Pharmacy cost groups for the German morbidity-based risk compensation scheme.

Introduction: To ensure fair competition and prevent risk selection by sickness funds, Germany employs a morbidity-based risk-adjustment scheme, primarily using diagnostic data to record insured persons' morbidity. However, concerns about the manipulability and quality of diagnostic coding have sparked discussions. This study proposes and evaluates an alternative risk-adjustment model based on pharmaceutical data, assessing its potential as an extension or an alternative to the diagnosis-based status quo.

Methods: We adapted an existing pharmacy-based model to German conditions and simulated various models. In order to create comparability to the status quo, we constructed a representative sample for the German statutory health insurance (SHI), using claims data of about 4.5 million insured persons. We evaluated the sample by assessing the standardized differences of the weighted means of the relevant covariates. For a quantitative assessment of the models we used the coefficients of determination (R²), Cumming's Predictive Measure (CPM), and the mean absolute prediction error (MAPE). Under- and overcompensation within different risk groups were also analysed.

Results: The sample closely matched SHI data (overall effect size after matching < 0.0001). Substituting diagnostic data with pharmacy cost groups (PCGs) showed comparable model quality, but worsened under- and overcompensation for groups vulnerable to risk selection. Conversely, integrating PCGs into the status quo improved nearly all performance measures.

Conclusion: Introducing pharmacy-based models into the German risk compensation scheme demonstrates significant potential. Extending the current model with PCGs enhances statistical performance, improves morbidity measurement, and offers a viable approach to mitigate coding manipulation incentives.