A short form of the positions on nursing diagnosis scale : development and psychometric testing

Português / Inglês www.scielo.br/reeusp Received: 07/12/2012 Approved: 10/02/2012 590 RESUMO O Posi on on Nursing Diagnosis (PND) é uma escala que utiliza a técnica do diferencial semân co para medir a tudes em relação ao conceito diagnós co de enfermagem. O obje vo deste estudo foi desenvolver uma forma abreviada da versão em espanhol desta escala e avaliar suas propriedades psicométricas e efi ciência. Foi u lizada uma dupla abordagem teórico-empírica para obter uma forma curta do PND, o PND-7-SV, que era equivalente à original. Usando um desenho transversal através de pesquisa, foi avaliada a confi abilidade (consistência interna e confi abilidade teste-reteste), validade de constructo (análise fatorial exploratória, técnica de grupos conhecidos e validade discriminante) e de critério (validade concorrente), sensibilidade à mudança e efi ciência da PND-7-SV em uma amostra de 476 estudantes de enfermagem espanholas. Os resultados endossaram a u lidade do PND-7-SV para medir a tudes em relação ao diagnós co de enfermagem de maneira equivalente à forma completa da escala e em um curto tempo.


INTRODUCTION
The nursing diagnosis is the clinical judgment made by the nurse about the responses of the individual, family or community that provides the basis for the selec on of interven ons directed to achieving the results for which the nurse has responsibility (1) . Therefore, these results will be condi oned by the ability to iden fy the problems that create a need for care. A good ability in the diagnos c process is essen al for accurate detec on of the pa ent's problems (2) to encourage the provision of adequate care.
Based on the above, the acquisi on of knowledge and skills necessary to perform the diagnos c process acquires a vital importance in the forma on of future nurses; therefore, these competencies have become part of the curricula of the undergraduate nursing educa onal program (3) . However there are factors, external to the educa on, which may be determinants for students to put into prac ce the nursing diagnosis in their future professional prac ce. One of these factors would be a tudes. The a tudes that people have about the concepts play an important role in implemen ng behaviors related to these (4) . Therefore, the a tude toward nursing diagnosis is a major factor in their u liza on.
Despite this, more a en on has been paid to research on formative factors (5) than the a tudinal, and few studies have focused on knowing the students' a tudes toward the nursing diagnos c ac vity. The limited evidence available may be due to the diffi culty of measuring the a tudes, as they are latent variables inferred through the conduct or oral statements of people, so are not directly measurable (4) . One of the most widely used techniques for measurement is the use of instruments that allow inference on the a tudes of the subjects star ng from the responses that are given to a series of proposi ons.
The Posi ons on Nursing Diagnosis (PND) is a scale that was developed to measure the a tudes of nurses towards nursing diagnosis (6) . This instrument uses the seman c differen al technique of the words to measure and encourage the expression of people's a tudes towards concepts. The logic underlying this type of scale is that the ideas a person has about something can be broadly represented and communicated through adjec ves (7) . The seman c diff eren al measures primarily connota ve aspects of meaning, that is to say, the implica ons that the concept evaluated have for the person responding to the instrument. A tudes are composed of several indicators; therefore, a series of items will describe rela vely complex constructs be er than one alone. That is why, to measure a tudes, it is always be er to construct a scale than to be limited to a single ques on. In addi on, the separate analysis of the items can help to determine which aspects of the a tude highlighted in posi ve or nega ve sense, to be useful when designing interven ons directed at modifying it. The PND has been translated into several languages and validated in several interna onal samples, which have shown adequate psychometric properes (6,(8)(9) . However, the results of the tests of internal consistency suggest some redundancy in items (10) , with extremely high Cronbach alpha values in all valida on studies (6,(8)(9) . This emphasizes the desirability of elimina ng this redundancy, by elimina ng some items, developing an abbreviated form of the scale that maintains similar psychometric proper es to the full version and that, also, reduces the comple on mes. The use of brief ques onnaires is the most eff ec ve method for increasing the response rate and reducing poten al bias because of the loss of subjects in studies that u lize them (11) .
Taking into considera on the important role of the attudes of students in the implementa on of nursing diagnosis in their future professional prac ce and the benefi ts that its implementa on would be possible to infer, it would be worth the eff ort to develop and validate a version reduced of the PND to facilitate their study, reducing the me spent by respondents and enhancing therefore, its use in many research context. The abbreviated form of the scale might be useful for genera ng knowledge that serves as a star ng point for the implementa on of strategies to produce a change in a tudes that result in increased diagnos c ac vity of nurses.
The objec ves of this study were: 1) to develop an abbreviated form of the Spanish version of the Posi ons on Nursing Diagosis scale, and 2) to evaluate its psychometric proper es (reliability, validity and sensi vity to change) and its effi ciency in a sample of Spanish nursing students.

