Background The purpose of this study was to explore the variability in safety culture dimensions within and between Swiss and US clinical areas.
Methods Cross-sectional design. The 30-item Safety Attitudes Questionnaire (SAQ) was distributed in 2009 to clinicians involved in direct patient care in medical and surgical units of two Swiss and 10 US hospitals. At the unit level, results were calculated as the percentage of respondents within a unit who reported positive perceptions. MANOVA and ANOVA were used to test for differences between and within US and Swiss hospital units.
Results In total, 1370 clinicians from 54 hospital units responded (response rate 84%), including 1273 nurses and 97 physicians. In Swiss hospital units, three SAQ dimensions were lower (safety climate, p=0.024; stress recognition, p<0.001; and perceptions of management, p<0.001) compared with US hospital units.
There was significant variability in four out of six SAQ dimensions (teamwork climate, safety climate, job satisfaction and perceptions of unit management) (p<0.001). Moreover, intraclass correlations indicate that these four dimensions vary more at the unit level than hospital level, whereas stress recognition and working conditions vary more at the hospital level.
Conclusions The authors found differences in SAQ dimensions at the country, hospital and unit levels. The general emphases placed on teamwork and safety climate in quality and safety efforts appear to be highlighting dimensions that vary more at the unit than hospital level. They suggest that patient safety improvement interventions target unit level changes, and they support the emphasis being placed on teamwork and safety climate, as these vary significantly at the unit level across countries.
- Patient safety
- safety culture
- cross-cultural comparison
Statistics from Altmetric.com
The last 10 years have given rise to an intense focus on patient safety in healthcare quality improvement initiatives,1 ,2 which has led to an increase in use of patient safety culture assessments as a core element of hospital quality of care.3 Patient care quality is often measured using clinical outcomes, such as adverse events,4 which in turn are associated with patient safety culture.5 ,6 In general, hospital organisational culture is associated with patient safety culture7 which is most diagnostic and actionable at the local unit level.8–10
A number of instruments have been developed to capture safety culture and its dimensions.11–13 These instruments elicit patient safety-related assessments of healthcare workers, such as nurses, physicians and other allied health professionals, mainly at the clinic, clinical area or unit level.8 ,14
Given the increasingly common use of safety culture questionnaires for research, quality improvement and accreditation, an investigation of the variability by country, hospital and unit, would better equip professionals to apply patient safety improvement strategies across national health systems in a meaningful, valid and methodologically consistent way. The results from these surveys enable organisations and professionals to more reliably target priority areas for patient safety interventions and, over time, to demonstrate the impact of those interventions.15–17 A growing body of research on patient safety culture in a variety of countries and healthcare settings is available.18–25 To date however, direct comparison of variation in safety attitudes across similar healthcare workers between countries is largely unexplored. What best practices can we learn from benchmarking and cross-cultural comparisons? The answer to this question requires the application of a validated safety culture instrument, for example, the Safety Attitudes Questionnaire (SAQ), to make such a cross-cultural investigation scientifically sound. Moreover, there is an opportunity to link quality improvement strategies with rigorous safety culture assessment (pre- and post-) that targets specific clinical areas,16 ,26 ,27 through interventions related to both patient safety (Executive WalkRounds,17 ,28 Learning From Defects,29) and teamwork (TeamSTEPPS,30 ,31 Situation, Background, Assessment, Recommendation (SBAR)32 ,33 or Briefings34 ,35). The purpose of this study was, therefore, to explore the variability in safety culture dimensions at the clinical area and country level in Swiss and US hospital units.
Design, setting and sample
We used a cross-sectional design. Ten US hospitals from a not-for-profit health system and two Swiss university hospitals were included in the study. The Swiss sample was collected as part of a psychometric validation of the German translation of the SAQ in the two Swiss hospitals, (Unpublished data, Master Thesis, Zimmermann, N, 2011) and included only nurses and physicians from medical, surgical and mixed medical/surgical inpatient units. In the US hospitals, healthcare workers including nurses, physicians, respiratory therapists, nurses aides, pharmacists, medical technicians and others, received the SAQ as part of their routine monitoring of patient safety and quality. In the US hospitals, the targeted respondents were all healthcare workers with at least four weeks of exposure to their current clinical area. Physicians were assigned to the single clinical area where they spent a significant amount of time, and their minimum inclusion criteria for a given clinical area were that they admitted at least two patients a month, or performed at least five procedures a month.
