Author: Hiwot Tesfaye, Greg Horne
Publisher: SAS
Publication Year: 2022
Summary: The following podcast episode discusses how algorithms in healthcare are often used for accessing risk. Pitfalls come from not understanding how the healthcare system has worked historically and how different populations interact with healthcare systems. Accuracy of algorithms at a global level and the individual subpopulation level is discussed as well. It is not a new problem; even medical devices used for decades are biased. Tools have the power to affect the course of people’s lives on a great scale. The episode encourages listeners to harness a sense of empathy and recognize that you do not know how algorithms are actually going to perform in the real world. It is important to ask, “Is it appropriate in this context to try to automate this process?” If it is, at every step of the analytics lifecycle, analysts need to ask themselves questions, such as “Is the data representative of the population I’m trying to effect change in?” and “Is the target variable distributed equally across various groups?”