Professional Documents
Culture Documents
Benitez - Factorial Structural Analysis of The Spanish Version of WHOQOL
Benitez - Factorial Structural Analysis of The Spanish Version of WHOQOL
net/publication/293476178
CITATIONS READS
0 305
2 authors:
Some of the authors of this publication are also working on these related projects:
FONDECYT 1180315 Factores mediadores del efecto de la discriminación sobre la salud física, mental y el bienestar en inmigrantes colombianos en el norte y centro de
Chile View project
All content following this page was uploaded by Alfonso Urzua on 04 March 2016.
Alfonso Urzúa-Morales
123
2206 Qual Life Res (2014) 23:2205–2212
Based on the previous assessment instrument, the same Table 1 Distribution of the sample by country of origin (n = 1,972)
WHOQOL group designed a short version of the ques- Country Frequency (%)
tionnaire for use in limited-time situations. Using the data
collected by the 15 international centers for the WHOQOL- Costa Rica 285 (14.5)
100 [2], the items for the abbreviated version (WHOQOL- Peru 140 (7.1)
BREF) were selected based on three criteria: their capacity Mexico 198 (10)
to explain a significant portion of the variance within their Cuba 170 (8.6)
facet and domain; their relation to the general WHOQOL Paraguay 191 (9.7)
model; and their discriminant validity [4, 5]. Psychometric Argentina 155 (7.9)
studies provided sufficient evidence of the instrument’s Colombia 187 (9.5)
cross-cultural validity both in its first validations [4–6] and Spain 226 (11.5)
in its validations in other languages and countries [7–11], Chile 420 (21.3)
as well as in its applications to specific populations [10,
12–19].
However, some studies called into question the factor indicated by Asparouhov and Muthén [23], is that it allows
structure of the WHOQOL-BREF. Yao et al. [20] exam- the researcher to propose a simple structure for the mea-
ined the content validity of the WHOQOL-BREF using surement model, given that he is incorporating the previous
quantitative methods examining the respondents’ perspec- substantive knowledge in the form of certain restrictions in
tive, and they found that the structure of the WHOQOL- the measurement model, thus yielding a more parsimonious
BREF adapted for Taiwan was not completely consistent model. Nevertheless, unlike in the classic exploratory
with the original factor structure (12 out of the 24 items did factor analysis (EFA), the CFA approach of fixing many or
not yield acceptable content validity). Similarly, other all of the cross-loadings to zero (because every item only
studies failed to replicate the original structure of the loads onto one factor) can result in the researcher speci-
WHOQOL-BREF in the version adapted for Brazil [18], fying a model that is more parsimonious than is required by
and other authors also found that some items did not dis- the data and can lead to goodness-of-fit indexes that are not
criminate well between domains and that other items cor- completely adequate [23].
related more highly with a domain that differed from its As indicated above, the use of CFA is an effective way
original one [5]. to estimate measurement structures [23]. However, a
The few studies conducted with the Spanish version of modification of the CFA algorithm was recently proposed
the WHOQOL-BREF [17, 21, 22] found it to have to provide SEM with an exploratory quality that was
acceptable psychometric properties. However, due to the lacking in its original application. This exploratory struc-
scarcity of the literature on the matter, the main purpose of tural equation modeling (ESEM) proposed by Asparouhov
the present study was to analyze the factor structure of the and Muthén [23] is based on a new conceptualization of the
Spanish version of WHOQOL-BREF in a sample of indi- exploratory factor loading matrix, which uses of parts of
viduals from different Spanish-speaking countries (i.e., the model through EFA with rotated factor loading matri-
Costa Rica, Peru, Mexico, Cuba, Paraguay, Argentina, ces in addition to, or instead of, parts of the measurement
Colombia, Spain, and Chile) using sufficiently large sam- model through CFA. That is to say, unlike CFA, the
ple size, to establish a systematic analytical approach that measurement model through EFA does not require strict
can be used to clearly identify the instrument’s limitations restrictions to the cross-factor loadings. Furthermore,
within a Spanish-speaking population, including cross- similarly to the classical SEM approach, it grants access to
cultural aspects. the usual factor loading parameters that allow us to make
Studies analyzing factor structure are generally based on common interpretations in terms of the saturation of each
confirmatory factor analyses (CFA), a methodology that observable indicator in relation to the non-observable latent
utilizes algorithms from the measurement model of struc- factor.
