# paper.pdf ## Metadata - PDFFormatVersion=1.4 - Language=en - IsLinearized=false - IsAcroFormPresent=false - IsXFAPresent=false - IsCollectionPresent=false - IsSignaturesPresent=false - Title=Construct validity of the Short Warwick-Edinburgh Mental Well-Being Scale in a diverse urban population - Creator=Chromium - Producer=Skia/PDF m147 - CreationDate=D:20260601174349+00'00' - ModDate=D:20260601174349+00'00' ## Contents ### Page 1 [](https://doi.org/10.21203/rs.3.rs-7166621/v1) # Construct validity of the Short Warwick-Edinburgh Mental Well-Being Scale in a diverse urban population - ###### Monique Millington CUNY Institute of Implementation Science in Population Health - ###### Sasha Fleary City University of New York Graduate ###### Research Article ###### Keywords: SWEMWBS, Rasch, factor analysis, measurement invariance, differential item functioning ###### Posted Date: October 21st, 2025 ###### DOI: [https://doi.org/10.21203/rs.3.rs-7166621/v1](https://doi.org/10.21203/rs.3.rs-7166621/v1) ###### License: This work is licensed under a Creative Commons Attribution 4.0 International License. [Read Full License](https://creativecommons.org/licenses/by/4.0/) ###### Additional Declarations: No competing interests reported. ###### Version of Record: A version of this preprint was published at Social Psychiatry and Psychiatric Epidemiology on May 26th, 2026. See the published version at [https://doi.org/10.1007/s00127-026-03109-0](https://doi.org/10.1007/s00127-026-03109-0) . ## ### Page 2 ## Abstract Although the Short Warwick-Edinburgh Mental Well-Being Scale (SWEMWBS) has been validated in samples across several countries it has yet to be validated in the US. This study examines the validity of the SWEMWBS within a diverse urban population of US adults. A population representative sample of adults living in NYC was obtained from the NYC Neighborhood Wellness Survey (NWS). Respondents answered questions about their mental health, emotional and social support, demographics (i.e., age, sex at birth, gender identity, sexual orientation, and race/ethnicity) and completed the SWEMWBS. Using a combination of Rasch analyses and classical test theory (CTT), we assessed the SWEMWBS for unidimensionality, monotonicity, local independence, differential item functioning, and measurement invariance. Ninety-five percent of eligible participants completed all seven items of the SWEMWBS (N=41,484). The measure showed high internal consistency (α = 0.88), and all items were highly correlated with the overall score. Assumptions of unidimensionality, monotonicity, and local independence were supported. Differential item functioning was identified between age groups (18-24 vs 65-74; 18-24 vs 75+; 25-29 vs 65-74; 25-29 vs 75+; 30-44 vs 75+) for SWEMWBS item 1 (I've been feeling optimistic about the future). Measurement invariance was assessed for key demographic variables, but no significant differences between configural and metric models were identified. Despite differences in item functionality by age, this assessment shows that the SWEMWBS is a useful tool for evaluating mental well-being in a diverse population and should be studied further among smaller groups and rural communities in the US. ## 1 Introduction Mental health research often focuses exclusively on mental illness, the implication being that the absence of mental health conditions is a consequential indicator of mental wellbeing. However, mental health can also be conceptualized as a construct comprised of traits that influence mental well-being and modulate the severity of mental health conditions. The World Health Organization’s (WHO) definition of mental health, “a state of mental well-being that enables people to cope with the stresses of life, realize their abilities, learn and work well, and contribute to their community,” notably excludes any mention of mental health conditions [1]. They further distinguish the two by noting that while mental health conditions often negatively impact mental well-being, persons living with mental health conditions do not necessarily have low mental well-being. A study conducted in Canadians aged 15 and older found that up to 4.5% of individuals with high mental well-being, and 4.7% of individuals with moderate mental well-being had a diagnosed mental health condition, demonstrating that positive mental health traits can coexist with mental health conditions [2]. Positive mental health traits have been shown to both protect individuals from developing mental health conditions and to aid in recovery from mental health conditions [3], [4]. Thus, improving our understanding of population mental health requires the study of mental well-being and positive mental health traits alongside mental illness. Several researchers have proposed dual-continua models of mental health [5]–[8]. In contrast to the bipolar model which has mental health and mental illness at opposite ends of a single spectrum, these dual continua models divide mental illness and mental well-being into distinct continua while acknowledging their impact on each other [6]. For example, Greenspoon and Saklofske advocated for a mental health model that separated psychopathology from subjective well-being, arguing that an assessment of mental health solely based on the absence or presence of psychological conditions would be unable to distinguish between individuals describing different levels of satisfaction with life [5]. They divide the population into four groups: low psychopathology-low subjective well-being (vulnerable); high psychopathology-low subjective well-being (struggling); low psychopathology-high subjective well-being (symptomatic but content); and high psychopathology-high subjective well-being (complete mental health) [5]–[8]. Masse et al. described mental health as a combination of an individual’s level of psychological distress and well-being. Psychological distress was indicated by markers of common mental health conditions such as anxiety and depression, while mental well-being was associated with life satisfaction, morale, and positive emotions among other traits. Another model developed by Keyes describes the state of an individual’s mental health, either “languishing” or “flourishing”, as a product of their emotional well-being and positive functioning. Emotional well-being is comprised of traits such as positive affect, lack of negative affect, generally positive feelings about life, and satisfaction with life among others. Positive functioning was further broken down into psychological well-being, self-acceptance, positive relations with others, personal growth, purpose in life, environmental mastery (individuals are able to modify their environments to satisfy their needs), and autonomy (individuals are capable of self-determination) [9]. While their descriptions differ in terms of how they conceptualize positive mental traits, they generally function to separate mental illness from mental well-being, and studies support the structure of these models [6]. The separation not only allows for the co-existence of mental health conditions and mental well-being but also allows them to be measured independently from one another. Several scales have been developed to measure traits of mental well-being. For example, Keyes’ 40-item Mental Health Continuum-Long Form (MHC-LF) and the 12-item Mental Health Continuum-Short Form (MHC-SF) measure emotional, social, and psychological well-being [10], [11]. While Keyes’ scales are meant to be an overall measure of mental health, Lukat et al.’s Positive Mental Health (PMH; 9 items) scale focuses only on mental well-being. It measures overall mood, enjoyment of life, confidence, and ability to deal with difficult stages in life.