For the purposes of the current study, the US sample was matched to the Swiss sample by selecting the nurses and physicians of the adult medical, surgical and mixed medical/surgical units only to include reasonably similar healthcare workers and clinical areas. Sample unit sizes ranged from 20 to 44 beds with 25 to 50 nurses and physicians (head count), respectively.
The Safety Attitudes Questionnaire (SAQ-Short Form) was used to generate a multidimensional safety culture profile of each unit. The 30 SAQ items garnered assessments of safety and quality-related behaviours and norms specific to a given unit. There are six dimensions covered by the SAQ, including teamwork climate (6 items), job satisfaction (5 items), safety climate (7 items), stress recognition (4 items), working conditions (4 items) and perceptions of unit management (4 items). The SAQ uses a five-point Likert scale with response choices of disagree strongly (1), disagree slightly (2), neutral (3), agree slightly (4), and agree strongly (5). The SAQ item scores reflect the respondent's level of agreement with individual item statements. In other words, units with higher proportions of per cent agreement had more reports of positive safety norms and behaviours. A single composite score comprised of the six SAQ dimensions would not reflect the multidimensional nature of safety culture in a specific clinical area. Though statistically equivalent, for example, a clinical area that was mediocre on all six dimensions is actually clinically and operationally distinct from a clinical area in which three dimensions are extremely positive, and the other three are extremely negative. Ongoing research using cluster analyses and culture profiles strongly supports the continued use of multidimensional safety culture scores over single composite indices.24 Initial evidence for the reliability and factor structure of the SAQ and its dimensions has been reported previously,11 ,24 and confirmed recently in hospital settings.18 ,26 ,36 There is growing evidence that the SAQ elicits attitudes that are responsive to interventions associated with clinical outcomes.13 ,37–42 Cross-cultural use of translated SAQ versions have demonstrated acceptable to good psychometric properties.8 ,19 ,22 ,24 For the Swiss sample a validated German language version of the SAQ was used, wherein the psychometric evaluation used multilevel confirmatory factor analyses achieving good model fit (RMSEA=0.045; CFI=0.941; TLI=0.934), scale reliability (Cronbachs's α >0.87), and test-retest analysis showed high correlations across all factors (r=0.662–0.788). Intraclass correlations and within-group agreement demonstrated that the German SAQ generates information that is meaningful at the unit level (Unpublished data, Master Thesis, Zimmermann, N, 2011).
The data collection took place in the fall of 2009. The SAQ was administered to all nurses and physicians providing direct patient care in medical, surgical and mixed medical-surgical inpatient units of two Swiss and 10 US hospitals. Completed questionnaires were returned via the internal hospital mailing system (US hospital), or in sealed mailings (Swiss hospitals) direct to the investigators. All SAQ questionnaire data were entered into spreadsheets of the electronic databases for further analysis. Based on the voluntary nature of these surveys, the study was independently approved by the respective cantons ethics committees in Switzerland, and by the Duke University Health System Institutional Review Board, USA.
Descriptive statistics were expressed as percentages and means (±SD) for respondent demographics and SAQ scale and item scores. SAQ scale scores were calculated for individual respondents by taking the average of the scaled items. At the unit level, results were calculated as the percentage of respondents within a unit, who reported positive perceptions (ie, those who agreed slightly or strongly).43 ,44 Mean values of safety dimensions between the US and Swiss hospital units were compared using general linear models (GLMs). Given the hierarchical data structure we considered multilevel analysis in two steps. First, MANOVA was used to test, overall, for significant between-unit-variability on the six safety dimensions over the entire sample and within-country samples controlling for country and hospital levels. Second, for each of the six safety dimensions, ANOVAs were performed to test for significant between-unit-variability for the entire sample and within the two countries controlling for hospital level.
In addition, we included an examination of the SAQ scale threshold of scores below 60%. The 60% threshold came from our experience using the SAQ, in which units with <60% of respondents reporting positive safety attitudes had the most to gain from quality improvement efforts, and were substandard in clinical and operational outcomes, such as ventilator-associated pneumonia rates, and ICU length of stay.13 ,28 ,41 ,43
Statistical significance was defined as P≤0.05. Statistical analyses were performed using IBM SPSS (V.19.0) and STATA (V.11.0.2).
