The Mental Health Continuum-Short Form (MHC-SF) is widely used to measure positive mental health. This study is the first to examine the psychometric properties of the MHC-SF in persons with schizophrenia living in the community.
Two hundred thirty-one individuals with schizophrenia living in the community in South Korea filled out the MHC-SF. Confirmatory factor analysis (CFA), bifactor CFA, exploratory structural equation modeling (ESEM), and bifactor ESEM were undertaken to examine the factor structure of the MHC-SF.
showed that a three-factor bifactor ESEM model yielded better fit than the other alternative models. Indices of internal consistency reliability were acceptable.
The results suggest that bifactor ESEM is an appropriate analysis for examining the factor structure of the MHC-SF in individuals with schizophrenia. The findings support the use of the overall MHC-SF scale rather than the use of the subscales in people with schizophrenia.
For a long time, mental health has been recognized as having no mental disorder or mental illness. However, the recent concept of mental health refers to a condition in which an individual does not have mental illness and has positive mental health at the same time [
Positive mental health includes three types of subjective well-being: emotional, social, and psychological well-being. Emotional well-being is a state of experiencing positive emotions and being satisfied with one’s life, psychological wellbeing is a self-evaluation of optimal functioning in one’s personal life, and social well-being is a self-evaluation of optimal functioning in one’s social life [
While the pursuit of positive mental health is regarded as an important human goal, there has been little interest in positive mental health among individuals with schizophrenia. Recently, Saperia et al. [
Recently, positive psychology interventions (PPIs), designed to improve the positive mental health of individuals with depression, have been applied to individuals with schizophrenia. PPIs for individuals with schizophrenia were developed to target negative symptoms, impaired cognition, and social impairment, on which traditional intervention methods did not show much effect for people with schizophrenia. Also, PPIs are intended to enhance positive mental health which is expected to play a role in the recovery of schizophrenia. PPIs studies in individuals with schizophrenia or psychosis [
Currently, many assessing tools are being developed to measure positive mental health. The Mental Health Continuum-Short Form (MHC-SF) is one of the scales for assessing an individual’s positive mental health. It covers most aspects of positive mental health (emotional well-being, social well-being, and psychological well-being) currently described in the literature. The MHC-SF is a proven reliable and valid tool for assessing positive mental health in normal adults [
The psychological properties of the MHC-SF were already examined not only in nonclinical samples but also in clinical samples [
To date, there are no adaptations of the MHC-SF in individuals with schizophrenia and its psychometric properties have not been examined in individuals with schizophrenia, even though positive mental health is important for individuals with schizophrenia, and treatments have begun to elevate the positive mental health of individuals with schizophrenia. The aim of the present study was to validate the Korean version of MHC-SF in a sample of individuals with schizophrenia. Internal consistency and construct validity of the MHCSF subscales and for the overall MHC-SF were examined.
Two hundred thirty-one individuals with schizophrenia in South Korea participated in this study. The data was obtained from the two independent study projects that were aimed to examine the determinants of positive mental health in individuals with schizophrenia [
Positive mental health was measured via the MHC-SF [
After receiving written informed consent from the participants, they were asked to complete the MHC-SF using a pencil or pen. A researcher was present beside the participants to answer their questions. It took about 20–30 minutes to complete the scales.
