Helfrich CD (IHD-QUERI), Li Y
IHD-QuERI developed Readiness-To-Change (RTC) survey scales to assess factors thought to determine organizational readiness to implement evidence-based changes to clinical practice. The objectives are to assess psychometric properties of the RTC scales and to recommend revisions to improve reliability and parsimony.
Data were 113 survey responses from 17 VAMCs collected for three quality improvement and research projects. The same 77-item RTC scales were used in each. The RTC scales were based on the Promotion Action in Research Implementation in Health Services (PARiHS) model. Survey items were divided among three scales, each representing one of three main constructs of PARiHS: Evidence, Context and Facilitation. The corresponding scales include four, six, and nine subscales, each comprising 2 to 6 items. We conducted item analyses and scale reliability to assess convergent and divergent validities of subscales. The underlying factor structure of scales was assessed using exploratory factor analysis.
Reliability was high for the Context (alpha = 0.87 – 0.94) and Facilitation subscales (alpha = 0.84 – 0.94), and modest for the Evidence subscales (alpha = 0.70 – 0.75). Item-rest correlations were 0.41 to 0.90, supporting their inclusion in subscales. High cross-scale correlations (e.g., Pearson’s r > 0.50) were seen for some conceptually related items, such as leadership culture of the Context scale, and leadership characteristics of the Facilitation scale. Exploratory factor analysis of subscales revealed three distinct factors largely corresponding to the three constructs of the PARiHS. However, the Facilitation "leadership" subscale of the Facilitation scale loaded onto the Context factor, and "resources" subscale of the Context scale failed to load on any factor. Factor loadings and uniqueness statistics suggested that “patient preference” subscale of the Evidence scale may be a distinct construct.
Most subscales of the RTC survey are reasonably reliable, and conform to the conceptual model they are based upon. Reliability and parsimony of the scales may be improved by revising items cross-loaded on multiple scales, and the "resources" and "patient preference" subscales warrant further development.
Theoretical models can only guide research and quality improvement activities to the extent that constructs can be operationalized. For this, well-validated instruments are essential.