<p>Morbidity-based risk adjustment schemes are of central importance to prevent incentives for risk selection and to ensure fair competition in health insurance systems with regulated competition, including in Germany. However, systematic under- and overcompensation for certain insured groups persists. This study is based on the hypothesis that individuals treated by specialists tend to be undercompensated, whereas those treated by general practitioners tend to be overcompensated. Using routine data from a nationwide sickness fund, we show that this is true and model an approach that extends the classification model of the German risk adjustment scheme by a differentiation criterion depending on general practitioner/specialist treatment. We demonstrate on the basis of three selected morbidity groups, that the introduction of such a split eliminates the misallocation and improves slightly model performance at the individual level. However, and more important, the effects at the group level vary depending on the type of morbidity group included. The findings suggest that in some cases a general practitioner/specialist differentiation can enhance the model performance by reducing risk selection incentives but raises questions regarding administrative complexity, neutrality of care and policy feasibility.</p>