SCHFI 6.2 Self-Care Confidence Scale - Brazilian version: psychometric analysis using the Rasch model

Objective: to evaluate the psychometric properties of the Self-Care Confidence Scale in heart failure in the Brazilian version of the Self Care Heart Failure Index, version 6.2, using the Rasch model criteria. Method: secondary study, of psychometric analysis, using the Rasch model, of the six items of the scale. The sample consisted of 409 patients with heart failure undergoing outpatient treatment [mean age 57.9 (standard deviation = 11.6) years, 54.8% male]. Results: of the total of six items, one (“De maneira geral, você está confiante sobre estar livre dos sintomas de insuficiência cardíaca?”) presented maladjustment to the model (Infit = 1.84 and Outfit = 1.99). After the exclusion of this item, the others showed a good fit, composed one dimension and explained 55% of the variance in the data; the categories of response to the items were adequate, the values of separation and reliability of person were 2.13 and 0.82, respectively, and Cronbach’s alpha was 0.87. Items of extreme difficulty were identified and there is no differential functioning of the items in relation to sex. Conclusion: with the exclusion of the first item, the Self-Care Confidence Scale showed good psychometric properties, with caution in interpreting the results of the six-item scale.


Introduction
Heart failure (HF) is considered the cardiac syndrome with the highest rates of morbidity and mortality and the main cause of hospital admissions, with worldwide prevalence rates of 1% to 7% (1)(2) . Adequate self-care in HF has positive effects on patients' clinical outcomes (3)(4)(5) , and the confidence that the person has that he/she can do self-care is fundamental for proper self-care (6)(7)(8)(9) .
The Self-Care Confidence Scale in HF is one of the Self-Care Heart Failure Index -Version 6.2 (SCHFI 6.2) (10)(11) , which has been adapted for use in Brazil (12) and, as its name suggests, it assesses the person's confidence that he/she can handle the self-care to control HF. The  (12) . Responses to each item vary in scores from one to four from "não confiante", "um pouco confiante", "muito confiante" to "extremamente confiante" in the Brazilian version (12) .
The scores obtained from each of the six items are added together to produce a total score that reflects the person's self-confidence to take care of themselves.
In the study of adaptation and validation of SCHFI 6.2 for Brazil, the authors used traditional analyses of psychometrics. However, more recent approaches, such as Rasch analyses, allow testing hypotheses about scales that cannot be tested by traditional psychometric analyses. For example, the Rasch model is useful for testing specific hypotheses about the dimensionality of items within a scale (13) .
In traditional psychometric analyses, it is assumed that all items in the set of items that make up a given scale measure the same dimension as the phenomenon in question. It is important to verify this assumption because, summing the items' scores can only be valid when all items measure the same dimension (13) . That is, if, in the Self-Care Confidence Scale, one of the items does not measure the same dimension as the others, the interpretations made about the scores obtained with the scale are not valid.
The experience of the authors of this article with the use of SCHFI 6.2 has shown that patients, in general, have difficulties in responding to the Self-Care Confidence Scale because they often express doubts about how to respond to their items. This experience motivated the realization of this study, which aimed to produce new evidence about the psychometric properties of the referred scale.
Considering that the Rasch model facilitates the identification of weaknesses in measurement instruments that cannot be detected by traditional psychometric analyses (14) , the objective of this study was to evaluate the psychometric properties of the Self-Care Confidence Scale in HF of the Brazilian version of SCHFI 6.2, using the Rasch model criteria. The tested hypotheses were that: 1) all items on the scale reflect the same dimension; 2) all items fit the Rasch model; 3) the scale allows good discrimination of degrees of confidence in self-care; 4) scale items do not vary in measurement between sex.

