Adaptation of the Safety Climate Survey: A Contribution to Improving Patient Safety

,


Introduction
Patient safety concerns the prevention of patient harm caused by healthcare-related errors [1].However, only 60% of care is based on evidence, or in accordance with guidelines, systems waste about 30% of all health expenditure, and 10% of patients experience harm [2].Te World Health Organization (WHO) also estimated that 1 in 10 patients in high-income countries is harmed while receiving hospital care, and in low and middle-income countries, 134 million adverse events occur, resulting in 2.6 million deaths [3].Tis indicates that it is crucial to establish high-quality and safe system designs to provide optimal healthcare services [1].
Te prevention of errors in healthcare services and the elimination or reduction of harm require a patient safety culture characterised by continuous improvement involving repetitive evaluations of the safety climate [4,5].For a safety culture to be adopted and embedded in institutions, healthcare professionals need to work with the wider health care team in the assessment and improvement of the safety climate [1], and strategies to involve them in decisions and activities that afect them would increase the likelihood of safety initiatives being successful [6].For example, it would be benefcial for healthcare organizations to formally assess patient safety processes at regular intervals (e.g., every six months) to identify the weaknesses in the system.One element of such an approach could be to examine safety climate [5].Moreover, developing an approach to organizational learning focused on what works well in complex healthcare systems would result in more rapid improvement than relying solely on conducting investigations when processes fail [7].
Te concept of a "safety culture" generally refers to the corporate values, and practices necessary to ensure safety are maintained, whereas the term "safety climate" focuses attention on employee perceptions of how patient safety is defned and managed [5,8].Studies which have examined the nature and scope of the safety climate have identifed a number of common dimensions.Tese include the following: leadership commitment to safety, prioritization of safety, teamwork, communication, and safety systems [9][10][11].Challenges to the delivery of efective, high-quality, and safe care tend to be the obverse of those listed above including lack of leadership support, staf work pressures, inadequate risk management, communication barriers, and limited resources [12].Terefore, it is essential to monitor the extent and impact of these factors using robust methods to gather reliable data on safety issues [5].Tools designed for this purpose include the Safety Attitudes Questionnaire [13], Patient Safety Culture in Healthcare Organizations [14], Hospital Survey on Patient Safety Culture [15], and the Safety Climate Survey [16].
In the present study, the Safety Climate Survey (SCS) was selected because it is unidimensional, yet takes account of complex nature of safety climate, and focuses on institutionwide improvement processes.Its unidimensional format with its 19 items was also a factor in its selection as this was likely to increase the likelihood that participants would complete the survey [5].A large number of the items with multiple subdimensions in a survey tool scale can be ofputting and result in participants failing to complete an instrument [17].Hence, the survey length may afect the reliability of the results obtained [18].Te fact that the scale has also been used successfully in diferent cultures/settings was also a factor in its selection [19].
1.1.Aim.Te aim of this study was to adapt the SCS and examine its validity and reliability features when applied to the Turkish nursing population.

Design.
Te study was conducted in two phases as given as follows: (1) Translation and language adaptation of the survey.(2) Administration of the survey to nurses working in the Turkish health care settings.Te translation and adaptation of instruments for cross-cultural research require rigorous planning and a robust methodological approach.We followed three sets of guidelines [20][21][22] for the reporting of the psychometric and psycholinguistic properties of the scale.

Setting.
To increase the likelihood of the generalizability of the fndings, a university hospital and two public hospitals were selected, as providers in other parts of the country were similar in size and structure.Te university hospital employed 748 nurses; one public hospital employed 376 nurses, and the other 761 nurses.

Sampling and Participants.
It has been recommended that to perform confrmatory factor analysis (CFA) efectively in validity and reliability studies, the number of participants (varying between 100 and 1000) should be at least seven times (in between 3 and 20) the number of items on the scale [23].For test-retest, the recommended number of participants is between 50 and 100 [24,25].Taking these recommendations into account, we aimed to recruit a minimum of 418 participants (20 × 19 items require 380 participants, plus a nonresponse rate of %10 � 418).Te fnal sampling frame was set at 434 nurses for the survey and 82 for the test-retest reliability element.

