Confirmatory factor analysis of the Appraisal of Self-Care Agency Scale - Revised 1

ABSTRACT Objective: to analyze the factor structure of the Appraisal of Self-Care Agency Scale-Revised (ASAS-R), adapted for Brazil. Method: methodological study conducted with 150 individuals with diabetes mellitus cared for by the Family Health Strategy, most of whom are elderly with low educational levels. The test of the hypothesis concerning the confirmatory factor composition of the ASAS-R was performed using latent variables structural equations. Results: the model’s goodness-of-fit indexes were satisfactory (χ2 = 259.19; χ2/g.l = 2.97, p < 0.001; GFI = 0.85; RMR = 0.07; RMSEA = 0.09); the factor loads were greater than 0.40; and most item-to-factor-correlations presented moderate to strong magnitude (0.34 to 0.58); total alpha value was 0.74, while the alpha of the three factors were 0.69, 0.38 and 0.69, respectively. Conclusion: the scale’s factor structure presented satisfactory validity and reliability results, with the exception of one factor. Application of this scale to samples of the general population is desirable in order to strengthen analyses of internal consistency and the dimensionality of the factor structure. This study is expected to contribute to further studies addressing the self-care agency construct and the development of the ASAS-R.


Introduction
Self-care agency, a central concept in Orem's Self-Care Deficit Theory (SCDT), is defined as one's ability to exercise self-care in order to maintain life, health and wellbeing. It is a complex ability acquired over the course of life, based on repeated behavior on a daily basis and is influenced by cultural aspects and background, personal skills and limitations, life experiences, health status, and resources available (1) .
The development of self-care agency enables an individual to discern between factors that need to be controlled and taken care of, decide what one can do and what one needs to do, recognize one's own needs, assess personal and environmental resources, and undertake actions that meet one's self-care needs (1) .
According to the SCDT, the relationship between one's self-care ability and needs is essential to establishing the actions one should perform and those actions one has the ability to develop in order to maintain health and prevent and manage diseases (1) .
This understanding is important to assessing the self-care ability of people with chronic diseases, especially diabetes mellitus, which is a condition that requires great responsibility and commitment, especially from those taking insulin (2)(3)(4) , to carry on with the therapeutic regimen that includes behavioral modification of daily activities (5) . From this perspective, assessing an individual's personal ability to perform selfcare has been widely studied to highlight the individual's performance in preventing and managing diabetes mellitus (4,(6)(7)(8) .
The Appraisal of Self-Care Agency Scale -Revised was adapted and validated in Brazil with a sample of individuals with DM taking insulin, though it is not a specific scale (9) .
The conceptual basis for the development of this scale was the Self-Care Deficit Theory developed by Orem (1) . The scale's items concerning the concept of selfcare agency were based on empowering traits or power components (specific personal abilities to perform selfcare) and operational traits (ability to organize personal and environmental resources that might be significant in self-care) (1,10) .
The revised version was chosen by Sousa (11) to be adapted to Portuguese with a sample of Brazilian individuals because it presents a better fit index, greater reliability (total Cronbach's alpha = 0.89) and better validation results in comparison to its original version (9) .
The changes performed in the revised version included the exclusion of nine items and the presentation of three factors that were not reported in the original version (11) .
The translation and adaptation process of the ASAS-R in Brazil followed the stages recommended in the literature (12) . After the translation and adaptation process, the scale was submitted to analysis of the distribution of frequencies of items' responses, reliability (internal consistency and product-moment correlation), reproducibility (test-retest and interobserver), and construct validity (convergent and discriminant) (9) .
The results of the psychometric analysis show that the one-dimension structure of the adapted scale is reliable (total Cronbach's alpha = 0.74), reproducible (test-retest p < 0.001 and inter-observer p < 0.001) and valid (confirmed the correlation hypotheses with the constructs depression and perceived health status and between distinct groups) (9 The ASAS-R contains 15 items assessed on a fivepoint Likert scale, of which only one alternative may be chosen. Scores range from 1 to 5, where: 1 -"totally disagree"; 2 -"disagree"; 3 -"I do not know"; 4 -"agree"; and 5 -"totally agree". Four out of the 15 questions refer to negative aspects (ASAS-R 4, 11, 14 and 15) (9,11) .
The total score ranges from 15 to 75; the higher the score, the greater one's operational self-care ability (9) . The three factors were denoted: "Having power www.eerp.usp.br/rlae Stacciarini TSG, Pace AE.
Pearson's product-moment correlation less than 0.30 was considered weak with poor clinical applicability; between 0.30 and 0.50 was considered moderate; and greater than 0.50 was considered strong (14) . The significance level was established at 0.05.
In the confirmatory factor analysis, overall fit of the hypothesized factorial model and estimation of the construct's effects on measured variables were considered. Hypothesis testing for the factorial composition of the ASAS-R scale was implemented using latent variable structural equations.
The following indexes were analyzed to verify the model's goodness of fit (15) : Chi-square test (χ 2 ), with significance greater than 0.05; Chi-square ratio At least three adequacy indexes with values greater than their references were considered in analyzing the goodness of fit of data to the proposed factors (16) . The estimation method used was maximum likelihood with a minimum of ten observations per item, which presented univariate normality of items (17) . Wald's test verifies the extent to which the removal of an item influences the model's Chi-square statistics.
Items can be removed without affecting future results when change is not significant (15) . The Lagrange's multiplier test verifies the need to reallocate an item to another factor to improve correlation among the items within the same factor. Similar to the Wald's test, it shows how much an item reallocated to a new factor will influence the Chi-square statistics (15) .
In the exploratory factor analysis, the Kaiser-  (17) .
In the analysis of the principal components, the factors that obtained eigenvalues (total variance explained for each factor) greater than one were selected and interpreted in a scree plot. The extraction of principal factors is performed after Varimax orthogonal rotation and Kaiser's criterion (17) .  Table 1. Table 1 shows that elderly individuals, retired, married with a low educational level and low income, predominate. Analysis of item reliability, as described in Table   2, revealed satisfactory internal consistency for factors

