Item generation
The initial questionnaire was developed through semi-structured interviews with nursing professionals. Participants were asked to explain about job-related low back pain and indicate the ways that could prevent low back pain in nursing personnel. The interviews were carried out by one of us (SSK) individually in the office of participants or the nursing office and in a quiet environment. Interviews continued until data saturation was achieved. As such 18 nurses, head nurses, educational supervisors, nursing managers, hospital managers, hospital affairs experts, quality improvement experts, and development and support managers were interviewed. For further certainty, two more interviews were conducted, but no new data was obtained. Accordingly, based on the interviews and literature review, the preliminary questionnaire containing 49 items was developed. Then the research team assessed the questionnaire and 9 items were deleted to ensure the clarity, objectivity and the ease of comprehension of the terms employed. Thus at this stage in all 38 items was remained.
Developing a preliminary questionnaire
At this stage, we assessed content and face validity. To determine the content validity, qualitative and quantitative methods were used (10). In the qualitative content method, 10 experts in health education, physiotherapy, educational management, and nursing were asked to check the quality of the questionnaire based on grammar, use of proper words, item allocation and scaling. Then the content validity ratio (CVR) and content validity index (CVI) was used to evaluate quantitative content validity. To determine the content validity ratio, the questionnaire was sent to 10 experts with different specialty (including health education, physiotherapy, nursing, education management) and was asked to respond where the items are essential, useful but not essential and no essential. Then responses were calculated based on the formula and matched to the Lawshe’s table (11). The same experts were asked to indicate whether items are simple, relevant and clear in order to calculate content validity index (12, 13). The CVR for the questionnaire was 0.84, which was well above the recommended value. The CVI for the questionnaire was 0.98. However, at this stage, 7 items were removed and the number of items was reduced to 31. Similarly, for face validity, qualitative and quantitative methods were used. The qualitative face validity determined by asking 10 nurses (the potential user of the questionnaire) to complete the questionnaire and to indicate whether any items were difficult, ambiguous or irrelevant. In the next step, Item Impact Score was calculated in order to assess the importance of each item indicated by the same nurses. For each item, a 5-point Likert scale was considered: strongly important, quite a lot important, moderately important, slightly important and not at all. The analysis showed that the impact score for all items was above 1.5 (the actual mean impact score was 3.73). There were no additional changes at this stage and thus the provisional version of the questionnaire with 31 items was subjected to psychometric evaluation (Additional file 1, and 2, 3).
Psychometric evaluation
A cross-sectional study was conducted to examine the psychometric properties of the questionnaire. The questionnaire was administered to a sample of nursing personnel recruited from teaching hospitals affiliated to Mazandaran University of Medical Sciences, Sari, Iran. Indeed, first we provided a list of all hospitals and three hospitals were randomly selected. Then in each hospital, individuals randomly selected based on a simple random number table and staff ID.
Statistical analysis
To assess construct validity, the exploratory factor analysis and item-scale correlation matrix were used. Indeed, the Kaiser-Meyer-Olkin (KMO) Index and Bartlett’s Test of sphericity were used to assess sampling adequacy (14). The factor structure of the questionnaire was extracted using Varimax rotation. The presence of an item in a factor was determined as approximately 0.4 as recommended (15). For item-scale correlation matrix, the Pearson correlation coefficient was used and coefficient values of 0.4 or above were considered acceptable (15).
The reliability of the questionnaire was assessed using internal consistency and stability. To compute the internal consistency, the Cronbach's alpha coefficient was estimated. To demonstrate the stability of the questionnaire the Intraclass Correlation Coefficient (ICC) was calculated. For this purpose, a subsample of 20 nurses completed the questionnaire twice with a two week interval.