Health professionals’ perception of patient safety culture in a university hospital in São Paulo: A cross-sectional study applying the Hospital Survey on Patient Safety Culture

ABSTRACT BACKGROUND: Patient safety culture is part of the organizational profile of healthcare institutions and is associated with better quality of care. OBJECTIVE: To assess patient safety culture in a university hospital. DESIGN AND SETTING: Hospital-based cross-sectional study conducted in a public university hospital in São Paulo, Brazil, between September and December 2015. METHODS: We randomly selected 68 sectors of the hospital, to include up to 5 employees from each sector, regardless of length of experience. We used the validated Brazilian version of the Hospital Survey on Patient Safety Culture (HSOPS) via an electronic interface. We calculated the percentage of positive responses for each dimension of the HSOPS and explored the differences in age, experience, occupation and educational level of respondents using the chi-square test. RESULTS: Out of 324 invited respondents, 314 (97%) accepted the invitation and were surveyed. The sample presented predominance of women (72%), nursing staff (45%) and employees with less than six years’ experience at the hospital (60%). Nine out of the 12 dimensions showed percentages of positive responses below 50%. The worst results related to “nonpunitive response to errors” (16%). A better safety culture was observed among more experienced staff, nurses and employees with a lower educational level. In the previous year, no events were reported by 65% of the participants. CONCLUSIONS: The patient safety culture presented weaknesses and most of professionals had not reported any event in the previous year. A policy for improvement and cyclical assessment is needed to ensure safe care.

are involved in the professionals' routine and their perceptions, along with the strengths and weaknesses of the culture of patient safety.Such assessments also make it possible to identify the sectors and processes that generate risks. 6Knowing the weaknesses in patient safety culture makes it feasible to establish interventions and improvements in the quality of care for users, thus changing the professionals' behavior. 7 assess patient safety culture, surveys with validated questionnaires are widely used. 8The Hospital Survey on Patient Safety Culture (HSOPS) and the Safety Attitudes Questionnaire have been widely cited in research that aimed to assess patient safety culture in hospital settings worldwide. 6The HSOPS, developed by the United States Agency for Healthcare Research and Quality (AHRQ) in 2004, 9 proposes 12 dimensions to assess patient safety culture from the professionals' perspective in a hospital setting (Table 1).This instrument has been translated and validated for use in several languages, [10][11][12] and in 2012, the Brazilian version was made available for use. 13In 2013, the Brazilian government issued regulations on patient safety actions, including identification, reporting and system improvement. 14,15

OBJECTIVE
Few studies have reported on use of the HSOPS in Brazil to characterize the level of patient safety culture in Brazilian hospitals.
In this scenario, studies that estimate patient safety culture are necessary.The objective of this study was to assess perceptions of patient safety culture in a university hospital.

Study design and context
This was a hospital-based cross-sectional study in which the HSOPS was used to assess patient safety culture from the professionals' perspective.It was conducted from September to December 2015 at Hospital São Paulo, the university hospital of Universidade Federal de São Paulo, located in the city of São Paulo, the largest city in Brazil.This hospital provides high-complexity care in all medical specialties and has more than 700 beds.The primary outcome was the percentage of positive responses for each dimension of the HSOPS.

Participants
All professionals who directly or indirectly were attending patients in the hospital, regardless of their length of experience at the institution, were eligible for participation in this study.Trainees, interns, dismissed employees and outsourced workers (cleaning, security and food service employees) were not eligible, because not including them would improve the homogeneity of the sample.

Sample size and sampling process
To calculate the sample size, we considered the population of approximately 5,000 employees at the hospital.We made a conservative estimate for the frequency of positive responses regarding the presence of patient safety culture ("strongly agree/agree" or "most of the time/always") of 50%.In the dimensions of the HSOPS, a precision rate of 7%, a value of 1.5 for the sampling effect and a possible loss rate of 10% were used.These parameters resulted in a need to survey a minimum of 312 professionals.
We randomly selected 60 primary and 20 secondary sectors out of the 106 sectors of the main building of the hospital and invited up to five employees who were present at the time of the visit to each sector, to be interviewed.

