In response to growing concerns about patient safety, the Institute of Medicine (IOM) in the USA and the Department of Health in the UK have recommended that healthcare organisations should consider adopting safety management techniques used in other industries.12 This shift in focus has been driven by the recognition that organisational, managerial and human factors rather than purely technical failures are often prime causes of accidents. High-risk industries, such as aviation, nuclear power and oil and gas, pay considerable attention to assessing safety and have increasingly focused their attention on the development of “leading indicators” such as safety audits or measurements of safety climate, rather than purely relying on measures based on retrospective data such as accident rates. It has been suggested that this supports safety condition monitoring, minimising the need to wait for failures to occur before identifying weaknesses in the system.3

To successfully implement patient safety initiatives, it is also increasingly recognised that there is a need to understand the safety culture of healthcare organisations.4 As described by Reason5 safety culture refers to the

“shared values (what is important) and beliefs (how things work) that interact with an organisation’s structures and control systems to produce behavioural norms (the way we do things around here).”

Different approaches may be used to explore these organisational characteristics. For instance, indepth interviews or focus groups can provide detailed insights into individual and group perceptions, although these methods are time consuming and resource intensive. In practice, self-report workforce surveys are often used to assess what is referred to as “safety climate”.6 Safety climate can be regarded as the surface features of the underlying safety culture discerned from the workforce’s attitudes and perceptions at a given point in time.78

Safety climate assessments have been developed in a range of high-risk industries.79 In healthcare, only limited attention has been paid to psychometric factors in the design of safety climate instruments, and most instruments have been derived from work in other industries with only limited consideration given to whether their use is valid in different settings.610 To our knowledge, no safety climate measure has been developed for use in community pharmacies. The aim of this study, therefore, was to first develop and then explore the component structure and internal consistency of a safety climate questionnaire for use in the community pharmacy setting.

METHODS

Community pharmacists were recruited by attending postgraduate training events on risk management provided by the Centre for Pharmacy Postgraduate Education (CPPE) in England. Participants were provided with freepost envelopes and asked to complete and return a self-administered questionnaire. All participants were voluntary and anonymous. The University of Manchester Senate Committee on the Ethics of Research on Human Beings approved the study.

The questionnaire measured respondents’ agreement with statements on nine themes, namely: commitment to patient safety (five items), communication in the pharmacy (five items), staffing and management (six items), education and training about safety (four items), team working (three items), perceptions of the causes of incidents (four items), incident reporting (three items), investigating incidents (five items) and learning following an incident (seven items). The themes and items were drawn from a qualitative self-assessment safety culture framework (MaPSAF, Manchester Patient Safety Assessment Framework) that we have previously developed through literature review and focus groups for use in community pharmacies.11 Responses were indicated on a five-point categorical scale: 1, strongly disagree; 2, disagree; 3, neither agree nor disagree; 4, agree; 5, strongly agree.

Data analysis

Principal components analysis (PCA) was performed on the questionnaire items using oblique rotation, communalities >0.4 and eigenvalues >1.12 PCA is commonly used with ordinal Likert scale data, provided that the scale has at least five categories. Other approaches to factor analysis of ordinal variables have been proposed, but from a practical point of view their use has been limited as they are computationally demanding and most analyses want to take account of many variables with many factors.13 PCA reveals the underlying structure of a scale and shows whether there are distinct components or themes being measured. It requires reasonably large datasets (sample sizes above 300 generally provide a stable component solution).1214 Tests of sampling adequacy (Kaiser–Meyer–Olkin (KMO)), multicollinearity (Bartlett test of sphericity p<0.05) and residuals were undertaken to check that the scale items were appropriate for PCA.1214

The items comprising each component were used to interpret the meaning of the components. Scales were constructed for each of the components on the basis of items that loaded significantly on the components and tested for reliability, using Cronbach α scores. This indicates the internal consistency of the scales from 0 to 1, with scores of 0 indicating no consistency (the items are unrelated to each other) and scores of 1 indicating that the items are practically identical. Opinions differ regarding the cut-off point for acceptable scores, but scores <0.6 are generally considered problematic. All calculations were performed using SPSS version 13.0.

