The demographic characteristics of the sample and subsamples are shown in Table 1. The mean age of the participants in years was 72.43 (SD = 8.48). In general, this is a sample composed of women (χ² (1) = 104.568; p < 0.001), who self-identified as white (χ² (1) = 459.744; p < 0.001), with incomplete elementary education (χ² (1) = 242.701; p < 0.001) and dwelling in the community (χ² (1) = 179.041; p < 0.001).

Table 2 summarizes the fit indices of the models tested with CFAs. Based on the ECVI, we can state that the three-dimensional model is the most adequate. Furthermore, all the fit indices (absolute, incremental and parsimonious) of this model are adequate. Except for the value of χ², df and p, the one-dimensional model also presents adequate fit indices. To achieve these adjustments, no respecifications were necessary.

Table 3 presents the factor loadings of the items of the three-dimensional model of the PGCMS-Br. We can observe that Z is significant for all items, with Item 2 (λ = 0.598; 95% CI = 0.384 ≥ 0.812) of the Attitude toward Own Aging factor and Item 17 (λ = 1.408; 95% CI = 1.209 ≥ 1.608) of the Agitation factor presenting the lowest and highest factor loadings respectively. The factor loadings are also significant and they correlate positively and moderately: Attitude toward Own Aging – Agitation (r = 0.382; 95% CI = 0.287 ≥ 0.470; p < 0.001); Attitude toward Own Aging – Lonely Dissatisfaction (r = 0.519; 95% CI = 0.437 ≥ 0.593; p < 0.001); and Lonely Dissatisfaction – Agitation (r = 0.509; 95% CI = 0.426≥0.584; p < 0.001)).

Regarding the internal consistency of the PGCMS-Br estimated with α, a score of 0.808 (95% CI = 0.776 ≥ 0.836) was observed, and when excluding item 2 this value increased to 0.810. It was also found that α of the factors (Table 3) varies between 0.623 (Attitude toward Own Aging) and 0.684 (Lonely Dissatisfaction). Only in the case of the Lonely Dissatisfaction factor, the exclusion of one item (Item 14) would improve the internal consistency by two thousandths.

A GLB of 0.881 (95% CI = 0.870 ≥ 0.910) was obtained for the PGCMS-Br, and the exclusion of item 7 increased this score by one thousandth. In the case of factors (Table 3), the GLB varies between 0.686 for Attitude toward Own Aging and 0.727 in the case of Lonely Dissatisfaction. Excluding Item 14 would also improve, based on the GLB, the internal consistency of this factor (0.733).

Regarding the validity evidence based on the relationships with external variables (Table 4), Agitation (M = 10.26; SD = 1.57; 95% CI = 10.01 ≥ 10.51), Attitude toward Own Aging (M = 8.15; SD = 1.51; 95% CI = 7.92 ≥ 0.39), Lonely Dissatisfaction (M = 10.79; SD = 1.41; 95% CI = 10.57 ≥ 11.01) and the Total PGCMS-Br (M = 29.20; SD = 3.66; 95% CI = 29.78 ≥ 28.63) correlate significantly, moderately and negatively with symptoms of depression tracked with the GDS-15. When correlating quality of life in old age and morale, we found that 26 of the 28 correlations between the factors [Lonely Dissatisfaction (M = 10.32; SD = 1.64; 95% CI = 10.08 ≥ 10.56), Attitude toward Own Aging (M = 7.71; SD = 1.52; 95% CI = 7.49 ≥ 7.93) and Agitation (M = 10.08; SD = 1.60; 95% CI = 9.85 ≥ 10.32)] and the Total (M = 28.12; SD = 3.76; 95% CI = 27.57 ≥ 28.67) of the PGCMS-Br and the WHOQOL-Old scores are significant and positive, three of them being weak and the others moderate. No significant correlation was observed between Agitation and Autonomy and between Lonely Dissatisfaction and Death and Dying.

Despite the fact that the morale mediates health perception (Kukihara et al., 2017) and constitutes an essential aspect for age well (Näsman et al., 2020), there is a lack of instruments in the Brazilian-Portuguese language to assess this construct in older adults. Thus, a Brazilian version of the PGCMS (Lawton, 1975) was developed, which, we should reiterate, is one of the most widely used measures to assess morale in old age (Doseděl et al., 2020; Klusmann et al., 2020; Marquez et al., 2020), called PGCMS-Br. This scale was applied to samples of community-dwelling and institutionalized older adults and some of its psychometric properties were examined.

Regarding the validity evidence based on the internal structure, AFCs identified that, like the original version (Lawton, 1975), the PGCMS-Br is a three-dimensional measure (Lonely Dissatisfaction, Attitude toward Own Aging and Agitation) with a second-order factor (Morale). This corroborates the studies of McDowell (2006), as well as those of Ma et al. (2009), Niklasson et al. (2015), Paúl (1992), and Pinar and Oz (2011).