Safety Climate Survey (SCS).
Te unidimensional SCS, developed by Sexton, Helmreich, Pronovost, and Tomas [16], consists of 19 items whose responses are rated on a fvepoint Likert-type scale.Te participants are asked to rate the items of the survey as follows: 1 (totally disagree), 2 (disagree), 3 (neither agree nor disagree), 4 (agree), and 5 (totally agree), 6 (have no idea/comment).Te option "I have no idea/comment" is not included in the calculations.Item 18 is reversely scored.Te total score is calculated by summing the scores given to all the items and dividing the result by the number of the items.At the end of the procedure, a mean score ranging from one to fve is obtained.A score ≥3.75 indicates a positive safety climate perception.Te reliability coefcient of the scale was 0.87 in the original study [16].We also included questions about age, sex, education, length of service, unit worked in, and weekly working hours in addition to the SCS items in order to identify the participants' characteristics.

Data Collection.
Nurses were invited to participate if they met the following inclusion criteria: working full-time providing direct patient care for at least one year postqualifcation.Tese criteria were necessary because some of the survey items require direct patient care and nursing experience.Nurses not providing direct patient care (e.g., polyclinics) or on sick/annual leave were excluded.A convenience sampling approach was used.During data collection, the researchers visited the hospitals according to the nurses' shifts -initial visit with two additional reminding visits-and distributed the survey to the staf nurses by hand in sealed envelopes.As there was no collection point, the researchers received the completed forms back in sealed envelopes, in person.Te returns were given sometimes just immediately after the frst visit.In others, it was in a week.Participation was voluntary.Te researchers gave 2 Journal of Nursing Management information on the survey form and asked participants to fll in it.Te sample who agreed and completed the survey form was 434 in total.
Analyses involved confrmatory factor analysis (CFA), splithalf method, and item-total correlation, as these have not been undertaken in previous studies.In the assessment of reliability of the instrument for use in the Turkish nursing setting, the test-retest method was used to measure stability, and Cronbach's α coefcient, Spearman̶ Brown coefcient, and item-total correlation were examined to measure internal consistency.For the validity analyses, content validity was assessed.Te CFA was performed for construct validity.Te statistical signifcance level was set at p < 0.05 (See Table 1 for a summary of the analytical approaches used).

Translation.
In the adaptation process, it is crucial to select idioms and sentence structure that are understandable in the target language but which are consistent with the meaning in the original instrument.Tis may involve the replacement of particular phrases to ensure they are suitable for the target culture.For this purpose, a pilot study is recommended following the forward and back translations [22].Forward-back translation was undertaken.Tree translators with a good command of English performed the forward translation from the original language (English) to the target language (Turkish).Following forward translation, the researchers (authors), as recommended in the guidelines, checked and modifed the items as necessary to eliminate the inconsistencies in the translations (For example, in item 6, "paid attention" was changed to "acted upon").During the back-translation process from Turkish to English, two different translators, who did not see the scale items beforehand, translated the scales.Te researchers then combined the two translations into a single form, and on receipt of approval from its authors [16], a pilot study was conducted with ten nurses to evaluate the items in terms of clarity.In the fnal part of the translation process, the researchers amended and revised the scale items based on the nurses' recommendations (for example, the wording of item 2, in which "unit" was changed to "clinical area").
2.8.Validity.We used the content validity index for content validity and confrmatory factor analysis for construct validity.

Content Validity.
It is recommended that at least 5-8 experts are involved in evaluating the extent to which a survey tool addresses the phenomenon of interest [22,26] and that the content validity index (CVI) is used [27].For CVI, the measurement tools are frst scored by experts (1 � not relevant, 2 � somewhat relevant, 3 � quite relevant, 4 � highly relevant).Ten, the number of experts who gave "three" and "four" points for each item is calculated and then divided by the total number of experts.Te result gives the CVI value.Te threshold CVI value is 0.78 and above for the scale items (0.75 and above with 10 or more experts) and ideally 0.90 and above for the total scale [28].In the present study, ten experts (eight nursing faculty members from diferent departments -experienced in methodological studies, and two clinical nurse managers -one from the ward, one from the critical care unit) assessed the scale items for their suitability.In line with these expert suggestions, the researchers reviewed the scale items' content and made changes in the words used and sentence construction before the survey was conducted [28].For example, the ffth item was changed after the recommendations, that is, from "Tere is a leadership approach that focuses on patient safety throughout the institution" to "Tere is a leadership approach in this institution that directs employees to focus on patient safety."Te CVI was between 0.80 and 1.00 for the scale items and 0.97 for the total scale.We found the CVI value calculated for experts to be 0.97.Te expert panel agreed that the scale items are related to the scope of the scale with an agreement rate of over 80% (Table 2).See Table 2 for the CVI indices.

