This study follows a psychometric methodological design.
Participants
The research population in this study included nursing students as well as staff nurses working in the Intensive Care Unit (ICU) and Oncology departments of teaching hospitals affiliated to Ahvaz Jundishapur University of Medical Sciences in Ahvaz, southwest of Iran. The participants were eligible to enter the study if they had at least 6 months of clinical work experience (for nurses in ICU) or were a final year bachelor's student or a student at a higher program in the field of nursing. Exclusion criteria were: unwillingness to continue participation in the research and not completing the questionnaire. The participants were selected from the research population through convenience sampling. According to the rule of thumb, the number of participants in the studies that conduct factor analysis is 10 participant per item [12]. Based on this, a total of 360 nurses and nursing students were invited and studied.
Data Collection
In order to collect data, a demographic and occupational profile questionnaire (age, gender, working or student status, work experience, educational level) as well as PCEP-GR questionnaire, which was designed in form of a 5-point Likert scale (completely disagree to completely agree) were used. Prior to data collection, written informed consent was obtained from the participants who were randomly selected thereafter, and their information was recorded anonymously. Also, permission to use the PCEP-GR tool was obtained from the tool developers.
Translation Process
At first, the translation of the initial English version of the tool was done in a forward-backward manner by two independent translators fluent in English. The translators were selected in such a way that one of them was fully familiar with the concepts and terms of medical sciences, while the other had no knowledge in this field but had a very good command of English. The two independent Persian translations obtained were examined and revised, and after merging them, a single Persian version of the PCEP-GR questionnaire (PPCEP-GR henceforth) was prepared. Finally, the final Persian version was translated into English and sent to the developer of the tool to be compared with the original version of the tool, and it was eventually approved.
Face Validity Of The Ppcep-gr
The face validity of the PPCEP-GR was measured by quantitative and qualitative methods.
In the quantitative validity analysis, 10 people from the target group (staff nurses and nursing students) were asked to express the importance of the instrument's items based on a 5-point Likert scale (from completely understandable to not understandable at all). Then, using the quantitative method of item impact, the scores of each item were calculated according to the following formula.
Impact Score = Frequency (%) × Comprehensiveness
Frequency is the number of people who gave 4 and 5 points to each item, and Comprehensiveness is the average score obtained from the responses of the participants to the above-mentioned Likert scale. A score higher than 1.5 was considered desirable for each item. The qualitative analysis of face validity was on items that had an impact score of less than 1.5, and it involved face-to-face interviews with the target group on item difficulty and relevance, and ambiguity in understanding the items [13].
Content validity
The content validity of the PPCEP-GR tool was assessed by quantitative and qualitative methods.
In the qualitative content validity evaluation phase, PPCEP-GR was given to 10 qualified experts (professors of nursing and experienced professors in end-of-life care). They were requested to evaluate the questionnaire after a qualitative review, and provide the necessary feedback based on criteria such as grammar, using appropriate words, placing items in their proper place, and giving appropriate points [13]. In this step, 11 items were corrected in terms of grammar and their position. Also, at this stage, the number of items rose to 38.
Quantitative content validity assessment was based on the opinion of 10 experts, and it involved two indexes of content validity ratio (CVR) and content validity index (CVI). In the CVR index, the necessity of an item was checked. The purpose of the content validity ratio is to select the most important and accurate content. For this purpose, the PPCEP-GR was given to 10 experts, and they were asked to examine and score each item based on a 3-point scale (1. Not necessary; 2. Useful but not necessary; 3. Necessary). Then, if the score obtained by the experts was greater than 0.62 according to the Lawshe table (to determine the minimum value of the index), it would indicate that the relevant item should be included in the questionnaire with a statistical significance level P < 0.05 [13].
The following formula was used to calculate CVR:
where nE is the number of experts who have considered the item in question as necessary, and N is the total number of experts.
CVI is calculated to ensure whether the items are designed in the best way to measure the constructs. For this purpose, in the current research, relevance of items was assessed using a 4-point Likert scale for each of the items (from not relevant to completely relevant) by an expert panel of ten members [14]. After the experts were consulted, Waltz & Bausell index was used to calculate CVI according to the following formula:
Ne is the number of experts who have chosen option 3 and 4, and N is the total number of experts. If the score of an item is more than 0.79, that item remains in the questionnaire. If the CVI score is between 0.70–0.79, the item is questionable and needs to be revised, and if the score is less than 0.70, the item is unacceptable and should be removed [13].
Construct Validity
In order to check the construct validity of PPCEP-GR, the first step involved extracting the number of latent factors based on exploratory factor analysis (EFA). To check the adequacy of sampling, the Kaiser-Meyer-Olkin (KMO) test and Bartlett test were performed. A KMO value more than 0.5 is acceptable, and KMO values between 0.7–0.8 and more than 0.9 are considered good and marvelous, respectively. Then extraction of latent factors was done based on the maximum likelihood ratio method, using varimax rotation and scree plot. The presence of an item in the factor was determined to be approximately 0.3 based on the following formula: CV = 5.152÷√(n-2). According to Steven, a valid latent variable is a factor that has at least 10 items with a loading of 0.4 when there are 150 research units. According to the three indicator rule, there should be at least 3 observed variables (items) for each latent variable. Commonalities with a value less than 0.5 were removed from EFA [14].
In the second step, confirmatory factor analysis (CFA) was performed based on the most common goodness-of-fit indicators, taking into account the accepted threshold and using maximum likelihood ratio. The assumption of normality was checked based on the skewness index of ± 3 and kurtosis of ± 7. According to Hooper et al. (2008), there is no gold rule for evaluating the fitness model, and it is smart to report several indicators [12]. Following Jaccard et al. (1996) and Meyers et al. (2005), we examined chi-square value (CMIN), Root Mean Squared Error of Approximation (RMSEA), and chi-square value to degree of freedom (CMIN/DF) [15].
Reliability
In order to evaluate the internal consistency of PPCEP-GR, Cronbach's alpha was calculated in two ways. First, alpha was calculated for all questions based on positively and negatively worded questions, and then it was calculated for each extracted factor. The obtained alpha was over 0.7, which indicated that internal consistency of the tool was acceptable [16]. Although the optimal value of alpha should be at least 0.9 according to some studies [17], some experts believe that an alpha value more than 0.9 is a sign of having too many questions, which should be reduced [18]. Then, by examining the variance ratio of observed variables to latent variables in confirmatory factor analysis, composite reliability and omega coefficient were calculated [19].