tural equation modeling (SEM). CFA has generally proved
efficient enough to estimate construct validity such that the
majority of analysis involving the psychometric structure Method
of psychological assessments revolves around this type of
technique and its several derivatives. As widely known, Participants
statistical approximation technique SEM has been exten-
sively used for estimating factor structures and is prevalent The data were based on a sample of 1,972 undergraduates
in psychometric studies of complex psychological phe- (63 % were women) with an age range between 17 and
nomena. One of the advantages of the use of CFA, as 34 years (mean = 21.21 years; SD = 3.403), from nine
123
Qual Life Res (2014) 23:2205–2212 2207
Spanish-speaking countries. The distribution of the sample with the original structure, assuming the orthogonality
by country is presented in Table 1. between factors and invariance of the structure across
All the participants who did not fully complete the countries. Second, and as consequence of the poor fit of the
measure or did not understand the task were excluded from CFA, we analyzed the possibility of a new factor structure
the analysis. No statistically significant differences were through both EFA and ESEM, fixing the invariance of the
found between countries with respect to their age and sex factor matrix through the countries. All the analyses were
distributions or concerning their socio-demographic condi- conducted with the whole sample to test the construct
tions, which leads us to assume certain homogeneity across validity of the questionnaire.
the nine subsamples assessed. In recent study adopting ESEM, Marsh et al. [25]
demonstrated the power and flexibility of ESEM. Simi-
Instrument larly, Marsh, Muthén, Morin, Lüdtke, Asparouhov, Trau-
twein, and Nagengast [26] presented ESEM as a possible
We administered the Spanish version of the WHOQOL- alternative for the factor analysis of the Big Five person-
BREF, an abbreviated version of WHOQOL-100 devel- ality model.
oped simultaneously in 15 international centers [5, 22]. The To compare the models described above, the following
WHOQOL-BREF includes 26 items: one item measures goodness-of-fit indexes were analyzed: comparative fit
Overall QOL (item 1); one item measures satisfaction with index (CFI) and Tucker-Lewis index (TLI), obtained from a
health (item 2); and the remaining 24 items (items 3–26) free-distribution estimation due to the ordinal scale of the
assess four additional aspects of QOL, namely Physical observed variables, and in both cases, values greater than
Health (7 items), Psychological Health (6 items), Social .95 indicate a good fit of the model [27], although initially,
Relationships (3 items), and Environmental Health (8 CFI values greater than .90 were considered to indicate of
items). Study participants were asked to indicate how sat- good fit [28]; root mean square error of approximation
isfied they felt with each aspect during the previous (RMSEA), where values smaller than .05 indicate a good
2 weeks using a Likert’s scale ranging from 1 to 5. The fit of the model and values up to .08 represent a reasonable
scores for each domain were converted according to a error of approximation to the population [29]; and stan-
correction table due to the different number of items in dardized root mean square residual (SRMR), where values
each domain [24]. Examination of the internal validity of smaller than .08 indicate a good fit of the model [27].
the Spanish WHOQOL-BREF [22, 31] revealed the fol- The data were analyzed using the IBM SPSS statistical
lowing psychometric characteristics in relation to the package (version 21.0) and Mplus software (version 5.1).
