The analysis showed that significant improvements were found in knowledge and help-seeking intentions among all students involved, and stigma change was relatively minor. This is consistent with what has been found in many MHL interventions 6, 8. Our observation is that student attitudes towards mental health were already very positive (indicating lower stigma), and stress levels were relatively low at baseline, suggesting a potential ceiling effect for improvement.
Past research demonstrates that effectiveness of an intervention depends not only on how it is designed and implemented (its mechanism), but also on contextual factors such as age, gender/sex, ethnicity, and English proficiency 3, 9, 18. We found male students achieved significant gains of knowledge and help-seeking intentions, while female students showed significant gains only in knowledge acquisition, which is consistent with similar school-based studies 11. No significant improvements were found in stigma, stress, or well-being among both male and female students. Previous studies identify inconsistent gender differences in MHL outcomes 12, 14. For instance, Wong et al. 14 demonstrated that female participants believed medications to be relatively more harmful compared to male counterparts, whereas Wang et al. 15 found no gender differences on help-seeking behaviors, stigma, or knowledge among Asian-American high school students. However, these cross-sectional studies lacked long-term observations.
By using a longitudinal design, the present study found that, except for help-seeking, both male and female students improved their MHL and general health outcomes at a similar pace (Table 3). Furthermore, ANCOVA analyses revealed that gender was the most significant predictor of change among three attitudinal/emotion-related MHL outcomes (stigma, stress, and well-being); however, this had little to do with knowledge acquisition (Tables 4 and 5). This is an important finding given that the role of gender in the effectiveness of school-based MHL interventions was not reported in available systematic reviews on this topic 8, 33–34.
The impact of students’ ethnicity on the effectiveness of the intervention was also investigated. Only knowledge improvement was statistically detectable in the students with Chinese heritage. In contrast, non-Chinese students demonstrated statistically detectable improvements in knowledge, stigma, and help-seeking. As such, in order to thoroughly assess the effectiveness of a school-based MHL intervention in multicultural contexts, participants’ ethnicity needs to be considered. More information is needed to further investigate why students with Chinese heritage performed differently than non-Chinese counterparts.
We found that ESL students achieved no significant improvement in stigma, help-seeking, or well-being, but rather demonstrated significantly increased stress at post-test (Table 3). Their mental health knowledge was the same as Chinese students with significant gains (Tables 3 and 5). This may suggest that both ethnicity and English proficiency impacted some of the intervention outcomes; however, English proficiency exhibited more influence than ethnicity, particularly in the case of knowledge acquisition where changes were solely predicted by English proficiency (Table 4). Moreover, findings from the ANCOVA analysis suggest that a lack of English proficiency during the intervention may have increased the stress levels of ESL students (Table 3). This may inform schools to adapt the intervention and its delivery to accommodate ESL students; perhaps by using lay English, or by relying less on language-dependent learning styles and pedagogies (e.g., audio and visual teaching methods, contextual clues, English-only scaffolding opportunities). The analysis further indicated that non-Chinese students performed similarly to non-ESL students in improving MHL outcomes. The similarity may be due to a large overlap effect; almost 82% of non-Chinese students were non-ESL students. Overall, the intervention was most effective among students who were both non-Chinese and non-ESL and was least effective among those who were both Chinese and ESL (Tables 3 and 5).
The impact of students’ language proficiencies on mental health knowledge acquisition warrants further investigation. Contemporary theories of learning claim that learning involves acquisition of knowledge, attitudinal change, and shifts in belief, emotion, and behavior. MHL interventions can be viewed as learning processes that instigate and measure changes of knowledge, attitudes, emotions, and behaviors 3, 35. One’s cognition is socially constructed and context-dependent 20, and language plays a fundamental role in our cognitive development, knowledge acquisition, thinking, emotions, behaviors, and interactions with others 36. When assessing the effectiveness of MHL interventions, English proficiency, as a cognitive factor, can provide powerful insight into MHL outcomes. The present study is the first attempt to examine MHL interventions from social and cognitive theories of learning.
The present data show that, after controlling for previous MHL experience, English proficiency was the sole significant predictor for students’ knowledge acquisition, whereas ethnic background was the only significant predictor for the change in the attitude-related MHL outcomes. As stated, attitude is a multifaceted construct that may be more resistant to change than forms of knowledge 37, which could help explain why stigma and help-seeking intentions changed relatively less than knowledge. This also suggests that change in mental health knowledge may be a prerequisite of the changes in attitude-related components of MHL considering the theory about cognitive influences on attitude 34. Presumably, delayed effects on attitudinal and behavioral changes could exist and be worthy of future exploration.
Table 4 shows that there were no significant interaction effects of ethnicity and English proficiency on MHL and general health, which rejected our third hypothesis. Regardless, MHL intervention designers and practitioners should consider students’ ethnicity and English proficiency when designing, implementing, and assessing school-based MHL interventions in multicultural and multilingual settings. Mansfield et al. 38 advised that adolescent MHL research should develop more theoretically and psychometrically reliable, valid and feasible measures that better suit adolescents’ cognitive development. Given that language as a cognitive tool plays a central role in development and learning 39, it should be considered an important variable by researchers and practitioners. English proficiency is malleable and can improve, which may lead to improved MHL outcomes and enhanced effectiveness of school-based MHL interventions.
Finally, our findings shed light on issues that may be worth further inquiry. First, as an evolving concept, the widely accepted MHL definition is multi-faceted and embraces various constructs: cognition, attitudes, emotions, and behaviors 3. Spiker and Hammer 35 called for reconceptualizing MHL as a theory containing interrelated and independent constructs, rather than a single construct containing different dimensions. Following this logic, Mansfield et al. 38 proposed that a multi-construct theory of MHL may help us understand ways in which MHL constructs relate, enabling the formation of better integrated theories to improve adolescent mental health. Our study suggests further empirical and theoretical investigation of interrelationships among MHL constructs. Specifically, the present study applied social and cognitive theories of learning to a school-based MHL intervention that reconceptualized English proficiency as a cognitive factor underlying the effectiveness of the intervention. Considering the theoretical rationale of our hypotheses and analysis of distinctions between Chinese and non-Chinese Canadian students, future research should also involve other ethnic minority groups to test the generalizability of these hypotheses.
One limitation of the study must be noted. English proficiency was measured by students’ subjective self-reports and was coded as a binary variable. Compared to continuous measures, dichotomous measures are more likely to lose information; it is difficult to measure and disseminate nuance from such responses 40. Follow-up research may use standardized assessments of English proficiency as a continuous variable. Another limitation is that all the participants were from the same grade, making it difficult to ascertain how students’ developmental stage, personality, and socialization could impact effectiveness of the intervention.