Setting and respondents
Fifty-four hospital units (40 in the USA, 14 in Switzerland) from 10 US and two Swiss hospitals completed 1370 of 1637 surveys (84%). Unit response rates for questionnaires ranged from 56% to 100%. Clinical areas included 14 (26%) surgical, 14 (26%) medical and 26 (48%) mixed medical-surgical units. Respondents were mostly registered nurses 1273 (93%), with 97 physicians (7%). Healthcare workers with <1 year of experience were the smallest tenure group, whereas, the largest group had between 5 and 10 years of experience. Details on hospital units and participating healthcare worker characteristics are depicted in table 1.
Variation of percentage positive responses across hospital units
The percentages of healthcare workers who assessed each of the SAQ dimensions positively, varied considerably at the clinical area level within the US and Swiss samples. In the two SAQ dimensions, stress recognition and perceptions of unit management, all Swiss units were below the 60% threshold.
For teamwork climate across the US and Swiss units, respectively, positive responses ranged from 35.7% to 100% and from 50% to 95.8%; for safety climate from 33.3% to 90.5% and from 29.4% to 83.3%; for job satisfaction from 46.2% to 100% and from 47.1% to 95%, for stress recognition from 28.6% to 74.2% and from 8.3% to 55%, for perceptions of unit management from 16.7% to 91.7% and from 28% to 57.1%, and for working conditions from 21.4% to 100% and from 44.1% to 83.3%.
The variability of the safety attitudes dimensions among the US and Swiss hospital units are depicted in the six panels of figure 1. Overall, units with proportions of healthcare workers reporting SAQ dimensions below the 60% threshold (indicating need for ‘improvement’) varied in numbers of units (figure 1, table 2), and in the dimensions of stress recognition and perceptions of unit management most of the hospital units were below 60% agreement.
Differences in safety attitudes dimensions between the two countries
Tests of between-subjects effects on the 54 hospital units using GLMs indicated that three SAQ dimensions varied significantly between countries; stress recognition, perceptions of unit management and safety climate (see table 2).
The mean scale score values for three of the six SAQ dimensions were significantly different between the US and Swiss hospital units, respectively (stress recognition: 57.3; 31.4, perceptions of unit management: 60.1; 41.5, and safety climate: 68.7; 58.4), whereas, teamwork climate, job satisfaction and working conditions were not significantly different (table 2). Given the differences on the three safety dimensions between the US and Swiss hospital units, item-level analysis revealed more insights. First, the stress recognition dimension showed most between-country differences attributable to the items: ‘When my workload becomes excessive, my performance is impaired’ (USA: 76.7, CH (Confoederatio Helvetica): 51.1; p<0.001) and “I am less effective at work when fatigued” (USA: 78.0, CH: 60.9; p<0.001). Second, for the perceptions of unit management dimension, the item ‘Management supports my daily efforts’ attributed most to the between-country difference (USA: 78.8, CH: 64.9; p=0.005). Third, for the safety climate dimension it includes the items ‘In this clinical area, it is difficult to discuss errors’ (USA: 19.8, CH: 7.8; p<0.001) and “I receive appropriate feedback about my performance” (USA: 78.0, CH: 62.9; p<0.001). (See online appendix for more details).
Between-unit variability of the six safety attitudes dimensions (between and within the two countries)
After overall testing of the six SAQ dimensions, MANOVA revealed a Wilks' Lambda, (Λ=0.177; F=10,17; p<0.001) indicating significantly different safety culture dimensions between US and Swiss hospital units. Within the two countries, healthcare workers' perceptions in the six SAQ dimensions differed significantly for the US hospitals (Λ=0.159; F=11,67; p<0.001), and for the Swiss hospitals (Λ=0.082; F =12,98; p<0.001), respectively. In other words, there was significant variability at the clinical area level, overall, and at the clinical area level within each country.
After testing of each of the six SAQ dimensions between units across the entire sample, ANOVA revealed significant differences among healthcare worker safety attitudes for teamwork climate, safety climate, job satisfaction and perceptions of management (but not for working conditions and stress recognition, see table 2).
