Improving risk equalization using information on physiotherapy diagnoses

BACKGROUND: Worldwide, risk-equalization (RE) models in competitive health insurance markets have evolved from simple demographic models to sophisticated models containing diagnosis and pharmacy-based indicators of health. However, these models still have important imperfections; adding information on (diagnoses of) physiotherapy treatment may further improve RE-models. Therefore, a new risk-adjuster based on physiotherapy costs in the prior year was introduced in the Dutch RE-model of 2016. METHODS: Physiotherapy claims-data (2012) and administrative data on costs and risk-characteristics (2013) for 94% of the Dutch population (N = 15.8 million) are used to evaluate the current risk-adjuster based on physiotherapy costs and to assess the effects of replacing it by different modalities of a risk-adjuster based on physiotherapy diagnoses. Of the 89 diagnoses in the claims-data, 62 are dropped because they relate to temporary health problems. The 27 retained diagnoses are added to the Dutch model in 4 modalities: 27 separate risk-classes, 9 diagnosis-clusters based on main pathology category, 4 diagnosis-clusters based on residual costs, and the 4 clusters of modality 3 interacted with age. RESULTS: Although the cost-based risk-adjuster improves the model’s predictive power and removes the average undercompensation (€919) for enrollees with physiotherapy costs in the prior year, it is outperformed by all 4 diagnosis-based modalities. Of these modalities, modality 3 is preferred based on its simplicity and comparable predictive power. CONCLUSIONS: Adding information on physiotherapy can further improve the performance of sophisticated RE-models. Regarding the Dutch model, a risk-adjuster containing 4 risk-classes for clustered diagnoses based on residual costs is the preferred modality. ELECTRONIC SUPPLEMENTARY MATERIAL: The online version of this article (doi:10.1007/s10198-017-0874-x) contains supplementary material, which is available to authorized users.