Advertisement

Journal of Well-Being Assessment

, Volume 2, Issue 1, pp 1–19 | Cite as

The Structural Validity and Measurement Invariance of the Mental Health Continuum – Short Form (MHC-SF) in a Large Canadian Sample

  • Paige Lamborn
  • Kenneth M. Cramer
  • Amber Riberdy
Original Research

Abstract

The Mental Health Continuum-Short Form (MHC-SF; Keyes 2005a) is a 14-item questionnaire designed to measure three components of positive mental health: emotional well-being (EWB), social well-being (SWB), and psychological well-being (PWB). Previous studies have proposed various models of mental health using the MHC-SF: a single-factor model, a correlated two-factor model, a correlated three-factor model (EWB, SWB, and PWB), and a bifactor model with three specific dimensions and a general factor, as well as the use of Exploratory Structural Equation Modelling (ESEM) to examine model structure. The present study assessed the suitability of multiple models using confirmatory factor analysis and ESEM in a large Canadian sample (N = 43,020), taken from the Canadian Community Health Survey (CCHS; Statistics Canada 2012a). The bifactor ESEM model had the best fit. Measurement invariance testing revealed that the bifactor ESEM model showed strict invariance across gender and ethnic minority status, and weak invariance across four age groupings.

Keywords

Canadian Community Health Survey (CCHS) Positive mental health Structural equation modeling Mental Health Continuum-Short Form (MHC-SF) 

1 Introduction

According to the ‘Economic Burden of Illness in Canada’ report from the Public Health Agency of Canada (PHAC 2014a), the cost of illness in Canada between 2005 and 2008 was 11.4 billion dollars. Relatedly, researchers have attempted to quantify well-being in an effort to understand the factors that contribute to sub-optimal functioning, which seem to include aspects of both mental illness and reduced positive mental health (Keyes 2005b). Quality of life research has received much support in the past five decades (Bradburn 1969; Keyes 1998; Ryff 1989; Waterman 1993), including the two continua model of mental health, which posits that mental health is not only the absence of mental illness but also the presence of positive mental functioning (Keyes 2005b). As such, the study of well-being in the social sciences has moved toward the promotion of positive functioning, leading to increased productivity and prevented illness (Hubka and Lakaski 2013; Keyes 2002; Westerhof and Keyes 2010).

Recently, studies have shown that simply treating mental illness is not a reliable method for reducing the symptoms and occurrence of mental illness (Keyes 2002). Further, “the aim of positive psychology is to begin to catalyze a change in the focus of psychology from preoccupation only with repairing the worst things in life to also building positive qualities” (Seligman and Csikszentmihalyi 2000, p. 5). As such, the positive psychology movement aims to prevent or lessen the effect and duration of mental illnesses, such as depression, but also to encourage individuals to practice being positive in all aspects of life including self-regard, and social and emotional functioning. Moreover, the creation of reliable, valid scales for the measurement of positive functioning and feelings will spread awareness of the benefits to maintaining positive mental health.

The study of well-being has traditionally fell under two overarching models, hedonism and eudaimonia. Hedonic, or emotional well-being (EWB), refers to being happy, having low levels of negative emotions, and feeling satisfied with life (Bradburn 1969; Ryan and Deci 2001). Correlational and experimental data relate hedonic well-being to the experience of positive affect, carefreeness, and low negative affect (Huta and Ryan 2010). Social scientists have aimed to compose measures of subjective well-being that encompass factors important to mental health. For instance, the Satisfaction with Life Scale (SWLS; Diener et al. 1985) was designed to assess cognitive components of an individual’s subjective well-being by rating statements such as I am satisfied with my life. On the other hand, the Scale of Positive and Negative Experience (SPANE; Diener et al. 2009) and the Positive and Negative Affect Schedule (PANAS; Watson et al., 1988) seek to measure affective components of subjective well-being. However, these scales focus on somewhat narrow aspects of hedonic (emotional) well-being and do not encompass all aspects of positive functioning.

The study of eudaimonic well-being argues that although some ends may produce pleasure, not all such outcomes promote well-being (Ryan and Deci 2001). In contrast to hedonic well-being, eudaimonic well-being refers to living a fulfilling and productive life (Ryff 1989). Such functioning is characterized by seeking to reach one’s own unique potential and striving for balance (Ryff and Singer 2008). Eudaimonic well-being encompasses psychological well-being (PWB) and, more recently, social well-being (SWB; Keyes 2002). Before Keyes introduced his three-factor model of well-being, measurement of eudaimonic well-being relied on existing scales that faced various psychometric issues, such as overlap among variables and inadequacy in measurement (Jovanović 2015). The Ryff Scale of Psychological Well-Being (Ryff 1989) is one such scale that measures six proposed dimensions of PWB: self-acceptance, positive relations with others, autonomy, purpose in life, environmental mastery, and personal growth.

Across studies, the distinction between hedonic and eudaimonic well-being has been contested. For instance, some authors have argued that the two should be separated due to their distinct outcomes, as ratings of life satisfaction uniquely predicts feelings of pleasure (hedonic well-being) and personal growth predicts feelings of interest (eudaimonic well-being), but not vice-versa (Vittersø and Søholt 2011). However, the distinction has been questioned recently; a large, international study found latent factor correlations of r > .90 between the constructs and limited evidence of discriminant validity (Disabato et al. 2016). Additionally, although there is evidence that hedonic and eudaimonic orientations relate to different aspects of well-being, they appear to have similar links to feelings of satisfaction and vitality (Huta and Ryan 2010). Moreover, those with both hedonic and eudaimonic motives are more likely to have particularly high levels of well-being (Huta and Ryan 2010).

1.1 The MHC-SF

To combat the problems with the measurement of well-being, Keyes (1998) proposed a tripartite model of positive mental health, encompassing the previously proposed EWB and PWB models, plus his own social model derived from the ideologies of Marx’ class conscientiousness and Durkheim’s social cohesion. The social component of mental health includes five dimensions: social integration, social acceptance, social contribution, social coherence, and social actualization. All three components of positive mental health were integrated into one comprehensive measure of well-being -- the Mental Health Continuum (Keyes 2002), which includes items derived from Bradburn’s (1969) Affect Balance Scale, Ryff’s (1989) Scale of Psychological Well-Being, and Keyes’ (1998) model of social well-being. The scales demonstrate good convergent and discriminant validity and high internal consistency (α > .80, Keyes 2005a). The Mental Health Continuum-Short Form (MHC-SF) was put forth a few years later (Keyes 2005a). In 2013, Hubka and Lakaski examined mental health predictors and determinants in a large Canadian sample and consequently contributed to the debate about the standardization of measuring positive mental health as a factor in the Canadian Community Health Survey (CCHS). In 2012, Statistics Canada made official the use of the MHC-SF as the standard measure for positive mental health in the CCHS (Hubka and Lakaski 2013).

1.2 Previous Research with the MHC-SF

In order to understand the structure of well-being, many researchers have investigated the best fitting model of mental health according to the MHC-SF (e.g., Gallagher et al. 2009; Joshanloo et al. 2013; Jovanović 2015; Machado and Bandeira 2015). However, to our knowledge, the models have not been tested in a Canadian sample. Earlier studies examining the structure of the MHC-SF using confirmatory factor analysis (CFA) found support for a correlated three-factor model encompassing EWB, SWB, and PWB (Guo et al. 2015; Lamers et al. 2012; Lamers et al. 2011; Petrillo et al. 2015). However, several more recent studies have found support for a bifactor model, with one general factor upon which all items load and three specific factors corresponding to EWB, SWB, and PWB (de Bruin and du Plessis 2015; Echeverría et al. 2017; Hides et al. 2016; Jovanović 2015; Machado and Bandeira 2015). A few studies have examined the MHC-SF using Exploratory Structural Equation Modelling (ESEM; Asparouhov and Muthén 2009), which combines aspects of CFA and exploratory factor analysis (EFA) to allow items to load onto all possible factors. The less restrictive model often improves fit and provides more accurate estimates of cross-loadings and factor correlations (Marsh et al. 2014), especially for multidimensional measures with conceptual overlap, such as the MHC-SF (Joshanloo and Lamers 2016). Such studies have found excellent fit when investigating three-factor (Joshanloo and Lamers 2016) and bifactor (Longo et al. 2017; Schutte and Wissing 2017) solutions. In this research we aim to expand on previous studies and investigate the best fitting model of Keyes’ (2005a) MHC-SF in a large Canadian sample.

