Examining the relationship between the Short Warwick-Edinburgh Mental Well-being Scale (SWEMWBS) and EQ-5D-5L and comparing their psychometric properties

Study design

This study utilized the psychometric survey data of the E-QALY project collected in China [21, 22]. The E-QALY project aims to develop a new generic measure that covers a broader quality of life construct, which is relevant to health, social care, and public health sectors [23]. The online survey includes a set of demographic questions, health condition status and caring experience, followed by 64 candidate E-QALY items, EQ-5D-3L, EQ-5D-5L and SWEMWBS. The sample size was fixed at 500 [24,25,26] considering the primary purpose of the data was used to conduct factor analysis and spearman correlation coefficients analysis for developing EQ-HWB [22]. This is a sufficient sample size for this study given that published EQ-5D-5L and SWEMWBS validation and comparison studies used a sample size of 500 or less [16, 27,28,29,30]. This data was collected between April and July 2019 online by Accent, a U.K. online survey company. Quotas and inclusion criteria were applied to recruit a sample of 500 participants who lived in China and were aged above 18, in which there were similar numbers of individuals with GAD, HIV/AIDS, CHB, or depression, or without any of those 4 chronic conditions. The study was approved by the Ethics Committee of University of Sheffield, United Kingdom (Approval letter number 025524) and the IRB of Jinan University, China (Approval letter number JNUKY-2020–001). Informed consent was obtained from all participants prior to the online survey.

The online survey began by giving an outline of the research purpose. Participants were then asked to report their disease history. Eligible respondents reported their background information including education level, gender, age, etc. Next, respondents were asked to respond to the core survey that includes the E-QALY candidate items, two versions of EQ-5D descriptive systems, EQ-VAS (only completed once) and SWEMWBS. This study utilized the background information, EQ-5D-5L, EQ-VAS and SWEMWBS data collected in the psychometric survey in China. The order of completing the SWEMWBS and EQ-5D-5L was also randomized with half of sample completing SWEMWBS first and the other half completing EQ-5D-5L first.

Instruments

The EQ-5D-5L is a generic preference-based HRQoL instrument developed by the EuroQol Group. It was translated into simplified Chinese following a strict translation process [4] and its validity and reliability have been demonstrated in different health conditions [5,6,7,8,9] in China. It consists of a five-item descriptive system and a visual analog scale (EQ-VAS) [31, 32]. The descriptive system has five health dimensions, i.e., mobility, self-care, usual activities, pain/discomfort, anxiety/depression, and five response levels (1 = no problems, 2 = slight problems, 3 = moderate problems, 4 = severe problems and 5 = unable/extreme problems) for each dimension. An important characteristic of EQ-5D-5L is it allows the calculation of health utility values that reflect the desirability of a health state. In this study, EQ-5D-5L health utility values were calculated using the value set of China [33]. The EQ-VAS records the respondent’s current self-rated health on a 20-cm-long vertical thermometer-like scale from 0 (‘Worst imaginable health state’) to 100 (‘Best imaginable health state’).

The SWEMWBS is the short version of the Warwick-Edinburgh Mental well-being Scale (WEMWBS), which was developed to measure the mental well-being of the general population. The SWEMWBS consists of seven questions: I’ve been feeling optimistic about the future (OP), I’ve been feeling useful (USE), I’ve been feeling relaxed (RE), I’ve been dealing with problems well (PR), I’ve been thinking clearly (CL), I’ve been feeling close to other people (CLO), and I’ve been able to make up my mind about things (MI) and each question includes five frequency options (1 = none of the time, 2 = rarely, 3 = some of the time, 4 = often and 5 = all of the time) [9]. The Simplified Chinese translation was obtained from the developer of WEMWBS, which was translated by Dong et al. [34]. Raw level summary score (LSS) was summed and converted to metric total score using the SWEMWBS conversion table [9].

Note the response levels reversed between SWEMWBS and EQ-5D-5L on item level, with a higher response indicating better results for SWEMWBS but worse results for EQ-5D-5L. On aggregate level, higher score suggests better results for both EQ-5D-5L utility value, EQ-VAS and SWEMWBS overall score. In addition, the recall periods differed as EQ-5D-5L uses ‘today’ and SWEMWBS uses ‘over the past two weeks’.