In the US hospitals, healthcare workers perceive safety culture issues differently as a function of the unit in which they work. ANOVA indicates that each SAQ dimension varied significantly at the US unit level, with the exception of stress recognition and working conditions (table 2). The Swiss healthcare workers perceive safety culture issues differently except for stress recognition and perceptions of unit management (table 2). In other words, there was significant variability in four of the six SAQ dimensions at the clinical area level, overall, and in four of the six dimensions at the clinical area level within each of the countries.
Given the lack of significant variability by unit for two of the six dimensions, we explored data clustering at the hospital level. To this end, we fit a three-level multilevel model (respondents within units, units within hospitals and hospitals within countries) for continuous responses,45 and calculated intraclass correlations (ICC) for the unit and hospital level (unit-respective hospital variance divided by the total variance).
Our analyses revealed that for stress recognition (ICC at hospital level=5.4%; ICC at unit level=0.3%) and working conditions (ICC% at hospital level=6.9%; ICC at unit level=1.1%), more variance was allocated at the hospital level than at the unit level (table 3).
This study explored the variation in six distinct safety culture dimensions in two countries, 14 hospitals and 54 clinical areas. Overall, our results revealed more empirical and meaningful safety culture variability within than between countries, which is not unlike previous studies13 ,18 ,20 ,22 ,24 ,26 ,41 ,46 in which the overarching source of variability was the unit or clinical area level. This underscores the empirical necessity of both, capturing safety culture assessments at the unit level, and improving quality using interventions at the unit level. Overall, four out of six safety culture dimensions varied significantly at the clinical area level for the entire sample and within each of the countries. ICCs revealed that hospital-level variation may be more meaningful for stress recognition and working conditions. Perhaps, improvement efforts targeting these two dimensions should address the hospital level more deliberately. Five of the six safety culture dimensions from the Swiss sample fell within the clinical area-level range of the US sample. Despite some significant differences between countries, it was the clinical area level that consistently garnered the statistical significance. Indeed, item-level analyses between the US and Swiss samples indicated that the significant differences between countries were more an issue of statistical power than of clinical meaningfulness.
Consistent with previous studies using the SAQ, relatively high response rates were achieved.26 ,41 ,46 With these interpretable response rates, we demonstrated the feasibility of administering, collecting, aggregating, analysing, interpreting and feeding back safety culture data from a valid translation using standard guidelines per the original version. This bodes well for future research using not only valid translations, but also consistent survey user guidelines and recommendations that were developed and refined elsewhere.
Nevertheless, there were consistent and significant differences between the US and Swiss samples on stress recognition, and we also found clustering at the hospital level for stress recognition. Stress recognition per cent positive scores had the lowest minimum and lowest maximum values, overall, and within each country. We recently found stress recognition to vary significantly between 144 clinical areas, and between 22 departments within Johns Hopkins Hospital.26 Indeed, stress recognition per cent positive scores in that study ranged from 0% to >90%, with response rates that ranged from 75% to 79%. We still have much to learn about stress recognition in healthcare quality improvement, but for now it is safe to say that the overall distribution at the clinical area level is lower, and there appears to be some clustering at the hospital level. Indeed, one robust result of our study is that teamwork climate and safety climate vary more at the clinical level than at the hospital level. Teamwork and safety climate domains are the most thoroughly researched, utilised and targeted for specific improvement efforts. The other four SAQ dimensions are generally used to support the interpretation and action planning around teamwork climate and safety climate.
Our study provides a detailed look at the patient safety and quality-related attitudes and behavioural norms of healthcare workers in adult medical-surgical inpatient units. This level of detail achieves a threshold of 60%, and benchmarking, overall, and within each country is intended to facilitate and simplify the efforts of others. Researchers, risk managers, patient safety officers and other quality and safety leaders can collect, analyse, interpret and feedback results from a single adult medical-surgical inpatient unit with a readily accessible cross-cultural comparison using a validated multidimensional safety culture questionnaire. Moreover, the use of action plans and strategic responses to survey data are prevalent in healthcare, perhaps to the point at which we now over-rely on them to drive change at the clinical area level. One outcome of this study is the enhanced ability to interpret and respond to safety culture results by understanding what is normal, what is possible and what is typically associated with relatively high or relatively low scores.
