Psychological distress and problematic internet use among language teachers: a latent profile analysis

Teachers’ psychological distress

Psychological distress is a multifactorial construct that refers to non-specific symptoms of depression, anxiety, or stress more generally (Belay, Guangul & Asmare, 2020; Chen, Qu & Hong, 2022; Monteagudo et al., 2023). High levels of distress can serve as a marker of impaired mental health (Zhu et al., 2022). Teaching is one of the most stressful professions (Gadermann et al., 2023; Lizana & Lera, 2022). Johnson et al. (2005) reported that teachers’ psychological health, physical health, and job satisfaction were lower than other professions like researchers, police, and medical workers. The increasing prevalence of mental health issues among teachers over recent decades, along with their broad social implications, has garnered global attention, highlighting the critical importance of preventing PD and fostering mental well-being in the teaching profession (Agyapong et al., 2022; Liu, Li & Fang, 2024; Rojas-Andrade, Aranguren Zurita & Prosser Bravo, 2024; Yang et al., 2019).

Although all teachers face growing pressures and complexity in their professional lives (Day & Qing, 2009), language teachers may suffer greater PD than their counterparts in other subjects (Piechurska-Kuciel, 2011). A survey conducted in China revealed that the prevalence of depression among foreign language teachers (17.1%) significantly exceeded that of math teachers (14.4%), while anxiety rates reached 54.7%—higher than those of math teachers (48.8%), and teachers of other subjects (49.1%) (Fu et al., 2023). The phenomenon may be particularly pronounced in primary and secondary education (Ozamiz-Etxebarria et al., 2021a; Ozamiz-Etxebarria et al., 2021b), where teaching autonomy and curricular flexibility are more restricted compared to higher education faculty (Piechurska-Kuciel, 2011; Teven, 2007).

The rising PD among language teachers may originate from unique challenges inherent in language teaching, such as linguistic anxiety (Cardoso-Pulido, Guijarro-Ojeda & Pérez-Valverde, 2022). In addition to these challenges, the technology-driven era has brought additional PD risks related to technology. First, the digital revolution has fundamentally altered their professional landscape. The proliferation of online learning resources has eroded teachers’ traditional knowledge authority (Cardoso-Pulido, Guijarro-Ojeda & Pérez-Valverde, 2022; Mercer, Oberdorfer & Saleem, 2016), with many no longer perceiving themselves as the “sole source of knowledge and information” in their language classroom Mercer, Oberdorfer & Saleem (2016, p. 215). This paradigm shift constitutes what Cardoso-Pulido, Guijarro-Ojeda & Pérez-Valverde (2022, p.3) term a “powerful disruptive element” to language teacher well-being. Second, technological advancements, especially in artificial intelligence, present existential challenges to the profession (Gao et al., 2023; Rashid & Kausik, 2024), intensifying job insecurity. Compounding these pressures, global education reforms continue raising performance expectations (Jiang, 2017; Li & Liang, 2024), despite persistently low salaries (Fu et al., 2023; Mercer, Oberdorfer & Saleem, 2016). These cumulative stressors create a perfect storm for PD development among language educators. Language education is a crucial element of basic education (Qi, 2016; United Nations Educational, Scientific and Cultural Organization, 2019). The psychological well-being of language teachers can significantly impact their teaching practices as well as students’ language learning experiences and outcomes (Liu et al., 2025; Yang, 2022; Zhou et al., 2024). Therefore, the PD faced by language teachers warrants comprehensive and in-depth investigation.

However, the majority of quantitative studies that investigated the teachers’ PD, or teacher psychology more broadly, have incorporated a variable-centered approach, such as regression analysis, and structural equation modeling, among others (e.g., Cao et al., 2023; Lizana & Lera, 2022; Miconi et al., 2024), treating teachers as a homogenous group. While these studies have yielded valuable insights into the factor structures, antecedents, and consequences of teachers’ PD, they have largely overlooked the inherent heterogeneity within this professional group. The population of teachers is heterogeneous (Lassri, 2023), so individual differences may play an important role in the varied responses to PD items.

Previous research (e.g., Li et al., 2020; Lizana & Lera, 2022; Ozoemena et al., 2021) has documented that several demographic characteristics could predict the degree of teachers’ PD, such as gender, teaching experience, school stage, and so on. Female, novice, and primary school teachers have been recognized as being vulnerable to PD by most previous studies (e.g., Fernández-Berrocal et al., 2017; Ozoemena et al., 2021). However, some recent studies reported different observations. For example, Li et al. (2020) reported that male teachers with ages 60 to 100 years and female with ages 50 to 60 had the highest prevalence of anxiety. Ozoemena et al. (2021) found that old experienced teachers have shown more psychological symptomatology than younger teachers, which might be due to the digital gap (Ozamiz-Etxebarria et al., 2021a; Ozamiz-Etxebarria et al., 2021b; Song & Chen, 2019).

