Framing minimal important difference in measuring of quality of life: absolute or relative?

Article type
Authors
Zhang Y1, Zhang S1, Thabane L1, Zhou Q1, Furukawa TA2, Johnston B3, Guyatt GH1
1Department of Clinical Epidemiology & Biostatistics, McMaster University, Canada
2Department of Health Promotion and Human Behavior, Kyoto University Graduate School of Medicine/School of Public Health, Yoshida Konoe-cho, Sakyo-ku, Kyoto, Japan
3SickKids Research Institute & Department of Anaesthesia & PainMedicine, The Hospital for Sick Children, Toronto, Canada
Abstract
Background: Defining the minimal important different (MID) (the smallest improvement that would lead to consideration of an intervention) plays a crucial role in interpreting results of health related quality of life (HRQL) instruments. Some instruments frame the MID as an absolute difference (e.g. 10 points on a 100 point scale) and some as a relative difference (e.g. 15% change in score). Empirical evidence regarding themerits of the two approaches is unavailable.

Objectives: To determine, optimal method of framing minimal important difference in measuring of quality of life outcome.

Methods: We correlated the absolute and relative results with a global rating of change (GRC - a transition question that addressed the degree of improvement or deterioration from a prior visit, from much better to much worse). We judged that whatever standard (absolute or relative) on the target instrument (the HRQL measure of interest) had a higher correlation with the GRC would be preferable. We reviewed the pre-existing databases recording measures of HRQL available in the department of Clinical Epidemiology and Biostatistics at McMaster University and identified studies in which at least two serial measurements of HRQL and a GRC at follow-up were available. We extracted data from each eligible dataset and calculated the correlations between the target instrument and the GRC and generated a pooled estimate of correlations through a meta-analysis.

Results: We identified 320 databases from 1987 till 2004 of which, thus far, 7 proved eligible. The meta-analyses thus far suggest no significant difference between correlation coefficient of absolute and relative difference on HRQOL instrument with GRC on symptoms, emotional function, physical function, and cognitive function, but a significantly higher correlation with absolute differences in the fatigue subdomain. Additional data will be available at the Colloquium.

Conclusions: This project will provide insight into the merits of relative versus absolute measures in determing MIDs.