A comparison of the separation ratio and coefficient alpha in the creation of minimum item sets

Short-form outcomes measures are becoming common in response to demands for increased efficiency in health care. This study examines Rasch measurement as an aid to selecting items for short form tests. The focus of this paper is on maintaining test quality while reducing items. The separation ratio (SR) aids item reduction by indicating how removing items impacts measurement precision. Results of the SR and coefficient alpha are compared.

To demonstrate the use of Rasch measurement to shorten clinical outcomes measures and to compare the separation ratio and coefficient alpha in evaluating when item reduction improved efficiency without sacrificing measurement precision.

Retrospective analysis of existing health outcomes data.

A convenience sample of 58 patients receiving cataract surgery.

The 14 items of the VF-14 (a measure of visual functioning), the published subset of items from this test (the VF-7), and 5 other 7-item combinations of the items.

The largest coefficient alpha was obtained from the VF14 (.84) while the largest separation ratio (2.67) was obtained from the 7-item subtest with the reduced rating scale.

This study demonstrated one way that Rasch measurement can be helpful in selecting items for minimum item sets while maintaining test precision. Both alpha and the separation ratio provide information about how a sample performed with a given test although variations in measurement precision may not always be detected with alpha.