The Mental Health Continuum–Short Form (MHC–SF) in the Argentinean Context: Confirmatory Factor Analysis and Measurement Invariance

The present research aimed at studying the psychometric properties of the Mental Health Continuum–Short Form (MHC–SF; Keyes, 2005) in a sample of 1,300 Argentinean adults (50% males; 50% females). Their mean age was 40.28 years old (SD = 13.59). The MHC–SF is a 14 item test that assesses three components (i.e., emotional, social, and psychological) of well-being. Convergent and divergent evidence of construct validity was assessed by conducting confirmatory factor analysis, cross-validation, factorial invariance, and correlations with external criteria. Internal consistency was studied using Cronbach’s alphas. Results indicated an adequate fit of a three-dimensional model. This structure was also confirmed, and was invariant throughout sex and age. The emotional well-being scores converged with life satisfaction and positive affect measures; the psychological well-being scale had a positive association with the presence of meaning in life; and the social well-being scores showed a positive and strong correlation with an external measure of well-being. Also, all scores were negatively associated with negative affect, search of meaning in life, and presence of depression symptoms. Internal consistency was .89 for the MHC–SF. Furthermore, the findings supported the two - continua model of mental health.

are those who are stuck in life, feel empty, and lack interest and engagement. Individuals who are neither flourishing nor languishing are considered to have moderate mental health.
Regarding the importance of measuring population well-being, epidemiological studies using reports of flourishing-such as the MHC-SF-have shown that it is associated with superior physical, psychological, and psychosocial functioning (Keyes, 2005). By contrast, it was found that languishing is related to poor emotional health, frequent limitations of daily living, and more missed days of work (Keyes, 2002).
Given that many countries have adapted the MHC-SF, there is a considerable volume of evidence to support its utility, validity, and reliability. For example, factor analyses has confirmed the MHC-SF's three-factor structure of emotional, psychological, and social well-being-found in earlier studies in the United States-in samples from South Korea (Lim, 2014), Italy (Petrillo, Capone, Caso, & Keyes, 2015), Poland (Karaś, Cieciuch, & Keyes, 2014), South Africa (Keyes, Wissing, Potgieter, Temane, Kruger, & van Rooy, 2008), and France (Salama-Younes & Ismaïl, 2011). Additionally, Joshanloo and colleagues investigated the factorial structure and invariance of the MHC-SF across cultural groups from three nations-the Netherlands, South Africa, and Iran -and the three dimensional structure was supported in all three groups (Joshanloo, Wissing, Khumalo, & Lamers, 2013). Furthermore, the MHC-SF has shown excellent internal consistency and criterion validity in adolescents and adults across different countries, including the United States, the Netherlands, Egypt, Italy, and South Africa, among other countries (Keyes, 2006;Keyes & Grzywacz, 2005;Keyes et al., 2008;Lamers, Westerhof, Bohlmeijer, ten Klooster, & Keyes, 2011;Petrillo et al., 2015;Salama-Younes & Ismaïl, 2011). Petrillo et al. (2015) found that all the subscales of the MHC, on the one hand, have positively correlated with the corresponding aspects of functioning and well-being (e.g., satisfaction with life, positive emotions), showing convergent validity. On the other hand, the MHC's subscales have negatively and moderately correlated with measures of mental illness (e.g., General Health Questionnaire 12 -GHQ-), indicating divergent validity. In addition, Lamers and colleagues reported test-retest reliability of the MHC-SF at four-time points over nine months, in a Dutch sample, using item response theory (Lamers et al., 2011). Similar results were found by Petrillo et al. (2015), who reported moderate test reliability.
Using exploratory and confirmatory factor analysis, most of the studies mentioned above have supported the two continua model of mental health (e.g., Lim, 2014;Karaś et al., 2014;Keyes, 2005;Keyes et al., 2008;Petrillo et al., 2015). The model with two distinct, but correlated (oblique), factors demonstrated a better set of fit coefficients than did other models (e.g., models with a single factor or two orthogonal factors). These findings confirm that mental health and mental illness are not the ends of one bipolar continuum, but two distinct and related dimensions.
Based on the aforementioned reasoning, the present research aimed to analyze the psychometric properties of the Mental Health Continuum-Short Form (MHC-SF; Keyes, 2005) in a sample of Argentinean adults. Both validity and reliability evidence were studied using a series of diverse statistical analyses, such as confirmatory factor analysis, cross-validation, factorial invariance, correlations with external criteria to assess convergent and divergent aspects, and internal consistency using Cronbach's alphas. Also, this study aimed to obtain evidence for the two-continua model. Lupano Perugini,de la Iglesia,Castro Solano et al. 95 Method Participants A convenience sample of 1,300 Argentinean adults from the general population was studied. They were proportionally distributed by sex: 650 males (50%) and 650 females (50%). The mean age was 40.28 years old (SD = 13.59). The majority of the participants (83.5%; n = 1,088) lived in Buenos Aires, 3.9% (n = 51) lived in other provinces of Argentina, and only 3.5% (n = 46) reported living abroad temporarily. The remaining percentage (8.8%, n = 115) did not provide information regarding their residence. Regarding their education, 80.2% (n = 1,043) achieved higher educational degrees, such as college or postgraduate studies. The remaining 19.8% (n = 257) received at least an elementary and/or middle school education. Concerning their occupation, 57% (n = 741) reported being employed; 17.1% (n = 223) were self-employed; 9.7% (n = 127) were managers or employers; 3.8% (n = 49) were housekeepers or non-paid workers; and the remaining 12.1% (n = 157) were unemployed or did not work. Only 0.3% of the sample (n = 3) did not give any data regarding their occupation. When asked about their socioeconomic status, the highest percentage (73.1%, n = 951) described itself as belonging to the middle class, 16.6% (n = 216) to high or middle-high class, 10.2% (n = 132) to middlelow or lower class, and 0.1% (n = 1) did not provide any data.

