A methodology for refining AR-DRG

Introduction Previous reviews of AR-DRG, such as that by Aisbett, Wiley et al (2007), have shown prior versions of ARDRG to be among the world’s best in practice, and that further major improvements in grouper performance are unlikely to have occurred. More recent work, such as that by Aisbett, Aisbett, Sutch et al (2008), has shown that hospitals dealing with (age) restricted sub-populations may be disadvantaged by funding mechanisms based on AR-DRG. The understanding here is that DRG systems rely on population-sampling assumptions (as well as matching on influential variables and mathematical modeling) to reduce the risk of biased comparisons of health services. Aisbett’s methodology can be used to identify sets of medical conditions (and procedures) that are associated particularly with increased risk of bias. The research findings also encourage development of AR-DRG along the lines of age-dependent complication levels, so it is appropriate to examine how this knowledge can be implemented to achieve effective refinement. The ultimate aim of this work is to make changes to the current grouper that will lead to better performance as evaluated under the criteria used in the two publications referred to above.