Using Rasch measurement theory to explore the fitness for purpose of the genetic counseling outcome scale: a tale of two scales.

Abstract

The genetic counseling outcome scale (GCOS-24) is commonly used in clinical genetics to measure patient empowerment; however, there is inconclusive psychometric evidence about this scale. Using data from an urban Canadian clinic where the GCOS-24 is routinely administered before (T1) and 1month after (T2) genetic counseling, we used Rasch measurement theory (RMT) to test the ordering of response option thresholds, fit, spread of item locations, residual correlations, person separation index (PSI), and stability across time. Data from 379 participants showed that the original GCOS-24 items had poor fit to the Rasch model (χ2 = 367.8, p < 0.001). Two models emerged that demonstrated excellent fit to the Rasch model. In Model 1, the response scale options were collapsed and 8 items were removed, leading to an excellent fit to the Rasch model (χ2 = 112.4, df = 144, p = 0.975), good reliability (rp = 0.82), and responsiveness to change (mean = 0.75 logits, F = 125.68, p < 0.001). In Model 2, the response scale options were collapsed, 3 items were removed, and the scale divided into two sets (GCOS-Pos, GCOS-Neg). The GCOS-Pos set showed excellent overall fit to the Rasch model (χ2 = 92.5, df = 90, p = 0.407), good reliability (rp = 0.73), and responsiveness to change (mean = 0.74 logits, F = 80.12, p < 0.001). The GCOS-Neg set showed excellent overall fit to the Rasch model (χ2 = 84.55, df = 97, p = 0.81), but poor reliability (rp = 0.19) and small, but significant responsiveness to change (mean = 0.19 logits, F = 28.73, p < 0.001). These models show that there are psychometric issues with the GCOS-24 scale, and our study provides options for how to measure empowerment more robustly.