1.3 Group Differences in Positive Mental Health

Group mean-level differences on positive mental health and the MHC-SF raise questions about whether the measure operates consistently in diverse groups. Across studies, gender has been associated with differences in positive mental health. Past studies have found that females report higher levels of overall positive mental health (Peter et al. 2011) and positive relations with others (Ryff et al. 2003). However, studies using the MHC-SF have found inconsistent gender differences. One study reported that females scored higher on EWB, with no differences on SWB or PWB (Joshanloo and Lamers 2016). Yet, another found that men reported higher levels of SWB, EWB, and overall well-being (Petrillo et al. 2015). These discrepancies may be moderated by other characteristics, as a cross-cultural review found that women were generally happier than men, but the gaps were greater in wealthier, older, more highly educated, and urban groups (Graham and Chattopadhyay 2013). Despite the lack of a consistent pattern, gender remains a relevant factor in our understanding of mental health.

Ethnic minority status has also been associated with differences in positive mental health. Black individuals have been found to report higher levels of overall mental health, compared to Caucasians (Keyes 2007). Similarly, facets of eudaimonic well-being -- including self-acceptance, environmental mastery, and autonomy -- have been found to be higher in African American groups (Ryff et al. 2003). However, results relating ethnic minority status to positive mental health have not been entirely consistent. Peter et al. (2011) did not find any significant differences in positive mental health related to ethnic identity. Other researchers reported that, although there were only small effect sizes associated with differences in positive mental health across sociocultural groups, Caucasian respondents consistently had the highest levels of positive mental health (You et al. 2015).

In terms of age, Keyes (2002) found that adults between the ages of 45 and 74 are more likely to report higher levels of mental health than younger adults. In a more recent study, this was further differentiated to suggest that as age increases, levels of EWB increase, levels of SWB remain the same, and levels of PWB decrease (Westerhof and Keyes 2010). However, these relations may be moderated by older adults’ self-perceptions, as it has been shown that lower felt age (or the age that the respondent most feels like) is a strong predictor of flourishing positive mental health, independent from chronological age (Keyes and Westerhof 2012). Despite these factors, older age appears to consistently be associated with higher positive mental health. As positive mental health has often shown differences along age, gender, and ethnic lines, we aim to determine whether these differences are genuine, or due to inconsistent operating of the MHC-SF.

2 Methods

2.1 Sample

Data were taken from the CCHS (Statistics Canada 2012a), a cross-sectional national survey produced by Statistics Canada that collects health information from individuals 12 years and older from all provinces and territories. Those living in remote regions -- on First Nations reserves or Crown Lands -- in institutions and full-time members of the Canadian forces were excluded (Statistics Canada 2012b). Collection began in 2000 to provide health information to researchers and assist policy makers in implementing effective programs. The CCHS is designed to obtain information on healthcare utilization and health determinants in the Canadian population (Statistics Canada 2012b). Data were collected via computer-assisted interviewing (CAI) either in-person or by telephone. The need for interviews by proxy was determined case by case in the event the respondent was unavailable or unable to participate without assistance. This research utilized the 2012 Public Use Microdata File (PUMF), a component of the CCHS made available to the public at the end of each sampling period; the 2012 PUMF is a one-year file (as opposed to two-year) and contains data from 61,707 participants.

2.2 Measure of Well-Being

To assess mental well-being, the CCHS used the MHC-SF (Keyes 2005a). Respondents are asked on a 6-point scale ranging from ‘never’ to ‘every day,’ how often in the past month they experienced positive emotion and felt satisfied with their life (EWB; 3 items), thought that their society was good (SWB; 5 items), and had positive psychological functioning (PWB; 6 items). The overall internal consistency of the scale is good; in past studies it has shown α = .86 in an Italian population (Petrillo et al. 2015), α = .91 for the Polish adaptation of the scale (Karaś et al. 2014), and α = .74 in a South African population (Keyes et al. 2008). See the Appendix for the subscales, dimensions, and items of the MHC-SF.

2.3 Statistical Analysis

2.3.1 Data Cleaning and Preparation

The data were cleaned in SPSS version 20 (IBM Corp 2011). Given the extremely large sample size, cases with missing data were deleted, resulting in a final sample size of 43,020 (30.2% of cases had missing data). The data were then randomly divided into five subsamples with 8604 participants in each, to examine the replicability of the models and fit indices. This procedure allowed a balance between investigating the stability of the factor solutions across multiple random subsamples, with large enough samples to satisfy statistical and ecological validity, in keeping with Random Subsample Replication (Finifter 1972). In each of the subsamples, 54.6% to 56% of the cases were female and 14.2% to 15.3% were visible minorities. See Table 1 for a detailed description of each of the subsamples. The MHC-SF item histograms indicated a high degree of non-normality, with most participants endorsing high levels of positive mental health. Additionally, several of the items had skewness and kurtosis values beyond the published cut-off of ±2.3 for severe non-normality (Lei and Lomax 2005).
Table 1

Characteristics of samples (in Percentages)

Characteristic

Sample

1

2

3

4

5

Sex

 Male

45.3

44.3

44.0

44.1

45.4

 Female

54.7

55.7

56.0

55.9

54.6

Province of residence

 Newfoundland & Labrador

3.1

3.0

2.8

2.8

3.2

 Prince Edward Island

1.6

1.6

1.5

1.5

1.4

 Nova Scotia

3.9

4.0

3.9

3.6

3.6

 New Brunswick

4.3

4.2

4.6

4.6

4.3

 Quebec

19.7

19.1

19.5

19.5

19.9

 Ontario

34.2

34.6

33.1

33.9

33.3

 Manitoba

5.0

4.8

5.5

5.4

5.3

 Saskatchewan

5.3

5.6

5.2

5.0

5.4

 Alberta

8.6

8.7

9.3

8.8

9.1

 British Columbia

11.8

12.1

12.2

12.2

11.8

 Yukon, NWT, Nunavut

2.5

2.2

2.3

2.7

2.5

Cultural Identity

 White

85.4

85.8

85.6

84.7

85.5

 Visible Minority

14.6

14.2

14.4

15.3

14.5

Age

 12 to 24

17.4

16.4

17.1

17.5

17.8

 25 to 49

33.3

33.7

33.3

33.9

33.7

 50 to 69

35.3

35.4

35.6

35.0

35.5

 70 to 80+

14.0

14.4

14.0

13.6

13.0

Marital Status

 Married

42.5

44.5

42.1

42.9

43.4

 Single/Never Married

29.9

29.2

30.1

30.4

30.0

 Widowed/Separated/Divorced

18.3

17.3

18.2

17.7

16.9

Field of Occupation

 Manager/Art/Education

20.9

20.9

20.8

21.1

20.7

 Business/Financial

9.5

10.3

10.0

9.5

9.4

 Sales/Services

13.9

13.1

13.1

13.4

12.8

 Trades/Transportation

7.6

7.8

7.7

8.4

8.5

 Prime Industry

4.4

4.5

4.8

4.3

4.9

Education

 Less than secondary

9.0

9.2

9.5

9.2

9.0

 Secondary grad

12.7

12.1

12.6

12.8

11.9

 Other Post-Secondary

3.8

3.5

3.6

3.7

3.6

 Post-Secondary Graduate

74.5

75.3

74.3

74.4

75.4

Total Income

 0 or < 20,000

9.9

9.5

10.0

9.9

9.6

 20,000 to 39,000

19.7

19.2

19.5

18.6

19.0

 40,000 to 59, 000

18.6

19.0

18.3

18.9

18.3

 60,000 to 79, 000

14.7

15.4

15.3

15.1

15.6

  > 80,000

37.1

36.9

36.9

37.5

37.6

2.3.2 Confirmatory Factor Analysis and Exploratory Structural Equation Modelling

All CFA computations were completed using the lavaan package (version 0.5–22; Rosseel 2012) in RStudio Version 1.0.136 (RStudio 2015) and ESEM analyses were conducted in Mplus Version 7 (Muthén and Muthén 1998–2015). Given the extent of univariate non-normality, multivariate normality could not be assumed. Thus, to address non-normality, the robust scaled chi-square (χ2R; MLR estimator; Yuan and Bentler 2000) is reported for all CFA results, where a non-significant χ2R value is indicative of good fit. Similarly, MLR estimation was used in the ESEM analyses. The target rotation method was used for the three-factor ESEM solution and orthogonal target rotation was used for the bifactor ESEM analysis, which is consistent with Schutte and Wissing (2017).