Reflection on the safety culture dimensions
Overall, in only one out of three units did healthcare workers report 60% or more positive responses for ‘stress recognition’ which appears to be remarkably low compared with the dimension, ‘job satisfaction’, in which most units reported ‘good’ job satisfaction.
Healthcare workers reported low overall stress recognition levels relative to the other SAQ dimensions, and this was true for both US and Swiss hospital units. Perhaps healthcare workers are becoming desensitised to the higher levels of volume, acuity and the increasingly common presence of hassles and risks in their daily work, whereby threats to safety, such as fatigue, production pressures, new technology, new managers, working with inexperienced colleagues and working in hostile environments, are blurring the lines between perceptions of inconveniences and high risk. Another area of concern may be perceptions of unit management in which three out of five units reported below the 60% threshold, activities contributing to the perceptions of unit management to include issues, such as supporting staff in their daily efforts.
Teamwork climate reflects the perceived quality of teamwork and collaboration within a given clinical unit. A low teamwork climate may reflect persistent interpersonal problems among the members of these units. When teamwork climate is low, healthcare workers feel that their co-workers are not cooperative, that their voices are not heard. Under such circumstances, strategies and tools such as TeamSTEPPS13 may be considered to improve teamwork.
Safety climate is significantly related to both, caregiver safety, for example, needle sticks, back injuries and patient safety, and bloodstream infections, pressure ulcers.47 When healthcare workers report that they do not perceive a good safety climate, they essentially indicated that they do not see a real dedication to safety in their unit. Emphasising the importance of keeping lines of feedback and communication open may be critical to maintain a climate of high level of safety by letting employees know that it is essential to bring errors to the attention of managers and clinical leaders in order to learn from them, since, accordingly, they need to be responsive.41
The safety climate items that differed between the US and CH samples suggest that some of the more active aspects of safety-related norms, such as reporting, performance feedback and discussing mistakes were reflecting the variability between countries (see online appendix). Nevertheless, the country-level difference was only 10%, and was relatively small compared with safety climate variability between clinical areas within the USA (33–91%) and Switzerland (29–83%). Indeed, teamwork climate and safety climate varied significantly, overall, and within each country, which is noteworthy, given that they represent the two safety culture domains most commonly reported on in the published literature.
Job satisfaction was, in the majority of Swiss and US hospital units, above the 60% threshold, which indicates high employee morale, and generally positive emotional reactions to healthcare workers' work experiences, including high individual job engagement and performance. These results may also show the potential of a workforce to foster positive work environments which may contribute to patient safety at the clinical area level.
Stress recognition was low in more than half the US and in all the Swiss hospital units, suggesting that healthcare workers frequently do not recognise the link between their stress levels and their on-the-job performance.48 All stress recognition items were significantly different between the USA and CH, with the largest discrepancies in the two items that reflect acknowledgement of performance impairment due to excessive workload and fatigue (see online appendix). One could assume that for stress recognition, the lower overall score values from Swiss hospitals compared with US hospitals might reflect a relatively less stressful working experience from nurses in view of regular shift hours, as they are legally limited to 8.4 h, whereas, the 12 h working shift model is common in some US hospitals.
Significant research has attested to the detrimental effects of stress on performance26 ,49 ,50; recognition of this relationship is integral to healthcare workers' utilisation of practices and policies that reduce employee stress and work overload. To increase employee awareness of the stress-performance link, it is important to emphasise the many ways in which stress can be problematic. For instance, it would be beneficial to highlight the fact that stressed and fatigued workers often make basic procedural errors that can compromise the safety of even routine procedures.51 ,52
Unit management perceptions reflect healthcare workers' beliefs about the local leadership's dedication to patient safety, support of frontline staff and deliverance of timely information on policies that affect their work. Almost half the US, and in all the Swiss hospital units, perceptions of management appeared low suggesting that healthcare workers do not think management is particularly concerned about their well-being or the patients' well-being. Two of the four unit management items reached statistical significance, and phrasing of both directly reflected ‘unit management support’ of patient safety and the daily efforts of healthcare workers.
