Background: The Safety Attitudes Questionnaire (SAQ) is commonly used to assess staff perception of safety climate within their clinical environment. The psychometric properties of the SAQ have previously been explored with confirmatory factor analysis and found to have some issues with construct validity. This study aimed to extend the psychometric evaluations of the SAQ by using Rasch analysis.
Methods: Assessment of internal construct validity included overall fit to the Rasch model (unidimensionality), response formats, targeting, differential item functioning (DIF) and person-separation index (PSI).
Results: A total of 420 nurses completed the SAQ (response rate 60 %). Data showed overall fit to a Rasch model of expected item functioning for interval scale measurement. The questionnaire demonstrated unidimensionality confirming the appropriateness of summing the items in each domain. Score reliabilities were appropriate (internal consistency PSI 0.6-0.8). However, participants were not using the response options on the SAQ in a consistent manner. All domains demonstrated suboptimal targeting and showed compromised score precision towards higher levels of safety climate (substantial ceiling effects).
Conclusion: There was general support for the reliability of the SAQ as a measure of safety climate although it may not be able to detect small but clinically important changes in safety climate within an organisation. Further refinement of the SAQ is warranted. This may involve changing the response options and including new items to improve the overall targeting of the scale.
Trial registration: This study was registered with the Australian New Zealand Clinical Trials Registry, number ACTRN12611000332921 (21 March 2011).
Keywords: Patient safety; Rasch analysis; Safety culture.