Nursing Care Interpersonal Relationship Questionnaire: elaboration and validation 1

ABSTRACT Objective: to elaborate an instrument for the measurement of the interpersonal relationship in nursing care through the Item Response Theory, and the validation thereof. Method: methodological study, which followed the three poles of psychometry: theoretical, empirical and analytical. The Nursing Care Interpersonal Relationship Questionnaire was developed in light of the Imogene King’s Interpersonal Conceptual Model and the psychometric properties were studied through the Item Response Theory in a sample of 950 patients attended in Primary, Secondary and Tertiary Health Care. Results: the final instrument consisted of 31 items, with Cronbach’s alpha of 0.90 and McDonald’s Omega of 0.92. The parameters of the Item Response Theory demonstrated high discrimination in 28 items, being developed a five-level interpretive scale. At the first level, the communication process begins, gaining a wealth of interaction. Subsequent levels demonstrate qualitatively the points of effectiveness of the interpersonal relationship with the involvement of behaviors related to the concepts of transaction and interaction, followed by the concept of role. Conclusion: the instrument was created and proved to be consistent to measure interpersonal relationship in nursing care, as it presented adequate reliability and validity parameters.

The items were elaborated and content validation was carried out.
The theoretical dimensionality was defined based on the concepts that make up the Interpersonal System of the Imogene King's Interacting Open Systems Model, which proposes that interpersonal relationship is composed of five constitutive elements: interaction, communication, transaction, role and stress (1) . These elements were carefully analyzed and the constitutive definitions emerged from them.
After elucidating the constitutive definitions, the operational definitions and the items were elaborated, based on an integrative revision (11) and on six focus The inclusion criteria for the participants were: at the PHCU and at the ICHD, individuals aged> 18 years, who had been followed for at least one year in the service and who were waiting for the nursing consultation. In the WCUH, the inclusion criteria were individuals> 18 years of age, having been hospitalized for at least 24 hours in the wards. Those who did not communicate verbally and those who were in isolation for some contagious infectious disease that prevented interaction with the researcher were excluded.
The elaboration of the items followed the twelve criteria of psychometry (amplitude, balance, behavior, simplicity, clarity, relevance, precision, modality, typicity, objectivity, variety and credibility) (9)(10) . The Afterwards, the content validation of the NCIRQ was performed, with content and semantic analysis.
The content analysis was performed by nine nurses that are experts in interpersonal relationship. These were five academics and four clinicians with clinical experience, research and publications on the subject, from four Brazilian states. Initially, the Coordination for Improvement of Higher Education Personnel (CAPES in Portuguese) database was searched for people who studied "interpersonal relationship in nursing" and then other experts were identified and contacted. To determine the level of agreement, the Content Validity Index (CVI) was ≥0.78 (12) .
In order to perform the semantic analysis, the NCIRQ was applied in a pilot test to 66 people in the same locations of the focus groups and considering the same inclusion and exclusion criteria, with 28 people from primary care, 23 from secondary care and 15 from tertiary care. These people were distinct from focus group participants. The pilot sample was constituted considering the minimum parameter of 5% of the sample of the empirical phase. The difficulties in understanding the words and expressions present in the items were observed, participants were asked about the need for adjustments, and the adequacy of the response categories of each item was reviewed.
In the empirical pole, the planning and application of the NCIRQ was carried out in order to evaluate its psychometric properties through the TRI. This stage was carried out in six PHCUs that were randomly selected, in the ICHD and in the WCUH, considering the same inclusion and exclusion criteria delineated for the focus groups.
Participants in studies for the development of instruments via TRI should have a sample size enough to preserve heterogeneity and achieve respondents covering the entire latent trait, but there is no consensus on the ideal number. Simulations for the decision of sample size in TRI concluded that 500 subjects in relation to 40,000 bring results very close to those estimated in larger samples (13) . In this sense, an average of 20 respondents were agreed for each of the 44 items, totaling 880 people, plus 10% to repair losses, thus reaching 968 people. Considering the application of the instrument at the three health care levels, the NCIRQ was applied to 950 patients, 319 of which were in the PHCU, 335 in the ICHD and 296 in the WCUH.
