Fig. 3

Preference Weights and Relative Importance of Attributes. The line graphs at the top of the figure represent preference weights estimates of each attribute level with 95% confidence intervals (CIs). A more positive preference weight indicates a stronger preference for the level. For example, physicians preferred no fatigue to severe fatigue. The vertical distance between an attribute’s best and worst levels is a measure of the relative importance of the attribute. Relative attribute importance scores (with 95% CI) presented at the bottom of the figure were calculated by expressing this vertical distance as a percentage of the sum of vertical distances across all attributes. The larger the score, the greater the impact that variations in the attribute levels had on treatment choices