The sample comprised 1,264 adults with ID who had been previously diagnosed by a multidisciplinary team, according to the Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition, Text Revision (American Psychiatric Association, 2000). The inclusion criteria of the participants included having a formal diagnosis of ID and being 18 years or older. The participants’ age ranged from 18 to 66 years (M=31.36; SD=10.55). Seven-hundred and sixty-nine participants (60.84%) were diagnosed as having mild ID and 495 (39.16%) as having moderate ID. All adults had the verbal skills to answer the self-report measure. Six-hundred were female (47.47%) and 664 were male (52.53%). Less than one-tenth lived in their own home (n=93; 7.36%) and about one-fifth lived in care facilities (n=247; 19.54%), and the majority lived in family homes (n=924; 73.10%). Their daytime activity was classified as vocational training (n=481; 38.05%), occupational activities (n=558; 44.15%), and activities developed with the family members (n=87; 6.88%). The remainder adults were employed, having paid jobs (n=138; 10.92%).

Respondents on EPR report-of-others (n=333) were members of the support staff (n=224) or family members (n=109). The support staff had known the person with ID for at least two years. The age of the key institutional workers ranged from 22 to 64 years (M=41.46; SD=9.87). Regarding gender, 179 were female (79.91%) and 45 were male (20.09%). Support staff included monitors (n=129; 57.59%), psychologists (n=56; 25.00%), psychomotor/occupational therapists (n=16; 7.14%), social workers (n=12; 5.36%), and special education teachers (n=11; 4.91%). The age of the family members ranged from 21 to 89 years (M=57.37; SD=13.64). Seventy-four were female (67.90%) and 35 were male (32.10%). More than half were the respective mothers (n=56; 51.38%) and the remainder were fathers (n=24; 22.02%), siblings (n=15; 13.76%), and other relatives (n=14; 12.84%).

The QOL assessment was made by the EPR (Simões & Santos, 2014; Simões et al., 2015b). The EPR structure is divided into two sections: the self-report part (a set of items to be answered by people with ID) and the report-of-others part (the same set of items, but to be answered by proxies who had known the person for at least two years). Each part has forty-eight items including six items per domain. All items from the EPR were measured on a three-point Likert scale (i.e., 3=always, 2=sometimes, 1=seldom or never).

Similar to the POS (Claes et al., 2010; Van Loon et al., 2009), the scale is focused on the assessment of QOL of individuals with ID aged 18 and over. The EPR is a multidimensional tool based on the eight core QOL domains (see Table 1) proposed by Schalock and Verdugo (2002). The exploratory factorial analysis has supported this multidimensional structure of the EPR (Simões et al., 2015b). These indicators were cross-cultural evaluated through content validity procedures in a Portuguese sample of adults with ID (Simões & Santos, 2014). According to Simões et al.’s (2015b) scores, the EPR revealed adequate test-retest (i.e., r's ranged from .67 to .92), internal consistency (i.e., α=.87 for self-report and α=.90 for report-of-others), and inter-rater reliability (i.e., r's ranged from .40 to .88).

The Ethic Committee of Centro Hospitalar de São João approved the research and the World Medical Association's (2008) ethical principles were guaranteed. Oral and written information about the study were given to all participants. Participants who wished to participate signed an informed consent form prior to becoming participants of the study and were free to leave at any time. Agencies providing services to adults with ID were asked by email to collaborate in the research. The sample was conveniently obtained from 45 Portuguese organizations that agreed to participate in the study. Individuals with ID, support staff, and family members were contacted by key stakeholders. The EPR was administered through an interview by qualified professionals who had been previously trained to administer the scale. Training sessions were performed to practitioners who agreed to contribute as interviewers. Specific training was given concerning the ethical principles and the importance of QOL in the ID field, as well as administration guidelines of the EPR. Before the participants answered the questions, the interviewers read standard instructions and collected socio-demographic data. The scale takes approximately one hour to one hour and a half to administer and was applied in Portuguese organizations or in family homes.