internal consistency estimated through Cronbach’s alpha: The routines applied in the latter case followed the scheme
.69 for Physical QOL, .74 for Psychological QOL, .77 for proposed by Marsh, Muthén et al. [25] for estimating dif-
Environmental QOL, and .70 for Social QOL. The Cron- ferent invariance-of-models scenarios. The general idea is
bach’s alpha values of all the domains were acceptable, to generate a factor loading matrix (K) for all the samples
except that the value for Physical QOL was slightly low. (countries in our case) with all the parameters declared as
free and to test the statistical fit of each of the components
Procedure across the samples. The final result was an invariant factor
loading structure for all the samples, which assumed that
In every country represented in our sample, a group of test the majority of the original standardized variance came
administrators was trained. They conducted the scale from the observed variables.
administration, tabulation, and correction, and generated
each factor’s score, a total score, and the general scores for
the Overall QOL index and the satisfaction with health Results
index. These two index values were not the object of a
specific statistical analysis in this paper because the Confirmatory factor analysis
objective was to study the factor structure of WHOQOL-
BREF regardless of the total value of each factor or scale. The factor structure proposed, as has been mentioned,
included four factors: Physical QOL (Items 3, 4, 10, 15, 16,
Data analysis 17, 18); Psychological QOL (Items 5, 6, 7, 11, 19, 26);
Social QOL (Items 20, 21, 22); and Environmental QOL
We conducted a psychometric factor structural analysis of (Items 8, 9, 12, 13, 14, 23, 24, 25). This structure was
the WHOQOL-BREF through a series of 3 approximations. defined as orthogonal between factors. Therefore, it was
First, CFA was applied to the four factors (physical, psy- assumed that the matrix of correlations between factors
chological, environmental, and social QOL) in accordance was equal to the identity matrix. Additionally, we chose to
123
2208 Qual Life Res (2014) 23:2205–2212
Table 2 Cronbach’s alpha values for each factor and sample converge in their values and Spearman’s correlations are
(n = 1,972) preferable with known parametric distribution models. In
Countries Physical Psychological Social Environmental addition, we made the following assumptions in specifying
QOL QOL QOL QOL the CFA model: independence between measurement
errors across the 24 items; a standardized metric for the
Costa .67 .71 .74 .70
Rica four latent factors; and a parameter estimation technique
Peru .71 .77 .72 .74 based on free distributions (asymptotically free distribution
Mexico .78 .82 .76 .71 estimation) given that sample sizes prevented more exact-
Cuba .67 .54 .63 .69 ing estimations. The results of the CFA including good-
Paraguay .73 .72 .69 .76
ness-of-fit indices and model estimations are shown in
Argentina .82 .88 .79 .83
Table 3.
The previous results allowed us to clearly establish that
Colombia .65 .68 .66 .74
the model derived from the CFA yielded a relatively poor
Spain .78 .79 .72 .77
fit, and the invariance of the factor structure formulated in
Chile .76 .72 .76 .75
the theoretical model is hardly assumable. Indeed, neither
Total .68 .71 .77 .75
sample the fit indices (in general having values less than .95 and
ratios up to 5 with very small p values (p \ .001 in all
cases)) nor the residual values (in general values up to .07
define the latent factors in a standardized way (defining the and .08 for all the models except for the global sample)
variance of each factor to be equal to 1 (one) to guarantee a allowed us to accept the factor structure proposed. Thus, it
homogeneous metric structure). Preliminary data analysis seemed necessary to devise an alternative structure to the
indicated that all the samples yielded values indicating one proposed theoretically. Along the same lines, the
internal consistency, as observed in Table 2. analysis of standard factor loadings estimated in each
Likewise, we conducted an inspection of the observed factorial model also proved that the assumption of invari-
distributions of all the items in every sample and noticed ance in the structure is clearly compromised. Table 4
that it would be difficult to assume the condition of mul- shows the range of values for the factor saturations when
tinormality (84 % of the Kolmogorov–Smirnov tests for estimated under the assumption of invariance.
