Patient centered utilities allow the construction of endpoints which directly reflect patient preferences in outcome. Clinical scales like the modified Rankin scale (mRS) used in stroke studies have reasonable ordering, but do not necessarily reflect patient values (a change from 2 to 3 may be viewed differently by a patient than a change from 3-4). Clinical utilities allow an interpretable value to be placed on each item in the scale. More efficient statistical analyses may be based on mean utility than simply using the ordered values.
We will describe a utility weighted approach to mRS, including the agreement between multiple elicitations of that utility, and its use an adaptive Bayesian trials such as DAWN. We will examine added value of the utility both statistically and scientifically. For example, a standard analysis of mRS dichotomizes mRS into 0-2 and 3-6. This is mathematically equivalent to a utility of 1 on 0,1,2 and a utility of 0 for 3,4,5,6, which would clearly not be chosen by any patient.