Data collection
The instrument used for data collection was the Brazilian version of the HSOPS. 13The survey is composed of 42 items grouped  1).The HSOPS makes it possible to measure the beliefs, skills and behaviors involved in the safety culture of the organization from hospital staff perspectives.
Each dimension is composed of three to four items that are constructed in a positive or negative manner (Table 1).For each item, the respondent may choose a score on a five-point Likert scale with the response options of strongly agree, agree, neither agree nor disagree, disagree and strongly disagree, or response options of never, rarely, sometimes, most of the time and always, in relation to frequency. 9Two other items assess individual assessments of patient safety: the "patient safety grade", with response options of excellent, very good, acceptable, poor and failing, and the "number of events reported", with response options of no events reported, 1 to 2 events reported, 3 to 5 events reported, 6 to 10 events reported, 11 to 20 events reported and 21 or more events reported.
After reversing the sentences that were negatively worded, we calculated the percentage of positive responses regarding the presence of patient safety culture in each dimension by dividing the number of positive responses ("strongly agree/agree" or "most of the time/always") by the total number of responses (positive, neutral and negative) in the dimension.A percentage of positive responses above 75% was considered strong, and a percentage below 50% showed that there were issues that needed improvement.For items with reverse wording and that had a negative connotation, disagreement indicated a positive response.Thus, to calculate the percentage of positive responses among the answers, we needed to consider the strongly disagree/disagree or never/ rarely responses.
In the process of pretesting the survey, we modified the Portuguese-language wording of three items (A5 in the "staffing" dimension, C1 in the dimension of "feedback and communication about error" and G1 in the dimension of "number of events reported"), in accordance with previous recommendations, to improve comprehensibility (Table 1). 16The research group that suggested this wording has, furthermore, validated a new version of the HSOPS in an electronic interface. 17e survey was developed using a suite of tools for field data called KoBo Toolbox (www.kobotoolbox.org,Cambridge, MA, USA) and was administered in the workplace.Notices invited hospital staff to participate in the study and, after agreeing to do so and signing an informed consent form, staff members completed the survey using tablet electronic devices (Samsung Galaxy Tab 3).The device recorded the data online or offline and, after connecting to the internet, the surveys were automatically uploaded to the online platform.
Two trained survey administrators performed the data collection: a pharmacy undergraduate student and a pharmacist.

Statistical methods
The negatively worded items were reverse-coded to calculate the percentage of positive responses for each dimension.
The answers were recoded as follows: strongly disagree, disagree, neither agree nor disagree, always, most of the time and sometimes were assigned a score of 0; while agree, strongly agree, never and rarely were assigned a score of 1, in accordance with the HSOPS manual. 9e proportion of positive responses for each dimension was stratified according to respondent age, length of employment at the hospital (in years: less than 1; 1 to 5; 6 to 10; 11 to 20; or 21 or more), profession (doctor, nurse or other professional) and educational level (completion of high school, undergraduate level or postgraduate level).The differences were tested using the chi-square test and were considered significant if P < 0.05.
To assess the internal consistency of the survey, we calculated Cronbach's alpha for each dimension and item of the Brazilian version of the HSOPS.The calculations on the data were done using Stata 14.2.

Ethical issues
The present study was approved by the hospital's research ethics committee, under the number CAAE 48415315.3.0000.5505.All subjects signed an informed consent form.

RESULTS
We invited 324 employees from 68 sectors of the hospital to participate.A total of 314 professionals (97%) accepted the invitation and were included, while 10 (3%) refused to participate.
Most participants were women (72%); 41% had undergraduate and postgraduate educational levels.The majority had direct contact with patients (80%), 45% were nursing staff (nurses, nursing technicians and nursing assistants) and 60% had been working at the hospital for less than six years.As shown in Table 2, different professionals participated in the survey.
Nine out of the 12 dimensions showed positive response rates below 50% (Table 3).The dimension of "nonpunitive response to errors" had the worst result (16%).A total of 65% of the participants indicated that they had reported no events in the past 12 months.The internal consistency was adequate for eight dimensions and the other four showed lower consistency (Cronbach's alpha < 0.6).
Greater age and length of work experience were associated with higher perceptions of patient safety culture in the dimensions of "supervisor/manager expectations and actions promoting patient safety", "organizational learning and continuous improvement", "frequency of events reported", "feedback and communication about error", "staffing" and "management support for patient safety".On the other hand, the dimension of "nonpunitive response to errors" was only associated with age (Table 4).The dimension of "frequency of events reported" was significantly different according to professional category (higher perception among nurses than among other professionals and physicians) and educational level (lower perception among employees with higher education).The dimension of "management support for patient safety" was also inversely proportional to educational level.

