NIH Grant Supports Creation of Non-Biased Machine-Learning Algorithm

The National Institute on Minority Health and Health Disparities has extended a grant to Dascena to support the development of a non-biased machine-learning algorithm that can enhance treatment for acute coronary syndrome.

Dascena is an organization that develops machine-learning algorithms to help improve patient outcomes.

According to the American Heart Association, ACS is a significant cause of morbidity and mortality in the US, and about 805,000 Americans suffer from ACS annually. But, research has shown that ACS treatment and diagnosis are targeted to White men, leading to a lack of evidence in how to assist other populations.

With the new grant, Dascena plans to develop an unbiased prediction tool for diagnosing and treating ACS. The prediction tool will utilize EHR data to create the impartial algorithm.

The machine-learning algorithm will operate through various steps, beginning with reviewing EHR data to identify and eliminate what could be causing bias while sustaining parts of data leading to accurate patient measurements. It will also be trained on preprocessed data and assessed on how it performs across subgroups.

“Machine learning presents an incredible opportunity to reduce bias and disparities in how patients are diagnosed and treated,” said Jana Hoffman, vice president of science at Dascena, in the press release. “Through this grant, we aim to harness the power of machine learning to reduce bias in ACS detection – and ultimately apply our learnings to other conditions to create even greater opportunities for equitable care.”

Adjusting AI algorithms to eliminate bias is becoming a common practice.

For example, to ensure that artificial intelligence standards were ethical, Intermountain Healthcare created a new model that brings together experts from data analytics, applied math and statistics, behavioral science, and multiple other fields.

A study from October 2019 also explained how an algorithm widely used by the US healthcare system contained racial bias. Brian Powers, MD, a physician, and researcher at Brigham and Women’s Hospital and lead author of the study, acknowledged the need to address this issue. Further, the University of Chicago implemented a Center for Applied Artificial Intelligence to combat this problem.  

Researchers have also machine-learning algorithms used by insurance companies. Insurers are increasingly turning to this technology for assistance, using machine learning to identify individuals with complex health needs. Researchers suggest using auditing algorithms and diversifying the data used to train them to remove bias.