Measuring safety climate in acute hospitals: Rasch analysis of the safety attitudes questionnaire

Participants and setting

The sample for this analysis was derived from nurses that completed the SAQ as part
of the 6-PACK cluster randomised clinical trial (RCT). The sample and sampling procedures
for the 6-PACK project, including hospital and ward selection, have been described
in detail elsewhere 21]. In brief, six public hospitals in metropolitan and regional Victoria, and metropolitan
New South Wales, Australia agreed to participate in the RCT. Hospitals ranged in size
from moderate (200–500 bed) to large (500 beds). Each hospital identified acute surgical
and medical wards where the average length of stay was less than 10 days, where falls
commonly occurred, and had low levels of use of the falls prevention strategies being
tested in the 6-PACK project. Sixteen medical and eight surgical wards were included
in the RCT.

Nurses were invited to complete the SAQ if they had worked on the participating wards
for more than 7.5 h per week in the two months prior to the survey being administered.
Staff that did not meet the above criteria were excluded from this study because they
might have limited knowledge of, or exposure to the ward (and hospital) culture. If
nurses completed and returned the survey, it was assumed that they agreed and consented
to participate in this study. The SAQ was administered to 702 nurses from the 24 acute
wards.

The SAQ

The SAQ is a refinement of the Intensive Care Unit Management Attitudes Questionnaire,
which in turn was derived from the Flight Management Attitudes Questionnaire 7]. The original version of the SAQ consists of 60 items, with 30 core items that are
identical in all clinical settings 12]. A short form version of the SAQ that included the 30 core items and six additional
items of interest to senior hospital leaders was used in this study to measure safety
climate 7], 12]. The SAQ Short Form assesses six-safety related climate domains including teamwork
climate (6 items); job satisfaction (5 items), perceptions of management (6 items);
safety climate (7 items); working conditions (4 items); and stress recognition (3
items) 7]. Five of the items in the questionnaire are responded to separately for the hospital
and ward unit, yielding a total of 41 items 7], 12]. The SAQ Short Form has been used to compare safety climate within and between facilities,
and benchmarking data is available to allow organisations to evaluate their own climate
data 7].

As per Gallego et al. 15], the SAQ Short Form was modified slightly in this study to reflect the Australian
hospital workplace, e.g. substituting ‘clinical area’ for ‘ward’. The SAQ items were
also combined with questions relating to staff knowledge and perceptions of falls
prevention strategies (see Additional file 1). Whilst there may be a possibility that these additional questions may influence
responses to the SAQ items, the participant and logistical burden of administering
two different questionnaires outweighed this risk. Responses for the SAQ items were
recorded using a 5-point numeric scale to reflect the level of agreement with each
individual item. Scores within each domain were calculated and converted to a 100-point
numeric scale 7]. Higher scores indicate greater agreement that more positive attitudes exist towards
the particular safety domain assessed.

Statistical analysis

All data were analysed using SPSS v22.0 (SPSS Inc., Chicago, Illinois) with the Rasch
analysis, being completed using the RUMM2030 package using a partial credit model
(RUMM Laboratory Pty Ltd, Perth, Australia). The methods and criteria for assessing
the measurement properties of the SAQ using Rasch analysis are outlined in Table 1.

Table 1. Statistical tests and criteria for assessment to examine specific measurement properties
of the SAQ

Sample size

In order to obtain an accurate estimation from the Rasch analysis, it is generally
recommended that a minimum of ten categories per response option is available to ensure
that responses are appropriately distributed across the response categories 22]. The sample size required for analysis also depends on whether items in the scale
are targeted properly to participants in the sample 18]. If the scale is well-targeted, a sample size of 108 will be required for accurate
estimation but if it is not, a sample of 243 will be needed 23]. Given that a sample of 420 participants was available in this study, an appropriate
degree of precision can be expected from the Rasch analysis of the SAQ 18].