Those inconsistent findings may be caused by neglecting the heterogeneity of teaching staff, warranting the necessity of using the personal-centered approach to clarify the predictive roles of these demographic variables on language teachers’ PD. LPA is such a technique, that focuses on finding out the unobserved heterogeneity within a population and ultimately provides us with latent sub-populations within the population (Asendorpf, 2015). This technique provides two key methodological advantages for examining PD among language educators: first, it reveals nuanced symptom patterns that traditional variable-centered approaches often fail to detect (Spurk et al., 2020), enabling a more sophisticated understanding of distress manifestations. Second, it facilitates the identification of distinct high-risk subgroups within the teaching population (Howard & Hoffman, 2018), which is crucial for developing precisely targeted intervention strategies. These capabilities make LPA particularly valuable for investigating the complex psychological experiences of language teachers, who face unique occupational stressors that may produce heterogeneous distress profiles. While Lassri (2023) has utilized this technique to identify the latent profiles of PD among general teachers, this proposition has yet to be investigated among language teachers to the best of our knowledge. Therefore, this study employed LPA to identify the unobserved latent subgroups in a sample of Chinese primary and secondary school language teachers and further explore the demographic characteristics of different latent groups. We aspired to provide a more detailed understanding of the PD experienced by language teachers and offer some practical insights for educators and stakeholders, supporting the professional well-being of language teachers and ultimately benefiting practical language teaching.

Teachers’ psychological distress and problematic internet use

Problematic internet use, also internet addiction (IA), refers to an inability to control one’s use of the internet which leads to negative consequences in daily life (Spada, 2014). In this study, the construct is conceptualized as the extended, compulsive utilization of the internet (He & Chen, 2024), rather than being classified as a pathological condition. It characterizes an uncontrollable urge to engage with the internet, perceiving periods, devoid of the internet access as insignificant, and manifesting excessive irritability and aggression when access is impeded (Young, 2004). PIU specifically encompasses issues such as excessive social media use, smartphone addiction, and problematic gaming behavior. Prior studies have highlighted the detrimental effects of PIU on individuals’ lives and work, including poor sleep quality, reduced physical activity, neglect of family responsibilities, job performance decline, and withdrawal from social interactions (He & Chen, 2024; Kapus et al., 2021; Xu et al., 2021). These consequences are particularly concerning within the teaching profession, where PIU may disrupt pedagogical processes and negatively impact students’ learning experiences and academic achievement (He & Chen, 2024).

While existing research has extensively documented PIU among adolescents (Cao & Su, 2007; Lozano-Blasco, Robres & Sánchez, 2022), far less attention has been paid to PIU among teachers, despite their critical role in shaping students’ digital behaviors (Szymkowiak et al., 2021). This gap is particularly striking in the case of language teachers, whose professional context may heighten their vulnerability to PIU (He & Chen, 2024). Two key factors underscore the urgency of investigating PIU in this sub-population. First, as one of primary role models, language teachers’ digital habits can influence students’ technology use patterns (Lai, Li & Wang, 2017; Szymkowiak et al., 2021). Second, the profession itself entails unique risk factors: (1) the highly interactive nature of language teaching, coupled with its emotional and cross-cultural demands, often leads to psychological fatigue (Horwitz, 1996; İpek et al., 2018; Mercer, Oberdorfer & Saleem, 2016; Wang & Song, 2022), potentially driving compensatory overuse of online platforms; and (2) in today’s technology-integrated educational landscape, the increasing reliance on digital tools (Mercer, Oberdorfer & Saleem, 2016), such as vocabulary apps, video-based materials, and online professional development resources, obscures the line between essential and excessive use, while simultaneously complicating the differentiation between productive use and problematic overuse. Given their dual role as behavioral exemplars and a high-risk occupational group, language teachers represent a critical yet understudied population in PIU research. Addressing this gap is essential not only for safeguarding educators’ well-being but also for ensuring their effectiveness in fostering healthy digital habits among students.