Item reduc on
To develop the reduced version of the PND, it was studied the psychometric proper es demonstrated by each of the 20 items of full scale in a previous study (9) , whose objec ve was the transcultural adapta on and valida on of the PND in a sample of 621 Spanish nurses.
Those that demonstrated a reduced psychometric performance were eliminated based on criteria that are shown in the following table (Table 1).
To check that the factorial structure of the retained items was analogous to that of the complete scale, an exploratory factor analysis (EFA) was performed, using principal component analysis (PCA) without rotation, on the data obtained in the study outlined above (9) . Additionally, the correlation between score of the complete scale, and the sum of the scores obtained for the retained items to verify the equivalence of both scales The use of brief questionnaires is the most effective method for increasing the response rate and reducing potential bias because of the loss of subjects in studies that utilize them. was evaluated. In addition, a committee of experts in nursing methodology reviewed the retained items to evaluate face and content validity as a new scale. The abbreviated form of the PND scale (Positions on Nusing Diagnosis-7-Spanish Version) was obtained from the above process, hereafter PND-7-SV.

Rev Esc Enferm
Chart 1 -Selection of the eliminated items, criteria utilized, theoretical foundation, analysis performed and statistical references

Design
The design used was a transversal observa onal validaon of a measurement scale.

Par cipants
An opportunistic sample of undergraduate nursing students was recruited from four Spanish universi es: University of Alicante, University of Cádiz, Valencia University and University of Granada. For the recruitment, the collabora on of all of the centers, departments and professors was requested.
It was considered that the sample should be made up of a minimum of 200 par cipants, the sample size recommended if you intend to perform factor analysis (12) . However, the eff ort was made to recruit the largest possible number of subjects, since the use of larger sample sizes reduces the sampling error and produces more stable factorial structures.

Instruments
The ques onnaire used for data collec on included: Sociodemographic and academic data. Informa on for demographic and academic data were collected on one sheet for the par cipants, including age, sex, university, course and degree of contact with the nursing diagnosis in clinical prac cum, rated on a 4-point Likert scale where 1 corresponded to nothing and 4 corresponded to a lot. Version (PND-7-SV). As discussed previously, the PND is a 20-item scale that uses the semantic differential method to assess attitudes toward nursing diagnosis (6) . Each item is composed of a pair of opposing adjectives that represent opposing properties of the diagnostic process. The order of the positive and negative descriptors is random to avoid acquiescence. In each item, two adjectives were joined by a line divided by 7 equidistant points. A score ranging from 1 to 7 is assigned to each one of these points, so that 1 corresponds with the point closest to the unfavorable adjective and 7 to the favorable adjective. The respondents must place a mark on the line, to coincide with the point that best expresses their feelings towards the concept. The scores obtained on each item are summed to obtain an overall score on the scale, which can vary between 20 and 140, with the attitude rated at the highest value being more positive with to the concept. A score of 80 corresponds to a neutral stance, which is equivalent to obtaining a mean of 4 points on the 20 items. The PND original and its adaptation to Brazilian Portuguese and Spanish, have shown adequate psychometric properties in different samples that have been evaluated (6,(8)(9). For this study the Spanish version of the instrument, the PND-SV, was used (9) . The PND-7-SV is an abbreviated form of PND-SV composed of 7 items that were developed based on the previously described method.
Comple on me. The par cipants were asked to esmate the me invested in comple ng the complete and abbreviated versions of the scale.

Data Collec on
Par cipants completed an electronic version of the instrument that fully respected the structure and layout of the original scale in paper form. Through the virtual campus, a web pla orm to support teaching, invita onal messages were sent to par cipate in the study containing a le er, informing about the purposes of this study, along with a hyperlink that led to the collec on notebook of data. The student accessed the message through his/her personal area in the virtual campus, which has password protected access. To assess the stability of the scores a second link to the scale was sent in a two week interval to those parcipants who agreed, in this case, by email. All cases with incomplete data were eliminated. Collec ng these lasted from April to May 2012.

