Method

Results

This study involved 518 participants who lived in Brazil, in 14 of the 27 federative units. Table 1 shows the sample characterization.

Table 2 presents the scores and the results of the EFA of the six scales. The factorability of the correlation matrix between the items of the scales was confirmed by KMO and χ2, with values considered adequate for the six instruments. The Hull retention method indicated that each scale represents a unidimensional measure. The total variance explained was above 42% on all scales. The plausibility of unidimensionality was also confirmed by the UniCo, ECV and MIREAL indices. Higher than .915, the H-Latent index thus shows that the set of items on each scale represents a well-defined latent variable and is likely to remain stable between studies. Values greater than .90 obtained for NNFI, CFI, GFI and AGFI suggest, in isolation, the adequacy of the scales.

All items on the collective efficacy and proxy-efficacy scales saturated in the extracted factor and presented factorial loads between .445 and .960. In the AE-C19, only one item saturated below the borderline value of .30 (^{Field, 2013}) but, even so, was very close to it (.296). Additionally, the communality was above the minimum value of .20 (^{Child, 2006}) for almost all the items of the scales, except for the scales of self-efficacy, which showed lower values in two (.087 and .116), and family efficacy, which showed a lower value in one (.198) of the items. Analyzing the internal consistency considering Cronbach's alpha, the scales presented the following indices: .819 (AE-C19); .921 (EFam-C19); .968 (ECom-C19); .927 (ED-C19-Municipal); .957 (ED-C19-Estadual); and .962 (ED-C19-Federal).

Table 3 shows the corrected item-total correlation and alpha values if items were deleted. Two items (namely, items 1 and 2 of the AE-C19) did not reach the minimum acceptable value in the item-total correlation, i. e., greater than .30 (^{Kline, 2000}) and, if excluded, would raise the values of Cronbach's alpha, but not significantly. The removal of other items would not contribute to substantially increase the alpha in any of the measures.

Friedman's test revealed that the mean scores (Figure 1) of the six instruments differed statistically, χ² (518; 5) = 1345,544; p < .001. Multiple comparisons formed five homogeneous subgroups (p < .05) in decreasing order of efficacy level: AE-C19; EFam-C19; ED-C19-Municipal; ED-C19-Estadual and ECom-C19 (p = .281); and ED-C19-Federal (Figure 1).

Figure 1 Confidence interval (95%) of the scores of the efficacy to cope with COVID-19 scales. 

Among the 15 possible correlations between the scores of the efficacy scales, 12 were significantly positive, one strong, three moderate, eight weak and three null (Table 4). ECom-C19 and ED-C19-Municipal were the only measures that correlated with all the others. The three proxy-efficacy scales showed moderate or strong correlations between each other.

Discussion

Individually and collectively, therefore, health behaviors can be understood and, in part, predicted by individuals’ beliefs of efficacy (^{Bandura, 2004}). As a health behavior, mitigating the COVID-19 pandemic also evidently requires individual, collective (family, community, etc.) and proxy (governments, etc.) agency (^{Airhihenbuwa et al., 2020}; ^{Anderson et al., 2020}; ^{Prime et al., 2020}). Therefore, building measures of beliefs of efficacy to cope with this disease that contemplate these different levels of agency, as well as analyze its psychometric properties have evident and high scientific and social relevance.

When analyzing the psychometric properties of the efficacy for coping with COVID-19 scales, all proved to have a unidimensional structure, showing that each of the six beliefs about COVID-19 coping efficacy investigated constitutes a single construct. This converges with ^{Bandura (2006}), who proposes that measures of efficacy should have a specific object.

However, unsatisfactory results were observed for the one-dimensional model in some of the indexes. For example, the RMSEA was above that recommended (^{Damásio & Dutra, 2017}) in the three scales of proxy efficacy. Commonality, however, presented values below the minimum indicated (^{Child, 2006}) for some items of the scales of self-efficacy and family efficacy, which indicates that these items may present problems. Despite this, the explained variance, the NNFI, CFI, GFI, AGFI and H-Latent adjustment indexes suggest the adequacy of the model.

Based on these results and additionally considering the theoretical contribution that supported the items’ generation and the propositions of the TSC, we would suggest that the six instruments be considered one-dimensional and that all items should be kept, as parsimonious indications of evidence of validity based on the internal structure for the efficacy scales for coping with COVID-19.

Regarding the reliability estimates, the internal consistency indexes were considered good or excellent. Although there are warnings in the literature (^{Taber, 2018}) that very high Cronbach's Alpha values may denote redundancy between items, this does not seem to occur in the COVID-19 efficacy scales. The items were generated in such a way that, on the six scales, each item reflects a type of behavior necessary to cope with COVID-19. Thus, it is considered that the high correlation between them, which is reflected in the high internal consistency indexes, indicates that each scale actually measures a single construct.