Data analyses were performed using SPSS 21 (IBM Corp., Armonk, NY, USA) and Mplus 7.0 (Muthén & Muthén, Los Angeles, CA, USA). To investigate factor structure of the MHCSF for individuals with schizophrenia, multivariate normality for the analyses was evaluated by Mardia’s tests. The investigation of the factor structure of the MHC-SF was conducted using the maximum likelihood robust (MLR) estimation method and the weighted least square mean and variance adjusted (WLSMV) estimation method, which are robust to non-normality of the data. Although the items of the MHC-SF have been viewed as continuous in previous studies, floor effects were possible for the individuals with schizophrenia, so sensitivity tests were conducted using the WLSMV. Twelve alternative models were compared using confirmatory factor analysis (CFA), bifactor confirmatory factor analysis (bifactor CFA), exploratory structural equation modeling (ESEM) and bifactor exploratory structural equation modeling (bifactor ESEM): a two-factor model comprising emotional and eudaimonic well-being with CFA (model 1); a two-factor model comprising emotional and eudaimonic well-being with bifactor CFA (model 2); a two-factor model comprising emotional and eudaimonic well-being with ESEM (model 3); a two-factor model comprising emotional and eudaimonic well-being with bifactor ESEM (model 4); a two-factor model comprising social and hedonic well-being with CFA (model 5); a two-factor model comprising social and hedonic well-being with bifactor CFA (model 6); a two-factor model comprising social and hedonic well-being with ESEM (model 7), a two-factor model comprising social and hedonic well-being with bifactor ESEM (model 8); a three-factor model with CFA (model 9); a three-factor model with bifactor CFA (model 10); a threefactor model with ESEM (model 11); and a three-factor model with bifactor ESEM (model 12). From model 1 to model 4, the social well-being and psychological well-being items were specified to load on eudaimonic well-being. From model 5 to model 8, the emotional well-being and psychological wellbeing items were specified to load on hedonic well-being. In the CFA models, each item of the MHC-SF had a non-zero loading on the target factor and a zero loading on the other factors. In the bifactor-CFA models, each item of the MHCSF had a loading on both the general factor and on the specific factor, no cross-loadings were allowed, and the factors were orthogonal to one another. In the ESEM models, cross-loadings were estimated through oblique target rotation. In the bibifactor-ESEM models, each item of the MHC-SF was defined by the general factor and by the specific factor and crossloadings were estimated through orthogonal bifactor-target rotation.
To examine the goodness of fit of the alternative models, the comparative fit index (CFI), the Tucker–Lewis Index (TLI), and root mean square error of approximation (RMSEA), and standardized root mean square residual (SRMR) were assessed. CFI and TLI values above 0.90 and RMSEA and SRMR values less than 0.08 indicate acceptable model fit [
In addition to fit indices, the mean and range of the target loadings from bifactor ESEM were examined. Mean target loadings should be equal to or above 0.50 and the lowest loading on a factor should be equal to or above 0.30 in a welldefined factor [
For the best fitting model, McDonald’s omega hierarchical (ωh) coefficients were calculated to evaluate the internal consistency reliability of the MHC-SF. The acceptable cut-off score of the McDonald’s omega hierarchical coefficients was 0.50 [
Because Mardia’s multivariate normality test showed that both multivariate skewness and kurtosis were significant (Mardia skewness=41.39, p<0.001; and kurtosis=326.60, p< 0.001), suggesting that the data failed to maintain normal distribution, the MLR estimator was employed for factor analysis. A sensitivity analysis was also conducted with the WLSMV estimator to compare with MLR, and no significant difference in results emerged.
As can be seen in
The improvement in model fit was examined using AIC and SSABIC. The two-factor bifactor CFA model (model 2) had an improved fit over the two-factor CFA solution (model 1) (ΔAIC=-39.290, ΔSSABIC=-35.469). Regarding the twofactor ESEM solutions of the MHC-SF, the two-factor bifactor ESEM model (model 4) had a better fit than the two-factor ESEM model (model 3) (ΔAIC=-38.534, ΔSSABIC=-35.258). Also, the fit of the two-factor ESEM solution (model 3) was better than the two-factor CFA model (model 1) (ΔAIC=-4.430, ΔSSABIC=-1.155), though the difference was modest. Concerning the two-factor bifactor solutions of the MHC-SF, the fit of the two-factor bifactor ESEM (model 4) was superior to the two-factor bifactor CFA model (model 2) (ΔAIC=-3.674, ΔSSABIC=-0.944), though the difference was negligible.
Regarding the two-factor ESEM solutions of the MHC-SF, the two-factor bifactor ESEM solution (model 8) was better than the two-factor ESEM solution (model 7) (ΔAIC=-38.534, ΔSSABIC=-35.258). Concerning the two-factor bifactor solutions of the MHC-SF, the two-factor bifactor ESEM (model 8) had better fit than the two-factor bifactor CFA model (model 6) (ΔAIC=-5.966, ΔSSABIC=-8.696).
In addition, the three-factor bifactor ESEM model (model 12) had a better fit than the three-factor ESEM model (model 11) (ΔAIC=-22.525, ΔSSABIC=-19.522). Also, the fit of the three-factor ESEM solution (model 11) was better than the three-factor CFA model (model 9) (ΔAIC=-9.500, ΔSSABIC=-3.494), though the difference was small.
