This paper deals with balancing groups with respect to prognostic fact
ors in clinical trials, as a means of increasing the power of a study.
We use the Mantel-Haenszel statistic to obtain a simple measure for a
ssessing the relative efficiency of different factors for stratificati
on and to find optimal cutpoints for the given factors. For any availa
ble database, the measure can be used to construct optimal strata for
a clinical trial. The paper also presents a practical approach for red
ucing the number of strata using discriminant regions with a method of
assessing the effectiveness of this approach. An example from a large
database of severe head trauma patients illustrates the utility of th
e measure.