Ethical considera ons
The Commi ee on Ethics in Research in Healthcare reviewed and approved the research (Record N. 12/09). Par cipants were guaranteed confi den ality and anonymity. The return of the completed ques onnaire was considered consent to par cipate. All necessary permits from the facul es of nursing to distribute the ques onnaires and from the authors for the reproduc on of the original scale were obtained.

Sta s cal analysis
Normality of the variables was examined using the Kolmogorov-Smirnov-Lilliefors test. Using descrip ve sta s cs, sociodemographic and academic variables were summarized, as well as the distribu on of the total scores and each item of the scale. The evalua on of reliability included tests of internal consistency and test-retest reliability. Internal consistency was assessed by calcula ng the Cronbach alpha coeffi cient for the en re scale and when each item was removed, as well as through the corrected item-total correlation by means of the Pearson's r test. Alpha coeffi cients > 0.80 were considered desirable while they did not exceed 0.90, indicating again the existence of redundancy among the items of the shortened form (13) . Addi onally, correla ons ≥ 0.40 for each item with the total score of the scale, when the item was removed, were considered adequate (14) . The test-retest reliability was assessed by correla ng the scores obtained by 101 par cipants between two administra ons of the scale separated by an interval of two weeks. For this the intraclass correla on coeffi cient (ICC) for the total and for each item were calculated. Intraclass correla on coeffi cient values ≥ 0.70 were considered to denote good stability of the scores (7) .
The construct validity, through the EFA and the technique of known groups, discriminant validity and criterion validity in terms of concurrent validity were evaluated. Through the EFA an a empt was made to ascertain whether the reduced version of the scale maintained the unidimensionality of the complete instrument (6,9) . To do this, the factorial structure of of the PNS-7-SV and the factor loadings of the items through a PCA without rota on was explored. To determine the number of factors to be extracted those that presented eigenvalues ≥ 1.00 were selected, and were posi oned above the elbow in the scree plot (12) . It was considered that the items with factor loadings ≥ 0.40 properly defi ned the factor (15) . Factoriability of the data was previously evaluated by means of the the Kaiser-Meyer-Olkin test (KMO) and Bartle test of sphericity (13) . Based on the known group technique, the scores in the scale of the par cipants that declared, in a general way, in favor of the nursing diagnosis were compared with those who showed they were against or indiff erent to the concept. The hypothesis was that the fi rst would obtain signifi cantly dis nct scores on the scale than the la er. To characterize both groups, the responses to the statement I am in favor of nursing diagnosis as a concept were u lized. To check this hypothesis, the Mann-Whitney test was performed. The discriminant validity was evaluated by means of the analysis of the receiver opera ng characteris c (ROC) curve to determine the ability of the scale to detect correctly among par cipants who belonged to one or the other of the groups described above. The area under the curve (AUC), the standard error (SE) and the confi dence interval at 95% (CI 95%) were calculated. The AUC allows the assessment of the ability of the scale to dis nguish between those who have a posi ve a tude of those who have a nega ve a tude or neutral based on their statements. An AUC of 0.80 to 0.90 was considered indica ve of a good precision (16) . Concurrent validity was es mated by evalua ng the concordance between the scores of the reduced and complete forms of the scale, which were determined through the Spearman´s Rho test. Correla ons > 0.75 were considered as proof of excellent concordance (17) .
The sensi vity to change of the PND-7-SV was evaluated in 25 students who received an interven on in the form of theore cal and prac cal educa onal ac vity about the nursing methodology, based on the resolu on of clinical cases by conduc ng care plans and imparted by a professor specializing in nursing diagnosis with extensive teaching and research experience in the fi eld. Previous studies (8,18) associate the use of nursing diagnosis and educa on in the subject with posi ve a tudes towards the concept. The ability of the scale to detect changes in the scores in the same subject, between the situaon before and a er comple ng the educa onal ac vity, was evaluated through the paired t-test, by checking the hypothesis of changes in a tudes. This sta s c is focused solely on the signifi cance of the observed change, so it was considered desirable to also use coeffi cients in which both the magnitude of the change and its variability intervene. For that reason, the standardized mean response (SMR) was also calculated, by dividing the mean change in scores between the standard devia on (SD), which refl ects the variability of change. Values of 0.20, 0.50 and 0.80 indicate, respec vely, instruments with limited, moderate and high sensi vity to change (19) .
The effi ciency of the PND-7-SV was evaluated by calcula ng the reduc on in the me of implementa on with respect to the complete scale.
Values of p<0.05 were considered sta s cally significant. All data were analyzed using SPSS version 17 (SPSS Inc., Chicago).