[12] While scales such as these are essential to identifying and measuring positive mental health traits and mental well-being and help us to understand their effects on mental and physical health, their utility is dependent on ability to measure these traits accurately in the environment in which they are deployed. It is therefore necessary that we examine the reliability and validity of any such scales before adapting them as tools to describe the population. One such scale that has been developed and adapted to accommodate different populations is the Warwick Edinburgh Mental Well-Being Scale (WEMWBS). This 14-item scale was developed in the United Kingdom by a panel of experts in psychology, psychiatry, public health, and other disciplines, to fulfil a need for instruments with the ability to measure positive mental health at the individual and population levels [13]. While the original instrument showed high internal consistency and test-retest reliability in convenience samples, an assessment of the internal validity of the scale in a sample of Scottish adults aged ### Page 3 16 to 74 using Rasch analysis found that seven of the fourteen items either did not fit the model or did not improve the fit and were subsequently removed [13], [14]. The shortened scale, dubbed the Short Warwick-Edinburgh Mental Well-Being Scale (SWEMWBS), has since been translated and validated in China, Norway, Denmark, Sweden, Finland, Oman, Qatar, Saudi Arabia and the Czech Republic [14]–[22]. Psychometric assessment of SWEMWBS in these populations have largely supported the factor structure and validity of the scale [15]–[22]. Despite not yet being validated in the United States (US), the SWEMWBS has been utilized in recent publications reporting on US-based samples [23], [24]. The validation of the SWEMWBS using a, population representative sample of New York City residents would, to our knowledge, be the first psychometric assessment of the scale in a US population. Additionally, the use of a survey with a large sample size and the diversity of the sample allows for the identification of measurement invariance and differential item functioning within demographic groups that would have limited sample sizes in smaller surveys. This assessment utilizes both classical test theory (CTT) methods and Rasch analyses. While both models assume unidimensionality, CTT methods such as factor analysis and reliability statistics rely on correlations within a given sample whereas with Rasch analyses, data are fitted to a linear probabilistic model whose distribution is not sample-dependent and fit statistics are produced to determine how well a particular item or response fits the model [25]–[27]. Additionally, with CTT methods we assume that all scale items are of equal difficulty and that scale responses are additive, but the underlying theory of Rasch is that items of lower difficulty are more likely to be endorsed than more difficult items, and that all items are more likely to be endorsed by individuals with higher ability than by those with lower ability [25], [27]. Rasch methods are able to separate item difficulty from person ability and can be used to calculate the probability of an individual endorsing an item based on both item difficulty and person ability [25], [26]. Rasch methods can also be used to assess floor and ceiling effects for individuals and items [26]. Use of both Rasch and CTT will allow for comparisons between validation of the SWEMWBS in this population to others as both Rasch and CTT methods have been utilized previously to validate the scale in other populations. ## 2 Methods ## 2.1 Data Collection & Sample The NYC Neighborhood Wellness Survey (NYC NWS) was a population-based survey mailed randomly to over 210,000 households in NYC beginning in May 2023. A collaboration between the NYC Department of Health and Mental Hygiene (NYC Health Department) and the City University of New York Graduate School of Public Health & Health Policy (CUNY SPH), the survey primarily assessed mental health and substance use needs, but included questions on caregiver experiences, interactions with police, social supports and stressors, and demographic characteristics. Households received up to five mailings notifying them to complete the survey: an initial invitation letter, the first reminder postcard, a second invitation letter, and two additional reminder postcards. Beginning with the second invitation letter, all survey respondents were provided with a passcode to complete the web-based version of the survey. Adults over the age of 18 from all five boroughs were invited to participate. In households with more than one adult, the adult with the nearest upcoming birthday was instructed to complete the survey. The initial survey invitation contained the invitation letter, paper surveys in English and Spanish, and a $5 incentive. The letter contained instructions for completing the survey in Traditional or Simplified Chinese, Haitian Creole, or Russian using the web-based survey. Respondents who completed the survey received an additional $20 cash incentive with their thank-you letter. ## 2.2 Demographics Respondents indicated their age in years and were asked to identify their birth sex (male or female), gender identity (man, woman, transgender man, transgender woman, non-binary or genderqueer person, a gender identity not listed (please describe), or not sure), and sexual orientation (gay or lesbian, straight or heterosexual, bisexual, another sexual orientation, or not sure). Race/ethnicity was established using two questions. Respondents were asked to identify if they were of Hispanic or Latino/Latina ethnicity (yes/no). They were then asked to select all racial/ethnic categories they would use to describe themselves. Gender, race, and ethnicity were asked as multi-select questions and included an “Other” option accompanied by a “please describe” text box, allowing participants to enter text responses. Where possible, these variables were upcoded through a process that involved both manual and code-based review of text responses. When text responses matched existing categories, the matching response was selected in the original question. Race