Using propensity score to create an equivalent comparison group ( Last revised : 12 / 4 / 2012 )

Background: It is often the case in research that groups of interest represent a small number of records or have other distinct characteristics. For example, the group may consist of those with extreme values (either high or low), such as Opioid problem users (Subramaniam, Ives, Stitzer, & Dennis, 2010), or have other unique characteristics such as those who are deaf or hard of hearing (Titus, 2008). There are several ways to adjust for these size and characteristic differences. One that has garnered a great deal of interest and use in recent decades is the use of a propensity score to either weight the control group or create a matching algorithm so that the control group more closely resembles the treatment group or to select a matching subset.