When the sample is very large, minor misspecifications may be exaggerated, leading to a significant χ2 tests (Cheung and Rensvold 2002). As such, fit indices were also used to evaluate model fit, including the Comparative Fit Index (CFI), Tucker-Lewis Index (TLI), Root Mean Square Error of Approximation (RMSEA), and the Standardized Root Mean Square Residual (SRMR). Conventional criteria for evaluating fit indices are as follows: CFI and TLI ≥ .95, RMSEA ≤.06, and SRMR ≤.08 (Hu and Bentler 1999). In the present study, the following models were tested:
  1. 1.

    The single-factor CFA model for which all items of the MHC-SF load onto one general factor of well-being (see Fig. 1).

     
  2. 2.

    The two-factor CFA model with two correlated dimensions: hedonic well-being (items 1 through 3, which comprise EWB) and eudaimonic well-being (items 4–14, which comprise SWB and PWB; see Fig. 2).

     
  3. 3.

    The three-factor CFA model with three correlated dimensions of well-being (items 1 through 3 on EWB, items 4 through 8 on SWB, and items 9 through 14 on PWB; see Fig. 3).

     
  4. 4.

    The three-factor ESEM model with three dimensions hypothesized to correspond to EWB, SWB, and PWB, allowing each item to also cross-load onto non-target factors (see Fig. 4).

     
  5. 5.

    The bifactor CFA model, with three specific factors and a general factor onto which all items load, with all factors kept orthogonal (see Fig. 5).

     
  6. 6.

    The bifactor ESEM model, with three specific factors (and all possible cross-loadings allowed) and an orthogonal general factor onto which all items load (see Fig. 6).

     
Fig. 1

Single factor model of the MHC-SF

Fig. 2

Correlated two-factor model of the MHC-SF

Fig. 3

Correlated three-factor model of the MHC-SF

Fig. 4

Three-factor ESEM model of the MHC-SF

Fig. 5

Bifactor CFA model of the MHC-SF

Fig. 6

Bifactor ESEM model of the MHC-SF

2.3.3 Measurement Invariance and Latent Mean Differences

In order to determine whether it is appropriate to use the MHC-SF to compare group scores, the best fitting model was tested for measurement invariance across three demographic groupings: males versus females, visible minorities versus Caucasians, and four age categories. Measurement invariance refers to the ability of a measure to produce scores that represent the same construct in the same way across two or more groups (Byrne and Stewart 2006; Meredith 1993). Invariance is tested through a series of increasingly restrictive models, wherein parameter estimates are constrained to be equal across the groups (Meredith 1993). Each model is compared to the previous, less restrictive model. If the decrement in fit after constraining the parameters is not of a meaningful magnitude, measurement invariance is supported (Meredith 1993).

First, the model is tested independently in each group, to ensure that the model has adequate fit in each (Byrne and Stewart 2006). Then, the same factor structure is imposed on both groups simultaneously, to test whether similar, though not identical, latent variables are supported in each group (Chen et al. 2005). This is referred to as configural invariance. The next level of invariance is referred to as weak or factor loading invariance, wherein the factor loadings are constrained to be equal across the groups. If supported, this indicates that the unit of measurement of the underlying factors is the same across groups (Byrne and Stewart 2006). Strong or intercept invariance constrains the items’ intercepts to be equal across the groups, supporting the existence of a common zero point across the groups for the scores on the scale (Byrne and Stewart 2006). The final stage of invariance testing is referred to as strict or residual invariance and is supported if the error terms are found to be equal across the groups. This would indicate that there is the same amount of random error in each of the observed variables between groups (Chen et al. 2005).

To formally evaluate measurement invariance, scaled (robust) chi-square difference tests (χ2D) are used to examine whether there is a significant difference in fit between two nested models (Asparouhov and Muthén 2010; Satorra and Bentler 2010). In invariance testing, a significant χ2D value between levels indicates that the added constraints significantly decrement fit, and thus provides evidence against invariance. However, the χ2D test is sensitive to sample size and may overstate non-substantive discrepancies (Cheung and Rensvold 2002). This is especially pertinent in the current study, due to the enormous sample size and substantial power. Thus, following the suggestions of Cheung and Rensvold (2002) and Chen (2007), changes in CFI > −.01 and RMSEA > .015 were used to indicate that invariance was not tenable.

Although there is some debate as to the level of invariance that must be supported in order for a measure to be considered truly invariant, it is accepted that invariance of factor loadings and intercepts is sufficient to permit group comparisons on latent variable means (Byrne and Stewart 2006; Chen et al. 2005; Meredith 1993; Thompson and Green 2006). To test latent mean differences, one group is chosen as the reference group; its latent mean is set to zero, while the latent mean of the other group is freely estimated. If the non-reference group’s latent mean is significantly different from 0, the groups are found to differ significantly on their latent means (Byrne and Stewart 2006).

3 Results

3.1 Best Model of MHC-SF

Table 2 presents the means and standard deviations for each scale item by subsample and Table 3 reports the fit indices across all samples and all models. The analyses revealed that the bifactor ESEM model was the best fitting model. The single-factor CFA showed poorest fit to the data, followed by the two-factor CFA model. The fit of the three-factor CFA model was just below acceptable levels. This improved to acceptable fit when the three-factor ESEM model was tested. The fit of the bifactor CFA model was comparable to, though generally slightly worse than, that of the three-factor ESEM model. The final model tested was the bifactor ESEM model, with three specific factors and one general orthogonal factor. This model showed excellent fit across all indices and samples. As such, it was selected as the best fitting model for further analyses. Examination of the factor loadings (see Table 4) shows that the general factor had consistently strong loadings, whereas the specific factors had somewhat variable target loadings. For instance, the SWB factor had target loadings ranging from −.06 (item 4) to .61 (item 8). Four of the items had loadings of less than .30 onto their target factor (items 4, 5, 12, and 14). However, each of these had strong loadings onto the general factor (.68, .61, .57, and .68, respectively). The cross-loadings of items onto their non-target factors were all relatively low, with none exceeding .15.
Table 2