Materials Mental Health Continuum-Short Form (MHC-SF)
The original 14-item Mental Health Continuum-Short Form (MHC-SF; Keyes, 2005) was developed by Keyes in response to demands for a brief self-rating assessment tool that combined the three components of wellbeing: emotional, social, and psychological. The emotional well-being subscale -EWB-(three items) is defined in terms of positive affect/ satisfaction with life. Social well-being -SWB-is assessed with five items that represent each aspect of the model described by Keyes (1998): social contribution, social integration, social actualization, social acceptance, and social coherence. Finally, six items represent each of the dimensions of psychological well-being -PWB-as described in Ryff's model (Ryff, 1989): self-acceptance, environmental mastery, positive relations with others, personal growth, autonomy, and purpose in life. Participants were required to respond to items on a 6-point Likert-type scale based on the experiences they had had over the last month (never, once or twice, about once a week, 2 or 3 times a week, almost every day, or every day). This response option was selected to provide a clear standard of mental health assessment and a categorization similar to the method used by the DSM (American Psychiatric Association, 2002 to assess a Major Depressive Episode (see the Results section for psychometric properties).

Satisfaction with Life Scale (SWLS)
The Satisfaction with Life Scale (SWLS; Diener et al., 1985;Pavot & Diener, 1993) was developed to obtain a measure of an individual's general satisfaction with life. Evaluation of quality of life-according to selfperception-is measured on a cognitive-judgmental level. It comprises five items that are judged using a 7point Likert-type scale, ranging from 1 (strongly disagree) to 7 (strongly agree). Cronbach's alpha coefficient from the sample of this study was .87.

Positive and Negative Affect Scale (PANAS)
The Positive and Negative Affect Scale (PANAS; Watson, Clark, & Tellegen, 1988) was selected to measure positive (PA) and negative (NA) affect. PA reflects the extent to which a person feels enthusiastic and active, whereas NA refers to being alert, or to being sad, anxious, and depressed. Both dimensions are assessed by 14 descriptors each (a total of 28 items). Responses were given in a 5-point Likert-type scale ranging from 1 (low frequency) to 5 (high frequency). In this research, Cronbach's alpha for PA was .76 and for NA was .85.

Meaning in Life Questionnaire (MLQ)
The Meaning in Life Questionnaire (MLQ; Steger, Frazier, Oishi, & Kaler, 2006; Argentinian adaptation: Góngora & Castro Solano, 2011) is an instrument to measure the presence of meaning in life. It refers to the extent to which people comprehend, make sense of, or see significance in their lives, accompanied by the degree to which they perceive themselves as having a purpose or mission in life. The adapted scale has two dimensions, presence of and search for, depending on whether people have goals and a purpose in life, or whether they believe they have to search for them. It comprises 10 items that are responded to on a 7-point Likert-type scale ranging from 1 (absolutely false) to 7 (absolutely true). The analysis of the adapted version has demonstrated adequate psychometric properties, such as reliability and convergent and divergent validity.
In this study, Cronbach's alpha for Presence was .88, and for Search was .87.

Well-Being Index (WBI)
The Cronbach's alpha for the total score was .88.

Self-Report Depression Scale for Research in the General Population (CES-D)
The Self-Report Depression Scale for Research in the General Population (CES-D; Radloff, 1977) is a short self-report scale designed to measure depressive symptomatology in the general population. The items of the scale are symptoms associated with depression that have been used in previously validated and longer scales.
Participants are asked to rate on a 20-item scale how often, over the past week, they experienced symptoms associated with depression, such as restless sleep, poor appetite, and feeling lonely. Response options range from 0 to 3 for each item (0 = less than 1 day, 1 = 1-2 days, 2 = 3-4 days, 3 = 5-7 days). For the present study, Cronbach's alpha reliability was .87.

