Improving risk equalization with constrained regression.

Area of Science:

• Health economics
• Biostatistics
• Health services research

Background:

• Current risk equalization models in health insurance exhibit systematic incentives for risk selection due to under- and overcompensation.
• Existing methods to mitigate these issues include adding risk adjustor variables, but suitable variables are not always available for all groups.

Purpose of the Study:

• To propose and evaluate a novel approach using constrained regression to reduce under- or overcompensation in risk equalization models when specific risk adjustor variables are unavailable.
• To quantify the trade-offs between reducing under/overcompensation for some groups and potentially increasing it for others.

Main Methods:

• The study proposes constraining estimated coefficients in risk equalization models to achieve a fixed level of under- or overcompensation for specific groups.
• Constrained regressions were empirically evaluated using individual-level data from the Netherlands (N=16.5 million).

Main Results:

• Constrained regressions can reduce under/overcompensation for certain groups compared to ordinary least-squares, but may increase it for others.
• The benefits of reduced under/overcompensation for targeted groups can outweigh the costs of increased disparities in other groups.

Conclusions:

• Constrained regression provides a valuable tool for developing more equitable and economically efficient risk equalization models.
• This approach can improve the performance of health plan payment schemes by addressing limitations in current risk adjustment strategies.