Regarding the CFA solutions of the MHC-SF, the three-factor CFA model (model 9) had a better fit than the two-factor CFA models (model 1 and 5) (ΔAIC=-33.464 and -193.507, ΔSSABIC=-32.919 and -192.961). Also, concerning the ESEM solutions of the MHC-SF, the three-factor ESEM model (model 11) showed a better fit than the two-factor ESEM models (model 3 and 7) (ΔAIC=-38.534, ΔSSABIC=-35.258). In addition, as to the bifactor ESEM solutions of the MHC-SF, the three-factor bifactor ESEM model (model 12) had a better fit than the two-factor bifactor ESEM models (model 4 and 8) (ΔAIC=-22.525, ΔSSABIC=-19.522).
These findings indicated that the bifactor models gave a more precise picture of the structure than the first-order models and three-factor models offered a more accurate representation of the data than the two-factor models. Of the eleven alternative models, the fit of the three-factor bifactor ESEM model was better than the other models.
As noted in
The current study is the first to examine the psychometric properties of the MHC-SF in individuals with schizophrenia using different analytic techniques. The results of the present study showed that the three-dimensional bifactor ESEM model yielded better fit than the other alternative models, indicating that the MHC-SF scores are better represented after the general factor is accounted for, and when cross-loadings are allowed to be estimated. Also, the results indicate that MHCSF scores show a bifactor structure with a strong general factor. For example, two-factor bifactor CFA fitted better than two-factor CFA, two-factor bifactor ESEM model showed a better fit than two-factor ESEM, and three-factor bifactor ESEM model had a better fit than three-factor ESEM model. These findings are consistent with the previous studies suggesting that the three CFA factors were highly correlated, and that the bifactor structure of positive mental health with a strong general factor were underlying the MHC-SF [
In the three-factor bifactor ESEM model, five items (items 4, 5, 9, 10, and 11) from specific factors had low loading values. Also, the general factor had sufficient ωh scores, but the specific factor ωh scores were insufficient in the three-factor bifactor ESEM model. The low loading values of specific factors and the coefficient omega specific suggest that MHC-SF is not sufficient to measure the three specific factors of positive mental health when controlling for the variance of the general factor. The results mean that the items of the MHCSF measure the general factor rather than the specific factors. Therefore, caution should be taken when interpreting the specific factor scores.
Regarding the fact that specific factors had low loading values and general factors showed strong loading values in studies with psychiatric patients, the previous researchers explained that it was because psychiatric patients could not distinguish the types of well-being due to their distress [
The study results showing a strong general factor in the MHC-SF provide a rationale for using the MHC-SF total score in a study on positive mental health of the individuals with schizophrenia. In addition, the findings of the current study raise the possibility that statistical problems may occur when the three factor scores of the MHC-SF are input into the analysis together [
There are some limitations to this study. First, there was no information on cognitive function, types of antipsychotic drugs, and severity of symptoms in the present study. Therefore, further research is needed to determine whether the same results could be obtained even after controlling the severity of symptoms, types of antipsychotics, and cognitive function. Second, the test-retest reliability was not examined in the current study. Because cognitive dysfunction, which is a core feature of schizophrenia, could be related with the test-retest reliability coefficients, future studies are also needed to examine testretest reliability.
In conclusion, overall, the current study shows that there is limited support for the use of the MHC-SF subscales in people with schizophrenia, and the overall scale score appears to be most appropriate. Also, additional research is warranted to examine the test-retest reliability of the MHC-SF in individuals with schizophrenia and to examine the relationship between the MHC-SF and the symptoms severity of schizophrenia, cognitive symptoms, and types of antipsychotic medications.
The datasets generated or analyzed during the study are available from the corresponding author on reasonable request.
The author has no potential conflicts of interest to disclose.
This research was supported by Daegu University Research Grant, 2020.
The author would like to thank Na E. and Seo M. for help in data collection and the anonymous reviewers and editor for their valuable comments.
