Healthcare, while considered safe, is a high-risk industry in which medical mistakes are likely to happen quite often. In European Union countries, the rate of adverse events that accompany healthcare recipients is around 8–12 % [1], and 50 % of them are preventable just by improving the culture for safety in hospitals [2].

Healthcare in Italy is provided by two types of trust: hospital trust and healthcare trust. Hospital trust is a tertiary care setting, which normally is teaching hospital. Healthcare trust, which is an integration of healthcare and social services, guarantees secondary acute care, primary and territorial care for residents in their territory. The national movement for patient safety started in 2004, with the publication of the report “Risk Management in Healthcare: The problem of errors” by the Ministry of Health. At the time, Italian hospitals were only at the beginning of their journey towards establishing patient safety as a basic, systematic and continuous care process. In January 2007, the Italian Ministry of Health established the National system for Patient safety, purporting to “build and foster a patient safety culture” [3] in the sense that hospital management boards and clinical risk managers need to monitor the safety performance, promote best practice, and have new methods to improve patient safety [4].

Surveying health workers using psychometric questionnaires is a way to investigate the existing safety culture in health facilities. To standardize this approach, a wide variety of psychometric questionnaires has been developed and published [5, 6]. These questionnaires often contain questions regarding several aspects of safety culture. Based on this information, healthcare managers could set up programs for their improvement [7].

The Safety Attitudes Questionnaire (SAQ) is a self-reported psychometric questionnaire developed to measure safety attitudes of front-line workers. In recent years, it has been the most commonly used questionnaire to measure safety culture [8]. The European Network for Patient Safety has recommended the SAQ as one of the three effective tools (alongside the Hospital Survey on Patient Safety Culture and the Manchester Patient Safety Assessment Framework) to be used in patient safety research [6]. It has been cross-culturally validated in different languages, including English [8], Norwegian [9], Turkish [10], Dutch [11], Chinese [12], Swedish [13], German [14], Portuguese [15], Arabic [16]. All these studies have showed that the SAQ possesses good psychometric properties in different languages. Moreover, when the questionnaire is externally validated, high scores of safety factors of the SAQ have demonstrated improvements of safety environment, such as lower rates of medical errors and inpatient mortality, shorter length of hospitalization [17, 18] or better working conditions, like lower rates of nurse turnover [8, 19].

So far, no evidence of the use of the SAQ in Italian language has been reported. Hence, this study is aimed at validating the Italian SAQ short form 2006 by evaluating its test–retest reliability, internal consistency, and the goodness-of-fit of the psychometric properties.

The study aims at cross-culturally validating the Italian version of the SAQ in two hospitals in Italy. Based on literature search, this is the first-ever study to do that. Response rate was 60 %, which is relatively lower than the international benchmarking of response rate (68 %) [8], or that of other validation studies [9, 10]. Overall, the Italian SAQ has shown satisfactory data on validity, displays a good internal consistency and a moderate correlation of test-retest reliability. The findings are comparable with the results of international benchmarking data [8], and other relevant studies using the SAQ in different languages [9, 10]. In line with previous SAQ validation studies, our study has supported the validity of the SAQ in Italian in measuring patient safety in hospitals.

Construct validity of the Italian SAQ, based on the goodness-of-fit indices, was not absolute but acceptable (p < 0.001). The p-value of the whole model is critical because it tends to accept complex models with many parameters and it disregards effect of sample size. On the other hand, other goodness-of-fit indices supported the fitness of the model (relative chi-square = 1.739, RMSEA = 0.05, CFI = 0.90). Cronbach’s alpha coefficients of all factors and items were within 0.70–0.90, confirming a very good internal consistency of the Italian SAQ [25]. The items were well located inside the safety factors that they are supposed to measure.

Correlation between latent factors achieved good values in our model, proving that they were well correlated with each other. All the correlations (except for Stress recognition) were more than 0.60, exceeding the international benchmarking for within-area correlations. The highest correlation observed between Teamwork climate and Safety climate (0.95, p < 0.001) was consistent with the results of the study by Sexton et al. [8].

