2.1. Design
This is a sequential exploratory mixed method study, conducted 2016-2019. It is part of a larger scale study that used a methodological approach in instrument development and psychometric validation.
2.2. Participants
A total of 22 articles obtained from an extensive review of the literature were initially analyzed. Thereafter, during a fieldwork phase, 13 participants including one member of a nursing board in Ministry of Health (MOH), three intermediary managers of medical universities in Tehran city, and nine faculty members of nursing schools were interviewed. They were selected with a purposive sampling method designed to ensure maximal variety in age, gender, specialty, and work experience in higher education centers [7, 14]. For the quantitative psychometric validation of the instrument, 10 managers, experts, and faculty members of the nursing schools participated in the face validation, 10 in the content validation, 30 in the primary reliability establishment and 194 faculty members in the construct validation and final reliability processes.
For construct validation, at least three participants were required for each item in the quantitative stage. 250 questionnaires were distributed to faculty members of nursing schools five of the biggest or highest schools in our ranking system selected with a convenient sampling method. The inclusion criteria were voluntary participation and at least two-year work experience as full-time faculty member in the school. Of these 250 questionnaires, 194 completed questionnaires were returned and these were analyzed for construct validation (response rate=78%).
2.3. Data collection
Concept analysis and determination of final operational definition of context-based shared governance using hybrid concept analysis was done [7]. To extract items from the available literature in the theoretical stage, the related papers were searched with the keywords “tool/instrument, faculty member, higher education, and shared governance” using the databases “Google Scholar, PubMed, Science Direct, and Eric” in the time interval 1990-2017. Simultaneously, Persian databases like MedLib, Scientific Information Database (SID), Magiran, Iran Doc, and Iran Medex were searched with Persian keywords for shared governance: these searches did not return a single related paper. Of 349 papers retrieved from the initial search, 22 articles related to exact research goals and related to shared governance instruments and guidelines. In the fieldwork phase, the first author performed semi-structured deep interviews with 13 participants to extract the intended items [7, 14]. After analyzing the data obtained from the theoretical stage and fieldwork phase, integrating the findings and extracting codes and categories using inductive-deductive approach, the items pool, consisting of 150 items including characteristics, antecedents, and consequences of the concept analysis stage, entered psychometric validation stage.
The instrument underwent quantitative and qualitative face and content validation by 10 faculty members; then, it underwent primary reliability establishment and item analysis and some items had been changed or omitted. Then, the questionnaire was distributed to faculty members with 70 items.
2.4. Ethical considerations
The Ethics Committee of Shahid Beheshti University of Medical Sciences approved the research (code: IR.SBMU.REC.2016.84). The participants joined the study under voluntary conditions and could leave the study at any stage. The researcher elucidated the research goals and procedures to the participants assured that principles of anonymity and confidentiality of information would be observed.
Data analysis
Data were analyzed by the method of Schwartz-Barcott & Kim [15]. Conventional content analysis by Graneheim & Lundman was used to determine the concepts and to extract codes and categories [16]. The full text of each paper in the theoretical stage and each interview in the field-work phase were considered as an analysis unit [17]. Then, each article was read several times to arrive at an agreed general content. The primary codes were then extracted as explicit and implicit concepts. In the next stage, similar codes (obtained from explicit and implicit concepts) were classified as subcategories, which in turn, were put into a group. The categories were subsequently labeled [16]. In the final stage, the data obtained from the theoretical stage and fieldwork phase were merged and the item pool was obtained. In the third part of the study, the methodological approach was applied to determine the psychometric properties of the first version of the instrument with 150 primary items. Having applied the corrections in the qualitative stage, quantitative face validity was established by measuring item impact. To make sure that the items measured the intended construct, content validation was done both quantitatively and qualitatively. To explore CVR, 10 experts in nursing management and instrument development were asked to assess the necessity of each item. To survey CVI, the experts were asked to express their opinions about the rate of relatedness of each item with the intended construct [18]. Regarding raters consensus on item relatedness, the modified Kappa 1 statistic [19] was used; this provides instrument developers with information on the degree of consensus without chance ratio. A corrected Kappa statistic greater than 0.74 was rendered as excellent, between 0.6 and 0.74 as good, and smaller than 0.6 as poor [20]. Colton & Covert (2007) mentioned item analysis (IA) as one way of construct validity assessment; to examine construct validity, the correlation between each item and other items and the whole instrument was performed. To investigate the primary reliability of the instrument before validation, the internal consistency coefficient (Cronbach’s alpha) was used. In this study, factor analysis in construct validation [18] and exploratory factor analysis (EFA) with maximum likelihood by the use of oblique rotation of the Promax type was utilized to determine the degree to which the developed instrument measures the concept of shared governance. Also, EFA was used to determine the relations among the latent and observed variables and then the significance and severity of these relations [21]. By doing this, the three criteria of Kaiser’s criterion, Scree plot and cumulative variance percentage determined by extracted factors were used. The KMO statistic was estimated to test sample volume sufficiency in which 0.8 and greater was rendered as suitable [21]. Moreover, scree plot was plotted to determine the number of factors. Lying of factors out of the formed horizontal line was rendered as reference [22].
In this study, three samples per item were rendered as sufficient in factor analysis [23]. The study sample consisted of faculty members at nursing schools of major Iranian medical universities selected based on inclusion criteria as mentioned before. The developed instrument was completed via self-report. Considering that the instrument contained 70 items, given the ratio of 3 samples per item, completion of at least 210 questionnaires was necessary{Plichta, 2013 #20}. On this basis, considering a probable rate of deficient or unreturned questionnaires, 250 written questionnaires were distributed to faculty members of nursing schools, of which 56 questionnaires were not returned. Hence, 194 completed questionnaires (response rate 78%) were collected. SPSS20 was used in data analysis. Finally, the reliability of the instrument was established using both internal consistency and stability. To estimate internal consistency, Cronbach’s alpha and conventional odd/even split-half reliability were used. Additionally, test-retest reliability and intra-class correlation coefficient with 2-week interval was used on 30 faculty members to examine the relative consistency. Estimation of ICC is used to assess the consistency of the intended measured variable by the use of an instrument for similar individuals in two different situations. A 2-week to 1-month interval between two tests is suitable [18]. In this study, 30 participants were used to determine consistency of the developed instrument so that they completed the final instrument twice with a 2-week interval. Koo & Li have rendered ICC less than 0.5 as weak, between 0.5 and 0.75 as moderate, between 0.75 and 0.90 as good, and greater than 0.90 as excellent consistency [24] . Ultimately, SEM was used to determine the accuracy of test measurement. SEM is one of indices of measurement accuracy that shows an estimation of acceptable expected deviation from real values in a group of measurements in a specific condition, i.e., SD of scores distribution [18]. Moreover, the weight and significance of each item in each factor were determined by the use of results of factor analysis and loading of each item onto that factor [25]. The following formula briefly presents the estimation method:
WIij=Wfi x PfI
PfI=factor loading of jth item in ith factor
WIij=weight of jth item in ith factor
Scoring of instruments is possible on the basis of Likert rules and/or linear transformation method so that the choices open to each item ranged from very much =5 to very little =1. To determine whether weighting created any changes in the items of the instrument, the mean rank of each item was estimated by paired t-test before and after weighting. Then, the hierarchical position of the item in the instrument was determined in two states based on these mean ranks [25].