Risk equalization in The Netherlands: an empirical evaluation

Abstract

The Netherlands relies on risk equalization to compensate competing health insurers for predictable variation in individual medical expenses. Without accurate risk equalization insurers are confronted with incentives for risk selection. The goal of this study is to evaluate the improvement in predictive accuracy of the Dutch risk equalization model since its introduction in 1993. Based on individual-level claims data (n = 15.6 million), we estimate the risk equalization models that have been successively applied in The Netherlands since 1993. Using individual-level survey data (n = 8735), we examine the average under-/overcompensation by these models for several relevant subgroups in the population. We find that in the course of years, the risk equalization model has been substantially improved. Even the current model (2012), however, does not eliminate incentives for risk selection completely. To achieve the public objectives, further improvement of the Dutch risk equalization model is crucial.

Keywords::

• claims data
• health insurance
• risk equalization
• risk selection
• survey data

Acknowledgements

The authors thank the following persons for their valuable comments on previous versions of this article: three anonymous reviewers, the members of the Risk Adjustment Network (RAN), in particular Florian Buchner, and the Advisory Group (‘Begeleidingsgroep’).

Disclaimer

The opinions and views expressed in this paper are those of the authors and do not necessarily reflect the opinions or views of the NPCF or those of the persons mentioned above. The responsibility for the content of this article fully rests with the authors.

Financial & competing interests disclosure

The authors gratefully acknowledge the Dutch Federation of Patients/Consumers (NPCF) for financing this study in part. The authors have no other relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript apart from those disclosed.

No writing assistance was utilized in the production of this manuscript.

Key issues

• • The Netherlands relies on risk equalization to compensate competing insurers for predictable variation in individual medical expenses.

• • The predictive accuracy of the Dutch risk equalization model substantially improved over the years, for example, for people who reported a chronic condition in year t-1 the current risk equalization model (2012) reduces the average undercompensation in year t by about 70% compared with no risk equalization at all.

• • Even the current model, however, undercompensates for particular subgroups in the population, for example, for people who reported a chronic condition in year t-1 it leads to an average undercompensation in year t of more than €400 per person per year.

• • Since Dutch insurers are not allowed to risk rate their premiums, undercompensation of identifiable subgroups confronts them with incentives for risk selection, which threatens quality of care as well as solidarity between low-risk and high-risk individuals.

• • Further improvements in risk equalization are necessary to safeguard public objectives.

• • Promising potential improvements with the currently available data are i) extending pharmacy-based cost groups (PCGs) and/or diagnoses-based cost groups (DCGs), ii) allowing for a restricted set of multiple DCGs and/or interactions among DCGs, iii) developing a health classification based on the prior use of durable medical equipment and iv) developing a health classification based on the prior use of physiotherapy.

• • As long as the risk equalization model is imperfect, it may be desirable to reduce remaining incentives for risk selection by risk sharing and/or allowing some degree of risk rating.

• • In case of imperfect risk equalization, decisions on risk sharing and risk rating require complex tradeoffs between risk selection, efficiency and solidarity.

Notes