Questions and answers #725

DRG splitting criteria

Added by Kristiina Kahur about 1 month ago. Updated 28 days ago.

Status:ActiveStart date:2020-09-09
Priority:NormalDue date:
Assignee:-Spent time:-
Initiator:Nordic Casemix Centre Target year:2021
MDC:GEN Owner / responsible:Nordic Casemix Centre
Target Grouper: Old forum status:


During the Expert group meeting on September 4th, 2020, the DRG splitting criteria were discussed and agreed as follows:

-For splitting a DRG, there should be at least 20% difference in cost between CC and MCC group as well as between non-CC and CC group
-As a rule, the minimum number of cases per DRG is 50. In case of rare clinical pathology, an exception is allowed
-Age can be used as a splitting criteria if there is a special form of disease for children and a separate DRG is needed
-In all cases when the splitting is considered, the medical description should be taken into account when making the splitting decision regardless of cost difference and number of cases

If there is any modification or clarification needed regarding the criteria, this will be discussed on this ticket.


#1 Updated by Mats Fernström about 1 month ago

Mats Fernström, the National Board of Health and Welfare, Sweden 2020-09-17
I found an old document, written 2008 by Rikard Lindqvist, the first chief of CPK, Centrum för patientklassificering (Centre for Patient Classifications) which later was replaced by NPK, Nätverket för Patientklassificering (Network for Patient Classifications). In his document, Rikard describes the overall criteria for changes in the grouping logic. I think these criteria are still valid and therefore I reproduce the following text from Rikard's document:

“Changes in DRG may be initiated by problems with either cost heterogeneity or clinical relevance, according to the basic idea of all DRG systems: Patient cases are to be assigned to clinically relevant groups with least possible variance in cost.
Main criteria for changes in NordDRG
• The change has relevance for the whole system and is supposed to be functional even in the future
• The change is motivated either by a need for better clinical relevance or by a need for better cost homogeneity, or both
• If the change is motivated by the need for better cost homogeneity, it may not threaten clinical relevance
• If the change is motivated by the need for better clinical relevance, it may not lead to reduced cost homogeneity
• The change may not lead to an uncontrolled increase in the number of groups
For all types of changes (splitting DRG, merging DRG, partial or total re-assignment) there are specified statistical criteria that have to be evaluated with cost-per-case data from at least one of the Nordic countries. Changes can sometimes be made even if not all criteria are met, but in those cases a clear rationale has to be put forward.”

More specified criteria in Sweden since at least 2010 are as follows.
To motivate a new DRG there should be a minimum of cases per year on a national level like:
• 100 inpatients or
• 1000 outpatients
When splitting a DRG:
• the number of cases in each group should be at least 3 % of the number in the original group
• the differences in average costs between the new groups should be at least 20 %
• the coefficient of variation in the original group should not increase
I think that these criteria were copied from the Australian DRG system.

#2 Updated by Kristiina Kahur 28 days ago

Nordic Casemix Centre/Kristiina Kahur 22-9-2020

Thank you Mats for additional information. The criteria you provided complement the ones we discussed in EG Autumn meeting. The difference is in the minimum number of cases - we discussed about 50 and Swedish criteria concern 100 inpatients or 1000 outpatient cases. When it comes to difference in average cost, it's the same - 20%. The rest are whether complementing or very similar.
I suggest we discuss the criteria once again in Spring 2021 meeting and come up with new complemented list of splitting criteria. Nevertheless, for 2022 changes those criteria can be used anyway while considering the splitting of DRG.

Also available in: Atom PDF