The questionnaire
The SAQ short form 2006 (available at https://med.uth.edu/chqs/surveys/safety-attitudes-and-safety-climate-questionnaire/) was translated independently into Italian by two native Italian translators. An Italian version with consensus on language was then issued through a reconciliation process and sent for back-translation. Following discussions between translators, an agreement was reached on the target language version; the back-translated version was then compared with the English one to make sure that the meanings were equivalent. A cognitive briefing was done with two risk-management nurses. The translation process followed World Health Organization (WHO) guidelines for linguistic validation of a questionnaire [20]. The Italian SAQ is found in additional file 1.
In the background information in the questionnaire, some participants’ qualifications were modified to reflect staff positions in Italian hospitals. The position “Environmental support (housekeeper)” was omitted because it was generally outsourced. “Physician Assistant/Nurse Practitioner” was translated as “Medico/infermiere tirocinante” which means doctor/nurse in training. “Attending/Staff Physician” and “Fellow Physician” were both translated as “Medico Specializzando”, a title given to physicians during their post-graduate specialization courses from the first to the last year.
Questionnaire administration and participants
The study was carried out at teaching hospitals of the University of Verona and the Healthcare Trust 20 of Verona (so-called “G. Fracastoro hospital”) in Veneto region in Italy in April-May 2011. The participants were permanent staff working in Geriatrics, Surgery, Internal medicine and Obstetrics departments. The reason of selecting only permanent staff was to make the results comparable between the two hospitals, as at teaching hospitals there are healthcare professionals in training at all stages and in all specializations, while at G. Fracastoro hospital there are few, if none, medical professionals in training.
An anonymous questionnaire was delivered either during mid-day shifts or weekly meetings. Two weeks were given to participants to complete the questionnaire. Upon completion, participants returned the questionnaires in envelopes to the chief nurses. In case of non-response, a reminder was made by the chief nurses or the department directors, and another week was given to responders. In total, 433 questionnaires were sent out in the two hospitals.
The retest was performed two weeks after the test, in Geriatrics, Obstetrics and Internal medicine departments at G. Fracastoro hospital with the same procedure. The retest group was informed in advance. Of note, 57 questionnaires were sent out for retest purpose.
A covering letter in Italian language with clear statements on the aims of the study, information confidentiality, voluntary participation and purpose of the retest, was provided to respondents. The information was clearly explained to them by researchers. The study was approved by Head of the Unit of Hygiene and Preventive Medicine at the University of Verona, and Medical Boards of the hospitals.
Hypothesized psychometric model of the questionnaire
The Italian SAQ contains 41 questions (or items) divided into seven safety attitudes factors which were: (i) Teamwork climate (perceived quality of collaboration among personnel), (ii) Safety climate (perceptions of a strong and proactive organizational commitment to safety), (iii) Job satisfaction (positive attitude about the work experience), (iv) Stress recognition (how stressors influence over performance), (v) Perceptions of hospital management, (vi) Perceptions of unit management and (vii) Working conditions (perceived quality of the work environment and logistical support). Items 2, 11 and 36 had reverse wording structure.
The hypothesized structural model was based on previous validation study in Norwegian language [9]. To the best of our knowledge, the structure suggested by the Norwegian team was the first validated structure of the SAQ short form 2006 ever published. This structure diverged to the one proposed by the Texas Center of Health Quality and Safety [21] on one specific point: the 29th item “The levels of staffing in this clinical area are sufficient to handle the number of patients” belongs to Working conditions factor instead of Perception of management. In fact, in the literature, this item was shown to be located in Working conditions in the SAQ ICU version [8], and in the Norwegian and Turkey SAQ short form version [9, 10]. Items 14, 33, 34, 35 and 36 did not belong to any safety factors. The model is described in Fig. 1.
Data analyses
The SAQ used a 5-point Likert-scale (1 = Disagree strongly, 2 = Disagree slightly, 3 = Neutral, 4 = Agree slightly and 5 = Agree). Missing and “Not Applicable” answers were coded separately. The scale was treated as interval in the analysis [22]. Test-retest reliability was evaluated by Pearson’s correlation coefficient between items and factors at two points in time.
The psychometric hypothesized structure of the Italian SAQ was evaluated by Confirmatory Factor Analysis (CFA). In details, the following indices were considered: (i) the chi-square goodness-of-fit: the model is acceptable if the p-value of chi-square is not significant, (ii) relative chi-square, which is the chi-square divided by degree of freedom (d.f): the ratio ranges from 3 to 1 is acceptable for a model fit [23]; (iii) Comparative Fit Index (CFI) should lie within 0.90–1.00 for a fit model; and (iv) Root Mean Square Error of Approximation (RMSEA): a value of about 0.05 or less would indicate a close fit of the model in relation to d.f [24]. The Polychoric correlations between latent safety factors were calculated by the CFA.
The internal consistency was represented by Cronbach’s alpha (cut-off = 0.70) [25]. The value of Cronbach’s alpha is maximized when all items within one factor measure the same construct. Besides Cronbach’s alpha, item-rest correlation that displays correlation between one item and a scale computed from the rest items in the factor was also calculated for each item.
Once the structure was confirmed, the mean scores of safety factors of each respondent were calculated both in Likert-scale and 0–100 point scale to see the percentage of respondents having positive attitudes toward each safety dimension (≥ 75), equivalent to 4–5 points on the Likert-scale, using the formula:
Mean score of a respondent = (mean of the factor items-1) × 25 [8].
A pair-wise criterion was used: if a respondent has more than two missing answers in a factor, their score is excluded from the score analysis for that factor [12]. The response options of reverse questions were converted into a positive scale before the mean of the factors was calculated. ANOVA test or Student’s t-test were performed to compare the mean scores of safety factors of different groups (hospitals, departments, working positions).
Analysis was performed in STATA InterCool 12.1 (Texas, TX, USA) and AMOS 19.0.0 (Meadville, PA, USA). P-value for significance is < 0.05.