Factor/Item Item-factor correlation
Cronbach's α of excluded item  For the exploratory factor analysis, Bartlett's sphericity test rejected the null hypothesis that the data correlation matrix was an identity matrix (p < 0.001), while Kaiser-Meyer-Olkin (KMO) was 0.643.
These results show good fit of the data matrix to the factor analysis, indicating that the analysis of principal components could be performed.
The analysis of the principal components using a scree plot resulted in three factors that explained 48.6% Table 3 -Analyses of the exploratory factor loads, communality (h 2 ), eigenvalues and variances for the total and each factor of the adapted version of the Appraisal of Self Care Agency Scale-Revised (n = 150). Uberaba, MG, Brazil, 2012 According to the exploratory factor analysis, the reallocation of items in the factors were as follows: factor 1 "items ASAS-R 1,2,3,5,8,10"; factor 2 "items ASAS-R 4,9,13" and factor 3 "items ASAS-R 6,7,11,12,14,15".
Factor 2 "Developing power for self-care" presented the least variance for each factor (8.74%) and the smallest number of items allocated in the factor theoretically proposed ( Table 3).
The factor loads were greater than 0.40 for all the items. Items ASAS-R 2 and 10 should be disregarded because they obtained a high load in more than one factor, though they were allocated to the factor with the highest load. Item ASAS-R 14 presented the greatest communality; that is, 70.0% of its variance was explained by the factors (Table 3). to-factor correlation coefficients of factor 2 "Developing power" remained unsatisfactory.

