We analyse a model of patient decision-making where anxiety about the future characterizes the patient's utility function. Anxiety corresponds to fear of bad news and results in the patient being averse to information. First, the patient chooses the accuracy of a signal which discloses information on his health status. Then he up-dates his beliefs according to Bayes's rule and chooses an action. We show that the choice of imprecise information can be optimal because it allows the patient to trade off the damage deriving from complete ignorance with the anxiety raised by the news about his health level.