Physicians preferred the items in normal wording to the reverse one. In our study, the reverse items had lower factor loadings than others, and the correlation of test-retest was negative (item 11th). Reverse wording is a method to control the agreement of responses so that responders must read carefully the questions before providing the answers [26]. However, studies have found that they may cause an unfavorable effect on psychometric properties [27], and measurement problems for the questionnaire [26]. Meanwhile items in normal wording structure could have higher correlation in test-retest reliability and higher internal consistency [28]. In our opinion, in order to keep the generality of the SAQ in different languages, reversed items should be retained as they are in the English SAQ. A brief reminder to participants to read carefully before each and every item is suggested to avoid unexpected effects of reverse wording.

The high rate of missing answers in Perceptions of hospital management and Perceptions of unit Management (7–11 %) was similar to the results of other studies using the SAQ in European countries [29]. The item with lowest percentage of positive answer (item 24th) also belongs to this safety factor. The publication of the SAQ international benchmarking likewise had the lowest mean score in Perceptions of management among all the safety culture factors [8]. In general, the existing hierarchical structure, i.e. the top-down management model, in hospitals in Europe could prevent unit staff from speaking out the issues or discussing them with the management [30]. Moreover, another possible explanation is the vague role of hospital managers to unit staff when the managers always work distantly. Thus, a “code of silence” is created in which the front-line staff cannot comment on perception of hospital management, safety events are not properly reported and risks are underestimated. This, in turn, affects patient safety in hospitals.

Working positions contributed significantly to difference in safety attitudes in hospitals (except for Stress recognition) in the present study. Low mean scores of six dimensions belonged to nurses. In general, nurses who spend more time in contact with patients suffer more stress. Meanwhile physicians who work more independently are more recognized by their fellows, and chief nurses have their own leadership and empowerment in the wards. Different studies in different medical settings and countries also exhibit different results. For example, mean scores in Working conditions and Perceptions of management were found to be lower in nurse compared with the scores in physicians in ICUs [18]. On the contrary, in the ambulatory setting in the United States, Modak et al. [31] found that attitude scores in Teamwork climate, Safety climate, Job satisfaction and Working conditions were similar in all health providers [31]. And Etchegaray et al. [32] found no difference of the mean scores in Job satisfaction between nurses and physicians in ICUs [32].

Stress recognition is the only factor in the SAQ that is put in terms of self-behavior of the respondents. The highest mean score of Stress recognition shows evidence of the awareness of workers about the effect of stressors on their performance [33]. However, this factor has been disputed since stress is a weakening factor to safety practice, and it has negative correlation with other safety factors. In the present study, in line with other studies, Stress recognition has negative correlation with other factors, presenting that it goes against safety practice in hospitals. The measurement of stress in this case does not contribute positively towards safety climate as it is supposed to [34], and is even removed from the SAQ Chinese final version [12].

The study findings are limited by the size and representativeness of the sample. The sample size was limited due to time constraints of the study. A convenience sampling was used for this collaborative study in order to define target departments in participating hospitals. Several methods were implemented to obtain a high response rate at G. Fracastoro hospital (71.1 %); for example, the questionnaires were delivered during mid-day shifts or physicians weekly meetings. Meanwhile, at teaching hospitals there were no such possibilities, making their response rate relatively lower (41.5 %) and no information was available for non-responders. Finally, external validation was not covered in the frame of the study; therefore no clinical relevance with the SAQ results could be obtained. While the external validation requires more hospitals participating [9], studies in the future should examine it with larger sample sizes.

The Italian short form SAQ, when being administrated in one tertiary care acute hospital and one secondary care hospital in Italy with different safety cultures and different patient safety pathways and stories, had a moderate correlation of the test-retest reliability, acceptable goodness-of-fit, high internal consistency within factors, and high correlation between factors. The study found similarities and differences with previous validation studies of the SAQ in the world, which constituted a good evidence of questionnaire validation. The fact that the hospitals did not differ in their answers respect to the SAQ dimensions, except for Perception of unit management, has confirmed that the Italian version of the SAQ is suitable to be used in diverse hospital settings and a valid tool for measuring safety culture in Italian hospitals. External validation study using the Italian SAQ should be performed when it is possible.