Scale reduc on
Thirteen items of the complete scale complied with at least one of the criteria for exclusion, so they were eliminated (Table 1). The remaining seven items were selected to be retained, forming the proposed abbreviated version of the scale (Table 2). This was subjected to analysis to assess its factorial structure and the correla on of their scores with those of the complete form. The PCA revealed that the shortened version consisted of a single factor, equivalent to the full scale (6,9) , which explained 66.2% of the variance with high factor loadings of all items on the same (0.70 to 0.86). An extremely high correla on was also found between the score on the complete scale and the sum of the scores of the seven retained items (r s =0.97; p <0.001). Based on the posive results obtained by the proposal for a shortened form in the previous analysis, it was reviewed by a commi ee of experts in nursing methodology who determined that it seemed to adequately measure the construct and the items that conformed were representa ve and contemplated the primordial aspects of it.

Sample descrip on
Five hundred and sixty-three ques onnaires were sent obtaining a response rate of 84.5%, represen ng a fi nal sample of 476 students. The mean age of par cipants was 23.46 years (SD: 7.45) with 74.6% of them being women. In the sample, 68.1% acknowledged having li le or no contact with the nursing diagnosis in clinical prac cum performed in health ins tu ons as part of their educa on. Table 1 shows in detail the above data with the distribu on of par cipants per course and university.

Distribu on of scores
The mean score of the scale for the sample was 36.43 (SD: 7.76). The range of scores observed was between 7 and 49 points, coinciding with the minimum and maximum that theore cally could be obtained in the scale. All of the items received high mean scores, shown along with their SD in Table 2.

Reliability
Internal consistency. The Cronbach's alpha coeffi cient was 0.888. Table 2 shows the values of alpha when each item was eliminated and the corrected item-total correlaons. High correla ons were observed between the scores of all items with the total scale, except for item 7 that correlated in a signifi cantly lower manner than the rest.
Test-retest reliability. The ICC for the total score of the scale reached the value of 0.92 (95% CI=0.88 -0.95, n=101) evaluated between the fi rst and second administra on of the PND-7-SV. The ICC for each one of the items ranged between 0.74 and 0.90.

Construct validity
Exploratory factor analysis. The signifi cance of the test of sphericity (X 2 =1797.24, df=21, p<0.001) and the the size of the measure of sampling adequacy (KMO=0.91) revealed a common variance of items suitable for performing the EFA. The PCA revealed the presence of a unique factor with an eigenvalue greater than 1.00 (4.28), which explained 61.1% of the variance. Table 2 shows the elevated factor loadings of the items on said factor.
Discriminant validity. Figure 1 shows the ROC curve obtained for the PND-7-SV. The AUC reached a value of 0.89 (SE=0.02; 95% CI=0.84 -0.93) it was, also, signifi cantly dis nct from that obtained the condi on of zero discriminaon (p<0.05). The cutoff point of 30 points had a sensi vity of 86.1% and a specifi city of 72.9%.

Criterion validity
Concurrent validity. The reduced form correlated in an excellent manner and highly signifi cantly with the complete version of the scale (r s =0.93; p<0.001).

Sensi vity to change
The diff erence in mean score on the scale between baseline and the the comple on of the educa onal ac vity reached a value of 5.68 (SD=4.97; 95% CI=3.63 -7.73; n=25). It was also highly signifi cant (t=5.71; p<0.001). The SMR for the scale was 1.14 (95% CI=0694 -1589).

Effi ciency
The me of comple on of the abbreviated form (=1.67 minutes; SD=1.14) was reduced by 2.08 minutes on average over the complete form (=3.75 minutes, SD=2.84).