and ethnicity were combined to create a single variable. Missing values for age, sex at birth, and race/ethnicity were imputed. Additional information on upcoding, imputation, and survey methods can be found in the NYC NWS Methodological Report. Age was categorized into six groups for analysis (18–24, 25–29, 30–44, 45–64, 65–74, 75 and older). Race/ethnicity was categorized into eight groups (American Indian, Native, First Nations, Indigenous Peoples of the Americas, or Alaska Native; Asian; Black or African American; Hispanic or Latino, Latina, or Latinx; Middle Eastern or North African; White; Multiple Races; and Other). Due to small cell sizes, gender identity was collapsed into three categories (cisgender, transgender, and another gender identity). Respondents answered questions about emotional support (How often can you count on someone to provide you with emotional support, such as talking over problems or helping you make decisions?), social isolation (In the past 30 days, how much have you been feeling cut off or distant from other people?), and ability to function (In the past 30 days, how difficult have any problems with your emotions, nerves, or mental health made it for you to do your work, take care of things at home, or get along with other people?) and were assessed for non-specific psychological distress (NSPD) using the Kessler-6 scale [28]– [30]. ## 2.3 SWEMWBS Of the 43,606 respondents who completed the survey, 95% (n = 41,484) completed all seven items of the SWEMWBS and were eligible for inclusion in the analysis. ### Page 4 Response options for the SWEMWBS are on a five-point Likert scale ranging from 1 (“None of the time”) to 5 (“All of the time”) allowing for a total possible score between 7 and 35. The raw scores for the SWEMWBS were transformed into metric scores using the SWEMWBS conversion table [31]. Most studies found that the SWEMBWS had high internal consistency (0.77–0.89) [15], [16], [19], [21], [22], [32]. RMSEA values ranged from 0.03–0.094 and CFI values ranged from 0.96–0.995 [15]–[19], [22], [33]. Variance explained by the models ranged from 48%-61% [15], [16], [22]. The assumption of unidimensionality was supported in most studies, with the exception of a Finnish study which found that a bi-factor model (RMSEA/CFI: 0.063/0.995) showed better fit than the one factor model (RMSEA/CFI: 0.065/0.995) [15]–[20], [22], [33], [34]. A study that compared results from the SWEMWBS to the PHQ-9 and GAD-7 in a clinical population found negative correlations between SWEMWBS score and both scales [32]. Partial measurement invariance by country and gender was identified in a study using a version of the scale that had been translated into Arabic [21]. Differential item functioning was thought to have been identified in a Swedish study, however, after adjustments to sample size, significant differences were reduced [19]. ## 2.4 Statistical analyses Descriptive statistics were calculated using SAS v9.4. One-way analysis of variance was used to examine differences in average SWEMWBS score by age, sex, gender identity, sexual orientation, race/ethnicity, and NSPD status. Correlations between SWEMWBS score and emotional support were used to assess convergent validity, while correlations with social isolation and difficulty functioning were used to assess divergent validity. Person and item parameters for the Rasch partial credit model were estimated in Winsteps v5.8.5 using conditional maximum likelihood estimation. Fit statistics for item parameters were used to assess model fit. Assumptions of unidimensionality, local independence, and monotonicity of the latent trait were assessed. Mean squares for item parameters between 0.5 and 1.5 indicated good fit while mean squares less than 0.5 or between 1.6 and 2.0 were not ideal, but not detrimental to the model. For item parameters with a mean square greater than 2.0, mean-squares standardized fit statistics with a value greater than 2.0 were indicative of item misfit [25], [26], [35]. Item fit was also assessed using point measure correlations with correlations greater than 0.4 indicating that the item is highly correlated with person ability. Rasch score measures and methods described by Ekstrand et al. were used to transform the SWEMWBS score creating the NWS transformed score in Table 1 [36]. The full score transformation table, including score transformations from the University of Warwick, can be found in Supplemental Table 1. Differential item functioning (DIF) was assessed with pairwise comparisons between groups using the formula: where d1 represents item difficulty for group 1, d2 represents item difficulty for group 2, se1 represents the standard error of item difficulty for group 1, and se2 represents the standard error of item difficulty for group 2. Due to the impact of sample size on the significance t, standard errors were normalized using the following equation [37]: where N represents the total number of individuals in the group. The resulting equation for tn using normalized standard errors is: Where the absolute value of the difference between d1 and d2 was between 0.55 and 1.3 logits, tn greater than 1.96 was indicative of differential item functioning [37]: Confirmatory factor analysis (CFA) and multi-group CFA were conducted using the lavaan package in R v4.4.2. Weighted least square mean and variance adjusted estimation and oblique rotation were used. CFA was used to test the internal structure of the scale. Internal consistency was assessed using Cronbach’s coefficient α with values greater than 0.7 indicating good internal consistency. Eigenvalues were plotted to assess dimensionality. Item residual correlations ( ) were produced in Winsteps to calculate Yen’s Q3 statistic ( ) using the following formula: Where represents the value of the largest item residual correlation, and represents the average of the absolute values of all item residual correlations. values closer to 0 support the assumption of local independence. values greater than 0.2 above indicate a violation of local independence [38], [39]. Factor loadings were used to assess the relationship between items and the factor with a value of 0.4 or greater signifying a strong relationship. Multi-group CFA was used to assess measurement invariance. Models were estimated for age, sex, gender identity, sexual orientation, and race/ethnicity. All models were assessed for invariance at four levels, each with increasingly stringent parameters. For configural invariance testing, factor loadings, intercepts, and residuals were freely estimated. For metric invariance testing, factor loadings were held to be equal across groups. Scalar invariance testing held both factor loadings and intercepts to be equal across groups, and residual invariance testing held factor loadings, intercepts, and residuals to be equal across groups. Testing for each model only progressed if the previous model showed evidence of measurement invariance. Because