Construct Validity.
Construct validity is the level of representation of the factors/items that are related to the construct in the measurement tool [27].Examining the accuracy of these factors in the scale structure helps determine the suitability of the subdimension items.Two forms of factor analysis (exploratory and confrmatory) and structural equation modeling are the main methods for determining the level of construct validity.While variables in exploratory factor analysis (EFA) produce loadings on all factors, only factors assigned to the model produce loadings in confrmatory factor analysis (CFA).Terefore, CFA is regarded as the approach of choice for the cultural adaptation of measurement tools [29,30].Acceptable values for ft indices obtained as a result of CFA are as follows: Chi-square/Degrees of freedom: 2 ≤ 3, Normed Fit Index-NFI: ≥0.90, Comparative Fit Index-CFI: ≥0.95, Incremental Fit Index-IFI: ≥0.95, Goodness of Fit Index-GFI: ≥0.90, and Root Mean Squared Error of Approximation-RMSEA: ≤0.08 [31].
2.9.Reliability.We used the test-retest method to measure the stability of the scale.Cronbach's α coefcient, split-half analysis, and item-total correlation were selected to determine its internal consistency.2.9.1.Test-Retest.Tis involves the measurement of the tool using the same sample group twice a specifed interval [27].A recommendation for the sample size is between 50 and 100 [24,25], and the interval between the two measurements should be at least 10-14 days to avoid the possibility of participants remembering the items [32].Te intraclass correlation coefcient (ICC) was used for interpretation with the ICC values being evaluated in four classifcations (<0.5: poor, 0.5-0.75:moderate, 0.75-0.9:good, >0.9: excellent) [33].
2.9.2.Cronbach's Alpha.Te alpha coefcient (or Cronbach's alpha) is normally calculated to evaluate internal consistency [27].Te range value varies between 0 and 1. High values support the internal consistency of the scale [27].Ideally, threshold values should be equal to or above 0.70 [34].

Split-Half Method.
It is another method for determining levels of internal consistency.It is calculated by dividing the total of items into halves.Te Spearman̶ Brown coefcient is used [35], and 0.70 and above are regarded as acceptable [34].
2.9.4.Item-Total Correlation.It indicates the items' suitability (whether they will change or not) and the correlation values with the score of each item that is examined.An acceptable level value is above 0.20 [36].

Ethical Considerations.
Ethical approval for the conduct of the study was provided by a University Research Ethics Committee (Decision No: 2015/02-20).Agreement to recruit nursing staf was obtained from the executive nursing management of the respective hospitals, and permission via e-mail to use the SCS was confrmed by its developers.

Results
Te mean age of the participants was 35.29 (SD � 7.12), and the majority (92.9%) were female.Two-thirds of the nurses (66.6%) held an undergraduate degree.Te distribution of nurses by hospitals was similar in numbers.Te average length of professional experience was 13.38 (SD � 7.85) years, and the average weekly working hours were 45.42 (SD � 6.51) (further information about the respondents' socio-demographic details and work-places can be found in Table 3).

Psychometric Properties
3.1.1.Validity.Table 4 shows the confrmatory factor analysis results, presenting the ft indices for the current study fndings and acceptable level indices.Te scores indicate that the model satisfes the threshold values.Additionally, all the factor loadings of the items were positive and standardized loadings ranged from 0.31 to 0.78, which are above the accepted cut-of value of 0.30.

Reliability.
We evaluated the scale's stability with testretest (two weeks apart).During the analysis, we marked two-way mixed as the model and absolute agreement as the type.We then identifed the intraclass correlation coefcient as 0.84 (95% confdence interval, lower limit 0.71, upper limit 0.91) between frst and post-test total mean scores (n � 82).Te results obtained indicate that the time invariance is good.We tested the internal consistency with Cronbach's alpha, split-half reliability, and item-total correlation.Te score for the Cronbach alpha was 0.90 for Spearman-Brown coefcient 0.83, which are above the acceptable level.Table 4 presents the item-total correlations.
We determined the item-total correlation coefcients to be between 0.33 and 0.70 (acceptable level).Te lowest scores are in the frst and nineteenth items, and the highest score is in the ffth item (Table 5).Te original survey items presented in Table 5 are reproduced from [16].

