Abstract:
Statistical Graphics
to Assess Patient Classification Systems

Abstract (English)

With regard to the introduction of a SwissDRG system in Switzerland, the question arises how well such patient classification system would represent paediatric cases. Against this background, appropriate methods for the analysis of DRG systems and for the suggestion of improvements were being sought.

Data covering 21,194 cases in 2005 were available of eight children's hospitals and paediatric departments of acute hospitals, combined in the «DRG work group» of the «Swiss Paediatric Society» (SGP). The data sets comprised APDRG codes, age, length of stay and partly also LEP times (LEP® = Workload Measurement in Nursing). For this study, 20,436 data sets could be used (of which 15,344 with LEP times). Altogether, 445 APDRGs were occupied. Out of them, there were 124 APDRGs with 30 or more paediatric cases. Additionally, reference could be made to overall Swiss data issued by the «Federal Statistical Office» (FSO, in German: BFS) of 2003 and 2004, with more than 800,000 faultlessly grouped cases from acute hospitals and paediatric clinics.

To obtain an overview of these data, a spoke plot as well as treemaps comprising the case proportions per DRG were compiled. As a demonstration of homogeneity, the spread of care days and LEP times were visualised by box plots and multiple density diagrams. The DRG key figures, that is, case weight, expected length of stay and trim points were recalculated on the basis of the SGP's data and compared with the standard values provided by the APDRG-Switzerland association (APDRG-CH). In this, the technique of «difference diagrams» developed by the author was used.

The 124 selected APDRGs were furthermore checked on whether paediatric age splits on the basis of the SGP's data could be suggested. To this end, each APDRG was divided as per the following age groups: 0–10 days, 11–28 days, 29 days – 1 year, 1–3 years, 3–6 years, 6–10 years, 10–14 years, 14–18 years. For an assessment of the age splits, the proportions of explained spread with regard to length of stay and LEP times were calculated. A list of assessment values was compiled, on the basis of which the candidates for analysis were determined. For an assessment of these DRGs, one page was programmed with various diagrams and a semi-automated comment was provided.

This study introduces methods, by means of which DRG homogeneity can be examined and ways of suggesting improvements can be found. Numerous APDRGs with significant spread were detected. (Some of them should be examined more closely as candidates for systematic age splits.) In case of many APDRGs which are already split by «age below 18 years», the values for expected length of stay and case weight calculated on the basis of the SGP's data were above the APDRG-CH values.

The results indicate that it is necessary to examine both the DRG key figures (such as length of stay and case weight) and homogeneity of the DRGs from a paediatric point of view. In this, the diagrammed evaluations – the treemaps, difference diagrams and multiple box plots, in particular –, help gain a survey of the existing data and also indicate in which areas there is a need for action. The question is to be asked whether any further criteria other than the existing ones used for the DRG classification are to be sought to achieve a sensible paediatric APDRG refinement.

Refinement of a DRG system is not the only possibility to achieve a more adequate reimbursement. Modular systems might turn out to be not only more appropriate but could also be more transparent than continuously refined DRG systems.