By correlating the PGCMS-Br scores with symptoms of depression tracked by the GDS-15 (Almeida & Almeida, 1999), evidence of discriminant validity was obtained for that scale, since its scores (total and factors) correlated significantly, moderately and negatively with the symptoms in question. These results corroborate those obtained by Almevall et al. (2021), Niklasson et al. (2017) and Pourtaghi et al. (2019) and signal, as highlighted by Kukihara et al. (2017), the contributions of morale to mental health.

Evidence of convergent validity was also obtained for the PGCMS-Br, more precisely with quality of life. However, although morale (total and factors) correlated significantly, positively and moderately with the total WHOQOL-Old (Fleck et al., 2006), the correlations between the factors of the two measures sometimes were weak or lacked significance although, in general, the investigations by Miki and Kawabata (2020), Pinar and Oz (2011) and especially by Quiroz et al. (2013) were corroborated.

Quiroz et al. (2013) also used the WHOQOL-Old and, as in the case of the study reported here (that had a correlation of r = 0.607), the correlation between the totals of the two measures was significant, positive and moderate (rs = 0.624). However, the scores of the PGCMS and WHOQOL-Old factors were not correlated. It should be noted that only the scores of Lonely Dissatisfaction and Agitation of the PGCMS-Br did not present a significant correlation with the quality of life domains, Death and Dying and Autonomy respectively. Thus, although these two morale components correlated positively with all the other WHOQOL-Old factors, their association with quality of life should be scrutinized in further research. The results indicate that morale and quality of life, although positively correlated, are distinct constructs, especially with regard to some domains. The magnitude of the correlations and the lack of correlation between some factors support the hypothesis stated above. Therefore, the PGCMS and its Brazilian version should not be considered measures of quality of life.

Although parsimoniously, we can state that evidence of convergent validity with quality of life was obtained for the PGCMS-Br. However, the evidence of discriminant validity with depression is more robust.

In addition to the validity evidence based on the internal, convergent and discriminant structure, reliability estimates were obtained for the PGCMS-Br, more precisely internal consistency. When considering “α”, it was observed that, in general, this scale presents satisfactory internal consistency (α > 0.65) (Vaske et al., 2017), converging with the outcome of Niklasson et al. (2015) and Pinaz and Oz (2011) investigations. However, both factors Attitude toward Own Aging (α = 0.623) and Agitation (α = 0.643) did not reach the cutoff point. This circumscription in the internal consistency of the Attitude toward Own Aging factor is in line with the results obtained by Ma et al. (2009) and Niklasson et al. (2015). We ought to consider, as a hypothesis that explains these unsatisfactory internal consistencies, the limitations of the estimator used, that is, α (Xiao & Hau, 2023).

When computing the GLB, satisfactory internal consistency (α > 0.65) was observed (Vaske et al., 2017) for both the PGCMS-Br and its factors. It is important to reaffirm that this estimator is the most appropriate option when dealing with nominal measures (McNeish, 2018). It is worth noting, again, that excluding an item would increase the internal consistency of the scale, but not significantly. The same would occur in the case of the Lonely Dissatisfaction factor. Thus, we can state that the PGCMS-Br has adequate overall internal consistency.

It should be noted that, in order to obtain these “good” psychometric properties for the PGCMS-Br, it was not necessary to modify the number of items on the scale thus maintaining the same 17 items from the original version, nor to substantially modify their wording. It should be noted that versions for other countries (Liang et al., 1987; McDowell, 2006; Paúl, 1992; Pinar & Oz, 2011) excluded or included items. Despite contributing to obtaining “good” psychometric properties, this type of procedure makes it difficult or even impossible to conduct cross-national studies.

Although it did not compromise the internal consistency and adjustment of the three-dimensional model, it was found during data collection that item 9 caused some confusion when answered. This probably happened because it has a negative formulation that should be answered negatively or affirmatively. Initially, the option was to keep it with a wording that was as similar as possible to that of the PGCMS (Lawton, 1975). Therefore, additional care should be exercised when the PGCMS-Br is self-completed; in addition items adequacy should be re-examined in future studies.

Although the psychometric properties of the PGCMS-Br are initially quite adequate, our investigation has limitations that should be addressed in other studies. External validity is the main one. Having a larger, randomized sample that is more representative of the elderly population in Brazil is essential, for example, to establish cut-off points. In this connection, Lawton (1975) already recommended the development of standards for different population groups.

Regarding internal validity, research should be designed to obtain, for example, evidence of criterion validity for the PGCMS-Br and test its factorial invariance between community-dwelling older adults and LTCF for older adults and between self-completion and interviews. Furthermore, it is suggested that the examination of psychometric properties with Classical Test Theory be combined with IRT (Item Response Theory), using, for example, Rasch Analysis to examine item difficulty.

Due to these and other limitations, it is suggested at this point, that the PGCMS-Br be used sparingly in professional practice. More research is still needed to make available to psychologists and other health professionals a measure of morale that, as indicated by Lawton (1975), be multidimensional, easy to apply and, consequently, can be applied to older adults who have very different profiles (e.g., those with cognitive decline). This is a fundamental prerequisite for measuring this generalized feeling of well-being in old age.