Means and standard deviations for MHC-SF items, by subsample

Item

Sample 1

Sample 2

Sample 3

Sample 4

Sample 5

M

SD

M

SD

M

SD

M

SD

M

SD

1

5.14

0.92

5.14

0.93

5.13

0.92

5.13

0.92

5.14

0.92

2

5.49

0.88

5.49

0.88

5.49

0.86

5.47

0.89

5.49

0.86

3

5.29

1.05

5.30

1.03

5.29

1.02

5.28

1.04

5.29

1.04

4

4.54

1.56

4.52

1.55

4.50

1.55

4.51

1.57

4.50

1.57

5

4.57

1.76

4.56

1.76

4.52

1.77

4.57

1.75

4.54

1.76

6

3.94

1.80

3.93

1.80

3.92

1.79

3.93

1.79

3.93

1.80

7

4.93

1.21

4.94

1.21

4.94

1.20

4.91

1.22

4.92

1.22

8

4.08

1.65

4.10

1.67

4.08

1.66

4.11

1.65

4.09

1.65

9

5.33

0.95

5.34

0.93

5.33

0.94

5.34

0.93

5.35

0.93

10

5.40

0.87

5.42

0.87

5.39

0.89

5.38

0.89

5.40

0.87

11

5.49

0.91

5.48

0.93

5.49

0.90

5.47

0.93

5.49

0.90

12

5.01

1.27

4.99

1.29

5.00

1.26

5.00

1.28

4.98

1.30

13

5.33

1.01

5.33

1.01

5.31

1.02

5.31

1.03

5.33

1.00

14

5.25

1.17

5.25

1.16

5.23

1.17

5.24

1.19

5.25

1.16

Table 3

Fit indices across models and samples

Sample

χ2R (df)

CFI

TLI

RMSEA 95% [CI]

SRMR

Model 1: Single factor CFA

 1

4279.28*** (77)

.823

.791

.104 [.101, .106]

.062

 2

4434.59*** (77)

.818

.785

.107 [.104, .110]

.063

 3

3946.96*** (77)

.833

.803

.101 [.098, .103]

.061

 4

4397.21*** (77)

.831

.801

.105 [.102, .108]

.061

 5

4354.88*** (77)

.817

.784

.105 [.102, .108]

.063

Model 2: Two correlated factors CFA

 1

2914.87*** (76)

.882

.859

.085 [.083, .088]

.052

 2

2998.93*** (76)

.879

.855

.088 [.085, .090]

.053

 3

2872.38*** (76)

.880

.857

.086 [.083, .088]

.053

 4

2946.63*** (76)

.889

.867

.086 [.083, .089]

.051

 5

2935.01*** (76)

.879

.855

.086 [.083, .089]

.053

Model 3: Three correlated factors CFA

 1

1749.42*** (74)

.932

.916

.066 [.063, .069]

.042

 2

1768.95*** (74)

.931

.915

.067 [.064, .070]

.042

 3

1610.75*** (74)

.935

.920

.064 [.061, .067]

.042

 4

1789.12*** (74)

.935

.920

.067 [.064, .069]

.042

 5

1790.54*** (74)

.928

.912

.067 [.064, .070]

.044

Model 4: Three-factor ESEM

 1

896.10*** (52)

.963

.935

.043 [.041, .046]

.024

 2

1004.51*** (52)

.959

.928

.046 [.044, .049]

.025

 3

838.90*** (52)

.965

.939

.042 [.039, .044]

.024

 4

933.13*** (52)

.964

.937

.044 [.042, .047]

.024

 5

1041.49*** (52)

.956

.923

.047 [.045, .050]

.026

Model 5: Bifactor CFA

 1

1186.97*** (63)

.958

.939

.056 [.053, .059]

.028

 2

1210.63*** (63)

.957

.937

.058 [.055, .061]

.029

 3

1089.41*** (63)

.962

.944

.053 [.051, .056]

.028

 4

1213.82*** (63)

.961

.944

.055 [.053, .058]

.027

 5

1374.98*** (63)

.952

.931

.059 [.057, .062]

.030

Model 6: Bifactor ESEM

 1

379.35*** (41)

.985

.967

.031 [.028, .034]

.013

 2

420.63*** (41)

.984

.963

.033 [.030, .036]

.014

 3

295.43*** (41)

.989

.975

.027 [.024, .030]

.012

 4

372.70*** (41)

.986

.970

.031 [.028, .034]

.013

 5

418.56*** (41)

.983

.963

.033 [.030, .036]

.014

χ2R - Scaled chi-square test; CFI - Comparative Fit Index; TLI - Tucker-Lewis Index; RMSEA - Root Mean Square Error of Approximation; CI - Confidence Interval; SRMR - Standardized Root Mean Square Residual

***p < .001

Table 4

Standardized factor loadings, error variances, R2, and reliability estimates for bifactor ESEM model

Item

Standardized factor loadings

Error variances

R2

General

EWB

SWB

PWB

1

.55

.47

−.01

.06

.48

.53

2

.61

.47

−.06

.07

.41

.59

3

.63

.52

.02

.09

.33

.68

4

.68

−.11

−.06

−.12

.51

.49

5

.61

−.09

.12

−.14

.59

.42

6

.54

−.05

.52

−.09

.42

.58

7

.44

.02

.39

.15

.63

.37

8

.42

.01

.61

.03

.45

.55

9

.53

.11

.11

.36

.56

.44

10

.46

.15

.03

.39

.61

.39

11

.53

.08

.01

.34

.60

.40

12

.57

−.13

−.01

.20

.62

.39

13

.52

−.04

−.07

.37

.59

.41

14

.68

.11

−.01

.27

.45

.55

ω

.91

.63

.52

.52

  

Factor loadings above .30 are bolded. EWB - Emotional Well-Being; SWB - Social Well-Being; PWB - Psychological Well-Being

3.2 Reliability

The inter-item correlations for all five sub-samples are low to moderate, indicating that the items are in fact testing for unique constructs. The average inter-item correlation for a set of items should be between .20 and .40, suggesting that while the items are reasonably homogenous, they do contain sufficiently unique variance so as to not be isomorphic with each other (Piedmont 2014). All inter-item correlations by sub-sample can be found in the Appendix. Latent variable reliability estimates (omega; ω) were calculating using McDonald’s (1970) formula of ω = (Σ|λi |)2 / ([Σ|λi |]2 + Σδii), with λi representing the factor loadings and δii the error variances (see Table 4). This calculation follows Perreira et al.’s (2018) discussion of how ω can be applied to bifactor models and is consistent with other studies of bifactor ESEM models, including Schutte and Wissing’s (2017) investigation of the MHC-SF. In the current study, the general factor showed strong reliability, whereas the specific subscales showed more moderate reliability. However, using the guideline of ω > .50 suggested by Perreira et al. (2018) for bifactor models, the subscales showed satisfactory reliability.

3.3 Measurement Invariance

As the bifactor ESEM model had the best fit, it was selected for measurement invariance testing. The total sample was split three ways; first, by gender, creating a male group and a female group. Then, the sample was split into Caucasians and visible minorities. Finally, the sample was split based on age into four groups; ages 12 to 24, 25 to 49, 50 to 69, and 70 to 80 or older. Table 5 presents the results for all invariance analyses. As can be seen for males versus females, the greatest ΔCFI across model comparisons was −.003 and the largest decrement in fit according to ΔRMSEA was .001, indicating that the MHC-SF shows strict invariance when comparing men and women. That is, the factor structure, factor loadings, intercepts, and item residuals were all equivalent between men and women. Similarly, invariance of the MHC-SF was supported for Caucasians versus visible minorities, up to the level of strict invariance, as the largest ΔCFI = −.002 and the greatest decrement to ΔRMSEA = .001. Thus, the MHC-SF appears
Table 5

Measurement invariance testing

Model

Description

χ2R

df

χ2D (Δdf)

CFI

ΔCFI

TLI

RMSEA [90% CI]

ΔRMSEA

SRMR

Gender

 1a

Males only

725.94**

41

.986

.969

.029 [.028, .031]

.013

 1b

Females only

1052.86**

41

.985

.966

.032 [.031, .034]

.013

 2

Configural

1778.71**

82

.985

.967

.031 [.030, .032]

.013

 3

Weak (factor loadings)

1897.25**

122

66.55* (40)

.985

.000

.977

.026 [.025, .027]

−.005

.021

 4

Strong (intercepts)

2238.20**

132

343.22** (39)

.982

−.003

.975

.027 [.026, .028]

.001

.025

 5

Strict (variances)

2431.22**

146

69.25** (14)

.980

−.002

.975

.027 [.026, .028]

