Roughly, overdiagnosis (ODx) occurs when people are harmed by receiving diagnoses (often accompanied by interventions) that do not benefit them, usually because the diagnosed conditions do not pose a threat to their health. ODx is a theoretical as well as a practical problem as it relates to definitions of disease. Elsewhere, it has been argued that disease is a vague concept and that this vagueness may contribute to ODx. In response, we develop a stipulative or précising definition of disease, for the specific purpose of decreasing or preventing ODx. We call this diseaseODx , aimed at distinguishing cases where it would be beneficial to identify (and treat the condition) from those where diagnosis is more likely to harm than benefit. A preliminary definition of diseaseODx is that X is a diseaseODx iff there is dysfunction that has a significant risk of causing severe harm. This paper examines the 3 concepts in this definition, using a naturalistic account of function, a Feinbergian account of comparative harm, and a probabilistic understanding of risk. We then test the utility of this approach using examples of clinical conditions that are currently overdiagnosed.