normality yielded significant values (p \ .001). The values in Table 4 showed some saturations that
In view of the previous data, we chose to estimate were sufficiently high enough to be considered adequate
Spearman’s correlation coefficients to try to account for according to the theoretical model proposed. However, in
having ordinal response options and the limitation in some cases, the values of those saturations were particu-
assuming multinormality. We used Spearman’s correla- larly low, indicating inadequate fits between the observable
tions instead of polychoric correlations due to the sample indicator and the latent factor. Despite the fact that their
size; with large sample sizes, the correlation estimations use may be disputed from a methodological standpoint
Table 3 CFA goodness-of-fit indices applied to the factor structure of WHOQOL (df = 299 for every partial model)
Models v2 Ratio v2/df p v2BL Model CFI TLI RMSEA SRMR 90 % CI
Costa Rica 2,208.893 7.38 \.001 15,284.755 .868 .841 .066 .139 .11–.15
Peru 1,666.109 5.57 \.001 df = 276 .903 .871 .060 .106 .09–.12
Mexico 1,185.092 3.86 \.001 Ratio 55.379 .933 .901 .052 .081 .07–.09
Cuba 2,142.55 7.17 \.001 .921 .899 .061 .071 .06–.08
Paraguay 1,123.77 3.76 \.001 .911 .877 .071 .082 .07–.09
Argentina 1,978.53 6.61 \.001 .933 .876 .059 .071 .06–.08
Colombia 2,435.12 8.14 \.001 .899 .899 .066 .074 .06–.08
Spain 1,943.55 6.49 \.001 .901 .901 .082 .110 .10–.12
Chile 2,145.12 7.17 \.001 .915 .855 .068 .093 .08–.11
Total sample 1,124.66 3.75 \.05 .943 .922 .051 .045 .04–.05
v2 chi-square test, v2BL Model chi-square test of null model, CFI comparative fit index, TLI Tucker-Lewis index, RMSEA root mean square error of
approximation, SRMR standardized root mean square residual, CI confidence interval of SRMR
123
Qual Life Res (2014) 23:2205–2212 2209
Table 4 kij factor loadings for each item and latent factor by country Table 5 Matrix of factor loadings estimated through EFA under the
(factor saturation for the whole sample) assumption of cross-country invariance
Items Physical Psychological Social Environmental Items Factor 1 Factor 2 Factor 3 Factor 4
QOL QOL QOL QOL
Item 3 -.442** -.030 -.103 -.144
Item 3 .12 to .14
Item 4 -.712** -.056 -.088 -.126
(.13)
Item 5 .827** .091 -.010 .003
Item 4 .23 to .32
(.29) Item 6 .496** .474** -.054 .082
Item 5 .33 to .49 Item 7 .114 .793** -.057 .025
(.49) Item 8 .177 .551** -.053 .003
Item 6 .38 to .55 Item 9 -.025 .487** .216 .055
(.54)
Item 10 .024 .208 .450** .117
Item 7 .35 to .49
(.42) Item 11 -.076 .089 .587** .176
Item 8 .11 to .27 (.26) Item 12 -.028 .006 .572** .013
Item 9 .09 to .33 (.31) Item 13 .018 .133 .025 -.622**
Item 10 .32 to .42 Item 14 .121 .106 .128 -.883**
(.40) Item 15 .137 .580** .188 .139
Item 11 .41 to .62 Item 16 .163 .557** .103 .200
(.59)
Item 17 .537** .086 .157 .028
Item 12 .16 to .41 (.40)
Item 18 .228 .388 .251 .175
Item 13 .21 to .32 (.31)
Item 19 .038 .227 .407** .192
Item 14 .09 to .23 (.22)
Item 20 .035 .176 .415** .213
Item 15 .19 to .28
(.26) Item 21 .432** .166 .156 .284
Item 16 .23 to .44 Item 22 .348 .305 -.016 .218
(.44) Item 23 .205 -.093 .658** .048
Item 17 .33 to .38 Item 24 .168 .047 .618** .108
(.36)
Item 25 .095 -.014 .509** .054
Item 18 .39 to .51
(.51) Item 26 .180 .209 .061 .409**
Item 19 .26 to .41 Variance explained 32.11 % 21.10 % 8.11 % 7.49 %
(.39) total = 68.81 %
Item 20 .12 to .27 ** Most statistical significant loadings (p \ .001)
(.27)
Item 21 .21 to .44
(.43)
Item 22 .18 to .31 Analyses of exploratory factor analysis
(.31)
Item 23 .10 to .41 (.41)
After demonstrating that the original theoretical model was
Item 24 .23 to .51 (.51)
not viable, we estimated a factor structure for four factors
Item 25 .19 to .54 (.52)
based on the original 24 items. In this case, we chose to
Item 26 .31 to .44
