Mental well-being is a public health concern worldwide; adequate mental well-being is associated with better health-related quality of life and longer life expectancy [1]. In recent years, the mental well-being of employees in several occupations has gained substantial attention [2-6]. A meta-analysis revealed that numerous health care workers had various psychological problems [7]. It is well known that medical staff experience many work-related stresses (eg, prolonged and irregular working hours, night shifts, high-intensity work, emotional exhaustion, chronicity of care, and moral conflicts), which may negatively influence their mental well-being, causing depression, anxiety, sleeping disorders, and other problems. Impaired mental well-being can affect health care providers’ professionalism, quality of care delivery, efficiency, and overall quality of life [8,9].

Moreover, it has been reported that the overall mental health status of Chinese medical staff is unfavorable [10,11]. This finding suggests that the mental well-being of medical staff is critically important to public health [12,13]. For this reason, it is crucial to measure the mental health status of medical staff and identify work-related risk factors to protect their well-being [14].

The Warwick-Edinburgh Mental Well-being Scale (WEMWBS) is a relatively new, short, acceptable scale that has been translated into several languages [15-18]. It has demonstrated excellent reliability, good validity, and internal consistency [19]. Studies of public mental health have confirmed the WEMWBS’s ability to offer rigor in psychological evaluations [20]; it focuses on protective and promoting factors that can provide a rational basis for the orientation of policy makers formulating interventions [21].

Previous studies have reported the psychological performance of the Chinese-language version of the WEMWBS in clinical and nonclinical settings in China, but all have had limitations [15,22]. Research by Liu et al [23] appears to be the earliest psychometric analysis of the Chinese WEMWBS; however, 2 issues need addressing. First, their paper was written in Chinese, making it burdensome to read for non–Chinese-speaking investigators and impeding comparisons of China with other countries. Second, the age of the study participants ranged from 60 to 97 years, resulting in information and selection bias. The generalizability of the findings from Dong et al [22] is problematic, because the 191 patients with chronic heart failure in that study came from 1 hospital in a Chinese city. A study by Fung [24] and an earlier study by Dong et al [15] were limited because all respondents were university students recruited from either a single university or a single hospital nursing internship program in a Chinese city; this could have caused pervasive information and selection bias in these studies’ assessment of the psychometric properties of the WEMWBS. A study by Waqas et al [25] explored the reliability and validity of the WEMWBS in Pakistan; Taggart et al [26] investigated the WEMWBS in a targeted sample of minority ethnic groups living in the UK who self-identified as Chinese or Pakistani by background. Additionally, no previous investigation has combined a graded response model (GRM), item response theory (IRT), and classical test theory (CTT) to evaluate the psychometric properties of the WEMWBS. It is necessary to find a comprehensive method and a better representative sample that covers participants from southern and northern areas to assess the performance of the Chinese WEMWBS.

We administered the Chinese WEMWBS to medical staff to evaluate their psychological characteristics and explore and popularize this questionnaire on mental well-being, which is suitable for Chinese national conditions. We aim to provide theoretical support for improving the mental well-being of medical staff.

From May 15 to July 15, 2020, purposeful sampling was conducted to recruit 572 medical staff online, mainly from 6 hospitals in mainland China (the First Affiliated Hospital of Wenzhou Medical University, the Second Affiliated Hospital of Wenzhou Medical University, the Second Hospital of Dalian Medical University, the Second Affiliated Hospital of Zhongguo Medical University, Lishui People’s Hospital, and Chenzhou Third People’s Hospital).

All participants provided informed consent before participation, and the Medical Ethics Committee of the Second Affiliated Hospital of Wenzhou Medical University approved the study (LCKY2019-288).

Data were collected via a self-administered online questionnaire. The first section collected sociodemographic characteristics, including age, marital status, gender, body weight (in kilograms), height (in meters), professional status, and education level. The second section examined lifestyle habits, including working hours, night shifts per week, smoking history, drinking history, consumption of vegetables and fruit, physical exercise, and self-reported personality. The third section examined mental well-being using the WEMWBS and self-perceived quality of life (QoL). The WEMWBS is a 14-item sequential scale that measures 3 aspects of mental well-being: positive psychological function, emotion, and interpersonal relationship satisfaction. All items were scored on a 5-point Likert scale, including 1 (never), 2 (occasionally), 3 (yes), 4 (often), and 5 (always). The total score ranged from 14 to 70, with higher scores representing stronger subjective well-being. The third section of the questionnaire used the 36-Item Short Form Health Survey, Version 2 (SF-36 v2) to assess self-perceived QoL. The SF-36 v2 is a 36-item structured scale that comprehensively summarizes respondents’ QoL across 8 dimensions: physical functioning (10 items), role-physical (4 items), bodily pain (2 items), general health (5 items), vitality (4 items), social functioning (2 items), role-emotional (3 items), and mental health (5 items). The physical component summary and the mental component summary are 2 subscales of the 8 dimensions. In addition to the 8 dimensions listed above, the SF-36 v2 includes another health condition, reported health transition, which measures overall changes in health status over the past year.

