On the generalizability of statistical expressions of health related quality of life instrument responsiveness: a data synthesis.

Responsiveness statistics have been suggested as criteria for selecting health-related quality of life (HRQoL) instruments or for calculating the required sample size. This paper raises questions about the stability of estimates of instrument responsiveness. The questions are examined by use of data synthesis. A literature search identified instances where two HRQoL instruments were used concurrently on two or more occasions. An effect size difference between the instruments was calculated. The distribution of the differences in the effect size between pairs of instruments used conjointly in different applications was used as an indicator of variation in responsiveness. The effect size for all possible published uses of generic and disease-specific HRQoL instruments was also examined. Responsiveness was found to be subject to considerable variation. Disease-specific instruments were found to be consistently more responsive than generic instruments. The authors suggest use of responsiveness data should be restricted to expanding pilot study results.