top of page

Algorithmic Bias in Health Care

AI/ML is proving to drastically improve patient outcomes by speeding up diagnosis and mapping out more effective treatment plans. 🔍

However, AI can suffer from bias, which has striking implications for health care. The term “algorithmic bias” speaks to this problem.

What is algorithmic bias? ⬇️

It can be described as systematic and repeatable errors in a computer system, such as AI/ML, that create unfair outcomes and compounds existing inequities, typically based on socioeconomic status, race, ethnicity, religion, gender, disability or sexual orientation.

Bias can sneak into the process at any point in the process, from study design to data selection, algorithm and model choice, even to how data are interpreted and presented.

In order to address the issue of AI/ML bias in the healthcare system, there is the legal approach. Utilizing class action lawsuits and legislation could incent or punish companies that do or do not take steps to mitigate potential bias in their healthcare systems.

The more practical and timely approach, however, would seem to be for healthcare organizations (payors, providers, researchers, clinical trials, PBM’s, government health agencies) to create more inclusive, diverse teams that represent a wider spectrum of backgrounds to oversee AI/ML projects, and not just data scientists.

Healthcare organizations should also develop checklists, safeguards and data standards that identify potential bias not only early on in the AI/ML process but throughout the end-to-end process as well.

Preventing algorithmic bias in your AI/ML models can start with two steps:

1️⃣ Creating a professional, diverse, inclusive data science team that includes both technical and clinical expertise.

2️⃣ Develop checks and balances for every step in the AI/ML end-to-end process.

Awareness and action are key❗️

Does your data science team have proper representation and diversity in thought and background? 📝

Do you have steps in place to address algorithmic bias? 📝

Learn more ➡️


bottom of page