I apply a Bayesian persuasion framework to study information transmission in the physician-patient relationship. I highlight each party’s financial incentives in this transmission and the subsequent determination of medical treatment outcomes. I estimate the model using a large database of claims from China, exploiting a policy change with independent sources of random variation in physicians’ and patients’ financial incentives. For a diagnosis for which surgical treatment is somewhat discretionary, fully informing patients, equivalently eliminating physicians’ financial incentives, reduces surgery rates by almost half and raises total welfare by 60 percent. This effect is greater when patients are better insured.