Mental well-being, often synonymous with positive mental health, encompasses a multifaceted understanding of self-acceptance, personal growth, resilience, autonomy, and mastery of one’s environment (1). Rather than solely focusing on mental illness, the World Health Organization (WHO) increasingly emphasizes the significance of positive mental health for overall population well-being (2). Over the past two decades, mental well-being has garnered substantial attention. The WHO defines positive mental health as a “state of well-being in which individuals recognize their own abilities, effectively cope with life’s normal stresses, contribute productively, and engage with their community” (3). This definition underscores that mental health extends beyond the mere absence of illness (4).

Based on the initial concept of mental well-being proposed by scholars such as Jahoda, Kayes et al., Waterman et al., and others (1, 5–8), Tennant and colleagues proposed a two-dimensional model of mental well-being consisting of hedonic and eudaimonic aspects (9, 10). The hedonic aspect refers to the individual’s subjective feelings of happiness and life satisfaction, whereas the eudaimonic aspect relates to psychological functioning and the actualization of an individual’s potential, capacity, and positive relationship with self and others. These early models significantly influenced the development and validation of the Warwick-Edinburgh Mental Well-being Scale (WEMWBS) (9, 10).

Since the scale’s publication in the UK, the WEMWBS has been translated into many languages and used in various countries and regions. Interested readers can visit the official website for a list of translations (https://warwick.ac.uk/fac/sci/med/research/platform/wemwbs/using/translations/). Numerous validation studies have been conducted on both the original 14-item scale (11–16) and the 7-item Short Warwick-Edinburgh Mental Well-being Scale (SWEMWBS) (17–28). The Chinese version of the SWEMWBS was first translated by Ng et al. in 2014 (17) and has been used in several studies since then (21–23, 29–33).

In terms of validating the SWEMWBS, many studies have utilized classical psychometric evaluations to examine reliability and validity, including factor analyses for structural validity and conventional techniques for convergent and divergent validity (20, 23, 25, 26, 34–36). Some validation studies also included measurement invariance assessments to provide further evidence for the scale’s validity (20, 34, 35). The exploratory and confirmatory factor analysis results of most studies affirmed that a single-factor model with global goodness-of-fit indices fit best to the data (20, 23, 25, 26, 34). Reliability assessments indicated reasonable reliability for the single-factor scale, with Cronbach’s alpha values ranging from 0.77 to 0.88 (20, 26). Sarasjärvi et al. reported a reliability measure of McDonald’s omega at 0.92 (25). For convergent validity, Ng et al. found moderate correlations between the scale and the General Health Questionnaire (GHQ-12) (r = -0.52, p < 0.001), Happiness Index (r = 0.53, p < 0.001), and EQ-VAS (r = 0.40, p < 0.001) (23). Koushede et al. reported moderate to high correlations with WHO-5 (r = 0.75, p < 0.05), Self-rated Health (r = 0.48, p < 0.05), PHQ-4 (r = -0.58, p < 0.05), and PSS (r = 0.65, p < 0.05) (20). Melendez-Torres et al. found correlations with Life Satisfaction (r = 0.50, p < 0.001), Somatization (r = -0.53, p < 0.001), School Pressure (r = -0.30, p < 0.001), and Bullying Victimization (r = -0.30, p < 0.001) (35). More recently, Sarasjärvi et al. found high correlations with GHQ-12 (r = -0.758, p < 0.001), MHI-5 (r = 0.731, p < 0.001), BDI-6 (r = -0.866, p < 0.001), and EUROHIS-QOL8 (r = 0.718, p < 0.001) (25). However, low to moderate correlations were found between SWEMWBS items and Need Satisfaction (r = 0.544, p < 0.001), Need Frustration (r = -0.359, p < 0.001), and Depression (r = -0.33, p < 0.001) in Zayed et al.’s study (26). Criterion validity was assessed in some studies (20, 33). For example, Ng et al. found that participants with lower self-rated health were more likely to have lower than medium well-being (OR = 9.1, 95% CI = 8.05-11.23) (20).

The Rasch Model and Item Response Theory (IRT) have been used as alternative approaches for validating WEMWBS for a long time (37, 38). These methods have become more frequent in recent validation studies on SWEMWBS (14, 28, 39). Results from these studies suggest that all items have high discrimination ability with relatively large discrimination parameters. For difficulty parameters, the scale appears to measure and discriminate respondents better with a negative latent trait, indicating a lower level of mental well-being (28, 39). Overall, the scale showed good functioning along the latent trait continuum, particularly within the range of about -2.0 to 2.0 (28, 39).

Few studies have employed both Classical Test Theory (CTT) and IRT to examine the structural validity of the SWEMWBS using factor analyses. Smith et al. evaluated the scale with factorial analyses and also utilized the Graded Response Model (GRM) of IRT to examine age differences in the latent factor (Theta). No statistically significant differences were found between middle-aged and young adults (14).

