Risk adjustment: beyond patient classification systems

Diagnosis related groups (DRGs) are widely used in several countries. Their various versions aim to value the cost of hospital production. In Europe, the patient classification systems and standard weights used are usually the American originals.

The objective of this study was to analyse the extent to which DRGs and DRG-weights explain patient cost variability. Different components of patient cost (severity, comorbidities, complications and socioeconomic status), which are not well explained by DRG and which can be approximated by using administrative data, were also analysed.

A total of 35,262 discharges from two public hospitals in Barcelona were analysed. The Health Care Financing Administration (HCFA)-DRGs and the All Patient Refined (APR)- DRGs were calculated. Severity was adjusted by Disease Staging, and comorbidities and complications were calculated using Elixhauser and Charlson comorbidities groupings. An ecological socioeconomic status indicator was used. Linear regressions were estimated to explain per-patient cost variability.

We found that Medicare's DRG-weights explained only 19% of cost variability. Cost-based weights explained nearly 40% (38-42%, depending on the DRG classification used). Exclusion of outliers increased explanatory power to R² = 47-48%. The remaining adjustment variables increased R² to 49-51%.

Medicare's DRG-weights are not well-suited to Europe. Cost-based DRG-weights and outlier trimming have significantly greater explanatory power. The remaining clinical and socioeconomic variables have considerably less explanatory power but were statistically significant and behaved as expected. Spanish and other European health authorities should adapt DRG-classification systems to their environments for use in hospital production cost valuation.