DISCUSSION
Patient safety culture in this hospital was fragile, considering that 9 of the 12 dimensions of HSOPS were rated at below 50%.Twothirds of the respondents did not report any events in the last 12 months, thus indicating that potential safety problems may be going unrecognized and are not being addressed properly.The low rate of positive responses for the dimension of "nonpunitive response to errors" has also been found in other studies, [18][19][20] and this may also explain the behavior of not reporting events.
The dimensions with higher levels of positive responses, i.e. "supervisor/manager expectations and actions promoting patient safety", "organizational learning and continuous improvement" and "teamwork within units", did not represent strengths in patient safety culture, since they fell below 75%. 9 Within their work units, professionals may seek to carry out their activities in a team with supervised support and to look for improvements to patient safety. 21Teamwork is a critical point and is important because it relies on collaboration and mutual respect. 21Such values lead to opportunities to adopt improvement programs.
Investigations conducted by different researchers have found similar results. 18,22,23study that applied the HSOPS to 26 hospitals in Iran 20 observed that there was better perception in the dimension of "organizational learning and continuous improvement".In teaching hospitals, professionals are willing to improve their understanding and knowledge.It has been observed that in the dimension of "organizational learning and continuous improvement", the percentage of positive responses improves as the amount of work experience increases. 24 The "staffing" dimension needs improvement, which may be an effect caused by a situation of an insufficient number of professionals with heavy workloads.This imbalance increases the risk relating to the assistance provided. 10In units that perform activities under unfavorable conditions, professionals feel that the level of support that they can count on to carry out their tasks safely when they are confronted by a high volume of responsibilities is lower. 25e number of working hours can also be related to the results, since tiredness decreases attention and increases the incidence of errors. 26A number of factors affect the safety and quality of patient care, such as the organization of nursing units, structure, communication, stress and workload. 27A better distribution of professionals and appropriate working hours are paramount for improving healthcare quality.
The dimensions of "communication openness" and "feedback and communication about error" indicated that there was a need for to improve priorities.Ineffective communication increases the occurrence of adverse events. 24As observed in other studies, failure in communication is directly related to worsening of quality of care. 28,29Hospitals in which there is a channel for free communication between supervisors and employees to exchange suggestions, questions and feedback on improvements in patient safety tend to have better scores for quality and motivation, with regard to learning from errors. 18ofessionals with greater experience had a better perception of safety culture.Usually, such professionals have more responsibility or occupy leadership positions within their teams.This may positively influence their perception of patient safety, as observed in a study conducted in Finland that compared the perceptions of managers and registered nurses. 30The experience of a professional can positively influence the results, as shown in a Palestinian study in which the number of adverse events reported increased with a professional's length of experience. 19The participants in the present study were mostly composed of early-career professionals, which may explain the low rate of errors reported.More events were reported by nurses than by the medical team, which is similar to what was seen in a study conducted in the United States. 31ven that contextual limitations may have influenced the present results, we need to highlight that an employee strike had ended just before the time of data collection and that budget cuts occurred during the survey period.Despite the difficulties faced by these professionals, a good acceptance rate was obtained for the survey.The participants were a diverse group of professionals who were either directly or indirectly involved with patient care.Examining the hospital as a whole improves the representativeness of the results. 32We also chose to approach employees in person instead of via remote strategies, which are more prone to give rise to a less diverse sample population and a lower response rate.The institution surveyed here is a university hospital and its staff include a wide variety of professionals for the purposes of undergraduate education, residency and specialization.These data may suggest that high turnover exists, 33,34 and this may have been related to the low perception of safety among these professionals.
The reliability of the HSOPS version used in the present study was fair.Changes that had been made to improve comprehensibility 16 resulted in better consistency in the "staffing" dimension, such that it improved from 0.20 in the first Brazilian validation of the HSOPS 13 to 0.53 in the present study.A new validation of the HSOPS that featured better wording of these questions was performed and published after our survey was conducted and had high instrument reliability. 17e negative results found in the present study may be viewed as demotivating with regard to patient safety in the hospital.Measuring safety culture is the first step towards identifying the priorities that need to be addressed if a change in patient safety is to be achieved.
In Brazil, the regulations in this field are still evolving, and greater investment in patient safety strategies is required. 14,15In addition to ameliorating assistance, improvement of patient safety culture in university hospitals enriches undergraduate and postgraduate education.

CONCLUSION
Patient safety culture in this Brazilian hospital was shown to be fragile, and improvement is necessary in order to ensure safe care.Implementation of enhancement measures and further assessment of patient safety culture should be a cyclical process to drive effective changes in patient safety forward.

Table 1 .
Dimensions of the Hospital Survey on Patient Safety Culture: numbers of items and what is assessed 9

Table 3 .
Percentage of positive responses according to dimension (n = 314)