Organizational commitment, psychometric qualities and invariance of the Meyer and Allen Questionnaire for Portuguese Nurses 1

ABSTRACT Objective: to evaluate the psychometric qualities of the Portuguese version of the Organizational Commitment Questionnaire for the nursing context, through confirmatory analysis and invariance, aiming to evaluate the reliability, internal consistency, construct validity and external validity of the instrument. Method: confirmatory factor analysis of the Portuguese version of the questionnaire was carried out with a sample of 850 nurses, in hospital context. The analysis was complemented using specification search. Goodness of fit was evaluated through different indices. Reliability, internal consistency and construct validity were estimated. The invariance of the model was evaluated in two subsamples of the same sample, in order to confirm the external validity of the factorial solution. Results: the refined model demonstrated good overall fit (χ2/df=6.37; CFI=0.91; GFI=0.92; RMSEA=0.08; MECVI=0.62). The factorial structure was stable (λ:Δχ2(14)=18.31; p=0;193; Intercepts: Δχ2(14)=22.29; p=0.073; Covariance: Δχ2(3)=6.01; p=0.111; Residuals: Δχ2(15)=22.44; p=0.097). Conclusion: the simplified model of the questionnaire demonstrated adequate goodness of fit, representing a stable factorial solution. The instrument was fit to monitor and evaluate the organizational commitment of Portuguese nurses.


Introduction
The commitment of human resources to career and organization is crucial for personal, professional and organizational development (1) .
In the organizational context, commitment is a multidimensional construct related to a psychological state associated with the affective relationship between employee and organization, the perception of a moral obligation of permanence for loyalty, and the cost/benefit associated with leaving the organization. It influences satisfaction, performance, productivity and professional investment, as well as turnover, with a strong impact on organizational behavior (2)(3)(4)(5)(6) .
In theoretical terms, the reference model for the analysis of organizational commitment defines the construct in three components: affective, continuance and instrumental/normative (2,(7)(8) .

The Affective Organizational Commitment (AOC)
is related to the emotional attachment established between employee and organization, when the employee identifies himself with the company due to compatibility with human values and common goals. Therefore, it is related to personal perception of the objective and subjective characteristics of the organization, and it is fundamental for satisfaction, professional motivation and organizational success (2,(9)(10) .
Complementarily, the continuance organizational commitment (COC) is related to the employees' perception of the cost associated with leaving the organization. Intention to quit is commonly related to the rewards system, professional recognition/appreciation, and career development opportunities. Therefore, the employees' personal investment is adjusted to the return they expect from the organization (2)(3)(8)(9) .

Normative organizational commitment (NOC)
is associated with the employee's moral duty to the organization and the surplus value received, which arouse the feeling of obligation of permanence. This can promote a performance of activities with competence, according to the norms of the institution, but without great motivational involvement (2)(3)(4)8,11) .
According to this approach, employees with strong AOC are committed and remain in the organization on their own will, without considering quitting it. Regarding the COC, which is associated with the achievement of personal objectives, the employees remain in the organization by necessity, by the lack of alternatives and by the costs associated with the change. Finally, a strong NOC results from the sense of associated duty and moral responsibility with the organization (2)(3)12) .
In theory the conceptual model is based on the independence of the three components. However, scientific research has shown that there are correlations between its dimensions. These dimensions are influenced by employees' work experiences, and it is expected that different levels of commitment in each of the components will coexist, with a variable impact on the behavior in the work context (2)(3) .
Therefore, organizational commitment arises from the joint influence of factors such as personal, structural, and performance-related characteristics and work experiences. It is also a variable that determines intention to quit, professional performance and organizational success (6,13) .
In organizational terms, and specifically in the area of health, the organizational commitment of human resources plays a key role in individual and organizational performance, in an effective and efficient manner. High levels of organizational commitment are determinant for attaining organizational objectives, and therefore can confer a competitive advantage to the health institutions.
However, the socio-economic context of recent years has led to the need for cost containment, with an impact on the effective management of resources, particularly human resources. Thus, organizational and legislative changes have been necessary, which influenced the relationships of health professionals with organizations, with subsequent impact on motivation, organizational commitment, and on the quality of health care (6,(11)(12)(14)(15)(16) .
Regarding nursing, this is particularly relevant given its privileged role in the provision of health care and considering it is the professional group most present in health systems. However, increasingly precarious employment contracts and unfavorable working conditions influence nurses' organizational commitment and subsequent health outcomes. In Portugal, the ratio of nurses (6. (6,12,(14)(15)17) .
Scientific evidence points to the organizational commitment of nurses as predictor of performance. It is also determinant for commitment with the patient and www.eerp.usp.br/rlae 3 Neves T, Graveto J, Rodrigues V, Marôco J, Parreira P.
for the development of positive behaviors, minimizing negative aspects such as conflicts and professional exhaustion and having an impact on the quality of nursing care and on patient satisfaction (6,9,12,(18)(19) .  (8) . This is one of the instruments most used internationally in scientific research to assess organizational commitment, namely in Portugal and among health professionals, especially nurses (4,13,(20)(21)(22) .  IBM Company, Chicago, IL) to verify the adequacy of the data to the study model. In this study, all items of the original scale were included, although some items were eliminated in the previous validation study for the Portuguese context (3) .
Normal distribution was determined by the asymmetry (Sk) and kurtosis (Ku) coefficients, considering that |Sk| < 3 e |Ku| < 10 did not indicate significant deviations from the normal distribution, which made the analysis through the maximum likelihood method possible. The presence of outliers was evaluated by the Mahalanobis squared distance (D 2 ).
Omitted values were replaced by the mean of the series, given the small percentage in the study sample (less than 1.5%) (24) .
The modifications introduced for the adjustment of the model were supported by the modification indices (greater than 11; p<0.001) provided by AMOS, and by theoretical considerations (24) .
Considering the need to readjust the model in order to find the best subset of factors and reflective variables and to optimize the goodness of fit and the parsimony of the factorial solution, an additional exploratory strategy was used: a heuristic exhaustive search based on the previous analysis of the modification indices (24) .
In order to verify the external validity of the factorial model obtained, a cross-validation was performed, comparing the indices observed in the test sample with the indices obtained in another independent sample, extracted from the same population, through multi-group analysis. Therefore, the total sample was randomly divided into two approximately equal parts. The factorial invariance (configural, metric and scalar) of the measurement model was tested in both groups by comparing the free model in the two groups with a constrained model in which factor weights (loading), intercepts, residuals and variances/covariance are fixed in both groups. The statistical significance of the difference between the two models was determined by the Chisquare test (24,(27)(28) .

