BR, South, Chapman WW, Dankwa-Mullan I, Matheny M, and Quintana Y. 2021. “Challenges and Opportunities for Implementing Artificial Intelligence at the Speed of Technology Innovation During the COVID-19 Era”. In AMIA 2021 Virtual Informatics Summit. American Medical Informatics Association Conference.
amia_aiincovidera-08262020.pdf129.37 KB
Abstract
The COVID-19 pandemic has created multiple opportunities to implement Artificial Intelligence (AI) technologies in new ways that address the initial infectious curve (e.g., triaging patients and disseminating information during disease outbreaks), as well as the subsequent curves of pandemic sequelae (managing gaps in care of chronic conditions, addressing new and exacerbated mental health needs, and rectifying worsening health disparities. However,
numerous challenges limit scaling development and application of AI technologies in healthcare settings, especially in the context of a rapidly evolving public health emergency. Data representing diverse patient cohorts are necessary both to train and to test systems but often are labor intensive to create and deidentify. The need for new codes and concepts can delay data availability. Biases in data must be identified, evaluated, and managed to mitigate
downstream effects. System performance must be continuously monitored and validated as clinical information, such as disease transmission characteristics, become available. This panel will discuss these challenges and propose solutions that include ensuring adequate, equitable, and unbiased data sources are used for AI development, validation of AI in clinical settings, with the context of the rapidly evolving COVID-19 public health crisis as a discussion focus.
Last updated on 03/07/2023
numerous challenges limit scaling development and application of AI technologies in healthcare settings, especially in the context of a rapidly evolving public health emergency. Data representing diverse patient cohorts are necessary both to train and to test systems but often are labor intensive to create and deidentify. The need for new codes and concepts can delay data availability. Biases in data must be identified, evaluated, and managed to mitigate
downstream effects. System performance must be continuously monitored and validated as clinical information, such as disease transmission characteristics, become available. This panel will discuss these challenges and propose solutions that include ensuring adequate, equitable, and unbiased data sources are used for AI development, validation of AI in clinical settings, with the context of the rapidly evolving COVID-19 public health crisis as a discussion focus.