This article explores a three‐party contracting problem when the patient and the provider possess private information that is unobservable to the insurer. We show that for an insurance mechanism to be collusion‐proof, it suffices for the insurer to rely on the incentive for one side of the patient‐provider coalition. If the risk premium for the patient is smaller than the provider's informational rent, placing the incentive on the patient generates a lower social cost than placing the incentive on the provider. We show that if the provider's effort is highly valued by the patient, the insurer should rely on the patient's incentive to implement a collusion‐proof second‐best insurance. Interestingly, an altruistic provider may lead to a higher social cost than a self‐interested provider. However, even if the insurer does not know the degree of provider altruism, it may still achieve the second‐best outcome by assuming that the provider is self‐interested. The model can be further extended to allow for different objective of the insurer, provider's informational advantage over patient, and auditing.