Data were analyzed using SPSS 21.0 and AMOS 21.0. A confirmatory factor analysis (CFA) was conducted to verify the factor structure of the EPR. As the EPR is a 3-point numeric scale, we used the optimal scaling procedure. This method allows “multivariate analysis when the data do not satisfy the classic quantitative measurement requirements but are qualitative” (Meulman, Kooij, & Heiser, 2004, p. 67). First, the outliers were analyzed through the squared Mahalanobis distance (Byrne, 2000; Marôco, 2014). Next, a CFA with the Unweighted Least Squares (ULS) estimation method was conducted to confirm the measurement model (Verdugo et al., 2015). The validity of QOL constructs was estimated through the composite reliability, the average variance extracted (AVE), and the discriminant validity (Marôco, 2014). The appropriateness of the model was assessed through a variety of absolute and incremental goodness-of-fit indexes (see Results section).

Based in the literature (Claes et al., 2010; Van Loon et al., 2009), a hypothesized model was proposed on each EPR section (i.e., self-report and report-of-others), including eight first-order domains and forty-eight indicators. The CFA results showed that the factor loadings of eight items failed to exceed the cut-off point of .40 (Matsunaga, 2010). Table 2 presents the items removed after the CFA using the self-report measure, as well as additional reasons for deleting the items ensuring the QOL construct. Likewise, the same eight items with lower scores with self-report section yielded a cut-off point lower than .40 with report-of-others section. As Table 2 also shows, the item-reduction seems to be related to the existence of unnecessary questions (Hooper, Coughlan, & Mullen, 2008; Matsunaga, 2010). Consequently, these eight items were eliminated in both QOL sections, because they did not provide theoretical consequences since the measurement model has all indicators presented in Table 1. The reduced item scale-data (i.e., with forty items) became the measurement model.

As Figure 1 illustrates, all factor loadings of self-report measure were statistically significant (p<.001), ranging from .42 (item 34) to .93 (item 2, 18, 24, and 27). Table 3 presents the composite reliability, the AVE, and the squared correlation test of discriminant validity for self-report version. The composite reliability, similarly to Cronbach's alpha coefficient, indicates the overall reliability of a collection of similar items (Arias, Verdugo, Navas, & Gómez, 2013). Scores of composite reliability above .70 were deemed reliable (Hair, Black, Babin, Anderson, & Tatham, 2005). All composite reliability coefficients were within Hair et al.’s (2005) standards for acceptable internal consistency ranging from .75 to .91. The AVE indicates how accurately the construct is measured (Arias et al., 2013). With exception to the domain of physical well-being, the AVE scores were equal or above the recommended cut-off point of .50 (Fornell & Larcker, 1981; Hair et al., 2005; Marôco, 2014). Thus, convergent validity of the self-report measure was generally accepted. Table 3 also shows the square correlations among constructs supporting discriminant validity between eight first-order constructs, given that all AVE scores exceeded the square correlations for each associated domain (Farrell, 2010; Fornell & Larcker, 1981; Hair et al., 2005).

Figure 1.

The confirmatory factor analysis of eight first-order factors by the self-report measure.

Note. PD=Personal development; SD=Self-determination; IR=Interpersonal relations; SI=Social inclusion; R=Rights; EW=Emotional well-being; PW=Physical well-being; MW=Material well-being; E=Error.

(2.16MB).

In addition, the fit indexes of the first-order QOL construct on self-report measure are presented in Table 4. Due to the limitations of Chi-square (χ2) when using large samples, we relied on the ratio of χ2 to its degrees of freedom (df; Hair et al., 2005), and a good fit was assumed with scores below 5.00 (Hooper et al., 2008). The Root Mean Square Error of Approximation (RMSEA) less than .06 was indicative of good fit (Byrne, 2000). Self-report results corroborate Hooper et al. (2008) and Kline's (2010) recommendations that the Normed Fit Index (NFI) and the Relative Fix Index (RFI) should be ≥ .95. The Root Mean Square Residual (RMSR) analysis, which should be less than .10 and low as possible (Hooper et al., 2008; Kline, 2010), was within the required scores. Finally, the Goodness of Fit Index (GFI) and Adjusted Goodness of Fit Index (AGFI) were higher than ≥ .95 (Hooper et al., 2008; Kline, 2010; Schreiber, Nora, Stage, Barlow, & King, 2006).