Statistical analyses

We first described the characteristics of our sample and examined the relationship of the EQ-5D-5L and SWEMWBS. Spearman’s rank correlation was used to evaluate the association between the EQ-5D-5L dimensions and SWEMWBS dimensions. Exploratory factor analysis (EFA) was used to ascertain the number of unique underlying latent factors associated with the attributes assessed by the EQ-5D-5L and SWEMWBS. Secondly, we assessed their psychometric properties. The distributions of the EQ-5D-5L and SWEMWBS were reported. Specifically, items with over 70% of respondents reporting the best state and the worst state suggesting ceiling effect and floor effect respectively [35]. Known-group validity between healthy and each condition group was assessed for EQ-5D-5L utility, EQ-VAS and SWEMWBS score. Convergent validity was examined for EQ-5D-5L utility and SWEMWBS score using EQ-VAS as a benchmark. Data were analyzed using IBM SPSS Statistics for Windows, Version 22.0. Armonk, NY: IBM Corp (2013) and Mplus 8.3 Combo Version for Windows.

Association

Spearman’s rank correlation was used to evaluate the relationship between the EQ-5D-5L dimensions and SWEMWBS dimensions. Correlations were deemed as weak when scores fell between 0.10 and 0.29, moderate when between 0.30 and 0.49, and strong when greater than 0.5 [36,37,38]. Statistical significance was set at the 5% level. Since SWEMWBS measures mental well-being, we hypothesized that its dimensions have low correlations with EQ-5D-5L dimensions, except for anxiety/depression, which measures mental health.

Exploratory factor analysis

The purpose of exploratory factor analysis (EFA) is to reduce data dimensionality and to ascertain relatively few factors to describe the observed correlations among variables [39, 40]. In the EFA, data were sifted using the Kaiser–Meyer–Olkin (KMO) measure of sampling adequacy (> 0.5) and Bartlett’s test of sphericity (< 0.05) [41]. The KMO value ranged from 0 to 1, with greater than 0.60 considered suitable for factor analysis. The number of factors retained was selected according to the Kaiser Criterion [26], which claims retaining factors with eigenvalues bigger than 1 and using the scree plot to evaluate the suitability of this choice. The parallel analysis was run to ascertain the number of factors to be retained in model [42, 43]. We applied 1,000 random data sets to conduct the parallel analysis and then overlaid the results onto a single plot with the scree plot. Factors with eigenvalues in the observed data that are greater than the simulated data suggest “true” factors. Using an oblique Promax rotation allows for the potential that factors are correlated. Examining the rotated factor matrix, we identified items with pattern coefficients of 0.40 or greater as contributing to a factor and retained them. EFA was applied to all items from both the EQ-5D-5L (5 items) and the SWEMWBS (7 items).

Known-group validity

Known-group validity was evaluated by examining the mean, standard error (SE), median, and interquartile range (IQR) of EQ-5D-5L utility score, EQ-VAS and SWEMWBS score between healthy and each condition group. We hypothesized that the healthy group would have higher scores than the four disease groups. To investigate how the EQ-5D-5L and SWEMWBS perform in terms of discriminating between healthy and each condition group, the Mann–Whitney test was used to compare the distributions of the responses to the EQ-5D-5L and SWEMWBS dimensions. We listed the median values of each dimension as a reference. The efficiency of the EQ-5D-5L/SWEMWBS scores in differentiating between the known groups described above was tested using the F statistics based on the one-way analysis of variance [44,45,46]. F statistic has been used in previous studies as a way of comparing relative efficiency between two instruments [45,46,47]. The F statistic is defined as the ratio of intergroup variance dividing by intragroup variance, which is used for model-level significance tests in the linear regression model. When the model is significant, the value of the F statistic could be interpreted as the advantage of intergroup variance over intragroup variance. As the regression model is increasingly capable of capturing the change of regression target, the intergroup variance is increasingly dominant, and we will also expect a larger value of F statistic [48]. As a result, the index score with a higher F statistic would be supposed to be more efficient than its comparator because a greater value is much more likely to lead to statistical significance. As a complementary analysis, the efficiency of the EQ-5D-5L/SWEMWBS scores was also evaluated using the area under the receiver-operating characteristics curve (AUROC) [49]. The AUROC value ranges from 0.5 to 1.0, with a greater value suggesting better predictive ability.

Convergent validity

Convergent validity was examined for EQ-5D-5L utility and SWEMWBS score using EQ-VAS as a benchmark using the Pearson correlation coefficient, where the absolute value of Pearson correlation coefficient < 0.40 were considered as weak, moderate if between 0.40 and 0.70 and strong if > 0.70. Since the EQ-VAS fully evaluates the respondent’s overall state of health including physical health and mental health, we hypothesized that EQ-VAS has a positive correlation with the EQ-5D-5L utility scores and SWEMWBS scores.

留言 (0)

沒有登入
gif