BACKGROUND. Adequate risk-adjustment is critical to the success of market-o
riented health care reforms in many countries. A common element of these re
forms is that consumers may choose among competing health insurers, which a
re largely financed through premium-replacing capitation payments mostly ba
sed on demographic variables. These very crude health indicators do not ref
lect expected costs accurately.
OBJECTIVE. This study examines whether the demographic capitation model can
be improved by incorporating information on the presence of chronic condit
ions deduced from the use of prescribed drugs. The revised Chronic Disease
Score was used to incorporate this information in the model.
METHODS. A panel data set comprising annual costs and information on prescr
ibed drugs for 3 successive years from Dutch sickness fund members of all a
ges, is used for the empirical analyses (N = 55,907). The predictive perfor
mance of the demographic model is compared with that of a chronic condition
s and a Pharmacy Costs Groups (PCG) model, which is a demographic model ext
ended with information on clustered chronic conditions.
RESULTS. The predictive accuracy of the demographic model substantially imp
roved when the model was extended with dummy variables for chronic conditio
ns. The 23 chronic conditions could be clustered into six PCGs without affe
cting the predictive accuracy. Based on these PCGs 17% of the members were
bad risks with a mean predictable loss that exceeds the overall average exp
enditures.
CONCLUSIONS. The use of information on chronic conditions derived from clai
ms for prescribed drugs is a promising option for improving the system of r
isk-adjusted capitation payments.