Approaches to Multidimensionality
in DRG Development

The continuing DRG refinement process over the last 20 years has not led to satisfying homogeneity concerning resource intensity.1 Other techniques have to be developed to gain more homogeneity. Some hopeful approaches have appeared in several countries which could lead to simpler DRG systems or similar PCSs using multidimensional techniques.2

Publications by producers of DRG systems and similar PCSs were analysed for the purposes of this comparative study.

Different approaches to DRG refinement were found:

• Systematic grouping of secondary diagnoses.
• Separating ICU treatment.
• Diagnostic groups (status) and procedural groups (action) are defined separately.
• Allowing multiple groups to be assigned to one case.
• Cost weights split into per-case and per-day components.

Approaches using real multidimensionality are rare although they have the potential to reduce the number of case groups rapidly. In such approaches, cost weight calculations have become more difficult, but several algorithms have been developed.

To overcome the weak homogeneity of existing DRG systems, more efforts should be made to use multidimensional grouping techniques which lead to a small number of groups with a large differentiation potential.

In Australia, a "Complication and Comorbidity Level" (CCL) with 5 levels has been defined for each diagnosis dependent on Base-ARDRG, sex and discharge destination. For each hospital stay, a "Patient Clinical Complexity Level" (PCCL) with 5 levels is calculated. The definitive AR-DRG is an aggregation formed from the two quasi-dimensions "Base-ARDRG" and "PCCL". As a result, the "Base-ARDRGs" are split into up 4 different CC-categories.3

In Austria, the cost weight of each LDF is split into a "performance component" and a "day component". It is allowed to attribute more than one surgical LDF to each stay. The cost weight of a stay includes the "performance components" of all LDFs and the "day component" of the main LDF.4

The length of intensive care is used as a second PCS dimension.5

In France, a project called "Effeuillage Progressif" (EfP; "Progressive Defoliation") proposes assigning more than one Base-GHM to each stay. The cost weight is calculated on the basis of multivariate statistical methods.6

In the U.K., the "Healthcare Frameworks" (HCF) which are under construction consist of two-dimensional matrices. One dimension is for needs described by "Health Benefit Groups" (HBG) which group "conditions at risk", "presented conditions", "confirmed diseases" and "continued consequences of disease". The other dimension is for interventions of the categories: "prevention and health promotion", "investigation and diagnosis", "clinical management of the disease" and "continuing care". For this dimension "Healthcare Resource Groups" (HRG) are used wherever it was possible. The HBG dimension of needs is complemented by outcome indicators. The HRG dimension of interventions is complemented by structure and process indicators.7

In the USA, the RDRG, APR-DRG and IAP-DRG systems use explicitly defined Base-DRGs which are subdivided into a constant number of 3 or 4 resource intensity levels.

Two PCSs developed in competition with DRG systems allow the assignment of more than one case group and one cost weight to a stay: "Disease Staging" (D.S.)8 and "Patient Management Categories" (PMC)9.

The approaches to DRG refinement always started by taking all diagnoses into account (and not only the principal diagnosis and the most severe secondary diagnosis). The differences in the approaches lay in the way in which information about resource intensity derived from the diagnoses in the discharge record was encoded and aggregated. Two main approaches could be found:

1. Encoding and aggregation of "Comorbidities or Complications" (CCs) in an early stage of the algorithm, resulting in one DRG and one cost weight per stay:
   1. DRG list without explicitly defined Base-DRGs: HCFA-DRGs, AP-DRGs, GHMs, AN-DRGs.
   2. DRG list with Base-DRGs subdivided into a constant number of 3 or 4 subgroups: RDRGs, APR-DRGs, IAP-DRGs; Base-ARDRGs in combination with PCCLs.
2. Assignment of one or more than one case group to each stay with a more or less sophisticated algorithm to calculate one cost weight. (LDFs, D.S., PMCs, GHM-EfP).

Additional grouping criteria such as emergent admission, same-day patient, functional impairment, and discharge destination are seldom used. Such criteria have been used at best as split criteria of some DRGs and never as an additional grouping dimension. There is one exception: intensive care.

The best approach to construction of flexible systems is the use of dimensions which are defined independently of each other.10

Essential dimensions of classification are shown in Table 2.

Most important independent dimensions of PCSs in acute care should be:

• Diagnostic base groups based only on diagnoses (irrespective of procedures) without splits for age or CCs,
• Procedural base groups as case groups based only on procedures.
• Severity levels or scales pertaining to the diagnostic base groups, e.g. CC levels.
• Intensity levels or scales pertaining to procedural base groups, e.g. ICU, tracheostomy, multiple procedures.
• Type of stay to reduce the necessity of redefining the PCS while changing the ordinary type of stay from inpatient to outpatient for an increasing number of case groups.
• Temporal units/time frames to model the treatment course.

Dimensions to consider additionally are:

• Disabilities / functional status, which are especially relevant for chronic diseases and for handicapped patients.
• Emergency admission, which becomes a more relevant cost driver when carrying out more and more diagnostics before hospitalisation.
• Discharge destination (planned), which can be used as a kind of indicator for the objectives of care.

Additional options for PCS construction could be:

• Different aggregation levels for clinicians and managers because managers need less differentiation, and remuneration will be easier while maintaining a high explanatory power for clinicians.
• Multiple groups for one case with possibilities of aggregation and combined weighting (models are GHM-EfP and Disease Staging).
• Numbering system that traces the hierarchies and dimensions modelled.