While artificial intelligence (AI) is revolutionizing health care — from improving the accuracy of a diagnosis to predicting a patient’s response to treatment — AI-derived algorithms are only as good as the data they use. Some AI models may invisibly and unintentionally reproduce the biases of their training sets. By definition, there is less data available on minority populations which makes fairness sometimes more challenging to achieve. Using tools to audit machine learning models can find bias and discrimination, allowing for making informed and equitable decisions about data. Not only is having a fair model creation better for health outcomes, it also can help to create efficiencies, improve quality and compliance, and create new revenue. This session details a model that can be replicated across other organizations seeking to use AI responsibly.
Learning objectives are as follows: Understand the challenges posed by the application of AI methods in healthcare, hear examples of how analytics may be used to identify bias that leads to disparities, and how to critique AI methods and assess the trustworthiness and value of an algorithm.
One study Optum has conducted that can be used as an example of this work regarded convening a leadership group focused on the topic of Advanced Analytics and Emerging Technology. As part of this effort, a subgroup was convened to focus on the responsible use of artificial intelligence. This subgroup includes representatives from the fields of data science, analytic methods, technology, and the law.
The Responsible Use Committee developed guiding principles to inform the application of advanced analytics in the enterprise. As part of this initiative, the team conducted a formal review of the legal literature, and reviewed existing AI ethics policies and position statements of relevant healthcare associations and thought leaders (e.g., Alan Turing Institute, American Medical Association, NIST, FDA). Additionally, the group obtained input and review from internal subject matter experts, and held discussions with industry experts.
While artificial intelligence (AI) has tremendous potential to revolutionize healthcare delivery, it also poses new challenges that must be addressed. In this presentation, we will review how AI can be used to improve the efficiency and quality of healthcare delivery, and guidance for the private industry, government, and international players on the use of AI. Additionally, Dr. Meg Good will provide practical guidelines, tools, and frameworks for the responsible use of AI in health care, and describe the different types of bias that may arise during model development and how to assess fairness during model development.
Abstract & Bio
Responsible Use of AI in Health Care