Demographic characteristics of participants (N=231)
Value | |
---|---|
Sex | |
Female | 94 (40.7) |
Male | 133 (57.6) |
Unknown | 4 (1.7) |
Age (yr) | 41.17±10.17 (20–67) |
Illness Duration (yr) | 14.73±8.61 (0–43) |
Education (yr) | |
≤6 | 9 (3.9) |
>6 and ≤9 | 23 (10.0) |
>9 and ≤12 | 112 (48.5) |
>12 and ≤16 | 86 (37.2) |
Unknown | 1 (0.4) |
Employment | |
Employed | 126 (54.5) |
Unemployed | 101 (43.7) |
Unknown | 4 (1.7) |
Values are presented as mean±standard deviation (range) or N (%)
Goodness-of-fit from factor analysis with robust maximum likelihood estimation
Model | χ2 | df | CFI | TLI | SRMR | RMSEA | 90% CI | AIC | SSABIC |
---|---|---|---|---|---|---|---|---|---|
Model 1 | 140.848 | 76 | 0.955 | 0.946 | 0.039 | 0.061 | 0.045–0.076 | 9,530.021 | 9,541.760 |
Model 2 | 118.435 | 62 | 0.960 | 0.942 | 0.029 | 0.063 | 0.045–0.080 | 9,490.731 | 9,506.291 |
Model 3 | 125.026 | 64 | 0.957 | 0.939 | 0.033 | 0.064 | 0.047–0.081 | 9,525.591 | 9,540.605 |
Model 4 | 97.348 | 52 | 0.968 | 0.944 | 0.024 | 0.061 | 0.042–0.080 | 9,487.057 | 9,505.347 |
Model 5 | 245.839 | 76 | 0.881 | 0.858 | 0.053 | 0.098 | 0.085–0.112 | 9,690.064 | 9,701.802 |
Model 6 | 99.937 | 62 | 0.973 | 0.961 | 0.028 | 0.051 | 0.032–0.070 | 9,481.091 | 9,496.651 |
Model 7 | 125.026 | 64 | 0.957 | 0.939 | 0.033 | 0.064 | 0.047–0.081 | 9,525.591 | 9,540.605 |
Model 8 | 97.348 | 52 | 0.968 | 0.944 | 0.024 | 0.061 | 0.042–0.080 | 9,487.057 | 9,505.347 |
Model 9 | 117.196 | 74 | 0.970 | 0.963 | 0.035 | 0.050 | 0.032–0.067 | 9,496.557 | 9,508.841 |
Model 10 |
87.025 | 60 | 0.981 | 0.971 | 0.029 | 0.044 | 0.021–0.064 | 9,474.543 | 9,490.649 |
Model 11 | 97.348 | 52 | 0.968 | 0.944 | 0.024 | 0.061 | 0.042–0.080 | 9,487.057 | 9,505.347 |
Model 12 | 64.984 | 41 | 0.983 | 0.963 | 0.017 | 0.050 | 0.025–0.073 | 9,464.532 | 9,485.825 |
the latent variable covariance matrix (psi) is not positive definite.
χ2, chi-square goodness of fit test; df, degree of freedom; CFI, comparative fit index; TLI, Tucker–Lewis Index; SRMR, standardized root mean square residual; RMSEA, root mean square error of approximation; CI, confidence interval; AIC, Akaike information criteria; SABIC, sample size-adjusted Bayesian information criteria
Parameter estimates for the three-factor bifactor ESEM solution of the MHC-SF
Item | General factor | Factor I | Factor II | Factor III |
---|---|---|---|---|
1 | 0.659 |
0.656 |
-0.001 | -0.065 |
2 | 0.692 |
0.607 |
0.024 | 0.066 |
3 | 0.638 |
0.442 |
-0.043 | 0.060 |
4 | 0.714 |
-0.102 | 0.193 |
0.039 |
5 | 0.825 |
-0.050 | 0.042 |
-0.051 |
6 | 0.695 |
0.032 | 0.326 |
0.071 |
7 | 0.675 |
0.071 | 0.335 |
-0.160 |
8 | 0.659 |
-0.032 | 0.613 |
0.049 |
9 | 0.819 |
0.015 | 0.020 | -0.072 |
10 | 0.775 |
0.077 | -0.005 | 0.231 |
11 | 0.788 |
-0.053 | -0.073 | 0.176 |
12 | 0.788 |
0.015 | -0.023 | 0.366 |
13 | 0.755 |
-0.087 | 0.043 | 0.388 |
14 | 0.668 |
0.134 | 0.062 | 0.328 |
ωh | 0.895 | |||
ωhs | 0.385 | 0.135 | 0.095 | |
ECV | 0.779 | 0.104 | 0.067 | 0.051 |
target loadings are shown.
ESEM, exploratory structural equation modeling; MHC-SF, Mental Health Continuum-Short Form; Factor I, emotional well-being; Factor II, social well-being; Factor III, psychological well-being; ωh, omega hierarchical; ωhs, omega hierarchical subscales; ECV, explained common variance