DISCUSSION
The development of the PND-7-SV u lized a dual approach, empirical and theore cal. The reduc on of those items sought to maintain better psychometric properes demonstrated in suffi cient number to appropriately represent the construct. The seven items that were retained, based on sta s cal performance, were reviewed by a commi ee of experts who confi rmed the suitability of the set as an independent instrument.
The elevated internal consistency and repeatability demonstrated by the PND-7-SV in this study supported the reliability of the instrument. The results obtained in this sense are comparable to those described for the complete scale, both in its original version (6) and its adapta ons to Spanish (9) and Brazilian Portuguese (8) . The  alpha coeffi cient value of the PND-7-SV was situated at the range that indicates an adequate consistency of the items without any redundancy. All items showed a high correla on with the total score of the scale and decreased alpha coeffi cient if they were removed, except for item 7 (rou ne-crea ve) that showed an item-total correla on slightly lower than 0.40 and an increase in alpha above 0.90 if it was excluded. Although, in terms of reliability, it can be stated that this item had a lower performance than others, this might be due to the characteris cs of the sample on which this study was conducted. One of the adjec ves that compose the item 7 is rou ne, which refers to the quality of an ac vity that is conducted out of custom, by mere habit and without ra onale. For an act to convert to a custom it is required to be repeated on an ongoing basis, however, the vast majority of par cipants acknowledged having li le or no contact with the nursing diagnosis in clinical prac cum. The fact that the nursing diagnosis is not an habitual ac vity among students who par cipated in the study may be a distor ng factor that could have infl uenced the results obtained by this item in reliability tes ng. The item is considered to achieve a be er performance in a future psychometric valida on of the PND-7-SV in graduate nurses because the diagnos c ac vity is part of their daily prac ce. The data with respect to the test-retest reliability of the scale support the stability of their scores in a two-week interval, and is also comparable to that of the complete form (6,9) .
Adequate construct, discriminant and criterion validity demonstrated by the PND-7-SV in this study endorsed its general validity. The EFA results obtained in this study are similar to those obtained for the complete scale, both in its original version (6) and in the Spanish (9) . The PND-7-SV is composed of a unique factor that explains a signifi cant variance with high factor loadings of all items. The results derived from the technique of known groups supported the usefulness of the abbreviated scale to dis nguish subjects with a favorable a tude towards the concept from those with an unfavorable or neutral a tude. Similarly, the PND-7-SV demonstrated an AUC indica ve of good accuracy for discrimina ng correctly between the groups of a tude and 30 points seems to be the cutoff point with a sensi vity and specifi city adequate for that purpose. The score achieved by the par cipants in the abbreviated form correlated with that obtained in the complete scale, which supports their equivalence.
The sensi vity to change of a scale is the degree to which one obtains diff erent results in repeated applicaons of the instrument when there has been a real change in the phenomenon it purports to measure (20) . The PND-7-SV demonstrated excellent detec on of relevant changes in a tudes mediated by an interven on implemented with that objec ve. The change in the a tude detected by the scale in the students who a ended the educa onal ac vity, supports the results of previous studies (8,18) that found an associa on between having received educa on in nursing diagnosis and having a more favorable a tude towards it.
The reduc on in the me of applica on of the abbreviated form compared to the complete form facilitates the use of the scale in research contexts in which there is a limited me and it may reduce the losses.
The previous findings presented some potential limita ons. The item reduc on was based on classical psychometric proper es, with possible approaches from diff erent theore cal perspec ves. The use of a convenience sample limits the generalizability of the results, so addi onal studies are needed with diff erent representa ve samples to strengthen them. Despite the eff ort made to reproduce the original format of the scale in the electronic version used in this study, this can be considered as a poten al confounding factor when interpre ng the results. However, although there are diff erent views on the eff ect of the format of the ques onnaires on the measurement error, there is not suffi cient empirical evidence in either direc on (21) . For its technical diffi culty the order of administra on of long and short scales is not balanced, despite the general recommenda on to do so from one par cipant to another to avoid the order eff ect (22) . The absence of other measures of the same construct with proven evidence of their validity, that may have been used as the gold standard, forced to resort to the self-assessment of the par cipants about their general a tude to the concept as a reference criterion, for both the technique of known groups as for discriminant validity tests.
We have iden fi ed several poten al uses of the scale in research. The PND-7-SV can be useful to iden fy the students' a tudes toward nursing diagnosis and to determine whether there is an associa on between them and academic factors, as they could be learning styles. Likewise, the scale would be useful as a measurement of the change response and, therefore, the eff ec veness of mo va onal interven ons. In that sense it could help provide evidence to support the eff ec veness of educaon in changing a tudes and determine which teaching methods are most eff ec ve in achieving this objec ve. The PND-7-SV, developed based on the Spanish version of the PND, could be validated in other languages in which the instrument is available. Furthermore, by making linguis c adjustments and comple ng the per nent studies, the PND-7-SV could be used in other Spanish-speaking countries. Finally, addi onal studies are needed aimed at assessing its psychometric proper es in nursing professionals to enable their use in this collec ve.

CONCLUSION
The results of this study support the reliability, validity and sensi vity of the change of the PND-7-SV for the measure of a tudes towards the concept of nursing diagnosis in Spanish nursing students in a manner equivalent to the complete form of the scale and in a shorter me.