Method
This is a methodological study, of psychometric analysis, (15) of data from 409 patients with HF undergoing outpatient follow-up who responded to the Brazilian version of SCHFI 6.2 in another study (16) . The primary study sample was convenient and the inclusion criteria were: confirmed medical diagnosis of HF, with functional class I, II or III and clinical conditions that allowed participating in interviews. The exclusion criteria were the presence of psychiatric, oncological, infectious diseases, with repercussions in the general state, or endocrine-metabolic diseases without treatment (for example, Diabetes Mellitus or thyroid disease); recent postoperative period (up to 60 postoperative days) of any surgical intervention (16) .
The primary study obtained a favorable opinion from the Research Ethics Committee (Protocol No. 145,630 -2012) and all participants signed the Free and Informed Consent Term (FICT). Other methodological procedures are detailed in the article that reports the primary study (16) .
From the primary study, demographic data and responses to the SCHFI Self-Care Confidence Scale, version 6.2 adapted for use in Brazil (17) . There were no missing data in the responses to this scale.
Rasch analysis was applied using the Winstep 3.91.0 software and Andrich's model was chosen, considering the assumption that all items have the same structure on the rating scale. Next, the order of the analyses: 1) The polarity of the items on the scale was examined to check for inverted items. If there were inverted items, it would be necessary to adjust their scores before other analyses; 2) An analysis of the functionality of the response categories was carried out, determining the www.eerp.usp.br/rlae as suggested by Wright (24)(25)(26) . The fit in the Rasch model (fit statistics) indicates whether the data deviate or not from the model (23) ; provides the comparison between what was expected in the model and what was obtained with the sample (18,23) .  (11,14) ; 5) The reliability of the scale was assessed using the person reliability, Cronbach's alpha and the person and item separation index. The reliability of person is conceptually equivalent to Cronbach's alpha and a value above 0.80 is considered appropriate.
In the case of Cronbach's alpha, a value above 0.70 is appropriate. The person separation index indicates how many groups with different skill levels (confidence in self-care) the item allows to identify (27) . The item separation index indicates how many levels of difficulty the items are distributed in.
A clinically useful tool should divide participants into at least two skill levels (high and low), as well as items must be distributed across at least three levels of difficulty; 6) The presence of the differential item functioning (DIF) was investigated to test the measurement invariance between the sexes at the item level, that is, which items for men are easier or more difficult than for women to fulfill the item description (28) . The cutoff point commonly used to indicate substantial DIF is the DIF contrast >0.5 (29) , with a statistical significance <0.05 in the Rasch-Welch test; the Rasch analysis can be used as evidence of the instrument's construct validity (27) .
The unit of measurement used in the Rasch model is the logit (log-odds), which is a linear function of the probability of obtaining a score by a person who has a certain skill (21) . Zero on the logit scale represents, in Rasch's analysis, arbitrarily, the mean; the easiest items have negative values and the most difficult items have positive values on the scale (22) .

Results
Data from 409 HF patients with a mean age of 57.9 years (standard deviation = 11.6) were analysed, 248 The results of the Rasch analysis of the responses of 409 patients with HF to the six items of the Self-Care Confidence Scale in HF showed that all items had positive polarity, with item-total correlations between 0.56 and 0.80.
The analysis of the functionality of the response categories showed that they meet all the established criteria, as shown in Table 1.     It can be seen in Table 3 Table 4 shows the results of the analysis of the differential functioning of the items in relation to sex. It can be seen, in Table 4, that none of the six items showed substantial contrast (all were less than 0.5) between men and women using the Rasch-Welch test.
As for the unidimensionality, the analysis by main

Discussion
The analyses carried out made it possible to investigate the structural validity and reliability of the  (18) . In the case of this study, it can be said that item 1 is not related to the construct of Confidence in Self-Care as are the other items (31) .
This result is possibly due to the Portuguese version of this item. Although adaptation procedures have been rigorous (12,17) , the analyses that were carried In the study of adaptation of the instrument (17) , the author mentions that the statement of the Self-Care Confidence Scale changed after the evaluation by the committee of judges. The term "confident", from the original, would have been translated as "seguro" and then changed to "confiante", with no other item being changed by the committee of judges or due to the pre-test carried out with 30 patients (17) . The adapted SCHFI 6.2 was applied to 190 patients with HF and the data were submitted to internal consistency analysis.
Cronbach's alpha of the Self-Care Confidence Scale was 0.94, which is very good, but the results showed that if item 1 were excluded, this index would not change (17) , indicating that this item was no longer contributing to the internal consistency of the scale. Rasch's analysis, which offers more information on the suitability of items, revealed that this item needs to be improved if kept.
As for the other properties of the scale, the results show that the Brazilian version of the SCHFI 6.2 Self-Care Confidence Scale with six or five items is onedimensional, explaining 50% or 55% of the variance in the data, respectively, with the eigenvalue of the first contrast less than two. These results appear to be consistent with what was obtained in the validation study of the Brazilian version of SCHFI 6.2 (12) in which confirmatory factor analysis, used to test a model with three components, showed the Self-Care Confidence Scale as one of the three components of SCHFI 6.2.
However, there are different results in other studies that also used confirmatory factor analysis (11,32) . In a study that explored the dimensionality and reliability of SCHFI 6.2, in its original English version, with data from 629 adults with HF, it was demonstrated that the Self-

