Separating the true effect from gaming in incentive-based contracts in health care

This paper studies the effect of incentive regulation on health care. In the context of incentive-based health contracts, which might also introduce an incentive for the providers simply to report better treatment outcomes, evaluation of treatment using the information supplied by the providers (reported output) could be problematic. The systematic error on the output report is called providers' gaming behavior. This paper develops a general method for decomposing the effect of incentive-based contracts on performance into the true effect, which is the result of clinicians' improved effort induced by the contract, and the gaming effect, which is due to the change in the providers' reporting practice. The method follows the essence of linear structural relation (LISREL) models, and the true treatment output is modeled using a latent variable. Various output measures can be included in the structural evaluation model, but objective measure(s) (output measures not affected by providers' potential gaming) must be constructed based on available information to identify gaming through its correlation with the reported measures. The strengths of this method are that information from more than one output measure can be used, no monitoring system is required, and the construction of a gold-standard measure is not necessary. This method is applied to evaluate the impact of Maine's performance-based contracting on its public providers' substance-abuse services. Evidence of gaming is found in Maine's system, which remains robust in most of the sensitivity analyses. The methodology developed here can be used to evaluate the impact of a broad range of incentive-based contracts.