We used EpiData (version 3.1; EpiData Association) for double entry and data management. Data collection and analysis were carried out using SPSS (version 27.0; IBM Corp) and R (version 4.1.1; R Foundation for Statistical Computing). Means and SDs were calculated for continuous data and frequencies and percentages for categorical data.

The total sample of 572 medical staff had a mean score for the Chinese WEMWBS of 38.47 (95% CI 37.45-39.61; SD 13.23; skewness 0.449; kurtosis –0.486) and a median score of 37, indicating a latent skewed trait distribution (Figure 1). An independent-sample t test showed no difference between the total WEMWBS score and gender (t₁=–1.477; P=.14). A Pearson correlation analysis did not indicate any significant relationship between the score for mental well-being and age; therefore, further validation analyses did not include participant age.

As shown in Table 1, the values for specific items were significantly different in the high-score and low-score groups (P<.001), meaning that all 14 items could differentiate the 2 groups well, and that none should be discarded. The correlation coefficient between each item and the total score of the instrument ranged from r=0.727 to r=0.900. As seen in Table 2, none of the items reached a rate of 20%, suggesting that there were no ceiling or floor effects.

Internal consistency reliability was good (Cronbach α=.965). The corrected item-total correlation values of the items were all greater than 0.5, indicating a good correlation between items and reliability (Table 3). Two weeks after completing the questionnaire, 35 subjects completed it again; the test-retest reliability was measured at 0.810, indicating that the scale had good stability. The split-half reliability of the scale was λ=0.947 according to the first half and the second half of the serial number, while the value was λ=0.970 according to the odd-even status of the serial number.

A Kaiser-Meyer-Olkin value of 0.963 for the 14 items and a value of 7844.584 for the Bartlett sphericity test (P<.001) demonstrated that the data obtained were suitable for factor analysis. A principal component factor analysis was used with varimax rotation to evaluate construct validity. Table 4 shows factor loadings for the 14 items, which ranged from 0.714 for item 4 to 0.903 for item 8.

An analysis of mean average precision (MAP) showed that the WEMWBS had a 1D structure. The minor average squared partial correlation was 0.02221, and the most negligible average fourth-power partial correlation was 0.00100. According to the revised MAP test [31], the number of factors was 1.

We conducted a CFA test of the hypothetical single-factor structure of the Chinese WEMWBS and measured the goodness-of-fit of the single confirmatory factor model. Assuming that there was no correlation between the residuals, the initial model fit poorly. The χ²/df was 8.437; the comparative fitting index (CFI) was 0.927; the RMSEA was 0.114; for the normed fit index (NFI), delta 1 was 0.918; for the relative fit index (RFI), rho 1 was 0.903; for the incremental fit index (IFI), delta 2 was 0.927; for and the Tacker-Lewis index (TLI), rho was 2.914.

There was a significant positive correlation between the Chinese WEMWBS and the WHO-5, with a correlation coefficient of 0.746 (95% CI 0.722-0.794; P<.01).

A CFA showed that the AVE value was 0.674 (ie, greater than 0.5). The CR value was 0.966 (ie, greater than 0.7), suggesting that the sample had good convergence validity.

Table 5 shows the results of the GRM analysis. The discrimination difference indices of the items ranged from 2.026 to 5.098, which demonstrates that the Chinese WEMWBS scores of low-score individuals differed from high-score individuals, corresponding to latent trait sensitivity. The item difficulty of thresholds ≥2, ≥3, ≥4, and 5 ranged from 1.06 to 1.73, 0 to 0.23, 0.56 to 1.06, and 1.12 to 1.66, respectively.

The ICCs and IICs for the Chinese WEMWBS are shown in Figures 2 and 3, respectively. The ICCs demonstrated that the sequence of the categories’ thresholds for the 14 items was as predicted, meaning that all regimentations were sufficient in including respondents; this finding, in turn, suggests that all categories were adequate based on placing a participant on the scale. The IICs displayed multimodal distribution. The shape of item 8 was the most precipitous and provided more knowledge than the other 13 items. The shape of item 4 was the flattest, indicating that the item provided the least information.

This is the first study to combine CTT and a GRM incorporating IRT to evaluate psychometric properties of the Chinese-language version of the WEMWBS in a sample of medical staff. Our results confirm the initial hypothesis that the WEMWBS is 1D. Since its establishment in 2006, the WEMWBS has been used in trials of patients and the general population with commendable results according to CTT and the Rasch model [32,33]. Given the broad and complicated spectrum of psychometric processes other than CTT, each with new evaluations and fixed statistical analyses in diverse models [34], we adopted the GRM to evaluate the contribution of the 14 items and their responses to the assessment of subjective well-being (SWB).