(.43) maintain the same assumptions already formulated. How-
ever, we added a varimax rotation to establish orthogo-
nality between factors, and given the previous results, we
given that these are confirmatory models, modification chose a factor extraction through principal components
indexes were estimated for the non-specified factor load- analysis. The estimation was conducted by using the mul-
ings and significant indexes were obtained for 14 of the 24 tigroup strategy defined for Mplus (version 5.1) and
defined factor loadings (p \ .01). Given these findings, we assuming a free distribution (see Marsh et al. [25]). In this
concluded that the originally proposed structure should be manner, we estimated the best solution for the matrix of
revised to establish a new factor structure that would be factor loadings that was able to withstand the assumption
more in keeping with the distribution observed. As con- of invariance across countries and showed good fit to the
sequence, we analyzed the data with exploratory statistical total sample, assuming it was the same sampling popula-
techniques to estimate a model that would fit a new tion. Table 5 shows the best factorial solution.
invariant factor structure across the countries under It is clear that these results should be contemplated and
investigation. understood within the framework of factor orthogonality
123
2210 Qual Life Res (2014) 23:2205–2212
Table 6 Goodness-of-fit indexes obtained through ESEM of the Table 7 Factor loadings derived from the ESEM estimation under
exploratory measurement models assuming invariance of the WHO- the assumption of cross-country invariance
QOL-BREF factor structure and cross-factor orthogonality
Items Factor 1 Factor 2 Factor 3 Factor 4
Models v2 p RMSEA 90 % CI CFI
Item 3 -.451* -.017 -.088 -.163
Costa Rica 336.78 .11 .042 .03–.05 .955 Item 4 -.707* -.062 -.096 -.119
Peru 392.44 .13 .044 .03–.06 .941 Item 5 .825* .093 -.008 .004
Mexico 412.77 .10 .051 .04–.07 .934 Item 6 .494* .477* -.050 .079
Cuba 389.11 .16 .048 .03–.05 .951 Item 7 .117 .791* -.055 .025
Paraguay 402.39 .11 .042 .02–.05 .943 Item 8 .182 .547* -.058 .008
Argentina 422.19 .15 .055 .04–.07 .961 Item 9 -.028 .491* .220 .050
Colombia 399.12 .18 .061 .04–.08 .988 Item 10 .026 .207 .448* .118
Spain 445.89 .12 .043 .03–.05 .971 Item 11 -.080 .094 .589* .173
Chile 399.11 .14 .059 .03–.06 .945 Item 12 -.028 .007 .572* .013
Total sample. 289.77 .21 .032 .02–.04 .967 Item 13 .020 .134 .022 -.621*
configurational
Item 14 .124 .106 .128 -.882*
Measurement 437.89 .28 .039 .02–.05 .977
Item 15 .138 .578* .186 .141
Equal intercept 523.12 .23 .033 .02–.05 .978
Item 16 .163 .558* .103 .200
v2 chi-square test, RMSEA root mean square error of approximation, Item 17 .535* .090 .158 .027
CI confidence interval of RMSEA, CFI comparative fit index
Item 18 .225 .391 .249 .176
Item 19 .036 .230 .403* .196
Item 20 .032 .178 .418* .410*
imposed in the rotation, which was used to maintain that Item 21 .434* .163 .152 .488*
part of the theoretical model proposed previously. Esti- Item 22 .348 .307 -.017 .218
mating the most exploratory solution would mean estab- Item 23 .209 -.097 .659* .049
lishing the same estimation while assuming non- Item 24 .166 .048 .621* .106
orthogonality. To this end, we assessed the same structure Item 25 .100 -.021 .506* .059
by means of the aforementioned ESEM modification, Item 26 .181 .207 .063 .409*
which was based on the specification of all the factor Variance explained 38.12 % 24.33 % 9.46 % 8.19 %
loadings as free parameters within the SEM logic and the Total = 80.1 %
estimation of the different parameters linked to the mea-
* Statistically significant saturations (p \ .01)
surement model proposed. The previous EFA data allowed
us to analyze the viability of factors orthogonality.