Symptom Checklist-90-Revised (SCL-90-R)
The Symptom Checklist-90-Revised (SCL-90-R; Derogatis, 1977; Argentinian adaptation: Sanchez & Ledesma, 2009) is a 90-item checklist used to measure the nine sets of psychological symptoms: Somatization, Obsessive-compulsive, Interpersonal Sensitivity, Depression, Anxiety, Hostility, Phobic Anxiety, Paranoid Ideation, and Psychoticism. The SCL-90-R also contains a Global Severity Index (GSI), which is used to estimate the "general psychiatric status" of a patient. The SCL-90-R uses a 5-point scale (1 = no problem to 5 = very serious) to measure the extent to which they have experienced the listed symptoms in the last 7 days.
Studies with the Argentinean adaptation of the checklist showed an adequate internal consistency for all of the scales (all alpha > .75) and also replicated the original factorial structure (e.g., Sanchez & Ledesma, 2009).
Cronbach's alpha coefficient from the sample of this study was .97.

Procedure
Regarding sampling method, it was a non-randomized sample. All participants were volunteers, and no incentives were provided for participation. Written informed consent was obtained before participation.
Questionnaires were administered by psychology students from a university located in the city of Buenos Aires as part of their professional practices in research program. A senior researcher supervised their work.
The method used for translation of the MHC-SF was forward translation. Two researchers were involved in the translation process. Both have a Ph.D. in Psychology and a good command of the English language. They translated the MHC-SF independently and without any discussion. As judged by the translators, both versions did not differ noticeably. Afterwards, adjustments were conducted to ensure understandability, psychological equivalence, and the accuracy of the translation from English to Spanish.
Data were analyzed with SPSS 17.0 and EQS 6.2.

Descriptive Statistics
First, univariate statistics were calculated for each item (

Confirmatory Factor Analysis
Confirmatory factor analysis (CFA) was run to ascertain the factor structure of the MHC-SF. Due to the categorical nature of the items (Likert-type scale), the estimation method used was robust Maximum Likelihood Index), and RMSEA (Root Mean Square Error of Approximation). Different models were tested: (i) a single factor model with a unique measure of mental health, (ii) a two factor model were one latent dimension represented hedonic aspects of well-being -EWB-and the other dimension the eudaimonic aspects of wellbeing -both SWB and PWB-), (iii) a three factor with independent measures for EWB, PWB, and SWB as proposed by Keyes (2002Keyes ( , 2005, and (iv) a second order model were the three first order dimensions proposed by Keyes load in a general factor of mental health. Table 2 shows that the three-factor structure is the bestfitting model and it replicates the proposed latent structure of subjective well-being (see also Figure 1).

Cross-Validation
Next, cross-validation analyses were conducted to test whether the Three factor model fit prevailed in samples differentiated by sex and age. The sample was first divided in two: females (n = 650) and males (n = 650). As Lupano Perugini, de la Iglesia, Castro Solano et al. 99 presented in Table 3, model fit indices showed excellent fit in both samples. Then, the main sample was divided into two age groups: the first sample consisted of individuals between 18 and 40 years old (n = 678), and the second group consisted of those whose ages ranged from 41 to 86 years (n = 622). Again, model fit was assessed using the robust method for ML. Indices of fit showed the appropriateness of the measurement model (Table 3).

Factorial Invariance
Additionally, hierarchical models with constraints were studied in order to analyse factorial invariance. This was done twice: first, to test invariance between the female and male samples, and second, to test invariance between younger and older adults (18 to 40 years old vs. 41 to 86 years old).
Initially, model fit was tested without constraints (MODEL 1) for both male and female samples. Both CFI and RMSEA values were adequate (see Table 4). A second model (MODEL 2) imposed constraints in all factor loadings. Again, the model's estimates fit the data at an acceptable level. The third and last model tested (MODEL 3) also imposed constraints in covariances. Data provided an adequate fit for this model. When testing for factorial invariances in age-compared groups, the same procedures were run. First, a model with no constraints was tested (MODEL 1), then factor loadings were imposed (MODEL 2), and finally, covariances were constrained (MODEL 3). CFI and RMSEA values were adequate in all models. Model comparison showed no statistically significant difference in the Satorra-Bentler scaled statistic between MODEL 1 and 2, nor between MODEL 1 and MODEL 3. No changes were found in the CFI, either (Table 5).  Table 6 summarizes the descriptive and correlational statistics obtained on the total MHC-SF scale and its three subscales. The highest means were obtained for emotional and psychological well-being, whereas the lowest value was obtained for social well-being. These results are consistent with those found in previous research conducted in other countries (e.g., Keyes et al., 2008;Lamers et al., 2011). Also, the EWB and PWB subscales have a stronger association with each other than they do with SWB. Cronbach's alpha coefficients were calculated to study internal consistency. The alpha coefficients for the MHC-SF scales were satisfactory. Reliability values above .70 are referred to as acceptable, and above .80 as high (Kline, 2000;Nunnally & Berstein, 1995). The total MHC-SF scale yielded a high internal consistency (alpha = .89). The PWB subscale had the next highest measure of internal consistency (alpha = .84), followed by the EWB subscale (alpha = .82), and finally, the SWB subscale (alpha = 0.78).