For the Chinese SWEMWBS, of the eight studies identified in the literature (17, 21–23, 29–32), four validation studies employed the classical approach (17, 21–23). These studies utilized samples based in a single location, namely Hong Kong, except for one conducted in China. One study involved patients with mental illness (17), another university students (21), and the China study utilized small samples of healthy participants and patients with chronic diseases (33). Sun et al. conducted the only study on the general population (22). To the author’s knowledge, no study has been conducted on the Chinese SWEMWBS in the community employing both classical psychometric evaluation and IRT. Hence, this study aims to further validate the Chinese SWEMWBS using a community sample of Macau residents and adopting both Classical Test Theory, particularly factor analysis and measurement invariance, and IRT approaches.

The growing importance of mental well-being as a global indication of mental healthiness addresses the balance between the negative focuses on mental health problems and illnesses and the positive development of mental wellness. The availability of a well-constructed and validated instrument for the measurement and assessment of the well-defined construct of mental well-being is of paramount importance for the ongoing development of research and clinical practice in the discipline of mental public health.

The SWEMWBS, as one of the few instruments designed specifically for the assessment of mental well-being, has been properly translated into different languages. With the growing development in the field of mental well-being research, there is also an increasing demand for the scale to be used as the outcome measure of research and practice globally. Hence, more substantiating evidence on the reliability and validity of different language versions of the scale will certainly enhance the confidence of researchers and practitioners in the adaptation and usage of the scale. In recent years, an emerging trend in the interest of mental well-being, particularly intervention programs in improving the mental well-being of young people, in the East Asia region has been observed (42) The availability of a theoretically-based and well-designed scale with strong substantiating evidence in the psychometric properties for assessing mental well-being in the Chinese and other East Asian languages will further motivate research in this field.

This study is the first to employ both CTT and IRT approaches for validating the Chinese SWEMWBS. The factor loadings and the measurement invariance results of the factor analyses in this study were comparable to those obtained from the studies by Anthony et al. and Melendez-Torres et al. (34, 35). In terms of the model fit statistics, particularly the greater than the expected values of the RMSEA (0.06) in the full model and the models for males and females, a possible explanation is that there would be a few larger residuals involved that might have skewed the calculation resulting in a larger value. This phenomenon is worthy of more investigation, given that all other goodness-of-fit statistics indicated a reasonably good fit of the model to the data. Another possible explanation is that some other models, such as a two-factor model, may provide a better fit to the data. Considering the overall evidence provided by the model statistics, in conjunction with the volume of evidence from the literature, the single model is still the best option. For the measurement invariance, the results of the current study echoed those obtained by the two reports. In this study, the changes in CFI and RMSEA were less than or close to -0.01 across models. These results were similar to that obtained by Anthony et al. (34). These suggested that the model fits did not change significantly from the most unrestricted configural model to more restricted invariance Metric and Scalar models demonstrating the measure invariance between groups.

The IRT GRM provides further evidence for the scale’s validity with large values of discrimination parameters demonstrating the high discriminative ability of all items, particularly item 4. The difficulty parameters at each threshold also showed good performance of all items along the latent trait continuum. The five-category response set was also shown to be appropriate and adequate, as well as performed well based on the results of the CCC. Different peaks were shown clearly in the curves for all items. For the whole test, the functioning of the scale was demonstrated to provide good information in the TIF graph. In comparison to the study by Hanzlová among adolescents in the UK, these results were comparable and consistent (39). In contrast to the few validation studies of SWEMWBS identified in the literature using the IRT approach, further examinations of the scale by GRM with different gender groups were performed. The results indicated slight differences in the discriminative power of the scale between males and females in favor of the male counterpart. Similarly, differences were also found in the difficulty parameters at each threshold with larger values for males on the negative latent trait than females but larger values for females on the positive latent trait than males. This finding suggested that males and females respond differently to the scale, although the extent of differences may not be problematic. This has not been reported in the literature and can be considered unique. These results could be due to the characteristics of the respondents in the sample or a gender difference in the interpretation of the translated wording of the scale. As such, it is worthy of further examination in future studies. In terms of the impact of gender differences on the validity of the scale, given the small extension of these differences, there may not be a very significant bearing. However, there could be an implication to the utility of the scale for assessing the mental well-being of the general adult population. The study results suggested that the scale could be slightly more sensitive in measuring the mental well-being of males than females. Hence, users of the scale should take note of the possible gender differences in the result of the assessment. Using the standardized conversion score suggested by the original authors could probably help in interpreting the test results. Future validity studies could be conducted to further examine the gender differences in the responses to the scale.

Some limitations have been identified in this study. First, is the skewed gender and age distributions of the sample. The sample consisted of nearly 70% of females with 45% of respondents in the age group 18-34 years. These reflected some gender and age imbalance in sampling. This may pose some limitations to the generalizability of the results obtained. Second, although it is not the aim of the current study to implement a full Classic Test Theory validation study, there is a lack of data using other instruments in assessing a similar construct of mental well-being, such as WHO-5 and PHQ-9, in this study. As such, the convergency validity of the scale cannot be fully examined. In future validation studies, a more gender and age-balanced sample should be used to ensure the generalizability of results. Moreover, it is prudent to include other measures of mental well-being for a more comprehensive examination of the validity of the scale.

In conclusion, the psychometric properties of the Chinese version of the SWEMWBS have been further examined using the Classic Test Theory and the Item Response Theory approaches. Results indicated that as a single-factor scale, on both the scale and item levels, the quality of the instrument is good, with high discriminative power and an adequate response set for assessing a full range of the latent trait, namely mental well-being.