Evaluation of the Dutch Risk Equalization System: Are the Insurers Confronted with Predictable Losses for the Chronically Ill?

Abstract

Rationale: Since 2006 everyone in the Netherlands must buy private health insurance from one of the competing insurers. A risk equalization system should sufficiently compensate the insurers for the differences in portfolio mix regarding the age, gender and health status of their insured. The current (2007) risk equalization formula in the Netherlands is predominantly based on age, gender, urbanisation, disability, Diagnostic Cost Groups (DCGs) and Pharmacy-based Cost Groups (PCGs). The regulation requires open enrollment and community rating. In case of insufficient risk equalization insurers are confronted with predictable losses on the chronically ill. These losses create incentives for risk selection, which can have several adverse effects. Objective: To evaluate the risk equalization system. The following research questions will be answered: 1. Are there identifiable subgroups of consumers with predictable lossses? 2. If so: How large are these subgroups? And how large are the predictable losses? In particular we focus on subgroups of persons with a chronic condition or with above average utilization rates in previous years. Methodology: We used the following data base: all information in the files of a large insurance company over the period 1997 - 2004 combined with an individual health survey (held in 2001) with many questions about the insureds' health status. In total some 17,000 observations. We were able to apply the 2006 and 2007 risk equalization formulae to simulate predictions of the insureds equalization-payments for the years 2003 respectively 2004. By comparing these predicted expenditures with their actual expenditures we calculated the average profits and losses for many identifiable relevant subgroups. Results: Many subgroups of chronically ill that can be easily identified by the insurers, generate substantial predictable losses. The size of these subgroups ranges from less than 1% to 30%. The predictable losses are in the order of hundreds to thousands euros per person per year. The results indicate predictable losses also for subgroups of insureds whose disease was included as a risk adjuster in the risk equalization formula (e.g. heart problems, cancer and rheumatism). The improvements in the 2007-formula relative to the 2006-formula result in an average reduction of about 10% of the predictable losses. Conclusion: The current (2007) risk equalization in the Netherlands compensates the insurers insufficiently for many identifiable high risk groups. In particular those insurers who make themselves attractive for the chronically ill e.g. by providing or contracting the best care for them, are insufficiently compensated. Consequently these insurers have to raise their premium which worsens their market position, in particular relative to insurers who have a low premium because they are successful in risk selection. Although there is an open enrolment requirement, insurers in the Netherlands have many tools for subtle forms of risk selection at their disposal. This risk selection can have many adverse effects in terms of affordability, efficiency and quality of care. A substantial improvement of the risk equalization system needs a high priorty. With an insufficient risk equalization system the disadvantages due to risk selection may outweigh the advantages of a competitive market.