Fairness in Medical Algorithms - Threats and Opportunities

Explore the critical issue of fairness in medical algorithms through this 41-minute conference talk. Delve into the challenges and opportunities surrounding AI adoption in healthcare, focusing on the disproportionate impact on minority patients during the COVID-19 pandemic. Examine the variable performance of AI models on unseen datasets and the potential bias in outcome proxies like healthcare costs. Learn about practical approaches to implementing fair medical AI and understand the difficulties in operationalizing fairness. Discover the background and context for both fair and unfair consequences of AI algorithms in healthcare. Engage with a grand challenge that presents open questions in the field. Cover topics such as racial bias in pulse oximetry measurements, performance of ICU severity scoring systems across ethnicities, geographic distribution of US cohorts in deep learning algorithms, and strategies for fighting bias in training. Gain insights into fairness in AI reporting guidelines, algorithmic approaches to reducing pain disparities, and the recalibration of race in medical research.

Intro
Disclosures
Racial Bias in Pulse Oximetry Measurement
Performance of intensive care unit severity scoring systems across different ethnicities in the USA: a retrospective observational study
Geographic Distribution of US cohorts Used to train Deep Learning Algorithms
Current status
Towards FAIR Algorithms in RL
Fighting Bias in Training
Saliency maps
Fairness in Al reporting guidelines (review)
An algorithmic approach to reducing unexplained pain disparities in underserved populations
Recalibrating the use of Race in Medical Research
Call to Action