Validation of the Safety Attitudes Questionnaire for Assessing Patient Safety Culture in Critical Care Settings of Three Selected Ugandan Hospitals

Abstract

Background: The safety attitudes questionnaire (SAQ) short form (2006) is important for assessing patient safety culture in clinical environments. However, little is known about its validity and applicability in Uganda. This study validated the SAQ short form (2006) for use in assessing patient safety culture in critical care settings of hospitals in the Ugandan context. Methods: Using a sequential exploratory mixed-methods research design, the face, content and construct validity for the SAQ short form (2006) was assessed in a multi-phased approach. A panel of eight (8) purposively selected experts assessed the face and content validity in rounds 1 and 2, respectively, while construct validity was assessed in round 3 using data from a cross-sectional survey of 256 frontline health workers in critical care settings of the selected hospitals. Analysis of survey data followed confirmatory factor analysis. Cronbach’s alpha examined internal reliability. Results: Of the 36 items in the tool’s original version, 33 were rated clear, with a score of 100% on face validity. The use of contextual vocabulary and formatting issues arose as concerns. The S-CVI/Ave was 100%, and S-CVI/UA was 86.1%. Four new items added on effective communication as another dimension of patient safety culture. The survey had KMO=0.8605, the a priori-based model had a scale Cronbach’s alpha=0.8881, with unsatisfactory goodness of fit (RMSEA=0.051, 90% CI: 0.044–0.057, pclose=0.427; chi-square=694.28, p <0.001; CFI=0.884, TLI=0.871). The modified final model had a scale Cronbach’s alpha =0.8967 and satisfactory goodness of fit (RMSEA=0.030, 90% CI: 0.019–0.039, pclose=1.000; chi-square=424.98, p=0.002; CFI=0.966, TLI=0.960). Conclusion: In the tool’s original form, the face validity was lacking despite satisfactory scores on item clarity. Content validity was adequate, while construct validity required modifications in construct specifications. Reliability was adequate before and after specification modifications. The modified version has adequate psychometric properties for Uganda