RESULTS

Characteristics of the respondents

A total of 998 community pharmacists (607 women (60.8%); mean (SD) age 48 (10.4) years, range 21–76) completed the questionnaire. Respondents included pharmacy owners (159, 16%) and employee (402, 40%) and locum pharmacists (437, 44%) working in community pharmacies throughout England. The length of time that the respondents had worked in community pharmacies ranged from 1 year to 55 years, with a mean duration of 23 (11.1) years.

Principal components analysis

PCA was conducted on the items included in the Pharmacy Safety Climate Questionnaire (PSCQ). The KMO measure of sampling adequacy and the Bartlett test of sphericity were used to determine the appropriateness of conducting PCA. Typically, the value of KMO is required to be >0.5, and the Bartlett test result must be statistically significant (p<0.05). For the current dataset, KMO was equal to 0.97, and the Bartlett test was statistically significant (p<0.001).

Examination of the unrotated component structure revealed that a substantial proportion of the variance (37.2%) was explained by a single component that had significant loadings on eight of the 42 items. Subsequently, oblique (direct oblimin) rotation was applied and after deletion of items with low communalities and eigenvalues <1, seven components were retained which represented the model of choice, explained 58.3% of the data variance and incorporated 34 items. Table 1 shows the wording of the items, component structure and item loadings, along with the Cronbach α scores and variance explained by each component.

Although it has been suggested that components loading significantly on three or fewer items should not be retained,12 in this study it made conceptual sense to accept three components each containing three items. They were judged strong enough to warrant inclusion in the model in terms of (i) the loadings of the items on the components and (ii) acceptable internal reliability on the basis of Cronbach α scores. The seven components that defined the PSCQ were interpreted as relating to:

1. investigating and learning from incidents (eight items);

2. staffing and management (five items);

3. perceptions of the causes of incidents and reporting (six items);

4. team working (three items);

5. communication (six items);

6. commitment to patient safety (three items);

7. education and training about safety (three items).

The variance explained by the individual components ranged from 4.9% to 8.5%. All items had communalities above 0.4 and a component loading threshold of 0.4 was also imposed on the model to aid interpretation.14 Pearson correlation coefficients were calculated to investigate the inter-relationships between the PSCQ scale scores (table 2). The scales were moderately to highly inter-related; observed correlations ranged from 0.441 to 0.725, and all were statistically significant (p<0.001, two-tailed).

DISCUSSION

The main aim of this study was to develop and evaluate the component structure and internal reliability of a safety climate questionnaire suitable for use in community pharmacies. From the results of the PCA, we have developed a 34-item questionnaire in which a seven-component solution fit the data reasonably well, explaining 58.3% of the data variance. The identified components showed good internal reliability, and logical inter-relationships. For instance, positive attitudes to investigating and learning from safety incidents were found to be strongly related to good communication, provision of education and training about safety, and a more positive commitment to patient safety within the pharmacy.

The seven-component solution corresponded well with the patient safety dimensions included in MaPSAF, from which the questionnaire items were originally derived11 and which has recently been promoted for use in community pharmacies by the National Patient Safety Agency in the UK.15 Interestingly, the components also agreed well with common dimensions of patient safety climate, such as communication, reporting, risk perception, staffing and teamwork, which have been used in instruments applied in other healthcare settings.610 Unlike some of the other industry-derived instruments, we generated items for inclusion from first principles, through focus group discussions with community pharmacists and their support staff, in order to ensure that the resulting items had high face validity.11

In practice, we anticipate that the PSCQ will have a range of uses, which include: measuring pharmacy staff attitudes about seven safety climate domains; comparing attitudes within and between pharmacies; identifying in which of the seven safety climate components it would be most effective to focus attempts to improve patient safety; and evaluating safety interventions and tracking changes over time.4 In the hospital setting, safety climate instruments have been used to identify the strengths and weaknesses within clinical areas and prompt the implementation of appropriate interventions.16 Findings from this work suggest that the prevailing climate can be targeted and improved, and that these improvements are associated with reductions in medication errors and shorter lengths of hospital stay.16

This 34-item questionnaire provides a measure of safety climate in the community pharmacy setting, meeting criteria on component structure and internal reliability. However, further work will be needed to examine the instrument’s other psychometric properties including measures of test–retest reliability and predictive validity. In particular, more research is needed on the relationship between PSCQ scores and other variables such as the numbers of complaints received, medication errors reported, job satisfaction and staff absenteeism and turnover.