Discussion
Te analysis demonstrated that the scale has acceptable psycholinguistic and psychometric properties in the Turkish nursing context, which was confrmed on examination of CFA, split-half reliability, and item-total correlations.We discussed the results of the analysis under separate subheadings in the following sections.

Stability.
As a result of the test-retest performed at two-week intervals, the ICC coefcient in this study was 0.84.Kho et al. [37] determined the ICC value as 0.92 in their study conducted in four intensive care units in a tertiary medical center in Ontario, Canada.Te results obtained support the good level of time stability for both studies.

Internal Consistency.
In the current study, Cronbach's alpha value was 0.90, and the item-total score correlation coefcients were between 0.33 and 0.70.Kho et al. [37] found the Cronbach's alpha value to be 0.86, and in work with surgical residents in the Netherlands, it was 0.87 [38].Similarly, a study involving 523 physicians and 1321 nurses working in the operating rooms and surgical services of Swiss hospitals found a value of 0.86 for the German version and 0.84 for the French version [19].A Cronbach alpha value of 0.70 and above in these adaptation studies in diferent countries provides evidence for the internal consistency of the scale.
We could not locate an adaptation study comprising split-half reliability and item-total score correlation analysis, and so, these analyses in the present study are the frst to provide evidence of these elements of the internal consistency of the scale.

Content
Validity.CVI values were equal to or above 0.80, which exceeded the threshold value of 0.78 in this study.Martowirono et al. [38] focused on the "appropriateness" of the items, in terms of their scale and scope, rather than of the CVI, and resident physicians evaluated the scale items against this criterion.Tey concluded that the scale items were appropriate for determining the nature of the safety climate.In the same study, the percentage values of the responses of "(6) I have no idea/comment" were calculated for each scale item to support the content validity.Te percentage was below 10% for all scale items, and a value of ten percent was considered acceptable [38].

Limitations.
Since the data refect the participants' perceptions, it is important to take this into account when evaluating and interpreting the fndings.In addition, nurses working in private hospitals did not participate, and so, caution may be needed when generalising to other settings.

Implications for Nursing Management
For patient safety to become embedded in the culture of an organisation, all management levels and the healthcare teams need to prioritize patient safety, establish manageremployee cooperation, and regularly revise patient safety policies and procedures and implement them [39].Te present study demonstrates that the SCS is suitable for use in the Turkish healthcare setting and could be used as part of a continuous programme of quality improvement.It is unidimensional, has 19 items, is acceptable to nurses, and helps highlight issues that require action to improve patient safety.It is not a solution but could certainly be part of a comprehensive approach to prioritizing patient safety, leadership, and learning from mistakes [10,40].Te data  .
Te SCS is reliable, has good time stability, a good Cronbach's alpha value, and acceptable item-total correlation coefcients for use in Turkish health care.It is also valid in terms of language, content, and the model ft.Tese fndings reinforce the evidence for the suitability of the scale for use in a range of settings.However, further work is needed to investigate the impact of the use of the SCS on patient safety outcomes.

Table 1 :
Analyses used in the study.
[5,19,38]est that the SCS can be used successfully in a range of other geographical and health care settings[5,19,38].

Table 5 :
(17)-total correlation coefcients.Te culture of this clinical area makes it easy to learn from the mistakes of others 0.33 (2) Medical errors are handled appropriately in this clinical area 0.63 (3) Te senior leaders in my hospital listen to me and care about my concerns 0.52 (4) Te physician and nurse leaders in my areas listen to me and care about my Briefng personnel before the start of a shift (i.e., to plan for possible contingencies) is an important part of safety 0.57(13)Briefngs are common here 0.62 (14) I am satisfed with the availability of clinical leadership 0.61 (15) Tis institution is doing more for patient safety now, than it did one year ago 0.66 (16) I believe that most adverse events occur as a result of multiple system failures, and are not attributable to one individual's actions 0.41(17)Te personnel in this clinical area take responsibility for patient safety 0.66 (18) Personnel frequently disregard rules or guidelines that are established for this