As Figure 2 shows, all factor loadings for report-of-others were higher than .40 and were statistically significant (p<.001). The scores ranged from item 35 (λ=.41) to item 2 (λ=.95). Table 5 illustrates that the composite reliability of the first-order constructs were within the Hair et al.’s (2005) standards. The scores ranged from .72 (physical well-being domain) to .92 (personal development domain). Generally, proxies’ measure showed higher coefficients of composite reliability at the domain level, compared to self-report measure. Likewise, the same domain with lower AVE with self-report measure yielded lower AVE with report-of-others measure. The AVE results on proxies ranged from .31 to .66, therefore convergent validity was generally accepted. Analogous to people with ID findings, report-of-others scores supported the discriminant validity between eight first-order constructs (see Table 5).

Figure 2.

The confirmatory factor analysis of eight first-order factors by the report-of-others measure.

Note. PD=Personal development; SD=Self-determination; IR=Interpersonal relations; SI=Social inclusion; R=Rights; EW=Emotional well-being; PW=Physical well-being; MW=Material well-being; E=Error.

(2.14MB).

The fit indexes for the first-order of the EPR report-of-others measure are present in Table 6. Similar to the results of self-report measure, data analysis also revealed suitable goodness-of-fit scores on proxies’ measure based on conventional guidelines (Byrne, 2000; Hair et al., 2005; Hooper et al., 2008; Kline, 2010; Schreiber et al., 2006).

Overall, the measurement model of the two EPR sections was within the required criteria and showed adequate psychometric properties. Consequently, further analyses were conducted with alternative models of higher order factor structure in order to examine the construct of QOL.

Additionally, it was also evaluated the best hierarchical structure (i.e., Salamanca or Schalock) on self-report measure. Table 4 shows that the RMSEA was the same (.06) for the second-order analyzed models, which was a suitable score (Byrne, 2000; Matsunaga, 2010; Schreiber et al., 2006). Nevertheless, incremental fit indexes (i.e., NFI and RFI) were higher in the Schalock model. According to several guidelines (Byrne, 2000; Hair et al., 2005; Hooper et al., 2008; Kline, 2010; Schreiber et al., 2006), the model showed acceptable goodness-of-fit results. Moreover, in the Salamanca structure we observed problems with the estimation of model’ parameters, and the covariance matrix was not positively defined. The Akaike Information Criterion (AIC) was also analyzed to compare the models, and smaller scores suggest the better fit of the data (Hooper et al., 2008; Kline, 2010; Schreiber et al., 2006). As Table 4 shows, the second-order model with the lowest AIC was also the Schalock hierarchical structure, and was regarded more suitable. However, multidimensional model showed higher absolute and incremental goodness-of-fit indexes, compared to Schalock model. Furthermore, AIC criterion suggests a better fit for the multidimensional model.

As Figure 3 illustrates, in Schalock model the factor loadings between the second-order constructs and the eight domains ranged from .40 to .79. Also, the factor loadings between the domains (i.e., first-order constructs) and the items (i.e., observed variables) ranged from .41 to .95. A high correlation between the second-order constructs was observed between social participation and well-being (Φ=.86). All scores were statistically significant (p<.001).

Figure 3.

The second-order model of Schalock by the self-report measure.

Note. I=Independence; SP=Social participation; WB=Well-being; PD=Personal development; SD=Self-determination; IR=Interpersonal relations; SI=Social inclusion; R=Rights; EW=Emotional well-being; PW=Physical well-being; MW=Material well-being; D=Disturbance (structural error); E=Error.

(1.62MB).

Concerning to comparison of the second-order models (i.e., Salamanca and Schalock), the pattern of results on report-of-others measure was very similar to data collected through self-report ratings (see Table 6). In regard to the scores based on report-of-others: (1) the goodness-of-fit scores were higher in the Schalock model, (2) the Salamanca model showed problems with the estimation of model’ parameters, and (3) the multidimensional model was more suitable than the Schalock solution.

In report-of-others measure (see Figure 4), the factor loadings between the second-order constructs and the domains ranged from .48 to .85. The factor loadings between the domains and its respective items ranged from item 35 (λ=.38) to item 2 (λ=.94). The highest correlation was observed between social participation and well-being (Φ=.85). All scores were statistically significant (p<.001).

Figure 4.

The second-order model of Schalock by the report-of-others measure.

Note. I=Independence; SP=Social participation; WB=Well-being; PD=Personal development; SD=Self-determination; IR=Interpersonal relations; SI=Social inclusion; R=Rights; EW=Emotional well-being; PW=Physical well-being; MW=Material well-being; D=Disturbance (structural error); E=Error.

(1.68MB).