There were 18 participants lost due to incompleteness in the answers to the items, but this was within the expected range. In the analytical pole, the steps for the analysis of the psychometric properties of the NCIRQ were conducted. In the reliability analysis, internal consistency was verified by the Cronbach's and McDonald's Omega coefficients, whose reference values for these measures were: <0.6 -low; between 06 and 0.7 -moderate; and between 0.7 and 0.9 -high reliability (14) .
The McDonald's Omega coefficient was used to verify the maintenance of the Tau-equivalence principle.
This coefficient is a better measure of reliability when the Tau-equivalence principle is violated, that is, when the items do not show similar values in the coefficient matrix; its reading is similar to that of Crombach's Alpha and should be performed in comparison, since a low Alpha value followed by a high Omega value indicates such a violation, the latter being the coefficient that best demonstrates reliability (14) .
The dimensionality study was done on the polyclonal correlation matrix and the main components were analyzed, with oblimim rotation and parallel analysis (15) .
These analyzes were performed using the statistical packages "Rcmdr" (16) and "psych" of R (17) . In order to establish the presence of a dominant dimension in the NCIRQ, the convention was adopted that a variance explained by the first factor greater than 20% indicates essential unidimensionality (18) .
In the estimation of parameters, the onedimensional Gradual Response Model of TRI was applied and it was performed in the Multilog software to observe the estimates of the standard errors of the parameters, subsidizing the decisions of exclusion of items, and confirmed by using the package "mirt" of R (19) . Regarding the interpretation of parameter a (item discrimination), values above 0.6 are acceptable, and the higher the value of a, the greater the discrimination power of the item; for parameter b (difficulty/positioning), the values are acceptable in the range of -5 to +5 (20) .
The scale construction was performed based on the anchor levels of the categories of items with good discrimination (a> 0.6). The anchor levels are points on the scale selected to be interpreted and the anchor items are those selected for each of the anchor levels (21) .
For an item to be considered anchor at a given level of Rev. Latino-Am. Enfermagem 2017;25:e2962. the scale, it is expected to be positively answered by at least 65% of the respondents and by a proportion less than 50% of those with the immediately lower level. The difference between the proportion of these two levels should be at least 30% (20) . Because it was difficult to meet all conditions, item categories were positioned at the 60% response rate (near-anchor levels).
After estimating the parameters, the Test and 43 presented low rates of discrimination with high standard errors, indicating that they were not part of the latent trait and were eliminated.
The third attempt was made with 32 items, and convergence was achieved. However, item 32 presented a high standard error associated with the difficulty parameter of the item, being removed from the NCIRQ.
The fourth calibration attempt was performed with 31 items. In this attempt, the parameters of all the  Figure 2 shows the NCIRQ items, indicating items 13, 31 to 41 and 43 that were eliminated in the calibration process.
The estimates of the parameters of the items are shown in Table 1, which presents the 31 items of the final instrument and the parameters a, discrimination, and b2, b3 and b4, difficulty, for the category sometimes, most of the times and always on the adjectival scale.
The items that best discriminated the patients regarding the effectiveness of the interpersonal relationship with the nurse were 06, 07, 18 and 30, with higher discrimination parameter, a. It was observed that items 23, 29 and 44 were below the criterion adopted because of the low discrimination. However, considering that these items did not disturb the calibration, they were kept in the instrument and excluded from the scale interpretation.   Regarding the difficulty of the item, its measure is given by parameter b, which indicates the position in the scale in which the item has more information.
The higher the b, the greater the difficulty of this item.