Previous research has demonstrated a robust association between PD and PIU across diverse populations. Among adolescent and adult cohorts, individuals with internet addiction consistently exhibit elevated levels of depression, anxiety, and stress compared to their non-addicted counterparts (Ostovar et al., 2016). This correlation is particularly pronounced in smartphone overuse, where depressive and anxious symptomatology serve as significant predictive factors (Rho et al., 2019). Teacher population demonstrates similar vulnerability to this association. Studies from different contexts highlighted that teachers experiencing PD face an increased risk of developing PIU (Lee & Chen, 2021; Tsumura et al., 2018; Yi et al., 2021). For instance, research conducted in Japanese educational settings revealed that schoolteachers with PD reported significantly higher rates of PIU, with consistent patterns observed across genders (Tsumura et al., 2018). Parallel findings have emerged in Chinese contexts: studies by Lee & Chen (2021) and Yi et al. (2021) further corroborated strong correlations between teachers’ PD levels and PIU tendencies.

Evidence suggests a bidirectional relationship between these two constructs (Odacı& Çıkrıkçı, 2022). On the one hand, PD is a major contributor to PIU. The internet for those frustrated teachers may be a sanctuary from stress and cruelty in reality (Lee & Chen, 2021; Tsumura et al., 2018). Thus, teachers suffering from PD are more likely to engage in PIU (Kızılok & Özok, 2021; Young, 2017). On the other hand, overuse of the Internet will exacerbate emotional perception (Gao & Chen, 2006; Liao et al., 2023), making people more sensitive, and leading to sleep problems (He & Chen, 2024). Poor sleep, further, amplifies basic emotional reactivity, increasing negative mood states (e.g., anxiety, depression, suicidality) (Ben Simon et al., 2020), and intensifying the deterioration of individuals’ mental health. The educational implications of this vicious cycle are particularly concerning. As key factors in students’ proximal environment (Pekrun, 2006), language teachers’ mental health and behavioral patterns significantly influence adolescent development (Cao et al., 2023; Li, Duan & Liu, 2023). Their compromised psychological well-being may create ripple effects throughout the language classroom environment, ultimately affecting students’ socioemotional growth and language academic achievement (Liu & Li, 2023).

Despite these insights, the interplay between PD and PIU remains underexplored among language teachers, particularly in primary and secondary education. To address this gap, the present study employed one-way analysis of variance (ANOVA) to examine differences in the degree of PIU across latent groups with varying levels of PD. By providing empirical evidence, this study aims to deepen the understanding of the complex interplay between teachers’ PIU and PD. It also seeks to offer practical guidance for preventing and addressing PD and PIU among language teachers. In summary, our research questions are as follows:

1. What are the PD profiles of language teachers in primary and secondary schools?

2. What are the demographic characteristics of language teachers with different profiles?

3. What are the relationships between PD profiles and teachers’ PIU?

Participants

A non-probability sampling (convenience sampling) was used to recruit our participants. We recruited our participants from more than 400 primary and middle schools in Sichuan, Jiangxi, and Sichuan, China. In total, 1,848 questionnaires were collected. After deleting those with missing data and those completing the survey in less than 3 min (n = 589), the survey resulted in 1,259 valid responses. There were 664 primary teachers and 595 middle schools, with a predominance of female teachers (primary schools: n = 537, 80.87%; middle schools: n = 557, 93.61%). There was no native English speaker among our participants. Their mother tongue is Chinese with English as their second language. Demographic details are presented in Table 1.

Instruments

Data collection

Due to data availability, the study conducted a cross-sectional online survey targeting primary and middle school teachers in four provinces of China (Sichuan, Jiangxi, and Sichuan provinces), using Sojump, an online questionnaire platform. A non-probability sampling (convenience sampling) was used to recruit our participants. The data was collected between 25 May and 30 June 2020. Initially, we reached out to principals of primary and middle schools in the four provinces. Those who agreed to participate then distributed the survey link to their teachers. Prior to starting the survey, electronic consent was obtained from all participants. Only those who agreed to participate were given access to the main questionnaire. No data were collected or analyzed from individuals who declined to participate. This research project received approval from the Jiangxi Psychological Consultant Association’s Institutional Review Board (IRB ref: JXSXL-2020- J013) and followed the ethical principles outlined in the Declaration of Helsinki.

Data analysis

We first excluded 589 invalid responses, including (1) all cases with missing data (which were found to have completion times under 3 min) (n = 12), and (2) additional cases with completion times below 3 min but no missing data (n = 577). This resulted in 1,259 valid responses. Then, we utilized LPA to identify participants’ latent groups (or profiles). Based on the results of LPA, we conducted multivariate logistic regression to explore the demographic characteristics of different latent groups. After that, we employed one-way ANOVA to reveal the different levels of three types of PIU across various latent groups. The optimal number of profiles is determined by Akaike information criteria (AIC), Bayesian information criteria (BIC), entropy, and bootstrap likelihood ratio test (BLRT). Lower AIC and BIC better model fit. Higher entropy indicates better classification accuracy with a value above 0.80 indicating acceptable delineation of clusters (Scotto Rosato & Baer, 2012). A significant p-value (less than 0.05) in the BLRT suggests that a k class model improves the fit over the k-1 class model (Spurk et al., 2020). The number of profiles to be compared ranges from 2 to 6. We employed R (4.3.3) and SPSS (26.0; IBM Corp., Armonk, NY, USA) for statistical analyses.

