How Algorithms Can Exacerbate Racial Disparities - New Research Insights

Decisions that influence our daily lives are increasingly guided by artificial intelligence. While AI has great potential to evaluate data more efficiently, flawed assumptions in the underlying algorithms can perpetuate biases and exacerbate existing disparities.

A new study by NIHCM grantee Ziad Obermeyer and colleagues found significant racial bias in an algorithm that uses health care costs to identify patients with high health needs.

• Since black patients currently face bias in our health care system, they are sicker than white patients at the same cost level.
• As a result, the algorithm underestimates the needs of black patients and is less likely to identify them as candidates for care management programs.

This research highlights the need to assess the potential for bias in algorithms used throughout the health care industry and beyond.