Health care is a credence good, and its market is plagued by asymmetric information. Patients have limited knowledge about the reasons for their ill-being and the proper treatment. Physicians who possess this knowledge may have incentives to exploit the information asymmetry through providing more or more expensive treatment than necessary, which is difficult to verify ex-post. In this paper, we use a laboratory experiment to test the performance of two potential remedies discussed in the theoretical and applied literature. Firstly, we evaluate the effectiveness of two-part tariffs, in which physicians may charge diagnosis fees in addition to treatment fees. Secondly, we provide evidence on the effectiveness of separation of prescription and treatment activities. We observe a significant amount of overtreatment (and a smaller non-predicted amount of undertreatment) in our baseline environment. Contrary to theory, requiring a patient to commit to whichever treatment the physician prescribes does not have a significant effect on overtreatment frequencies. Allowing for two-part-tariffs yields a higher treatment take-up at the cost of lower diagnosis take-up, but does not reduce the extent of overtreatment. The most effective way to reduce overtreatment in our laboratory setting is to require a different than the treating physician to provide diagnosis and prescription for free. This effect, however, is partially offset by an increased frequency of undertreatment. Allowing prescription and treatment doctors to independently set prices for their services reduces efficiency due to coordination failures: in sum prices are often higher than expected patient benefit, who in turn do not attend to the doctor. Also contrary to theory, bargaining power did not play a significant role for the distribution of profits between doctors.