the models are nested, ### Page 5 comparisons between them can be used to determine if model fit improves or worsens at each stage. Fit statistics including chi-square (χ2), comparative fit index (CFI) and root mean square error of approximation (RMSEA) were used to assess model fit. If the overall model fit worsens, measurement invariance is not supported. A p-value < 0.001 was used to identify significant differences between models. CFI > 0.90 and RMSEA < 0.08 were indicative of adequate fit while a CFI > 0.95 and RMSEA ≤ 0.05 were indicative of good fit. Likelihood ratio tests, the difference in χ2 between nested models, and a reduction in CFI ≥ |0.01| were indicative of worse model fit [26]. ### Page 6 | Table 1. Demographic Characteristics of the NWS sample | | - | | Average SWEMWBS Score | | n (%) | Raw Score | UW Score a | NWS Score b | F statisticc | P- value | | | | 18–24 | 1,998 (4.8) | 24.8 (5.2) | 22.8 (4.8) | 22.8 (4.5) | 174.2 | < .0001 | | 25–29 | 2,923 (7.1) | 25.1 (4.6) | 22.9 (4.3) | 22.9 (4.0) | | 30–44 | 10,390 (25.1) | 25.4 (4.8) | 23.2 (4.5) | 23.2 (4.2) | | 45–64 | 13,701 (33.0) | 26.3 (5.3) | 24.2 (5.1) | 24.0 (4.9) | | 65–74 | 7,389 (17.8) | 27.2 (5.4) | 25.0 (5.3) | 24.8 (5.0) | | 75+ | 5,083 (12.3) | 26.9 (5.7) | 24.8 (5.5) | 24.7 (5.3) | | | | Male | 16,118 (38.9) | 26.1 (5.2) | 24.0 (5.0) | 23.9 (4.8) | 0.15 | 0.6985 | | Female | 25,366 (61.2) | 26.2 (5.3) | 24.0 (5.1) | 23.9 (4.8) | | | | Cisgender | 40,920 (98.6) | 26.2 (5.2) | 24.0 (5.0) | 23.9 (4.8) | 69.2 | < .0001 | | Transgender | 188 (0.5) | 24.8 (6.0) | 23.0 (5.5) | 22.9 (5.3) | | Another gender identity | 376 (0.9) | 23.3 (5.1) | 21.5 (4.4) | 21.6 (4.2) | | | | Gay or lesbian | 2,132 (5.4) | 25.2 (4.8) | 23.0 (4.4) | 23.0 (4.1) | 132.9 | < .0001 | | Straight or heterosexual | 34,711 (87.2) | 26.4 (5.1) | 24.2 (5.0) | 24.1 (4.7) | | Bisexual | 1,482 (3.7) | 24.2 (4.7) | 22.1 (4.0) | 22.1 (3.7) | | Another sexual orientation | 309 (0.8) | 23.6 (6.3) | 21.9 (5.6) | 21.9 (5.3) | | Not sure | 1,162 (2.9) | 24.6 (6.3) | 22.7 (5.8) | 22.7 (5.6) | | 1,628 | | | | American Indian, Native, First Nations, IndigenousPeoples of the Americas, or Alaska Native | 159 (0.4) | 26.0 (6.2) | 24.1 (5.9) | 24.0 (5.7) | 16.5 | < .0001 | | Asian | 6,656 (16.0) | 25.6 (5.3) | 23.5 (5.1) | 23.5 (4.8) | | Black or African American | 7,390 (17.8) | 26.4 (5.6) | 24.3 (5.4) | 24.2 (5.2) | | Hispanic or Latino/Latina | 9,881 (23.8) | 26.2 (5.7) | 24.2 (5.5) | 24.1 (5.2) | | Middle Eastern or North African | 500 (1.2) | 25.0 (4.9) | 22.9 (4.4) | 22.9 (4.2) | | Native Hawaiian or Other Pacific Islander | 61 (0.2) | 26.7 (6.0) | 24.7 (5.8) | 24.6 (5.5) | ### Page 7 | Table 1. Demographic Characteristics of the NWS sample | | - | | White | 15,323 (37.0) | 26.3 (4.7) | 24.0 (4.5) | 23.9 (4.2) | | Multiple Races | 964 (2.3) | 25.6 (5.1) | 23.5 (4.7) | 23.4 (4.5) | | Other | 550 (1.3) | 25.9 (5.9) | 23.8 (5.5) | 23.7 (5.3) | | | | None of the time | 6,878 (15.8) | 26.7 (6.8) | 24.9 (6.6) | 24.7 (6.4) | 673.6 | < .0001 | | A little of the time | 6,988 (16.1) | 25.1 (5.2) | 23.0 (4.8) | 23.0 (4.5) | | Some of the time | 9,533 (21.9) | 24.6 (4.8) | 22.5 (4.3) | 22.5 (4.0) | | Most of the time | 9,888 (22.7) | 25.9 (4.2) | 23.6 (4.0) | 23.5 (3.7) | | All of the time | 10,218 (23.5) | 28.2 (4.7) | 25.9 (4.9) | 25.6 (4.7) | | 101 | | | | Not at all | 26,840 (61.7) | 28.0 (4.8) | 25.8 (5.0) | 25.5 (4.8) | 6,329.2 | < .0001 | | Somewhat | 13,747 (31.6) | 23.7 (4.1) | 21.6 (3.4) | 21.7 (3.1) | | A lot | 2,903 (6.7) | 20.2 (4.9) | 19.1 (3.7) | 19.3 (3.5) | | 116 | | | | Not difficult at all | 27,189 (62.7) | 28.0 (4.8) | 25.7 (5.0) | 25.5 (4.8) | 4,444.4 | < .0001 | | Somewhat difficult | 13,616 (31.4) | 23.6 (4.0) | 21.5 (3.3) | 21.6 (3.1) | | Very difficult | 1,909 (4.4) | 20.3 (4.4) | 19.1 (3.3) | 19.4 (3.1) | | Extremely difficult | 665 (1.5) | 17.8 (5.6) | 17.4 (4.3) | 17.7 (4.2) | | 227 | | [28] | | Psychological Distress (score: 0–12) | 2,916 (7.2) | 19.6 (4.6) | 18.7 (3.4) | 18.9 (3.3) | 5,537.0 | < .0001 | | No Psychological Distress (score: 13–24) | 37,450 (92.8) | 26.7 (5.0) | 24.4 (4.9) | 24.3 (4.7) | | d | 1,118 | | 24.0 (5.0) | | e | 2,122 | | a SWEMWBS transformed score as published by the University of Warwick and University of Edinburgh.[31] | | b Raw score transformation using NWS data sample based on methods published by Ekstrand et al., 2022. Full NWS score transformation can be found in the supplemental table.[36] | | c One-way analysis of variance using SWEMWBS raw score | ### Page 8 | Table 1. Demographic Characteristics of the NWS sample | | - | | d Missing | | e Missing denotes individuals who did not answer all 7 SWEMWBS items and were | ## 3 Results The majority of participants were aged 45–64 (33%), assigned female at birth (61%), identified as cisgender (99%), and described their sexual orientation as straight or heterosexual (85%) (Table 1). The largest racial/ethnic group was non-Hispanic Whites (37%), followed by Hispanic (24%), Black or African American (18%), and Asian (16%). The average raw SWEMWBS score was 24.0 (SD = 5.0). One-way analysis of variance showed significant differences in mean for age group (F = 174.4, p < .0001), gender identity (F = 69.2, p < .0001), sexual orientation (F = 132.9, p < .0001), race/ethnicity (F = 16.5, p < .0001), emotional support (F = 673.6, p < .0001), social isolation (F = 6,329.2, p < .0001), difficulty functioning (F = 4,444.4, p < .0001), and serious psychological distress (F = 5,537.0, p < .0001). Raw SWEMWBS score was low among participants in the 18–24 age group, participants who identified as bisexual or were unsure of their sexual orientation, participants whose gender identity was other or unknown, and participants whose race/ethnicity was Middle Eastern or North African compared to other groups. SWEMWBS score was lowest among participants who indicated it was “extremely difficult” to function in the past 30 days, those who described experiencing “a lot” of social isolation, and those experiencing serious psychological distress. ## 3.1 Results Using Rasch Methodology Statistics for the Rasch partial credit model can be found in Table 2. Infit and outfit mean-squares were within 0.62 and 1.28, all within the range indicative of good fit. Point measure correlations were between 0.69 and 0.78 for all items, also indicative of good fit. Total variance explained by the model was 52.3%, with 39.4% explained by persons and 12.8% explained by the items. The eigenvalue for unexplained variance in the 1st contrast was 1.71, supporting unidimensionality. Values for ranged from 0.05 to 0.39. The resulting was 0.2 upholding the assumption of local independence. The slope of the test characteristic curve (Fig. 1) is positive which supports the assumption of monotonicity. Person separation and person reliability were 2.13 and 0.82 respectively indicating that the instrument is sensitive enough to distinguish between individuals with high and low ability [40]. Approximately 