Care Confidence Scale is one-dimensional, confirming
what had been assumed in other studies (32) . However, confirmatory factor analysis in another study, conducted with data from 659 Italian HF patients, revealed the Self-Care Confidence Scale with two factors, called Basic Self-Care Confidence and Advanced Self-Care Confidence. The factor of Basic Confidence in Self-Care included the most general items (for example: "...

seguir o tratamento recomendado"?), while Advanced
Confidence in Self-Care reflected more challenging behaviors (for example: "... estar livre dos sintomas de insuficiência cardíaca?"), which require specific guidance and experience (11) . The discrepancy between the two studies that used confirmatory factor analysis was attributed to the differences in method with which the analyses were performed and not to differences between the samples (11) . The results of the study in which the Self-Care Confidence Scale resulted with two factors were supported by more robust techniques (cross-validation) and resulted in models with better adjustment rates (11) . In addition, the final analyses were made considering the three scales separately (11) and not as a single model as in the other studies cited (12,32) .
The results of this study showed that the scale is onedimensional. However, as far as it is known, it is the only study that used Rasch analysis to investigate the properties of the Self-Care Confidence Scale. These controversial results indicate that the dimensionality of the Self-Care Confidence Scale needs further study.
Another property that Rasch's analysis allowed to verify is whether the response scale of each item is adequate. In the case of the Brazilian version of the 4="extremamente confiante") (12) . The data in Table   1 show that the four categories of responses (não confiante; um pouco confiante; muito confiante; extremamente confiante) represent increasing intensities of confidence, since the average of category 1 is lower than that of category 2, which in turn , is less than category 3, and less than category 4.
Other important information provided by Rasch's analysis was the DIF (Table 4) and the degree of difficulty of the items (Figure 1). The DIF evaluation allowed establishing that the instrument does not present bias due to the sex of the respondents. ), which was the easiest item. This order of difficulty seems reasonable, considering the challenges that the represented behaviors offered to patients with HF. Having confidence that they can do something to alleviate the symptoms seems to be more difficult than having confidence that you can assess the importance of the symptoms. Having confidence that they can assess the importance of symptoms seems more challenging than having confidence that they can recognize them.
Given these behaviors, it seems that having confidence that you can follow the recommended treatment is the least challenging behavior for patients with HF. Knowing the degree of difficulty of the items can serve as a guide for the organization of self-care promotion programs for people with HF.
The results regarding the distribution of persons ( Figure 1) suggest that there is a need for items to measure difficulties greater than those of item five and less than those of item 2, because there are parts of the distribution of persons that are outside the spectrum of difficulties of existing items. It is also observed that new items would fit with difficulties between items 2 and 6 and between items 3 and 5.
Further studies on the properties of the SCHFI 6.2 Self-Care Confidence Scale, in its various versions, using Rasch analysis, are necessary to refine an instrument that assesses a fundamental variable for knowledge about selfcare in HF. In the case of the Brazilian version, the first step would be to review the version of item 1 through content validity studies (14) . If, in a new version, this item fits the model, comparisons between samples of Brazilians and samples that used the original instrument will be favored.
The main limitation of the reported study is the fact that the data are from patients from a single public service specialized in Cardiology, which indicates caution in generalizing the results. Studies that analyse whether there is a differential functioning of items between patients with HF from specialized and non-specialized services would offer evidence on the magnitude of such limitations.

Conclusion
In summary, the analysis using the Rasch method of the Brazilian version of the SCHFI 6. scale items do not vary in measurement between sexes.