.000

.047

Ethnic Minority Status

 1a

Caucasians only

1459.58**

41

.986

.968

.031 [.029, .032]

.013

 1b

Minorities only

364.40**

41

.981

.958

.035 [.032, .039]

.015

 2

Configural

1846.69**

82

.985

.966

.032 [.030, .033]

.013

 3

Weak (factor loadings)

1920.59**

122

41.46† (40)

.984

−.001

.977

.026 [.025, .027]

−.006

.018

 4

Strong (intercepts)

2205.12**

132

320.42** (39)

.982

−.002

.975

.027 [.026, .028]

.001

.021

 5

Strict (variances)

2323.77**

146

42.88** (14)

.981

−.001

.977

.026 [.025, .027]

−.001

.032

Age

 1a

Ages 12–24 only

299.69**

41

.986

.970

.029 [.026, .032]

.013

 1b

Ages 25–49 only

715.01**

41

.985

.966

.034 [.032, .036]

.013

 1c

Ages 50–69 only

673.32**

41

.985

.967

.032 [.030, .034]

.013

 1d

Ages 70–80+ only

261.30**

41

.983

.961

.030 [.027, .034]

.015

 2

Configural

1957.54**

164

.985

.966

.032 [.031, .033]

.013

 3

Weak (factor loadings)

2541.79**

284

325.37** (120)

.981

−.004

.975

.027 [.026, .028]

−.005

.030

 4

Strong (intercepts)

4812.06**

314

2631.69** (30)

.962

−.019

.956

.036 [.036, .037]

.009

.041

χ2R - Scaled Chi-Square test; χ2D - Scaled Chi-Square difference test; CFI - Comparative Fit Index, ΔCFI - Change in Comparative Fit Index; TLI - Tucker-Lewis Index, RMSEA - Root Mean Square Error of Approximation; 90% CI - 90% Confidence Interval; ΔRMSEA - Change in Root Mean Square Error of Approximation; SRMR - Standardized Root Mean Square Residual

† p > .05, * p < .01, ** p < .001

to measure the same constructs in Caucasian and visible minority groups; the item scaling, zero-points, and error terms are all identical between the groups.

Invariance across the four age groupings was supported up to the level of weak invariance. When the intercepts were constrained to be equal across the groups, the ΔCFI was −.019, exceeding the threshold of ΔCFI of −.01 outlined by (Cheung and Rensvold 2002). However, the ΔRMSEA of .009 was within published cut-offs (Chen 2007). Because the verdicts of the two guidelines were inconsistent, it was conservatively concluded that strong invariance was not tenable. As such, across age groupings, the MHC-SF shows similar overall structure and loading patterns, but the age groupings appear to differ in terms of their intercepts, or the zero-point of the MHC-SF item scaling. Partial intercept invariance was examined across the four age groupings. However, even after 25% of the intercept constraints had been released, the decrement in fit between weak and strong invariance was still large enough to indicate that strong invariance was not supported according to ΔCFI (χ2R (300) = 3834.27, p < .001; CFI = .970, ΔCFI = −.011; RMSEA = .033 [.032, .034], ΔRMSEA = .006). Moreover, in a Monte Carlo study, Steinmetz (2013) found that even one unequal intercept can bias latent factor scores, further suggesting that neither full nor partial strong invariance of the MHC-SF across age groupings is tenable.

3.4 Latent Mean Differences

As both weak and strong invariance were supported between males and females, latent mean differences between the groups were investigated. With males chosen as the reference group, the results showed that females scored statistically significantly lower on EWB (M = −.09, SE = .013, p < .001), PWB (M = −0.07, SE = .015, p < .001), and SWB (M = −0.07, SE = .014, p < .001), where M refers to the unstandardized fitted mean. An effect size was computed for each of these differences, based on the formula for Cohen’s d (Cohen 1992). With strict invariance supported, the group variances were pooled for each factor. The differences between males and females on the specific factors correspond to Cohen’s d values of 0.08 for EWB, 0.07 for SWB, and 0.07 for PWB -- indicating that the differences are generally of a trivial magnitude. It was also found that females had significantly higher levels of general positive mental health than males (M = 0.05, SE = .011, p < .001). However, this corresponds to a Cohen’s d of 0.05, indicating that the magnitude of the difference between males’ and females’ overall positive mental health is negligible.

Latent mean group differences were also tested between visible minorities and Caucasians, with Caucasians set as the reference group. The results indicated that visible minorities had statistically significantly lower levels of EWB (M = −0.20, SE = .020, p < .001) and PWB (M = −0.10, SE = .022 p < .001), but higher levels of SWB (M = 0.26, SE = .020, p < .001). The Cohen’s d values for EWB and SWB were 0.20 and 0.26, respectively, which correspond to small effect sizes. The group difference effect size on PWB was 0.10, indicating that the effect was of less practical significance. There was no difference between Caucasians and visible minorities on overall positive mental health (M = 0.00, SE = .016, p > .05).

4 Discussion

The aim of the present study was to evaluate the structure of Keyes’ (2005a) MHC-SF in five subsamples of a large Canadian dataset (CCHS; Statistics Canada 2012a). Analyses revealed that the bifactor ESEM model, with one general factor and three specific factors corresponding to EWB, SWB, and PWB, had the best fit to the data. This is consistent with Schutte and Wissing (2017), who tested CFA and ESEM models with three-factor and bifactor solutions and found that the bifactor ESEM model had the best fit. This study also builds upon past research that found acceptable fit with bifactor CFA models (de Bruin and du Plessis 2015; Echeverría et al. 2017; Hides et al. 2016; Jovanović 2015; Machado and Bandeira 2015). The less stringent ESEM analysis identified consistently small but non-zero cross-loadings; allowing these to be freely estimated, rather than constrained to zero as in a CFA model, provided excellent fit to the data. Moreover, these findings are consistent with past ESEM examinations of the MHC-SF, which report generally strong target loadings with non-zero cross-loadings (Joshanloo and Lamers 2016; Longo et al. 2017; Schutte and Wissing 2017). The cross-loadings from the bifactor solutions reported in this and other investigations (Longo et al. 2017; Schutte and Wissing 2017) are smaller than those in the three-factor solution presented by Joshanloo and Lamers (2016), suggesting that the addition of the general factor accounts for some of the overlap between scale constructs.

In the current investigation, the general well-being factor showed strong reliability, with more moderate reliability estimates for the specific factors. This is consistent with Schutte and Wissing (2017), who found support for a strong general factor and less reliable specific subscales. It is important to note that the addition of the specific factors greatly improved fit over the first, unidimensional, model tested. However, as with other research supporting bifactor models of the MHC-SF, these results suggest that researchers should exercise caution when using and interpreting mean specific subscale scores (de Bruin and du Plessis 2015; Echeverría et al. 2017; Hides et al. 2016; Jovanović 2015; Machado and Bandeira 2015; Schutte and Wissing 2017). Broadly, the strength of the general factor is consistent with a comparison of hedonic and eudaimonic constructs, which found support for a more general conceptualization of well-being (Disabato et al. 2016).

The bifactor ESEM structure showed a very high level of measurement invariance across both gender and ethnic minority status, which allowed latent mean differences to be calculated for both groupings. The significance of the reported differences between males and females is likely due primarily to the large sample size; the effect sizes associated with these differences revealed little practical significance. However, ethnic minorities were found to have lower EWB and higher SWB, of small effect size magnitude. Although consistent patterns of positive mental health across ethnic status have not been established in past literature, nationally representative examinations of patterns of mental illness may provide some clues. One such investigation found that rates of mental illness were lower among immigrants who lived in communities with higher concentrations of other immigrants, suggesting a buffering effect of community involvement and SWB (Menezes et al. 2011). Another study found that Asian-Canadians were less likely to utilize mental health services compared to Caucasians, indicating that ethnic minority status may be associated with under-treatment of psychological difficulties, leading to diminished EWB (Tiwari and Wang 2008). However, the ethnic group differences reported here are of small magnitude, remain to be replicated, and likely result from a number of contributing factors. Only weak invariance was supported between the four age categories, indicating that respondents of different ages may not respond to the items from the same reference point, thus making age-group mean comparisons on the MHC-SF questionable.