7% of participants obtained the maximum extreme score, and less than 1% of participants obtained the minimum extreme score. Wright’s sample independent reliability statistic was 0.96 and scores differentiated five distinct levels of performance. Results of differential item functioning showed no significant differences by age group, sex at birth, gender identity, sexual orientation, or race/ethnicity for any of the scale items (Table 3). | Table 2. Results of the construct validity of the SWEMWBS using Rasch analyses and classical test theory analyses | | - | | INFIT | OUTFIT | | MEAS | SE | MNSQ | ZSTD | MNSQ | ZSTD | PT-MEAS COR | Factor Loadings | Item-total Correlations | Cronbach's α | | I've been feeling optimistic about the future | 0.48 | 0.01 | 1.12 | 9.90 | 1.28 | 9.90 | 0.69 | 0.66 | 0.55 | 0.88 | | I've been feeling useful | -0.01 | 0.01 | 0.91 | -9.90 | 0.95 | -6.62 | 0.73 | 0.75 | 0.65 | 0.86 | | I've been feeling relaxed | 0.58 | 0.01 | 0.75 | -9.90 | 0.84 | -9.90 | 0.77 | 0.77 | 0.67 | 0.86 | | I've been dealing with problems well | 0.12 | 0.01 | 0.62 | -9.90 | 0.69 | -9.90 | 0.78 | 0.82 | 0.72 | 0.85 | | I've been thinking clearly | -0.39 | 0.01 | 0.62 | -9.90 | 0.64 | -9.90 | 0.77 | 0.87 | 0.74 | 0.85 | | I've been feeling close to other people | 0.07 | 0.01 | 0.88 | -9.90 | 0.96 | -6.05 | 0.74 | 0.75 | 0.66 | 0.86 | | I've been able to make up my own mind about things | -0.85 | 0.01 | 0.93 | -8.54 | 0.90 | -9.90 | 0.69 | 0.79 | 0.64 | 0.86 | | Note: MEAS = item difficulty; SE = standard error; MNSQ = mean-square fit statistics; ZSTD = standardized fit statistics; PT-MEAS COR = point measure correlation. | ### Page 9 | Table 3. Results of differential item functioning testing | | - | | Item | | SWEMWBS1 | SWEMWBS2 | SWEMWBS3 | SWEMWBS4 | SWEMWBS5 | SWEMWBS6 | SWEMWBS7 | | SHIFT | t | SHIFT | t | SHIFT | t | SHIFT | t | SHIFT | t | SHIFT | t | SHIFT | t | | | | 18–24 vs 25–29 | -0.04 | -0.09 | 0.15 | 0.36 | -0.27 | -0.67 | 0.03 | 0.07 | 0.06 | 0.14 | -0.02 | -0.05 | 0.12 | 0.27 | | 18–24 vs 30–44 | -0.21 | -0.51 | 0.25 | 0.59 | -0.34 | -0.84 | 0.03 | 0.07 | 0.13 | 0.32 | -0.13 | -0.32 | 0.38 | 0.88 | | 18–24 vs 45–64 | -0.63 | -1.52 | 0.16 | 0.37 | -0.24 | -0.58 | -0.01 | -0.03 | 0.37 | 0.85 | -0.22 | -0.51 | 0.77 | 1.71 | | 18–24 vs 65–74 | -1.03 | -2.46 | -0.10 | -0.23 | 0.00 | 0.00 | 0.05 | 0.12 | 0.59 | 1.32 | -0.11 | -0.26 | 0.87 | 1.88 | | 18–24 vs 75+ | -1.18 | -2.85 | -0.30 | -0.70 | 0.20 | 0.47 | 0.18 | 0.42 | 0.52 | 1.16 | 0.15 | 0.34 | 0.69 | 1.51 | | 25–29 vs 30–44 | -0.17 | -0.42 | 0.10 | 0.23 | -0.07 | -0.17 | 0.00 | 0.00 | 0.07 | 0.18 | -0.11 | -0.27 | 0.26 | 0.61 | | 25–29 vs 45–64 | -0.59 | -1.42 | 0.01 | 0.02 | 0.03 | 0.08 | -0.04 | -0.10 | 0.31 | 0.72 | -0.19 | -0.46 | 0.65 | 1.45 | | 25–29 vs 65–74 | -0.99 | -2.37 | -0.25 | -0.58 | 0.27 | 0.65 | 0.02 | 0.05 | 0.53 | 1.19 | -0.09 | -0.21 | 0.75 | 1.62 | | 25–29 vs 75+ | -1.14 | -2.76 | -0.45 | -1.05 | 0.47 | 1.11 | 0.15 | 0.35 | 0.46 | 1.03 | 0.17 | 0.39 | 0.57 | 1.25 | | 30–44 vs 45–64 | -0.42 | -1.00 | -0.09 | -0.21 | 0.10 | 0.24 | -0.04 | -0.10 | 0.24 | 0.54 | -0.08 | -0.20 | 0.39 | 0.86 | | 30–44 vs 65–74 | -0.81 | -1.95 | -0.34 | -0.80 | 0.34 | 0.81 | 0.02 | 0.05 | 0.46 | 1.02 | 0.02 | 0.05 | 0.49 | 1.05 | | 30–44 vs 75+ | -0.97 | -2.34 | -0.54 | -1.28 | 0.54 | 1.27 | 0.15 | 0.35 | 0.38 | 0.86 | 0.28 | 0.64 | 0.31 | 0.67 | | 45–64 vs 65–74 | -0.40 | -0.95 | -0.26 | -0.60 | 0.24 | 0.57 | 0.07 | 0.15 | 0.22 | 0.49 | 0.11 | 0.25 | 0.10 | 0.21 | | 45–64 vs 75+ | -0.55 | -1.33 | -0.46 | -1.06 | 0.44 | 1.03 | 0.19 | 0.45 | 0.14 | 0.32 | 0.36 | 0.83 | -0.08 | -0.17 | | 65–74 vs 75+ | -0.16 | -0.38 | -0.20 | -0.46 | 0.20 | 0.46 | 0.13 | 0.29 | -0.08 | -0.17 | 0.26 | 0.59 | -0.18 | -0.38 | | 0.06 | 0.14 | 0.05 | 0.11 | -0.18 | -0.44 | -0.03 | -0.07 | -0.03 | -0.07 | 0.17 | 0.39 | 0.00 | 0.00 | | | | Cisgender vs Transgender | 0.00 | 0.00 | -0.05 | -0.12 | 0.28 | 0.69 | -0.20 | -0.48 | -0.06 | -0.15 | -0.06 | -0.14 | 0.09 | 0.20 | | Cisgender vs Another gender identity | 0.40 | 0.99 | -0.07 | -0.18 | -0.13 | -0.32 | 0.25 | 0.61 | -0.22 | -0.53 | -0.08 | -0.19 | -0.20 | -0.48 | | Transgender vs Another gender identity | 0.40 | 1.00 | -0.02 | -0.05 | -0.41 | -1.02 | 0.45 | 1.11 | -0.16 | -0.38 | -0.02 | -0.05 | -0.29 | -0.69 | | | | Gay or lesbian vs Straight or heterosexual | -0.20 | -0.50 | 0.02 | 0.06 | 0.04 | 0.09 | -0.03 | -0.07 | 0.10 | 0.24 | 0.07 | 0.17 | 0.00 | 0.00 | | Gay or lesbian vs Bisexual | 0.14 | 0.35 | 0.02 | 0.06 | -0.01 | -0.03 | 0.07 | 0.18 | -0.11 | -0.27 | -0.02 | -0.04 | -0.13 | -0.30 | | Gay or lesbian vs Another sexual orientation | -0.25 | -0.44 | -0.15 | -0.26 | 0.30 | 0.54 | -0.03 | -0.05 | 0.15 | 0.26 | -0.10 | -0.17 | 0.12 | 0.20 | | Gay or lesbian vs Not sure | -0.20 | -0.50 | -0.09 | -0.20 | 0.41 | 1.00 | 0.00 | 0.01 | 0.10 | 0.24 | 0.07 | 0.17 | -0.34 | -0.78 | | Straight or heterosexual vs Bisexual | 0.35 | 0.85 | 0.00 | 0.00 | -0.05 | -0.12 | 0.10 | 0.25 | -0.21 | -0.51 | -0.09 | -0.22 | -0.13 | -0.30 | | Straight or heterosexual vs Another sexual orientation | -0.04 | -0.07 | -0.17 | -0.31 | 0.27 | 0.48 | 0.00 | 0.00 | 0.05 | 0.08 | -0.17 | -0.30 | 0.12 | 0.20 | | Straight or heterosexual vs Not sure | 0.00 | 0.00 | -0.11 | -0.26 | 0.38 | 0.91 | 0.03 | 0.08 | 0.00 | 0.00 | 0.00 | 0.00 | -0.34 | -0.77 | ### Page 10 | Table 3. Results of differential item functioning testing | | - | | Bisexual vs Another sexual orientation | -0.39 | -0.69 | -0.17 | -0.31 | 0.31 | 0.56 | -0.10 | -0.18 | 0.26 | 0.45 | -0.08 | -0.14 | 0.25 | 0.42 | | Bisexual vs Not sure | -0.35 | -0.84 | -0.11 | -0.26 | 0.42 | 1.03 | -0.07 | -0.16 | 0.21 | 0.50 | 0.09 | 0.22 | -0.21 | -0.49 | | Another sexual orientation vs Not sure | 0.04 | 0.09 | 0.06 | 0.13 | 0.11 | 0.22 | 0.03 | 0.07 | -0.05 | -0.09 | 0.17 | 0.34 | -0.46 | -0.88 | | | | American Indian/Native American vs Asian | 0.51 | 1.24 | 0.54 | 1.26 | -0.28 | -0.67 | -0.18 | -0.43 | -0.41 | -0.95 | 0.22 | 0.51 | -0.57 | -1.28 | | American Indian/Native American vs Black or African American | 0.25 | 0.59 | 0.32 | 0.75 | -0.21 | -0.49 | -0.23 | -0.54 | -0.20 | -0.46 | -0.06 | -0.15 | 0.12 | 0.26 | | American Indian/Native American vs Hispanic or Latino/Latina | 0.43 | 1.03 | 0.46 | 1.07 | -0.45 | -1.07 | -0.16 | -0.37 | -0.34 | -0.77 | 0.13 | 0.30 | -0.17 | -0.37 | | American Indian/Native American vs Middle Eastern or North African | 0.50 | 1.22 | 0.57 | 1.37 | -0.47 | -1.15 | -0.16 | -0.38 | -0.41 | -0.96 | 0.33 | 0.80 | -0.51 | -1.17 | | American Indian/Native American vs White | 0.24 | 0.58 | 0.40 | 0.93 | -0.62 | -1.51 | -0.12 | -0.28 | -0.20 | -0.45 | 0.44 | 1.03 | -0.18 | -0.39 | | American Indian/Native American vs Multiple Races | 0.50 | 1.21 | 0.57 | 1.35 | -0.57 | -1.38 | -0.14 | -0.33 | -0.32 | -0.74 | -0.03 | -0.06 | -0.09 | -0.19 | | American Indian/Native American vs Something else | 0.04 | 0.09 | 0.79 | 1.82 | -0.34 | -0.82 | -0.11 | -0.26 | -0.18 | -0.40 | -0.13 | -0.30 | -0.07 | -0.15 | | Asian vs Black or African