Therefore, it can be said that, as a whole, the validity evidence based on the internal structure and the reliability estimates are satisfactory for the scales, although, as mentioned above, with parsimony for the former. Validity evidence and reliability estimates are considered minimum quality requirements for any psychological measure (^{American Educational Research Association, American Psychological Association, & National Council on Measurement in Education, 2014}). Furthermore, in times of social distance due to COVID-19, online psychological instruments that allow remote application are essential (^{Wright, Mihura, Pade, & McCord, 2020}).

It was observed that the levels of efficacy tend to decrease when the control of actions moves away from the individual. Self-efficacy was the highest, followed by family efficacy. The federal government’s proxy-efficacy had the lowest score. At the intermediate levels, however, the averages of ED-C19-Municipal and ED-C19-Estadual were slightly higher than that of ECom-C19. The lowest scores in federal proxy-efficacy may be related to the perception of the inadequacy of the posture adopted by this level of government, especially the executive, when facing COVID-19 (^{Campos, 2020}). By disowning science, the federal executive opposed the mitigation measures adopted by states and municipalities, which culminated in the unanimous decision of the Brazilian Supreme Federal Court that states and municipalities have autonomy to determine measures of social isolation (^{Lima, Pereira & Machado, 2020}).

The high level of self-efficacy accompanied by lower scores of beliefs of efficacy related to the community is worrying. ^{Germani, Buratta, Delvecchio and Mazzeschi (2020}) point out that shared goals, interdependence and sociability must be emphasized and promoted as a protective factor against COVID-19. The proxy agency, for example, played a major role in determining behavior in analogous situations, such as Ebola epidemics (^{Li, 2018}; ^{Webster et al., 2020}). Thus, even if the level of self-efficacy is high, as observed in this study, the adoption of health behaviors may be at risk, as the belief in the community's ability to deal with the situation is low. Therefore, to increase adherence to COVID-19 mitigation measures, it is necessary to take actions that favor the increase of the governmental proxy-efficacy, especially at federal level.

The correlations between self-efficacy and other beliefs of efficacy also tend to weaken the further away control over the actions necessary to achieve the objective (mitigation of COVID-19) is from the individual. Thus, the moderate positive correlation between self-efficacy and collective family efficacy can be explained by the more intense and direct participation of the individual in this group. The significant positive correlations between the ECom-C19 and ED-C19-Municipal scores and those of the other instruments are possibly due to the fact that the community and the municipality are intermediate instances; in other words, contexts with easier participation or access. The moderate or strong positive correlations between the three proxy efficacy scales were expected and showed expressive convergence between them, despite the fact that the average beliefs in the ability of the three levels of government to mitigate COVID-19 were distinct.

The existence of positive correlations between most measures, although predominantly weak or moderate, indicates that, even though different, they are associated constructs. Thus, it corroborates ^{Caprara et al. (2004}) who obtained similar results in research into self-efficacy and collective family efficacy.

Among the limitations of this study, those related to external validity stand out. The sample, which is predominantly white, female, highly educated and with high income, differs from the Brazilian population. As for internal validity, the online collection strategy and the fact that confirmatory factor analysis have not been carried out are circumscriptions. Therefore, other studies are recommended to test the efficacy for coping with COVID-19 scales in more representative and randomized samples. We would also suggest that other evidence of validity (based on the relationship with other variables, etc.) and more estimates of reliability (stability, etc.) for them be obtained.

Despite those and other limitations, it is possible to state that the measures of efficacy for coping with COVID-19 have adequate psychometric properties. They therefore represent a contribution to further research on factors that influence behavior in health emergencies, such as the current one, generating fundamental knowledge for the development of evidence-based interventions to preserve health and lives.

Credit authorship contribution statement

Andrêze Cristine do Nascimento Silva: Conceptualization; Data Curation; Formal analysis; Funding acquisition; Investigation; Methodology; Project administration; Visualization; Writing - Original Draft; Writing - Review & Editing. Juliana Célia de Oliveira: Data Curation; Formal analysis; Methodology; Validation; Visualization; Writing - Original Draft. Altemir José Gonçalves Barbosa: Conceptualization; Data Curation; Formal analysis; Funding acquisition; Investigation; Methodology; Project administration; Supervision; Validation; Visualization; Writing - Original Draft; Writing - Review & Editing. Roseane Ribeiro Mendonça: Validation; Writing - Original Draft.

Acknowledgments

To the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES), which granted a master scholarship for the first author.