Table 8 Correlation matrix across factors derived from the ESEM
Analyses of exploratory factor analysis (ESEM) solution
Factors Factor 1 Factor 2 Factor 3 Factor 4
Just as in the previous cases, the estimations were based on
Spearman’s correlations between indicators and a free- Factor 1 1
distribution parameter estimation was used in Mplus (ver- Factor 2 .524** 1
sion 5.1), according to the aforementioned scheme by Factor 3 .297* .306** 1
Marsh et al. [25] and applied by Chahin, Cosi, Lorenzo- Factor 4 .320** .318** .330** 1
Seva, and Vigil-Colet [30]. The result of this estimation is * p \ .01, ** p \ .001
presented in Table 6.
These results indicated that the resulting factor structure
can be assumed to be invariant across countries and sup-
ports the assumption of one population with an insignifi- estimation of the correlation between factors of the ESEM
cant country effect. The table of invariant factor loadings solution yields a non-orthogonal matrix, as shown in
derived from the ESEM estimation is shown in Table 7. Table 8.
This solution is clearly different from the EFA solution Although these values do not indicate particularly large
and is much more complex than the theoretically proposed effect sizes, they should not be discarded. As a result, we
CFA. It is fitted to the empirical behavior of the observed concluded that the strictly orthogonal solution might be a
distribution. Further, the ESEM solution calls into question somewhat rigid assumption in a Spanish-speaking
the assumption of factor orthogonality given that the population.
123
Qual Life Res (2014) 23:2205–2212 2211
123
2212 Qual Life Res (2014) 23:2205–2212
10. Von Steinbüchel, N., Lischetzke, T., Gurny, M., & Eid, M. 20. Yao, G., Wu, C., & Yang, C. (2008). Examining the content
(2006). Assessing quality of life in older people: Psychometric validity of the WHOQOL-BREF from respondents’ perspective
properties of the WHOQOL-BREF. European Journal of Ageing, by quantitative methods. Social Indicators Research, 85(3),
3(2), 116–122. 483–498.
11. Yao, G., Chung, C., Yu, C., & Wang, J. (2002). Development and 21. Espinoza, I., Osorio, P., & Torrejón, M. J. (2011). Validación del
verification of validity and reliability of the WHOQOL-BREF cuestionario de calidad de vida (WHOQOL-BREF) en adultos
Taiwan version. Journal of Formosan Medical Association, mayores chilenos. Revista Médica de Chile, 139, 579–586.
101(5), 342–351. 22. Lucas-Carrasco, R. (1998). Versión Española del WHOQOL.
12. Barros da Silva, A., Fleck, M., Pechansky, F., de Boni, R., & Ergón, D.L.: División de Salud Mental. Organización Mundial de
Sukop, P. (2005). Psychometric properties of the World Health la Salud. Madrid. ISBN 84-89834-32-6.
