Phase III: assessment of psychometric properties and the validity of the final version of the Validation study
3.1. Sample characteristics
The demographic data and clinical characteristics are presented in Table 1. There was a proportionate number of females and males in our sample, 53.4% and 46.6%, respectively. The mean scores of the BrP-MQOL-R for males were 5.69 (1.63) and for females 5.69 (2.20) (t= 2.75, P=0.007], respectively. The mean score on the BrP-MQOL-R for the total sample was 6.09 (SD=2.0). The median of all items was 6.17 [interquartile (IQR25-75) 4.67; 7.60].
![](https://myfiles.space/user_files/58653_1b1c6aeb34a62c68/58653_custom_files/img1604687437.jpg)
3.2. Psychometric properties of the MQOL-R- BrP
3.2.1. Internal consistency
The BrP-MQOL-R final 14-item had a satisfactory internal consistency (α=0.85). The mean (SD) for all items of the scale was 6.09 (2).
![](https://myfiles.space/user_files/58653_1b1c6aeb34a62c68/58653_custom_files/img1604687454.jpg)
The MQOL-R scale and subscales and the total result were scored by averaging across items. We checked whether the findings in subscales differ from one another; we conducted a one-way repeated-measures ANOVA. The data comply with the variance sphericity (Muychaly’s test: W=0.89, P=0.008). This result indicates that the results in the MQOL-R sub-scales differ from one another. The multiple comparison test by Bonferroni revealed that QOL in the subscale social [mean (standard deviation)] was highest in our sample [8.14, (1.87); P< 0.001 for all comparisons], followed by existential [6.36 (2.10) P< 0.001 for all comparisons], psychological [5.21 (2.60)], and physical [4.88 (2.01)]. Other comparisons returned no significant difference.
![](https://myfiles.space/user_files/58653_1b1c6aeb34a62c68/58653_custom_files/img1604687475.jpg)
3.2.2. Construct validity: questionnaire item selection, structural validity and cross-cultural- validity
Confirmatory factor analysis of the MQOL-R
We tested the internal structure of the MQOL-R using CFA, using the generalized least squares method. CFA revealed that all items were related to four specified factors, verifying the item's relationships and latent factors. Figure 2 shows the diagram and factor loading generated and presented in table 2, the fit indices for this model. The analysis elicited adequate model goodness of fit (table 2). The χ2 test (CMIN=117.38; df=73; p=0.001) suggests insufficient fit, although this statistical tool is too restrictive and often points to rejecting a model with high samples involved. The chi-square/degree of freedom (CMIN/df= 1.608) reached a satisfactory value under 5. Following the strategy of presenting fit indices suggested by Hu and Bentler (19) if the root means the square error of approximation (RMSEA= 0.065; confidence interval 0.042- 0.086) is 0.06 or below, and the standardized root-mean-square residual (SRMR) is 0.08 or below, thus, the model fitting is good. Comparative Fit Index (CFI = 0.934; RMSEA=0.065; 95% CI range 0.042, 0.086). The revised model has the following fit-indices: PCLOSE=0.131, Tucker-Lewis Index (TLI) rho2=0.918, incremental fit index (IFI) delta 2=0.936. A second-order factor model was specified (Figure 2) to support the derivation of an MQOL-R total score.
3.2.3. Convergence validity
The correlation between the BrP-MQOL-R total scale and subscale scores is displayed in Table 4. Convergence-related validity is also supported by significantly positively correlated with higher levels in the BrP-MQOL-R total scores and their subscales with both the KPS score and the SIS related to the quality of life. In contrast, the BrP-MQOL-R was conversely correlated with the pain scores in the NPS (0-10). Most of the time, patients with higher pain scores in the last 24h and after use pain medication showed a lower score in the BrP-MQOL-R, or vice-versa.
The pain scores on NPS (0-10) in two conditions, the pain level on most of the time in the last 24 h and relive of pain score when you take pain medication. The mean (SD) on the KPS was 59.18 (21.38). The mean score (SD) in the SIS measuring overall quality of life was 6.5 (2.54). The mean (SD) on the question of their pain level on most of the time in the last 24 h was 4.50 (3.72), and after taking pain medication was 2.06 (2.78).
![](https://myfiles.space/user_files/58653_1b1c6aeb34a62c68/58653_custom_files/img1604687504.jpg)
3.2.4. Responsiveness and criterion-group validity
The responsiveness of the BrP-MQOL-R can be seen by the mean (standard deviation) of the total score. Patients with KPS equal to or lower than 30% could be discriminated from those with scores on KPS higher than 30%; the score on the BrP-MQOL-R was 4.83 (1.77) vs. 6.36 (1.95) (P=0.00), respectively. Also, the scores of the QOL scale and subscales tend to be higher in subjects with the best functional status. That is means that this tool has properties to capture differences between patients in palliative care with the worst performance of those who have better functional status. We assessed the criterion validity by the screening accuracy to discriminate patients with KPS equal to or lower than 30% (n=25) those with scores on KPS higher than 30% (n=121) by non-parametric receiver operating characteristics (ROC) analyses an area under the curve (AUC) = 0.71, sensitivity=97% and specificity=92%).
A regression analysis was used to assess if sex, hospitalization, formal education, and age could influence the score in the BrP-MQOL-R. The variables retained in the model were sex and the hospitalization at the time of assessment, the beta-coefficient was -0.86 (95% CI; -1.49 to -0.23; P = 0.00) and 0.78 (95% CI; 0.15 to 1.42; P = 0.01), respectively. That is, females and, if they were at home at the time of the assessment, showed higher scores.
Separate regression analyses were performed to determine the global BrP-MQOL-R score and a combination of the BrP-MQOL-R subscales to predict the SIS. These models were adjusted by hospitalization at the time of assessment adjusted and sex. The total score predicted similar variance in the SIS (R2 adjusted =0.36; β=0.49, t = 4.50, p < 0.001) than those found in the MQOL-R subscales (R2adjusted = 0.36). A combination of two subscales was significant in predicting the SIS: Physical (β = 0.25, t = 2.43, p < 0.01) and Existential (β =0.21, t=2.29, p=0.02).