Discussion
The first version of the Appraisal of Self-Care Agency Scale (ASAS) was developed by a group of American and Dutch researchers, who belonged to the www.eerp.usp.br/rlae 7 Stacciarini TSG, Pace AE.
Nursing Development Conference Group (NDCG), to measure the central concept of Orem's Self-Care Deficit Theory in1986 (10) .
Even though the ASAS is based on the ten power components, it does not mention dimensions.
Measurement is taken in a global and nonspecific way and can be applied and compared to different age groups under various health conditions (10) . Since then, studies have been conducted to verify the factor structure and internal consistency of the scale's items in different countries, to meet criteria concerning construct validity (4,11,(18)(19)(20)(21) .
One study conducted with a sample of Americans with diabetes mellitus taking insulin verified that weak correlations found for some items suggested that the scale could have more than one dimension (4) . The authors continued the studies and decided to verify the exploratory and confirmatory factor structure of the ASAS with 24 items for a sample of 389 American individuals from the general population (11) .
The aforementioned study reports a new structure that obtained excellent goodness-of-fit index after excluding nine items and describing and listing three factors (11) . Comparison of the confirmatory factor analyses among the versions: ASAS 24 items with a single factor; ASAS 24 items with three factors; and ASAS 15 items with three factors, revealed that the last version presented the best goodness of fit, as well as the best construct validity, strongest factor loads, a high variance explained for all the items, and high reliability, in addition to showing strong linear correlation with the original (r = 0.98; p < 0.001) (11) .
Therefore, based on psychometric analysis of validation and reliability, a new version with 15 items called Appraisal of Self Care Agency-Revised (ASAS-R) was established. One of the conclusions reached by the aforementioned study was that there was a need to conduct further studies seeking to perform psychometric assessments among people with chronic diseases, especially diabetes mellitus (11) .
In this sense, based on the revised version, ASAS-R, applied to a sample of Brazilian individuals with diabetes mellitus (9) , this study sought to continue the validation process, analyzing correlations, internal consistency, and the results of the hypothesized model's overall fit, so that these results could be compared with those from the original version (11) .
The correlations among the items of each of the three factors in this study presented from moderate to strong magnitude, with the exception of the items from factor 2 "Developing power for self-care". The correlations reported by the study conducted with the original version were also of moderate to strong magnitude, though in this case, among the items of the three factors (r = 0.41 to r = 0.60, de r = 0.34 to r = 0.61 and from r = 0.40 to r = 0.57, respectively) (11) .
The results concerning the analysis of the total internal consistency of the items in the adapted version of ASAS-R (Cronbach's alpha = 0.74) and factors 1 "Having power for self-care" and 3 "Lacking power for self-care" (Cronbach's alphas of 0.69), were considered satisfactory, except for factor 2 "Developing power for self-care" (Cronbach's alpha = 0.38).
The results concerning the internal consistency of the ASAS-R original version were: total alpha = 0.89 and the alphas among factors were 0.86, 0.83 and 0.79, respectively (4) , the highest total alpha, compared to studies that used the ASAS version with 24 items (total from 0.59 to 0.80) (11) . Expecting to identify the items that could be affecting reliability and the quality of the model's fit, Wald's and Lagrange's multiplier's tests were performed together with exploratory factor analysis.
Even after reallocating or removing some items, however, the estimation of the factor loads and differences in the χ 2 statistics and goodnessof-fit measures obtained in a new confirmatory analysis were not significant enough to suggest any adjustment in the specified factorial model.
In the exploratory factor analysis, the scree plot test suggested three factors, the same number presented in the original version (11) , but factor 2 still presented weaker correlations and low internal consistency, as well as a low variance was explained for each factor (8.74%).
Given the preceding discussion, it is desirable for this scale to be applied in samples from the general population to advance its development and present www.eerp.usp.br/rlae 8 Rev. Latino-Am. Enfermagem 2017;25:e2856 more evidence to strengthen analysis of the internal consistency and dimensionality of the factor structure.
Additionally, we do not know the extent to which the sample's homogeneity, in terms of sociodemographic, cultural, clinical characteristics, or in terms of accessibility to public services, contributed to the reliability results or the goodness of fit of this scale.

Conclusion
Analyses of product-moment correlation and reliability of the factor structure of the adapted ASAS-R were satisfactory, except for factor 2 "Developing power