Further research using the long version of the MHC-SF should be conducted, because multidimensional constructs such as subjective well-being are difficult to record using only a few items. The MHC-SF uses one item for each of the SWB and PWB dimensions and only three items to measure EWB (Keyes 2005a), which may be inadequate to measure such complex constructs. Similarly, the use of a longer inventory may assist researchers in circumventing the difficulties with the specific MHC-SF subscales described above. Moreover, the strength of the general factor in accounting for variance, when compared to the specific factors, calls for a refinement of or elaboration on the specific factors.

This study is one of a few to use the MHC-SF to examine wellness factors in a Canadian context. Hubka and Lakaski (2013) studied mental health predictors as a way to encourage the Canadian government to begin the study of mental well-being in national research. The CCHS 2012 was the first edition to study the construct. However, to the best of our knowledge, ours the first Canadian study to examine the structure of well-being using Keyes’ (2005a) MHC-SF. This research is a testament to the use of the scale as a valid measure and further supports the inclusion of the MHC-SF in the CCHS. Another strength of this study is the very large sample (N = 43,020); structural equation modelling is well equipped to handle large datasets, especially through fit indices such as the CFI and TLI, as chi-square values can be biased in samples that are very large, as in the present study (Cheung and Rensvold 2002). Further, the sample is representative of the Canadian population, aside from a few regions that were not sampled by the CCHS, and the samples used to compare the models were divided equally.

Hubka and Lakaski (2013) used the PHAC definition of mental health and compared it to MHC-SF factors. PHAC describes mental health as “the capacity of each and all of us to feel, think, and act in ways that enhance our ability to enjoy life and deal with the challenges we face. It is a positive sense of emotional and spiritual well-being that respects the importance of culture, equity, social justice, interconnections and personal dignity” (PHAC, 2014b). The ability to maintain at least an adequate level of mental wellness will help to combat the overwhelming number of people with mental health issues. Research on the introduction and maintenance of positive mental health will benefit individuals as well as organizations and government agency. We can move from a system of treating illness to a system of mental health promotion such as resilience training, coping strategies and maintenance of support systems.

4.1 Limitations of the Study

One limitation of this research is the exclusion of a number of key populations from the CCHS (Statistics Canada 2012b), meaning that the data might not be completely representative of the Canadian population. First Nations reserves, full time members of armed forces, and institutionalized populations including those in jails, hospitals, and nursing homes are populations that intuitively house individuals at greater risk for mental health issues. Additionally, North America is made up of a mosaic of culturally diverse peoples. However, the samples used in this research consisted of 81% Caucasian participants. Exploration of a Canadian sample with greater emphasis on a multicultural sample would thus be of immense benefit, especially given the findings of positive mental health as a function of ethnic minority status.

4.2 Conclusion

The present study underscores the utility of the MHC-SF in the Canadian context and supports the bifactor ESEM solution as the most appropriate model for the measure. The results of this study suggest that positive mental health can be thought of as a general construct, that subsumes more specific facets of living well, including positive feelings and attaining an optimal level of life functioning (Keyes 2005b). These facets are reflected by the specific EWB, SWB, and PWB factors of the MHC-SF (Keyes 2005a). Additionally, positive mental health as a unitary structure can be contrasted with mental illness, suggesting the existence of an underlying general domain of positive mental health (Keyes 2005b). The MHC-SF has been shown to be a valid measure for examining these constructs in various groups. The results reported here indicate that the MHC-SF demonstrates strict invariance across gender and ethnic minority status, supporting group mean comparisons across these groupings. In sum, these results indicate that the MHC-SF is generally a strong measure, with a robust factor structure and measurement invariance when tested between groups that have shown differences in positive mental health.

Notes

Compliance with Ethical Standards

Conflict of Interest

On behalf of all authors, the corresponding author states that there is no conflict of interest. For this type of study formal consent is not required. Informed consent was obtained from all individual participants included in the study.

Supplementary material

41543_2018_7_MOESM1_ESM.docx (51 kb)
ESM 1 (DOCX 50 kb)