Organization Quality of Life instrument (WHOQOL-BREF) in 23. Asparouhov, T., & Muthén, B. (2009). Exploratory structural
alcoholic males: A pilot study. Quality of Life Research, 14(2), equation modeling. Structural Equation Modeling, 16(3), 397–438.
473–478. 24. World Health Organization. (1996). WHOQOL-BREF: Intro-
13. Berlim, M., Pavanello, D., Caldieraro, M., & Fleck, M. (2005). duction, administration, scoring and generic version of the
Reliability and validity of the WHOQOL BREF in a sample of assessment. Field trial version. Geneva: Programme on mental
Brazilian outpatients with major depression. Quality of Life health.
Research, 14(2), 561–564. 25. Marsh, H. W., Muthén, B., Asparouhov, T., Lüdtke, O., Robi-
14. Fang, C., Hsiung, P., Yu, C., Chen, M., & Wan, J. (2002). Val- tzsch, A., Morin, A., et al. (2009). Exploratory structural equation
idation of the World Health Organization Quality of Life modeling, interpretating CFA and EFA: Application to student’s
instrument in patients with HIV infection. Quality of Life evaluations of university teaching. Structural Equation Modeling,
Research, 11(8), 753–762. 16(3), 439–476.
15. Izutsu, T., Tsutsumi, A., Islam, A., Matsuo, Y., Sayuri, H., Ku- 26. Marsh, H. W., Muthén, B., Morin, A., Lüdtke, O., Asparouhov,
rita, H., et al. (2005). Validity and reliability of the Bangla ver- T., Trautwein, U., et al. (2010). A new look at the big five factor
sion of WHOQOL-BREF on an adolescent population in structure through exploratory structural equation modeling. Psy-
Bangladesh. Quality of Life Research, 14(7), 1783–1789. chological Assessment, 22(3), 471–491.
16. Jang, Y., Hsieh, C., Wang, Y., & Wu, Y. (2004). A Validity study 27. Hu, L. T., & Bentler, P. M. (1999). Cutoff criteria for fit indexes
of the WHOQoL-BREF assessment in persons with traumatic in covariance structure analysis: Conventional criteria versus new
spinal cord injury. Archives of Physical Medicine Rehabilitation, alternatives. Structural Equation Modeling, 6(1), 1–55.
85(11), 1890–1895. 28. Bentler, P. M. (1990). Comparative fit indexes in structural
17. Lucas-Carrasco, R., Laidlaw, K., & Power, M. J. (2011). Suit- models. Psychological Bulletin, 107(2), 238–246.
ability of the WHOQOL-BREF and WHOQOL-OLD for Spanish 29. Browne, M. W., & Cudeck, R. (1993). Alternative ways of
older adults. Aging and Mental Health, 15(5), 595–604. assessing model fit. In K. A. Bollen & J. S. Long (Eds.), Testing
18. Moreno, A., Faerstein, E., Werneck, G., Lopes, C., & Chor, D. structural equation models (pp. 136–162). Beverly Hills, CA:
(2006). Psychometric properties of the World Health Organiza- Sage.
tion abbreviated instrument for quality of life assessment in the 30. Chahin, N., Cosi, S., Lorenzo-Seva, U., & Vigil-Colet, A. (2010).
Pró-Saúde Study. Cadernos de Saúde Pública, 22(12), Stability of the factor structure of Barrat’s Impulsivity Scales for
2585–2597. children across cultures: A comparison of Spain and Colombia.
19. Trompenaars, F., Masthoff, E., Van Heck, G., Hodiamont, P., & Psicothema, 22(4), 983–989.
De Vries, J. (2005). Content validity, construct validity, and 31. Lucas-Carrasco, R. (2012). The WHO quality of life (WHOQOL)
reliability of the WHOQoL-Bref in a population of Dutch adult questionnaire: Spanish development and validation studies.
psychiatric outpatients. Quality of Life Research, 14(1), 151–160. Quality of Life Research, 21, 161–165.
123