The mean score for the Chinese version of the WEMWBS used in this study was 38.47 (SD 13.23), which is lower than WEMWBS scores in medical staff surveys in other countries (eg, the United Kingdom [35], Pakistan [25,36], and Northern Ireland [37]). This discrepancy may be due to the data having been collected during the outbreak of COVID-19, meaning that the SWB of the medical staff would have been impacted to a certain extent [38]. Moreover, with the aging population of China, medical staff are under a great deal of pressure and need to master multidisciplinary knowledge and skills even as their work intensity increases [39].

The original 1D structure of the WEMWBS, as confirmed by previous studies in other countries [24,27,29], was not fully supported by earlier research in mainland China. This outcome was expected; some studies [28,40] identified a 2D structure that differed from the original assumption.

Researchers have pointed to differences between Eastern and Western cultures to explain this: the original meaning of the individual items might be changed in translated versions, and this alteration could affect the perceived intentions of the target population [22]. Furthermore, previous studies [33] adopted the Likert ordinal interval for a comprehensive rating, in which the 14 individual item scores were added to produce a total score. Bartram [35] found that using only a CFA may lead to misunderstanding, because the total score has a serial order, and the intervals between each score are not necessarily equal. The unidimensional structure was not without problems in this study. First, the model fitting effect was insufficient, because the χ²/df was greater than 5, and the RMSEA was greater than 0.08. Only the NFI, TLI, and CFI values supported the unidimensionality of the model. However, the AVE was greater than 0.5, and the CR was greater than 0.8, suggesting a relevant result. Second, the 1D model’s factor loadings for the 14 items were similar to the 2-factor model. Third, considering that the number of factors according to the revised MAP test was 1 [31], we adopted the 1D structure. An exemplary configuration of the Chinese WEMWBS would be favorable for facilitating IRT analyses in the future. Administering the Chinese WEMWBS based on IRT could strengthen its sensitivity and precision, guaranteeing that the items reflect the participants’ SWB levels.

The proportion of participants selecting the options “sometimes” and “often” was high in this study, suggesting that most respondents had relatively good SWB. To test the accuracy of the results, we examined the 14 items for floor and ceiling effects; we did not find extreme ceiling or floor effects, indicating that the process was reliable. There have been no reports on the distribution of responses to the WEMWBS in mainland China. In addition, the Chinese version of the WEMWBS displayed outstanding reliability, with a Cronbach α of .96, more significant than other studies for Chinese and other language versions [18,19,28,29,41].

The GRM was the best-match IRT model in this study. No previous studies have used the GRM to evaluate the psychometric properties of the WEMWBS. Our study reinforces the use of IRT models and supports existing studies on the psychometric evaluation of the WEMWBS with IRT methods.

The GRM analysis demonstrated that the global performance of WEMWBS items was satisfactory. The ICCs showed that the feedback categories of all the items were ordered and that all categories were presumably at the same point on the continuum [42].

Mental health assessment has drawn increasing attention from the Chinese government. In 2017, the Chinese government released the first guidelines to improve mental health in schools, workplaces, and hospitals. The WEMWBS has proven to be a convenient and valuable psychometric tool for academics, medical professionals, and other prominent stakeholders to measure the SWB of medical staff [43,44]. The Chinese WEMWBS has good reliability and validity with comprehensive and understandable content [15,24,26,45].

There are some limitations to this study. First, our investigation concentrated on hospitals in Zhejiang and Hunan provinces, and most participants were nurses, suggesting some selection bias. Follow-up research needs a larger sample size that includes therapists, physicians, and surgeons to assess the psychometric properties of the Chinese WEMWBS. Second, the sample size was only 572, which is less than 1000; this may have caused ambiguity in evaluating the IRT model. A larger sample size is needed in future research to confirm our findings. Third, we did not discriminate between medical staff with anxiety or depression when calculating the psychometric properties of the Chinese WEMWBS, which may have caused difficulty in demonstrating the scale’s validity. The performance of the Chinese WEMWBS should be further assessed in distinct staff groups.

Detailed provisions were made for the Chinese version of the WEMWBS in this study, and its psychometric properties were evaluated in a group of medical staff. We found that the Chinese WEMWBS has good reliability and validity and that it could be used as a reliable tool to evaluate the SWB of medical staff. It is critical to adopt measures to enable decision-making departments of hospitals to reduce work pressure, improve the SWB of clinical medical staff, improve patient satisfaction, and promote the development of the medical industry in a favorable direction.