1.
Acredito que há muito poucas alternativas para poder pensar em sair desta instituição 2. Não me sinto "emocionalmente ligado" a esta instituição  for each factor, considering values greater than 0.50 as convergent validity indicators (24,26) ; the discriminant validity, considered evident when the AVE value of each of two factors is equal or higher than the square of the correlation between these factors; and factorial validity, which considers the standardized factor loading (λ) and individual reliability (λ 2 ), which are also indicators of the goodness of the local fit. Factor loadings greater than 0.50 and, subsequently, individual reliability greater than 0.25 (24,29) are usually considered appropriate; however, the area of social sciences sometimes considers lower values (30) . In a previous study of the OCQ, the authors propose factor loadings greater than 0.40 (31) . However, in this investigation, we chose to consider factor loadings equal to or greater than 0. 30  work organized by shifts (81.71%), or roulement, as described in Table 1.    identified (3,10,(32)(33)(34) .
In the initial work of translation and validation for the Portuguese context, two original items had already been eliminated, namely item 10, due to inadequacy to the factorial structure. This study also suggests, for future investigations, the elimination of item 19, because it is in the threshold of acceptability (3) .
A validation study for the Brazilian context also identified item 10 as an affective component (33) . It should be pointed out that the term loyalty in this item, which is classified as normative, may be, according to some authors, an indicator of affective commitment, which is probably its interpretation in the Portuguese culture (33,(35)(36) .
The divergences found between the different models may be due to the fact that the solutions resulting from the exploratory factor analysis using the orthogonal rotation of the factors consider that each item saturates only in one factor. Thus, when evaluating these models through confirmatory analysis, it is common for modification indices to identify items that are reflected in different factors, influencing the goodness of fit, a frequent occurrence in social and human sciences (24) .
The analysis of the reliability and internal consistency of the model revealed adequacy of the AOC and NOC scales, with a slightly lower value in the COC www.eerp.usp.br/rlae 9 Neves T, Graveto J, Rodrigues V, Marôco J, Parreira P.
scale, but still considered adequate by some authors.
These results are similar to the values of the original scale and values found in other studies (12,(31)(32)37) , but are lower than those of the previous study of this Portuguese version (3) .
Regarding However, discriminant validity is confirmed in all scales.
According to some authors, factor loadings equal to or greater than 0.30 or 0.40 are acceptable in exploratory analysis in the social sciences (30,38) .  (3,32) .
The low values verified in items 3 (λ=0.44) and 13 (λ=0.38) may be associated with problems of cultural interpretation resulting from the adaptation process to the Portuguese nursing context. It should be noted that item 1 was also removed from the simplified model because it has very low factor loading.
These items belong to the COC scale, which presented the greatest need for adjustment, with the removal of three items (1, 14 and 19). Therefore, for theoretical issues, items 3 and 13 were kept in the model, since they are important to ensure the evaluation of the latent construct of continuance commitment.
All modifications associated with the COC scale may be due to the fact that, in theory, this factor can be subdivided. The two-dimensional approach consists of two factors, one related to the lack of alternatives and employment opportunities and the other to the losses that would occur by leaving the organization.
However, these two factors are often strongly correlated.
The scientific evidence is inconclusive on the topic of the two-dimensionality of COC, since there are studies supporting both perspectives (unidimensionality and two-dimensionality) (39) .
In the context of the factorial structure, the authors of the scale (2,8) postulate that the affective, continuance and normative components are independent, which is an assumption supported by several studies. However, given the evidence of strong correlation between the AOC and the NOC found in most of the studies, including this one, the pertinence of their separate use is questioned, and there actually are studies that consider only the AOC and COC scales (7,32) . Despite