Improving Subjective Scaling of Pain Using Rasch Analysis

Abstract 

Pain management outcomes assessment depends on valid measurement of pain. However, the validity of single-item scales, such as numeric or faces scales, with the assignment of ordinal numerical values to response scale categories, is questionable. The universal assumption that equal distances between response choices represent equal distances on the dimension being measured is essentially erroneous. Herein we demonstrate that Rasch analysis can be used to expose and repair scale inequity and reengineer scale structure. Thirty-one subjects with severe ocular surface disease repeatedly completed a 7-category faces pain scale. Rasch analysis demonstrated that response category 5 was underutilized, leading to disordering of the response scale. Collapsing category 5 into either category 4 or 6 produced an ordered 6-category faces scale that could be recalibrated with average Rasch person measures to create linear measurement on a continuous latent variable. The value of further alterations to the scale was explored, and the implication for scale redesign discussed. Rasch analysis could be applied to any subjective pain measure post-hoc to create linear measurement or applied during instrument development to optimize design.

Perspective

Single-item scales like the faces scale or a 1–10 numerical rating scale are commonly used for the subjective assessment of pain. However, scores applied to response categories are arbitrary, so do not represent equidistant steps in the underlying latent variable (pain). Scale inequities are easily demonstrable and repairable with Rasch analysis.

Key words:  Calibration , data interpretation (statistical) , outcome assessment (health care) , pain , pain measurement , Rasch analysis