American | -0.27 | -0.64 | -0.22 | -0.50 | 0.07 | 0.17 | -0.05 | -0.11 | 0.21 | 0.48 | -0.28 | -0.66 | 0.69 | 1.50 | | Asian vs Hispanic or Latino/Latina | -0.08 | -0.20 | -0.08 | -0.18 | -0.17 | -0.40 | 0.02 | 0.05 | 0.07 | 0.17 | -0.09 | -0.21 | 0.40 | 0.88 | | Asian vs Middle Eastern or North African | -0.01 | -0.03 | 0.04 | 0.09 | -0.19 | -0.47 | 0.02 | 0.05 | 0.00 | 0.00 | 0.12 | 0.28 | 0.06 | 0.13 | | Asian vs White | -0.28 | -0.67 | -0.14 | -0.33 | -0.34 | -0.84 | 0.06 | 0.15 | 0.22 | 0.50 | 0.22 | 0.53 | 0.39 | 0.88 | | Asian vs Multiple Races | -0.02 | -0.04 | 0.04 | 0.09 | -0.29 | -0.71 | 0.04 | 0.11 | 0.09 | 0.22 | -0.24 | -0.58 | 0.48 | 1.09 | | Asian vs Something else | -0.48 | -1.16 | 0.25 | 0.58 | -0.06 | -0.16 | 0.07 | 0.17 | 0.23 | 0.53 | -0.34 | -0.82 | 0.50 | 1.11 | | Black or African American vs Hispanic or Latino/Latina | 0.18 | 0.44 | 0.14 | 0.33 | -0.24 | -0.58 | 0.07 | 0.16 | -0.14 | -0.31 | 0.19 | 0.45 | -0.29 | -0.63 | | Black or African American vs Middle Eastern or North African | 0.25 | 0.62 | 0.25 | 0.60 | -0.27 | -0.65 | 0.07 | 0.17 | -0.21 | -0.49 | 0.40 | 0.95 | -0.63 | -1.44 | | Black or African American vs White | -0.01 | -0.02 | 0.08 | 0.18 | -0.42 | -1.02 | 0.11 | 0.26 | 0.01 | 0.01 | 0.50 | 1.18 | -0.30 | -0.65 | | Black or African American vs Multiple Races | 0.25 | 0.61 | 0.25 | 0.60 | -0.36 | -0.88 | 0.09 | 0.22 | -0.12 | -0.27 | 0.04 | 0.09 | -0.21 | -0.46 | | Black or African American vs Something else | -0.21 | -0.51 | 0.47 | 1.08 | -0.14 | -0.33 | 0.12 | 0.28 | 0.02 | 0.05 | -0.06 | -0.15 | -0.19 | -0.41 | | Hispanic or Latino/Latina vs | 0.07 | 0.17 | 0.11 | 0.27 | -0.03 | -0.06 | 0.00 | 0.00 | -0.07 | -0.17 | 0.21 | 0.49 | -0.34 | -0.78 | ### Page 11 | Table 3. Results of differential item functioning testing | | - | | Middle Eastern or North African | | Hispanic or Latino/Latina vs White | -0.19 | -0.47 | -0.06 | -0.15 | -0.18 | -0.43 | 0.04 | 0.10 | 0.14 | 0.33 | 0.31 | 0.73 | -0.01 | -0.01 | | Hispanic or Latino/Latina vs Multiple Races | 0.07 | 0.17 | 0.11 | 0.27 | -0.12 | -0.30 | 0.02 | 0.05 | 0.02 | 0.05 | -0.15 | -0.37 | 0.08 | 0.19 | | Hispanic or Latino/Latina vs Something else | -0.40 | -0.95 | 0.33 | 0.75 | 0.10 | 0.25 | 0.05 | 0.12 | 0.16 | 0.37 | -0.26 | -0.61 | 0.10 | 0.22 | | Middle Eastern or North African vs White | -0.26 | -0.64 | -0.18 | -0.42 | -0.15 | -0.37 | 0.04 | 0.10 | 0.21 | 0.49 | 0.11 | 0.25 | 0.34 | 0.76 | | Middle Eastern or North African vs Multiple Races | 0.00 | 0.00 | 0.00 | 0.00 | -0.10 | -0.24 | 0.02 | 0.05 | 0.09 | 0.21 | -0.36 | -0.86 | 0.43 | 0.96 | | Middle Eastern or North African vs Something else | -0.47 | -1.13 | 0.21 | 0.49 | 0.13 | 0.31 | 0.05 | 0.12 | 0.23 | 0.53 | -0.46 | -1.10 | 0.45 | 0.98 | | White vs Multiple Races | 0.26 | 0.63 | 0.18 | 0.42 | 0.05 | 0.13 | -0.02 | -0.04 | -0.12 | -0.28 | -0.46 | -1.11 | 0.09 | 0.20 | | White vs Something else | -0.20 | -0.49 | 0.39 | 0.91 | 0.28 | 0.67 | 0.01 | 0.02 | 0.02 | 0.04 | -0.57 | -1.34 | 0.11 | 0.24 | | Multiple Races vs Something else | -0.46 | -1.12 | 0.21 | 0.49 | 0.23 | 0.54 | 0.03 | 0.07 | 0.14 | 0.32 | -0.10 | -0.25 | 0.02 | 0.04 | | Note: Shift represents the absolute value of the difference in item difficulty between each group in logits. | ## 3.2 Results Using CTT Methodology All seven items of the SWEMWBS were highly correlated with the overall score (item-total correlations: 0.55–0.74) and showed good internal consistency (α = 0.88). A scree plot of the eigenvalues (Fig. 2) showed a single factor with an eigenvalue greater than 1 which supports the assumption of unidimensionality. Although emotional support was significantly correlated with total SWEMWBS score and all individual scale items, the correlations were very weak, ranging from 0.038–0.223 (Table 4). Stronger, negative correlations were found between total SWEMWBS score and scale items for social isolation (-0.272 – -0.522), difficulty functioning (-0.266 – -0.456), and serious psychological distress (Kessler-6) score (-0.336 – -0.549). Results of CFA showed that the model explained 60% of the total variance and factor loadings ranged from 0.66 to 0.87 indicating high correlation between the items and the latent factor. Results of configural invariance testing (Table 5) for age, sex at birth, gender identity, sexual orientation, and race/ethnicity showed adequate model fit for CFI (0.935–0.940) but poor results for RMSEA (0.134–0.139). Results were similar for metric invariance testing with CFI (0.927–0.940) values again indicating adequate model fit and RMSEA (0.117–0.123) values indicating poor fit. While Δχ2 p-values were significant (p < 0.05) suggesting significant differences between configural and metric models, ΔCFI ranged from (0.000-0.008) indicating no difference. It was determined that the Δχ2 p-values were likely inflated to significance because of the large sample size thus there was no need to conduct scalar or residual invariance testing. | Table 4. Convergent and Divergent Validity | | - | | Emotional support | Social isolation | Difficulty Functioning | Serious Psychological Distress | | SWEMWBS Raw Score | 0.126 | -0.483 | -0.490 | -0.571 | | SWEMWBS1 | 0.119 | -0.272 | -0.266 | -0.336 | | SWEMWBS2 | 0.108 | -0.339 | -0.334 | -0.387 | | SWEMWBS3 | 0.039 | -0.384 | -0.417 | -0.515 | | SWEMWBS4 | 0.068 | -0.353 | -0.397 | -0.457 | | SWEMWBS5 | 0.056 | -0.382 | -0.440 | -0.490 | | SWEMWBS6 | 0.223 | -0.522 | -0.397 | -0.458 | | SWEMWBS7 | 0.038 | -0.321 | -0.380 | -0.418 | ### Page 12 | Table 5. Results of measurement invariance testing using multiple group CFA | | - | | Model | Number of parameters | 2 | df | Δ2 | p-value | CFI | ΔCFI | RMSEA | ΔRMSEA | RMSEA 95% CI | SRMR | ΔSRMR | | Age | | 1. Configural invariance | 210 | 12,852.53 | 84 | 0.940 | 0.134 | 0.132– 0.136 | 0.042 | | 2. Metric invariance | 210 | 3,841.92 | 119 | 810.80 | < .0001 | 0.932 | -0.008 | 0.120 | -0.014 | 0.118– 0.122 | 0.051 | 0.009 | | Sex at Birth | | 1. Configural invariance | 70 | 14,378.54 | 28 | 0.940 | 0.134 | 0.132– 0.136 | 0.043 | | 2. Metric invariance | 70 | 3,039.10 | 35 | 38.19 | < .0001 | 0.940 | 0.000 | 0.121 | -0.013 | 0.119– 0.123 | 0.043 | 0.000 | | Gender Identity | | 1. Configural invariance | 105 | 12,735.33 | 42 | 0.939 | 0.134 | 0.132– 0.136 | 0.043 | | 2. Metric invariance | 105 | 2,875.56 | 56 | 36.55 | 0.0009 | 0.939 | 0.000 | 0.117 | -0.017 | 0.115– 0.118 | 0.043 | 0.000 | | Sexual Orientation | | 1. Configural invariance | 175 | 13,096.89 | 70 | 0.936 | 0.137 | 0.135– 0.139 | 0.044 | | 2. Metric invariance | 175 | 3,038.52 | 98 | 221.72 | < .0001 | 0.934 | -0.002 | 0.118 | -0.019 | 0.116– 0.120 | 0.047 | 0.003 | | Race/Ethnicity | | 1. Configural invariance | 280 | 13,495.99 | 112 | 0.935 | 0.139 | 0.137– 0.141 | 0.045 | | 2. Metric invariance | 280 | 3,666.65 | 161 | 544.40 | < .0001 | 0.927 | -0.008 | 0.123 | -0.016 | 0.121– 0.125 | 0.053 | 0.008 | | Note:2 = Chi-square; df = degrees of freedom; CFI = comparative fit index; RMSEA = root mean square