ethics statement

The University of Guam Institutional Review Board (IRB) committee, which serves as the regional IRB board, approved this study (approval number for the study is CHRS #22–69). Both written parental consent and student assent were sought before students were allowed to participate in the survey. In addition, permission to conduct the study was granted by the Guam Department of Education (GDOE) and the principals of each of the 8 public middle schools.

3.2. Procedure

The present study conducted a secondary analysis of survey data collected for a pilot clinical trial that tested the efficacy of a school-based substance use prevention curriculum for Guam middle school students. We analyzed survey data from Wave 1 (W1; baseline) and Wave 2 (W2; posttest) of the primary study. These surveys were administered from September 30, 2022 through May 9, 2023. We conducted multivariate analyses of variance to test whether there were statistically significant differences in the mean scores between the intervention and control groups along all our key variables in W1 and W2. Because these results were not directly relevant to our main analyses, we do not include them in this study, but we have included them as a supplement (S1 Text).

These preliminary analyses revealed that there were no statistically significant differences (p < .05) in the mean scores between the intervention and control groups along all our key variables. However, we included group as a covariate in our models. The average class size for the schools that participated in the study was approximately 15–25 students. Participants were assured that their responses would remain anonymous as the survey did not ask for any identifying information. Surveys were self-administered and facilitated by research assistants. The survey took approximately 30 minutes to complete. As compensation for their time and effort, all students received a $10 gift certificate ($5 for each Wave) to a local store. The survey contained items that measured demographics, substance use, social media addiction, parental support, and depression. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

3.3. Participants

W1 data were collected during baseline and W2 data were collected one month after baseline. All 8 of Guam’s public middle schools participated in this study. There were a total of n = 538 completed surveys for W1 and n = 507 for W2. The overall response rates were roughly 76% for W1 and about 71% for W2, suggesting that around 5% of the sample dropped out during W2. The participant characteristics for Waves 1 and 2 are reported on S1 Table. Of note, for both Waves, the ages of participants ranged from 10-15 years, with roughly 52% male and 48% female for W1, and roughly 53% male and 47% female for W2 (for comparison, male students comprise of about 52% of the GDOE population as a whole) [45]. The grade levels for both W1 and W2 were roughly similar, although there were slightly less 6th graders in W1 (about 24%) than there were for W2 (about 26%), and slightly more 8th graders in W1 (about 46%) than W2 (about 44%). We should point out, however, that our sample may have over-represented 8th graders, as based on data compiled by the Guam Department of Education, the true population of the island’s public middle schools comprises of roughly 33% 6th graders, 34% 7th graders, and 33% 8th graders [46]. Regarding ethnicity for both W1 and W2 data, roughly 35% were CHamoru, about 31% were Filipino, about 32% were from the Freely Associated States (FAS), and about 2% were other. Comparatively, population-level data from the Guam Department of Education indicate that there are around 44% CHamorus, 22% Filipinos, and about 31% FAS students enrolled in Guam middle schools [46]. Thus, our sample may have over-represented Filipino students while underrepresenting CHamoru students.

3.4. Measures

3.5. Data analysis strategy

We used SPSS version 30 to analyze the descriptive statistics for the demographics and covariates and the mean scores and reliabilities of the key indicators. We used cross-lagged panel design to test the proposed hypotheses [49, 50, 51]. This analysis allows us to estimate the influences of the predictor on the outcome, with the influences of the lagged outcome taken into account. This procedure allows for making more precise predictive inferences than cross-sectional analyses. We took into account age, biological sex, grade levels, condition, and ethnicity (Filipino, Freely Associated States). The errors in the focal variables were allowed to covary within each wave. For all analyses, listwise deletion was used to handle missing data.

To evaluate model fit and test our hypotheses, we used R package, lavaan [52]. We examined the following fit indices – chi-square test, comparative fit index (CFI), standardized root mean square residual (SRMR), and root mean square error of approximation (RMSEA). We used the following criteria as acceptable fit – a value close to.95 for CFI, close to.06 for SRMR, and close to.06-.08 for RMSEA [53,54].

In addition, as part of our investigation for RQ1, we conducted a multivariate analysis of variance (MANOVA) on SPSS Version 30 to examine whether there were differences in the scores for the key variables in both Waves 1 and 2 between the three ethnic groups—CHamoru, Filipino, and FAS. We additionally conducted a multigroup path analysis on R with the three ethnic groups. We first fitted the model for each group to determine whether it was the same across the groups. Next, we constrained the path coefficients to be equal across the groups and tested whether the initial, configural and constrained models differed in model fit. If the constrained model fitted the data better, we freed each hypothesized path coefficient one at a time to determine which path coefficient(s) differed by ethnicity.