References

  1. Asparouhov, T., & Muthén, B. (2009). Exploratory structural equation modeling. Structural Equation Modeling: A Multidisciplinary Journal, 16(3), 397–438.  https://doi.org/10.1080/10705510903008204.CrossRefGoogle Scholar
  2. Asparouhov, T., & Muthén, B. (2010). Computing the strictly positive Satorra-Bentler chi-square test in Mplus. Mplus Web Notes, 12. Retrieved from: http://www.statmodel.com/examples/webnotes/webnote12.pdf.
  3. Bradburn, N. M. (1969). The structure of psychological well-being [monograph]. National Opinion Research, Retrieved from: http://www.norc.org/PDFs/publications/BradburnN_Struc_Psych_Well_Being.pdf
  4. Byrne, B. M., & Stewart, S. M. (2006). Teacher's corner: The MACS approach to testing for multigroup invariance of a second-order structure: A walk through the process. Structural Equation Modeling, 13(2), 287–321.  https://doi.org/10.1207/s15328007sem1302_7.CrossRefGoogle Scholar
  5. Chen, F. F. (2007). Sensitivity of goodness of fit indexes to lack of measurement invariance. Structural Equation Modeling, 14(3), 464–504.  https://doi.org/10.1080/10705510701301834.CrossRefGoogle Scholar
  6. Chen, F. F., Sousa, K. H., & West, S. G. (2005). Teacher's corner: Testing measurement invariance of second-order factor models. Structural Equation Modeling, 12(3), 471–492.  https://doi.org/10.1207/s15328007sem1203_7.CrossRefGoogle Scholar
  7. Cheung, G. W., & Rensvold, R. B. (2002). Evaluating goodness-of-fit indexes for testing measurement invariance. Structural Equation Modeling, 9(2), 233–255.  https://doi.org/10.1207/S15328007SEM0902_5.CrossRefGoogle Scholar
  8. Cohen, J. (1992). A power primer. Psychological Bulletin, 112(1), 155–159.  https://doi.org/10.1037/0033-2909.112.1.155.CrossRefGoogle Scholar
  9. de Bruin, G. P., & du Plessis, G. A. (2015). Bifactor analysis of the mental health continuum—Short form (MHC—SF). Psychological Reports, 116(2), 438–446.  https://doi.org/10.2466/03.02.PR0.116k20w6.CrossRefGoogle Scholar
  10. Diener, E., Emmons, R. A., Larsen, R. J., & Griffin, S. (1985). The satisfaction with life scale. Journal of Personality Assessment, 49(1), 71–75. Retrieved from: http://dx.doi.org.ezproxy.uwindsor.ca/10.1207/s15327752jpa4901_13
  11. Diener, E., Wirtz, D., Biswas-Diener, R., Tov, W., Kim-Prieto, C., Choi, D. W., & Oishi, S. (2009). New measures of well-being. In E. Diener (Ed.), Assessing well-being: The collected works of Ed Diener (pp. 247–266). Dordrecht: Springer.CrossRefGoogle Scholar
  12. Disabato, D. J., Goodman, F. R., Kashdan, T. B., Short, J. L., & Jarden, A. (2016). Different types of well-being? A cross-cultural examination of hedonic and eudaimonic well-being. Psychological Assessment, 28(5), 471–482.  https://doi.org/10.1037/pas0000209.CrossRefGoogle Scholar
  13. Echeverría, G., Torres, M., Pedrals, N., Padilla, O., Rigotti, A., & Bitran, M. (2017). Validation of a Spanish version of the mental health continuum-short form questionnaire. Psicothema, 29(1), 96–102.  https://doi.org/10.7334/psicothem2016.3.Google Scholar
  14. Finifter, B. M. (1972). The generation of confidence: Evaluating research findings by random subsample replication. Sociological Methodology, 4, 112–175.  https://doi.org/10.2307/270731.CrossRefGoogle Scholar
  15. Gallagher, M. W., Lopez, S. J., & Preacher, K. J. (2009). The hierarchical structure of well-being. Journal of Personality, 77(4).  https://doi.org/10.1111/j.1467-6494.2009.00573.x.
  16. Graham, C., & Chattopadhyay, S. (2013). Gender and well-being around the world. International Journal of Happiness and Development, 1(2), 212–232.  https://doi.org/10.1504/IJHD.2013.055648.CrossRefGoogle Scholar
  17. Guo, C., Tomson, G., Guo, J., Li, X., Keller, C., & Söderqvist, F. (2015). Psychometric evaluation of the mental health continuum-short form (MHC-SF) in Chinese adolescents–a methodological study. Health and Quality of Life Outcomes, 13(198), 1–9.  https://doi.org/10.1186/s12955-015-0394-2.Google Scholar
  18. Hides, L., Quinn, C., Stoyanov, S., Cockshaw, W., Mitchell, T., & Kavanagh, D. J. (2016). Is the mental wellbeing of young Australians best represented by a single, multidimensional or bifactor model? Psychiatry Research, 241, 1–7.  https://doi.org/10.1016/j.psychres.2016.04.077.CrossRefGoogle Scholar
  19. Hu, L. T., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6(1), 1–55.  https://doi.org/10.1080/10705519909540118.CrossRefGoogle Scholar
  20. Hubka, D., & Lakaski, C. (2013). Developing research and surveillance for positive mental health: A Canadian process for conceptualization. International Journal of Mental Health Addiction, 11, 658–671.  https://doi.org/10.1007/s11469-013-9443-4.CrossRefGoogle Scholar
  21. Huta, V., & Ryan, R. M. (2010). Pursuing pleasure or virtue: The differential and overlapping well-being benefits of hedonic and eudaimonic motives. Journal of Happiness Studies, 11(6), 735–762.  https://doi.org/10.1007/s10902-009-9171-4.CrossRefGoogle Scholar
  22. IBM Corp. (2011). IBM SPSS statistics for windows, version 20.0. Armonk: IBM Corp.Google Scholar
  23. Joshanloo, M., & Lamers, S. M. (2016). Reinvestigation of the factor structure of the MHC-SF in the Netherlands: Contributions of exploratory structural equation modeling. Personality and Individual Differences, 97, 8–12.  https://doi.org/10.1016/j.paid.2016.02.089.CrossRefGoogle Scholar
  24. Joshanloo, M., Wissing, M. P., Khumalo, I. P., & Lamers, S. M. A. (2013). Measurement invariance of the mental health continuum-short form (MHC-SF) across three cultural groups. Personality and Individual Differences, 55, 755–759.  https://doi.org/10.1016/j.paid.2013.06.002.CrossRefGoogle Scholar
  25. Jovanović, V. (2015). Structural validity of the mental health continuum-short form: The bifactor model of emotional, social and psychological well-being. Personality and Individual Differences, 75, 154–159.  https://doi.org/10.1016/j.paid.2014.11.026.CrossRefGoogle Scholar
  26. Karaś, D., Cieciuch, J., & Keyes, C. L. M. (2014). The polish adaptation of the mental health continuum-short form (MHC-SF). Personality and Individual Differences, 69, 104–109.  https://doi.org/10.1016/j.paid.2014.05.011.CrossRefGoogle Scholar
  27. Keyes, C. L. M. (1998). Social well-being. Social Psychology Quarterly, 61(2), 121–140. Retrieved from: http://www.jstor.org/stable/2787065.
  28. Keyes, C. L. M. (2002). The mental health continuum: From languishing to flourishing in life. Journal of Health and Social Research, 43, 207–222. Retrieved from: http://www.jstor.org/stable/3090197.
  29. Keyes, C. L. (2005a). The subjective well-being of America's youth: Toward a comprehensive assessment. Adolescent & Family Health, 4(1), 3–11.  https://doi.org/10.1037/0022-006X.73.3.539.Google Scholar
  30. Keyes, C. L. M. (2005b). Mental illness and/or mental health? Investigating axioms of the complete state model of health. Journal of Consulting and Clinical Psychology, 73(3), 539–548.  https://doi.org/10.1037/0022-006X.73.3.539.CrossRefGoogle Scholar
  31. Keyes, C. L. (2007). Promoting and protecting mental health as flourishing: A complementary strategy for improving national mental health. American Psychologist, 62(2), 95–108.  https://doi.org/10.1037/0003-066X.62.2.95.CrossRefGoogle Scholar
  32. Keyes, C. L., & Westerhof, G. J. (2012). Chronological and subjective age differences in flourishing mental health and major depressive episode. Aging & Mental Health, 16(1), 67–74.  https://doi.org/10.1080/13607863.2011.596811.CrossRefGoogle Scholar
  33. Keyes, C. L. M., Wissing, M., Potgieter, J. P., Temane, M., Kruger, A., & Rooy, S. (2008). Evaluation of the mental health continuum-short form (MHC-SF) in Setswana-speaking south Africans. Clinical Psychology and Psychotherapy, 15, 181–192.  https://doi.org/10.1002/cpp.572.CrossRefGoogle Scholar
  34. Lamers, S. M. A., Glas, C. A. W., Westerhof, G. J., & Bohlmeijer, E. T. (2012). Longitudinal evaluation of the mental health continuum short form (MHC-SF); measurement invariance across demographics, physical illness, and mental illness. European Journal of Psychological Assessment, 28(4), 290–296.  https://doi.org/10.1027/1015-5759/a000109.CrossRefGoogle Scholar
  35. Lamers, S. M. A., Westerhof, G. J., Bohlmeijer, E. T., Klooster, P. M., & Keyes, C. L. M. (2011). Evaluating the psychometric properties of the mental health continuum-short form (MHC-SF). Journal of Clinical Psychology, 67(1), 99–110.  https://doi.org/10.1002/jclp.20741.CrossRefGoogle Scholar
  36. Lei, M., & Lomax, R. G. (2005). The effect of varying degrees of nonnormality in structural equation modeling. Structural Equation Modeling, 12(1), 1–27.  https://doi.org/10.1207/s15328007sem1201_1.CrossRefGoogle Scholar