The results of latent profile analysis

Table 2 presents the fit indices for different models with varying numbers of latent profiles. AIC and BIC decreased until the number of latent profiles was set at 3. The entropy value was 0. 98, exceeding the threshold of 0.8. Besides, the BLRT test of the 3-profile model was significant (p < 0.001) indicating that the 3-profile solution fitted the data better than the 2-profile model. Although the smallest group comprised only 44 teachers (3.49% of the sample), this group represented the subgroup who have the most severe PD in our sample and revealed important heterogeneity in the data that has significant practical implications (Masyn, 2013). Therefore, we retained this subgroup and selected the 3-profile model as the optimal solution.

Guided by Fig. 1C, we named the three profiles based on the scores for depression, anxiety, and stress. The scores for all three indicators in the first class (C1) exceeded 10, which, when doubled, far surpassed the diagnostic thresholds (10 for depression, 8 for anxiety, and 15 for stress). Consequently, this class was labeled as the “moderate-to-severe PD profile” (n = 44, 3.49%). Teachers in the second class (C2) had scores just above the thresholds, indicating mild PD. Accordingly, this class was labeled as the “mild PD profile” (n = 262, 20.81%). The third class (C3) scored very low on all three indicators, and we named this class the “mentally healthy profile” (n = 953, 75.69%).

Further examination of the outcomes of the 3-profile model demonstrated its strong discriminatory capability. As shown in Table 3, the detection rate for the number of positive symptoms in the “moderate-to-severe PD profile” (C1) was 100%. Over 97% of participants in this latent group exhibited high levels of positive symptoms of depression, anxiety, and stress simultaneously. The detection rate among teachers with the “mild PD profile” (C2) was 92.25%, with teachers in this group typically reporting one or two mild levels of PD. In contrast, the detection rate for the “mentally healthy profile” (C3) was only 1.89%. Most teachers in this group did not report any positive symptoms. In rare cases where positive symptoms were present, they were mild and singular. These findings indicate that the 3-profile model effectively identifies abnormal cases, further supporting the 3-profile solution.

The results of multivariate logistic regression analysis

Based on the results of LPA, this study further explores the demographic characteristics of three different PD latent groups. We conducted a multivariate forward-stepwise logistic regression analysis with the “mentally healthy” group as the reference group. The dependent variable is the latent groups of teachers, and the independent variables include gender, teaching experience, school type (private school vs public school), school stages (primary school vs junior high school), and whether the teacher held the position of headteacher. The results indicated that only gender and teaching experience could enter the final model successfully and had a significant impact on the distribution of PD patterns among language teacher populations (see Supplementary Material).

The results (Table 4) indicated that novice teachers with less than or equal to five-year teaching experience are 0.68 times as likely as those experienced teachers with more than five-year teaching experiences to develop mild PD (OR = 0.68, 95% CI [0.50–0.94]). This means novice teachers are 32% less likely to experience mild PD than experienced teachers. As for the odds of developing moderate-to-severe PD, the two subgroups did not show a significant difference (p = 0.165). In terms of gender, male language teachers were 2.27 times more likely to experience moderate-to-severe (OR = 2.27, 95% CI [1.09–4.71]) and 1.51 times more likely to experience mild PD (OR = 1.51, 95% CI = [1.02–2.21]) than female language teachers. These findings highlight the potential influence of gender and teaching experience on PD among language teachers, but further research is needed to confirm these associations.

The results of one-way ANOVA

One-way ANOVA was employed to find the association between different PD profiles and PIU. The results (Table 5) indicated significant differences across PD groups in levels of smartphone addiction (F = 63.52, p < 0.001, η² = 0.09), problematic social media use (F = 53.98, p <  0.001, η² = 0.08), and internet gaming disorder (F = 115.55, p < 0.001, η² = 0.16). Post-hoc comparisons revealed that teachers experiencing PD (C1 and C2) reported higher levels of smartphone addiction, problematic social media use, and internet gaming disorder compared to mentally healthy teachers (C3). Furthermore, teachers with moderate-to-severe PD (C1) scored significantly higher in internet gaming disorder than those with mild PD (C2).