What is it to be healthy?

According to Christopher Boorse's Bio-Statistical Theory (BST), 'health' is statistically normal function in a reference class, and 'health' and 'disease' are empirical, objective and value-free concepts. I demonstrate that the success of the BST depends on its choice of reference classes; different reference classes result in different accounts of health. I argue that nothing in nature empirically or objectively dictates the use of reference classes Boorse proposes. Reference classes in the BST, and the concept of health, are therefore not value-free. Nor is there a reason to favour the BST over accounts of health that use different reference classes.