  37. Longo, Y., Jovanović, V., Sampaio, D. C. J., & Karaś, D. (2017). The general factor of well-being: Multinational evidence using bifactor ESEM on the mental health continuum-short form. Assessment.  https://doi.org/10.1177/1073191117748394.
  38. Machado, W. L., & Bandeira, D. R. (2015). Positive mental health scale: Validation of the mental health continuum - short form. Universidade de São Francisco, Programa de Pós-Graduação Stricto Sensu em Psicologia, 20(2), 259–274.  https://doi.org/10.1590/1413-82712015200207.Google Scholar
  39. McDonald, R. P. (1970). The theoretical foundations of principal factor analysis, canonical factor analysis, and alpha factor analysis. British Journal of Mathematical and Statistical Psychology, 23(1), 1–21.  https://doi.org/10.1111/j.2044-8317.1970.tb00432.x. CrossRefGoogle Scholar
  40. Menezes, N. M., Georgiades, K., & Boyle, M. H. (2011). The influence of immigrant status and concentration on psychiatric disorder in Canada: A multi-level analysis. Psychological Medicine, 41(10), 2221–2231.  https://doi.org/10.1017/S0033291711000213.CrossRefGoogle Scholar
  41. Meredith, W. (1993). Measurement invariance, factor analysis and factorial invariance. Psychometrika, 58(4), 525–543.  https://doi.org/10.1007/BF02294825.CrossRefGoogle Scholar
  42. Marsh, H. W., Morin, A. J., Parker, P. D., & Kaur, G. (2014). Exploratory structural equation modeling: An integration of the best features of exploratory and confirmatory factor analysis. Annual Review of Clinical Psychology, 10, 85–110.  https://doi.org/10.1146/annurev-clinpsy-032813-153700.CrossRefGoogle Scholar
  43. Muthén, L. K. & B. O. Muthén. (1998-2015). Mplus (Version 7). Los Angeles: Muthén and Muthén.Google Scholar
  44. Peter, T., Roberts, L. W., & Dengate, J. (2011). Flourishing in life: An empirical test of the dual continua model of mental health and mental illness among Canadian university students. International Journal of Mental Health Promotion, 13(1), 13–22.  https://doi.org/10.1080/14623730.2011.9715646.CrossRefGoogle Scholar
  45. Perreira, T. A., Morin, A. J., Hebert, M., Gillet, N., Houle, S. A., & Berta, W. (2018). The short form of the workplace affective commitment multidimensional questionnaire (WACMQ-S): A bifactor-ESEM approach among healthcare professionals. Journal of Vocational Behavior, 106, 62–83.  https://doi.org/10.1016/j.jvb.2017.12.004.CrossRefGoogle Scholar
  46. Petrillo, G., Capone, V., Caso, D., & Keyes, C. L. M. (2015). The mental health continuum - short form (MHC-SF) as a measure of well-being in the Italian context. Social Indicators Research, 121, 291–312.  https://doi.org/10.1007/s11205-014-0629-3.CrossRefGoogle Scholar
  47. Piedmont, R. L. (2014). Inter-item correlations. In A. C. Michalos (Ed.), Encyclopedia of quality of life and well-being research (pp. 3303–3304). Dorerecht: Springer.CrossRefGoogle Scholar
  48. Public Health Agency of Canada, Ministry of Health. (2014a). Economic burden of illness in Canada (Publication No. 130148). Retrieved from: http://www.phac-aspc.gc.ca/publicat/ebic-femc/2005-2008/assets/pdf/ebic-femc-2005-2008-eng.pdf.
  49. Public Health Agency of Canada, Ministry of Health. (2014b). Promoting mental health means promoting the best of ourselves. Retrieved from: http://www.phac-aspc.gc.ca/mh-sm/mhp-psm/index-eng.php.
  50. Rosseel, Y. (2012). Lavaan: An R package for structural equation modeling. Journal of Statistical Software, 48(2), 1–36. Retrieved from: http://www.jstatsoft.org/v48/i02/.
  51. RStudio (2015). RStudio: Integrated development for R (version 1.0.136) [computer software]. Boston. Retrieved from: http://www.rstudio.com/.
  52. Ryan, R. M., & Deci, E. L. (2001). On happiness and human potentials: A review of research on hedonic and eudaimonic well-being. Annual Review of Psychology, 52(1), 141–166.  https://doi.org/10.1146/annurev.psych.52.1.141.CrossRefGoogle Scholar
  53. Ryff, C. D. (1989). Happiness is everything, or is it? Explorations on the meaning of psychological well-being. Journal of Personality and Social Psychology, 57(6), 1069–1081.  https://doi.org/10.1037/0022-3514.57.6.1069.CrossRefGoogle Scholar
  54. Ryff, C. D., Keyes, C. L., & Hughes, D. L. (2003). Status inequalities, perceived discrimination, and eudaimonic well-being: Do the challenges of minority life hone purpose and growth? Journal of Health and Social Behavior, 44(3), 275–291.  https://doi.org/10.2307/1519779.CrossRefGoogle Scholar
  55. Ryff, C. D., & Singer, B. H. (2008). Know thyself and become what you are: A eudaimonic approach to psychological well-being. Journal of Happiness Studies, 9(1), 13–39.  https://doi.org/10.1007/s10902-006-9019-0.CrossRefGoogle Scholar
  56. Satorra, A., & Bentler, P. M. (2010). Ensuring positiveness of the scaled difference chi-square test statistic. Psychometrika, 75(2), 243–248.  https://doi.org/10.1007/s11336-009-9135-y.CrossRefGoogle Scholar
  57. Schutte, L., & Wissing, M. P. (2017). Clarifying the factor structure of the mental health continuum short form in three languages: A bifactor exploratory structural equation modeling approach. Society and Mental Health, 1–17.  https://doi.org/10.1177/2156869317707793.
  58. Seligman, M. E. P., & Csikszentmihalyi, M. (2000). Positive psychology: An introduction. American Psychologist., 55(1), 5–14.  https://doi.org/10.1037//0003-066X.55.1.5.CrossRefGoogle Scholar
  59. Statistics Canada. (2012a). Canadian Community Health Survey, 2012 [Data File]. Available from: http://odesi1.scholarsportal.info.ezproxy.uwindsor.ca/webview/.
  60. Statistics Canada. (2012b). Canadian Community Health Survey – Annual Component (CCHS) [extraneous documents] Retrieved from: http://www23.statcan.gc.ca/imdb/p2SV.pl?Function=getSurvey&SDDS=3226.
  61. Steinmetz, H. (2013). Analyzing observed composite differences across groups: Is partial measurement invariance enough? Methodology: European Journal of Research Methods for the Behavioral and Social Sciences, 9(1), 1.  https://doi.org/10.1027/1614-2241/a000049.CrossRefGoogle Scholar
  62. Thompson, M. S., & Green, S. B. (2006). Evaluating between-group differences in latent variable means. In G. Hancock & R. Mueller (Eds.), Structural equation modeling: A second course (pp. 119–169). Greenwich: Information Age Publishing.Google Scholar
  63. Tiwari, S. K., & Wang, J. (2008). Ethnic differences in mental health service use among white, Chinese, south Asian and south east Asian populations living in Canada. Social Psychiatry and Psychiatric Epidemiology, 43(11), 866.  https://doi.org/10.1007/s00127-008-0373-6.CrossRefGoogle Scholar
  64. Vittersø, J., & Søholt, Y. (2011). Life satisfaction goes with pleasure and personal growth goes with interest: Further arguments for separating hedonic and eudaimonic well-being. The Journal of Positive Psychology, 6(4), 326–335.  https://doi.org/10.1080/17439760.2011.584548.CrossRefGoogle Scholar
  65. Waterman, A. S. (1993). Two conceptions of happiness: Contrasts of personal expressiveness (eudiamonia) and hedonic enjoyment. Journal of Personality and Social Psychology, 64(4), 678–691.  https://doi.org/10.1037/0022-3514.64.4.678.CrossRefGoogle Scholar
  66. Watson, D., Clark, L. A., & Tellegen, A. (1988). Development and validation of brief measures of positive and negative affect: The PANAS scales. Journal of Personality and Social Psychology, 54(6), 1063.  https://doi.org/10.1037//0022-3514.54.6.1063.CrossRefGoogle Scholar
  67. Westerhof, G. L., & Keyes, C. L. M. (2010). Mental illness and mental health: The two continua model across the lifespan. Journal of Adult Development, 17(2), 110–119.  https://doi.org/10.1007/s10804-009-9082-y.CrossRefGoogle Scholar
  68. You, S., Furlong, M., Felix, E., & O'Malley, M. (2015). Validation of the social and emotional health survey for five sociocultural groups: Multigroup invariance and latent mean analyses. Psychology in the Schools, 52(4), 349–362.  https://doi.org/10.1002/pits.21828.CrossRefGoogle Scholar
  69. Yuan, K. H., & Bentler, P. M. (2000). Three likelihood-based methods for mean and covariance structure analysis with nonnormal missing data. Sociological Methodology, 30(1), 165–200.  https://doi.org/10.1111/0081-1750.00078.CrossRefGoogle Scholar

Copyright information

© Springer International Publishing AG, part of Springer Nature 2018

Authors and Affiliations

  • Paige Lamborn
    • 1
  • Kenneth M. Cramer
    • 1
  • Amber Riberdy
    • 1
  1. 1.Department of PsychologyUniversity of WindsorWindsorCanada

Personalised recommendations