The aim of this paper is to investigate how risk attitudes in medical decisions for others vary across health contexts. A lab experiment was designed to elicit the risk attitudes of 257 medical and non-medical students by assigning them the role of a physician who must decide between treatments for patients. An interval regression model was used to estimate individual coefficients of relative risk aversion, and an estimation model was used to test for the effect of type of medical decision and experiment design characteristics on elicited risk aversion. We find that: (i) risk attitudes vary across different health contexts, but risk aversion prevails in all of them; (ii) students enrolled in health-related degrees show a higher degree of risk aversion; and (iii) real rewards for third parties (patients) make subjects less risk-averse. The results underline the importance of accounting for attitudes towards risk in medical decision-making.