error of approximation. | ## 4 Discussion This study assessed the construct validity and reliability of the SWEMWBS in a diverse urban environment in the United States using both Rasch analysis and CTT methods. Both Rasch and CTT methods support the assumption of unidimensionality which is consistent with most other studies [15]–[22]. While these results were inconsistent with those of Sarasjavi et al. which favored a bi-factoral structure, their study was based only on CTT methodology, and the differences noted between the bi-factoral and one factor models as indicated by RMSEA and CFI were small [33]. Rasch analyses showed no evidence of item misfit, and the overall model fit was good. While no floor effects were noted, there was a small ceiling effect as 7% of participants obtained the maximum score indicating that the scale may not capture the full range of the latent variable. Although p-values indicated significant differences between configural and metric models, changes in both CFI and RMSEA indicated that model fit remained unchanged supporting evidence of measurement invariance. Overall, both Rasch and CTT methods showed that the SWEMWBS performed well in this diverse urban US population. There was no differential item functioning for any of the observed groups, except age. Notably, for SWEMWBS item 1 (I've been feeling optimistic about the future), as the differential item functioning is what we might expect given different age cohorts stage in life and future orientation and optimism. Major life events such as marriage, childbirth, illness or loss of a job greatly impact our view of the world and accumulate as we age [41]–[43]. In addition, certain unique experiences shared by generational cohorts such as living through an economic recession could lead to differences in perceptions and behavior by cohort [44]–[46]. Considering these potential challenges, the age-related differential item functioning on item 1 of the SWEMWBS is not a major concern. While the longer WEMWBS has been validated among specific populations in the US, to our knowledge, this study is the first to validate the SWEMWBS in the US and the first to validate the scale using a large and diverse urban population. New York City is home to over 8 million people of different ages, races, ethnicities, and backgrounds, with nearly 40% of residents born outside of the US and over 200 languages spoken [47]. As part as one of the largest metropolitan statistical areas in the US, New York City also has a larger LGBTQ + population than most other US cities [48]. As such, we were able to examine measurement invariance and differential item functioning for characteristics that were not explored in previous analyses. This study had a large sample size and missingness was limited as variables used in the sampling and weighting of the survey (age, income, race, ethnicity, and sex at birth) were imputed. Despite the sample size, some groups (e.g. transgender men, transgender women, and Native Hawaiian or Other Pacific Islander) were collapsed into larger groups due to small cell sizes. The large sample size also influenced the accuracy of results, as smaller differences were ### Page 13 more likely to be identified as significant. Calculations were standardized to avoid overestimation for differential item functioning, and ΔCFI was compared in place of Δχ2 p-values for measurement invariance testing. Overall, this assessment shows the SWEMWBS to be a useful tool for evaluating mental well-being across the populations considered. Future studies should seek to validate the scale among smaller groups as well as in rural communities in the United States. ## Declarations The authors have no competing interests to declare that are relevant to the content of this article. ## Funding: This study was conducted in collaboration with and was fully funded by the NYC Department of Health and Mental Hygiene NYC Department of Health and Mental Hygiene ## Author Contribution M.M. conducted analyses, wrote the main manuscript text, and prepared tables and figures. S.F. contributed to the discussion and was responsible for the analytic plan. All authors reviewed the manuscript. ## References - 1. World Health Organization, “Mental health,”. [Online]. Available: https://www.who.int/news-room/fact-sheets/detail/mental- health-strengthening-our-response. [Accessed: 09-Oct-2024]. - 2. H. Gilmour, “Positive mental health and mental illness,”, vol. 25, no. 9, pp. 3–9, Sep. 2014. - 3. M. Iasiello, J. van Agteren, C. L. M. Keyes, and E. M. Cochrane, “Positive mental health as a predictor of recovery from mental illness,”, vol. 251, pp. 227–230, May 2019. - 4. C. L. M. Keyes, S. S. Dhingra, and E. J. Simoes, “Change in level of positive mental health as a predictor of future risk of mental illness,”, vol. 100, no. 12, pp. 2366–2371, Dec. 2010. - 5. P. J. Greenspoon and D. H. Saklofske, “Toward an Integration of Subjective Well-Being and Psychopathology,”, vol. 54, no. 1, pp. 81–108, 2001. - 6. M. Iasiello, J. van Agteren, and E. M. Cochrane, “Mental health and/or mental illness: A scoping review of the evidence and implications of the dual- continua model of mental health,”, vol. 2020, no. 1, pp. 1–45, 2020. - 7. C. L. M. Keyes, “Mental illness and/or mental health? Investigating axioms of the complete state model of health,”, vol. 73, no. 3, pp. 539–548, Jun. 2005. - 8. S. M. Suldo and E. J. Shaffer, “Looking beyond psychopathology: The dual-factor model of mental health in youth,”, vol. 37, no. 1, pp. 52–68, Mar. 2008. - 9. C. L. M. Keyes, “The mental health continuum: from languishing to flourishing in life,”, vol. 43, no. 2, pp. 207–222, Jun. 2002. - 10. C. L. M. Keyes, “Promoting and protecting mental health as flourishing: a complementary strategy for improving national mental health,”, vol. 62, no. 2, pp. 95–108, Feb. 2007. - 11. S. M. A. Lamers, G. J. Westerhof, E. T. Bohlmeijer, P. M. ten Klooster, and C. L. M. Keyes, “Evaluating the psychometric properties of the Mental Health Continuum-Short Form (MHC-SF),”, vol. 67, no. 1, pp. 99–110, Jan. 2011. - 12. J. Lukat, J. Margraf, R. Lutz, W. M. van der Veld, and E. S. Becker, “Psychometric properties of the positive mental health scale (PMH-scale),”, vol. 4, no. 1, Dec. 2016. - 13. R. Tennant, “The Warwick-Edinburgh Mental Well-being Scale (WEMWBS): development and UK validation,”, vol. 5, p. 63, Nov. 2007. - 14. S. Stewart-Brown, A. Tennant, R. Tennant, S. Platt, J. Parkinson, and S. Weich, “Internal construct validity of the Warwick-Edinburgh Mental Well-being Scale (WEMWBS): a Rasch analysis using data from the Scottish Health Education Population Survey,”, vol. 7, no. 1, p. 15, Feb. 2009. - 15. S. S. W. Ng, “Translation and validation of the Chinese version of the short Warwick-Edinburgh Mental Well-being Scale for patients with mental illness in Hong Kong,”, vol. 24, no. 1, pp. 3–9, Mar. 2014. - 16. A. Haver, K. Akerjordet, P. Caputi, T. Furunes, and C. Magee, “Measuring mental