Negative perceptions can affect employee morale, and may be indicative of actual managerial problems. If a particular item is low, focus on behaviours related to that item, such as the ways in which management expresses an indifference to patient safety or staffing.28 ,53 Respondents from the Swiss units might be more critical to their unit's manager performance, or suspicious of their ability to run the unit more effectively than their US colleagues.
The working conditions dimension assesses healthcare workers' perceptions that new employees are adequately trained, problem personnel are adequately dealt with and information vital to patient care is properly disseminated.
The biggest concerns for Swiss units were stress recognition and unit management perceptions, given none of the units responded above the 60% threshold. Similarly for US units, stress recognition and unit management accounted for more than half and almost half the units falling below the threshold, respectively.
Assessing safety culture through the SAQ offers a powerful feedback tool to raise staff awareness towards patient safety issues, and to identify for staff and leaders priority areas for improvement.54 Since effective patient safety strategies are often generic in nature and implementable regardless of hospital type, as seen in a cross-country study,55 one by-product is the inconsistency with which they are implemented. In other words, all the hospitals had structures, plans and responsible people in place to manage patient safety, but the degree of comprehensive implementation and compliance of regulators and hospital leadership varied considerably.55 We found that hospital-level safety-culture metrics diluted the diagnosticity, meaningfulness, and value of the results, whereas, clinical area/unit-level metrics concentrated diagnosticity, meaningfulness and value.
The over-representation of US clinical areas relative to Switzerland may have limited our exposure to actual variability between countries. The 10 US hospitals from one not-for-profit health system, and the two Swiss universities, are not representative of their respective populations, so caution should be used in interpreting the cross-cultural results as the safety culture dimensions are also unlikely to be representative in either setting. SAQ use and administration strategies varied between the two countries with regard to its regular application in the US hospital sample, whereas in Switzerland, the SAQ was distributed for the first time. Moreover, in the US sample, a surgeon who admitted several patients per month to a medical surgical unit may not have been included in the medical surgical unit if they spent significantly more time in the operating room of that hospital (in the US sample, operating rooms, emergency rooms, ICUs, etc, were surveyed at the same time as the medical surgical units).
The majority of respondents were registered nurses with <10% being physicians, which may give a more accurate picture of nurses' safety attitudes. Also, the sample of Swiss university hospitals limit generalisability of findings to Switzerland, but simultaneously collected data from another study56 suggest that the two hospitals are not significantly different from a national Swiss norm.
In our study, a small fraction was physicians; their sample size reflects clinical realities as well as survey administration logistics in general medical, surgical or mixed medical surgical hospital units. It will be fruitful to expand the look at what is associated with differences in safety culture dimensions in larger samples and even more countries, to better understand variability of safety attitudes dimensions across countries and hospital units.
This study provides a detailed look at the patient safety and quality of care-related attitudes and behavioural norms of healthcare workers in adult medical-surgical inpatient units of US and Swiss hospitals. In a study that achieved high response rates, we found that despite some country-level variability, it is clinical area membership that is the more potent level for capturing respondent variance in dimensions of safety culture. Moreover, the two most commonly researched and published safety culture dimensions, teamwork climate and safety climate, were excellent examples of dimensions that may vary at the country level, but that variability might be dwarfed by the variance captured at the clinical area level. Our results suggest that a deliberate focus on the assessment and improvement of patient safety and quality at the unit or clinical area level is empirically sound, and opens a diverse set of potential interventions from different countries to be considered across countries while maintaining an emphasis on the critical importance of targeting the clinical area level. To date, patient safety and quality improvement interventions have focused conceptually on improving teamwork climate (eg, TeamSTEPPS, SBAR, briefings) and safety climate-related issues (eg, Executive Walkrounds, Learning from Defects, Comprehensive Unit-based Safety Program). We found that these two domains were clearly more diagnostic and meaningful at the unit level than at the hospital level. This suggests that the use of teamwork and patient safety-related interventions require an emphasis on the unit level, and that the use of these interventions could be explored across countries.
The SAQ is available for free on the internet for researchers, and those who wish can interact with Pascal Metrics to manage the survey administration for technical support and customer service.
Competing interests None.
Ethics approval Ethical Review Board of the cantons of Bern and Basel, Switzerland, and the IRB of the health system for surveying its employees.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement The cited unpublished data (Master thesis of Zimmermann) is submitted to a scientific journal.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.