Evidence of Two-Continua Model
The two-continua model suggests that mental health and mental illness constitute two distinguishable but related factors and that they are not two sides of the same continuum. In order to obtain evidence for the twocontinua model, two factor analyses (exploratory and confirmatory) were performed using the MHC-SF and SCL-90-R subscales, which assess psychopathological symptoms.
Firstly, for the exploratory factor analysis a forced two factor solution with Varimax rotation was performed. It variance for this solution was 71% (see Table 8).
The Argentinean Adaptation of the MHC-SF 102 Next, three models were studied using confirmatory factor analysis. The two-continua model is the best-fitting data model to these data, suggesting that the correlated two-factor model was the most tenable model of the structure of mental health and illness. RMSEA was right on the expected limits (Kline, 2000) but lower than those obtained in the other models. The correlation between the latent factors of mental illness and mental health was -.50 which indicated, as expected, a moderate correlation of two aspects which are related but are not the same construct. These results confirm the two-continua model (see Table 9).

Discussion
The Mental Health Continuum, or MHC, represents a clinical approach to the continuous assessment and categorical diagnosis of states of positive mental health (Keyes, 1998(Keyes, , 2002Keyes et al., 2008). The short form of the MHC scale, consisting of 14 items, is one of the instruments to assess well-being most widely used internationally. For this reason, the main objective of this study was to validate the MHC-SF within an Argentinian population. The current findings confirm that the Argentinian MHC-SF is a valid and reliable instrument that can be used for assessing well-being within the Argentinean population.
In this study, robust evidence of construct validity was found for a three-factor structure of emotional, psychological, and social well-being. The same structure was found in earlier studies in the United States, and more recently, in samples from South Korea (Lim, 2014), Italy (Petrillo et al., 2015), Poland (Karaś et al., 2014), South Africa (Keyes et al., 2008), and France (Salama-Younes & Ismaïl, 2011), among others. Although Item 4 and 5 had lower factor loading in comparison with other items, they were retained to maintain the original number of items and desirable for each factor. Considering the sample size used, some authors argue that .32 can be acceptable because it would be equivalent to approximately 10% of the variance explained (Tabachnick & Fidell, 2001).
In this research, the model fits well not only in the main sample, but also in the cross-validation study, indicating that the fit prevailed in the female and in the male sample. Moreover, excellent fit was also found when the sample was split into a group of younger and older adults. Additionally, several tests of factorial invariance performed in the mentioned samples showed that model fit was appropriate even when constraints in loadings and covariances were imposed. Data supports the adequacy of the three-dimensional model of the MHC for Argentinean adults, regardless of sex or age.
Furthermore, the MHC-SF has shown excellent internal consistency and criterion validity because it positively correlated with measures of well-being and was negatively associated with measures of mental illness. The data support the hypotheses on convergent and divergent validity. These results are consistent with the two continua model, which has proposed that mental illness and mental health are distinct, but related, dimensions (Keyes, 2005). Also, evidence for this model was obtained using a measure of psychopathological symptoms (SCL-90-R).
In the studied sample, Items 6 and 8 showed mean values lower than expected. Although this results are not uncommon as similar results were obtained in samples from other countries (e.g. Joshanloo et al., 2013), their interpretation in light of local aspects is not out of place as these items are related to social factors. It is possible that this is related to the insatisfaction Argentinians have with some aspects related to society, especially regarding security. This aspect can also be observed in the differential association among presence Some of the limitations of the present study are related to the use of self-reported data and the utilization of convenience samples. Future studies should employ non-self-report assessments and more representative samples. Finally, more research should be done to (a) investigate the predictive validity of the MHC-SF in terms of health and illness outcomes and work productivity, among other important public health outcomes (Karaś et al., 2014), and (b) explore the levels of well-being in specific groups of people (e.g., teachers; Petrillo et al., 2015) in countries like Argentina.