Thus, when reaching the items with greater value of b, the patient will have a more effective interpersonal relationship in the nursing care. Considering the items with discrimination above the reference value adopted,  At this level of interpersonal relationship, the time for performing care is sometimes assessed as sufficient. Transaction becomes more effective as the nurse calls the patient by his/her name or how he/she likes to be called. There is evidence of the beginning of the formation of therapeutic bond with the perception of affection received during the accomplishment of the care. In addition, patients feel that their care needs have sometimes been met by the nurse. In communication, the understanding of messages is enhanced by trust in each other's speech and in the nurses' nonverbal language. Transaction is marked by the nurse's understanding of the needs that the treatment imposes on the patient with the establishment of mutual commitment and respect for each other's opinions, as well as the patient's confidence in the clinical evaluation of the health situation and in the procedures performed. However, despite this level of effectiveness in the interpersonal relationship, the nurse's touch causes tension in the patient.

to 59
High effectiveness At this level, greater effectiveness of interpersonal relationship in nursing care is demonstrated with the intensification of the characteristics mentioned in previous levels. Interaction is enhanced by the sporadic emergence of fun times during nursing care. The whole process of communication is demarcated by face-to-face interaction with high non-verbal language effectiveness in all encounters. In transaction, the patient perceives that the nurse puts him/herself in the patient's shoes to better understand him/her, besides realizing that he/she receives an individualized care. Interaction is marked by the feeling of companionship by the nurse care. In addition, the nurse's touch is received with tranquility by the patient. At this level, for the performance of the procedure, the nurse requests authorization.   On parameter b, the NCIRQ results show almost all the difficulty parameters (b) as negative. It is supposed that, since the behaviors contained in the items are inherent to the human interactions and to the daily life of the people receiving nursing care, and thus practiced by those involved in the process without great efforts, it has culminated with negative parameters b, in its large majority. A similar result was found in a study involving the construct "comfort of relatives of critically ill people" (25) .
In the case of TIF, a reliability measure of TRI (26) , the NCIRQ presented better reliability to measure interpersonal relationship in low to medium levels of effectiveness. This does not detract from its relevance, since low-effectiveness interpersonal relationships in care generate greater concern for nurses than those with high effectiveness.
According to the constructed scale, the evolution of the effectiveness of interpersonal relationship in nursing care in five levels was evident. At the first level, the communication process begins. The following levels demonstrate qualitatively the points of effectiveness of the interpersonal relationship with the involvement of behaviors related to the concepts of transaction and interaction, followed by role. This interpretation of levels is a characteristic of the TRI models, which enables the creation of a plan of care for the patient, according to their individual score (21) .
It is worth mentioning that interpersonal relationship is a basic tool of care in nursing and, therefore, is a fundamental skill for the performance of the entire professional activity. Seeing the interpersonal relation from the perspective of the production of a measurement technology is to give subsidy to the profession as a way of evaluating its daily behavior, allowing space so that the knowledge of situations generates improvement in the interaction between nurse and patient.
The components of interpersonal relationship become paramount in the development of care with a view to its humanization, pointing to the need for constant training of nurses involved in the care process, not only in technical procedures, but especially in their better qualification for the development of safe interpersonal relationships, learned as professional care tools (27) . In this sense, the measurement of interpersonal relationship can be used both to evaluate competence and to strengthen these skills in groups or individuals, as these can be improved with instruction and modified over time (28)(29) . In addition, this instrument can be used to improve understanding of the communication process.
A study that developed an instrument to measure communication was shown to be important in different situations and to provide guidelines for individual or group Borges JWP, Moreira TMM, Andrade DF.
intervention with the purpose of improving relationships and well-being in the context of health services, as well as to reflect on the theme in an educational manner (29) .
Regarding limitations, not all concepts of theory remained represented in the final instrument. In addition, it is necessary to formulate items that anchor at the upper levels of the scale, improving the measurement at these levels.

Conclusion
The NCIRQ was built on the framework of the