well-being: A validation of the Short Warwick–Edinburgh Mental Well- Being Scale in Norwegian and Swedish,”, vol. 43, no. 7, pp. 721–727, Nov. 2015. - 17. O. R. F. Smith, D. E. Alves, M. Knapstad, E. Haug, and L. E. Aarø, “Measuring mental well-being in Norway: validation of the Warwick-Edinburgh Mental Well-being Scale (WEMWBS),”, vol. 17, no. 1, p. 182, May 2017. ### Page 14 - 18. Y. Sun, “The reliability and validity of the Chinese Short Warwick-Edinburgh Mental Well-being Scale in the general population of Hong Kong,”, vol. 28, no. 10, pp. 2813–2820, Oct. 2019. - 19. V. Koushede, “Measuring mental well-being in Denmark: Validation of the original and short version of the Warwick-Edinburgh mental well-being scale (WEMWBS and SWEMWBS) and cross-cultural comparison across four European settings,”, vol. 271, pp. 502–509, 2019. - 20. J. Melin, A. Lundin, and M. Johansson, “An off-target scale limits the utility of Short Warwick-Edinburgh Mental Well-Being Scale (SWEMWBS) as a measure of well-being in public health surveys,”, vol. 202, pp. 43–48, Jan. 2022. - 21. K. Zayed, “A validation study of the Arabic version of the Warwick-Edinburgh Mental Well-being scale among undergraduate students,”, vol. 11, no. 1, p. 399, Nov. 2023. - 22. R. Hanzlová and A. Kudrnáč, “Validation and psychometric evaluation of the Short Warwick-Edinburgh Mental Well-Being Scale (SWEMWBS) among Czech adolescents using Item Response Theory,”, vol. 22, no. 1, p. 66, Aug. 2024. - 23. G. Barbayannis, M. Bandari, X. Zheng, H. Baquerizo, K. W. Pecor, and X. Ming, “Academic stress and mental well-being in college students: Correlations, affected groups, and COVID-19,”, vol. 13, p. 886344, May 2022. - 24. M. A. Kirwin and A. K. Ettinger, “Working mothers during COVID-19: a cross-sectional study on mental health status and associations with the receipt of employment benefits,”, vol. 22, no. 1, p. 435, Mar. 2022. - 25. S. A. Fleary, K. M. Freund, and C. R. Nigg, “Development and validation of assessments of adolescent health literacy: a Rasch measurement model approach,”, vol. 22, no. 1, p. 585, Mar. 2022. - 26. P. L. Joseph and S. A. Fleary, “Construct validity and measurement invariance of the eHEALS in a diverse US high school sample,”, vol. 36, no. 3, pp. 188–198, 2024. - 27. W. J. Boone and A. Noltemeyer, “Rasch analysis: A primer for school psychology researchers and practitioners,”, vol. 4, no. 1, p. 1416898, Jan. 2017. - 28. R. C. Kessler, “Short screening scales to monitor population prevalences and trends in non-specific psychological distress,”, vol. 32, no. 6, pp. 959–976, Aug. 2002. - 29. J. Vaglio Jr, “Testing the performance of the ENRICHD Social Support Instrument in cardiac patients,”, vol. 2, no. 1, p. 24, May 2004. - 30. NYC Department of Health and Mental Hygeine, “NYC Health Opinion Poll.” 2019. - 31. “Raw score to metric score conversion table for SWEMWBS.” [Online]. Available: https://warwick.ac.uk/fac/sci/med/research/platform/wemwbs/using/howto/swemwbs_raw_score_to_metric_score_conversion_table.pdf. - 32. N. Shah, M. Cader, B. Andrews, R. McCabe, and S. L. Stewart-Brown, “Short Warwick-Edinburgh Mental Well-being Scale (SWEMWBS): performance in a clinical sample in relation to PHQ-9 and GAD-7,”, vol. 19, no. 1, p. 260, Nov. 2021. - 33. K. K. Sarasjärvi, M. Elovainio, K. Appelqvist-Schmidlechner, P. Solin, N. Tamminen, and S. Therman, “Exploring the structure and psychometric properties of the Warwick-Edinburgh Mental Well-Being Scale (WEMWBS) in a representative adult population sample,”, vol. 328, p. 115465, Oct. 2023. - 34. D. J. Bartram, J. M. Sinclair, and D. S. Baldwin, “Further validation of the Warwick-Edinburgh Mental Well-being Scale (WEMWBS) in the UK veterinary profession: Rasch analysis,”, vol. 22, no. 2, pp. 379–391, Mar. 2013. - 35. J. M. Linacre, “Help for Winsteps Rasch Measurement and Rasch Analysis Software,” Sep-2020. [Online]. Available: https://www.winsteps.com/winman/misfitdiagnosis.htm. [Accessed: 27-Aug-2024]. - 36. J. Ekstrand, A. Westergren, K. Årestedt, A. Hellström, and P. Hagell, “Transformation of Rasch model logits for enhanced interpretability,”, vol. 22, no. 1, p. 332, Dec. 2022. - 37. J. M. Linacre, “An Adjustment for Sample Size in DIF Analysis,”, Nov-2016. [Online]. Available: https://www.rasch.org/rmt/rmt203e.htm. [Accessed: 27-Aug-2024]. - 38. W. M. Yen, “Effects of Local Item Dependence on the Fit and Equating Performance of the Three-Parameter Logistic Model,”, vol. 8, no. 2, pp. 125–145, 1984. - 39. K. B. Christensen, G. Makransky, and M. Horton, “Critical values for yen’s Q3: Identification of local dependence in the Rasch model using residual correlations,”, vol. 41, no. 3, pp. 178–194, May 2017. - 40. J. M. Linacre, “Reliability and Separation Measures,” Sep-2020. [Online]. Available: https://www.winsteps.com/winman/reliability.htm#:~:text=Person%20reliability%20and%20separation%20are,between%20high%20and%20low%20perfo [Accessed: 27-Aug-2023]. - 41. K. A. Durbin, S. J. Barber, M. Brown, and M. Mather, “Optimism for the future in younger and older adults,”, vol. 74, no. 4, pp. 565–574, Apr. 2019. - 42. W. J. Chopik, E. S. Kim, and J. Smith, “Changes in optimism are associated with changes in health over time among older adults,”, vol. 6, no. 7, pp. 814–822, Sep. 2015. - 43. W. J. Chopik, E. S. Kim, T. Schwaba, M. D. Krämer, and J. Smith, “Changes in optimism and pessimism in response to life events: Evidence from three large panel studies,”, vol. 88, no. 103985, p. 103985, Oct. 2020. - 44. S. De Hauw and A. De Vos, “Perspective Psychological Contract Expectations: Does Recession Lead Lowered Expectations?,”, vol. 25, pp. 293–302, Mar. 2010. - 45. R. Kaniel, C. Massey, and D. T. Robinson, “Optimism and economic crisis,”, 2010. ### Page 15 - 46. C. Chaves, T. Castellanos, M. Abrams, and C. Vazquez, “The impact of economic recessions on depression and individual and social well-being: the case of Spain (2006-2013),”, vol. 53, no. 9, pp. 977–986, Sep. 2018. - 47. NYC Department of City Planning, “Language Access,” Jul-2013. [Online]. Available: https://www.nyc.gov/site/planning/about/language-access.page. [Accessed: 30-Jan-2025]. - 48. The Williams Institute on Sexual Orientation and Gender Identity Law and Public Policy, “LGBT Adults in Large US Metropolitan Areas,” Mar-2021. [Online]. Available: https://williamsinstitute.law.ucla.edu/publications/lgbt-us-msa/. [Accessed: 30-Jan-2025]. - 49. “Table 20.1 Complete score-to-measure table on test of all items: Winsteps Help.” [Online]. Available: https://www.winsteps.com/winman/table20_1.htm. [Accessed: 21-Mar-2025]. ## Figures ## Figure 1 Test characteristic curve of SWEMWBS score ## ### Page 16 ## Figure 2 Figure 3. Scree plot of confirmatory factor analysis for SWEMWBS items ## Supplementary Files This is a list of supplementary files associated with this preprint. Click to download. - [SupplementalTable1.docx](https://assets-eu.researchsquare.com/files/rs-7166621/v1/0d9db89836d7e97d171d1e2d.docx)