Crystal balls for PD care: How predictive models can help us see ahead

Abstract

Care teams and patients want to know what happens next, and researchers have put together a lot of tools, such as predictive models, to help them predict the future. While these researchers are well-intentioned, the tools they develop are not always helpful. Most researchers know enough to perform various tests of their predictive models, such as statistical tests that answer the question: “Are the predictions based on this model better than a coin flip?” We urge researchers to add another test to their existing lists: “Does this model tell care teams anything they don’t already know?”