Panelists:

• Yolonda Wilson, associate professor, Saint Louis University, Albert Gnaegi Center for Health Care Ethics
• Joanna Abraham professor anesthesiology, Washington University in St. Louis, Institute for Informatics
• Thomas Maddox, cardiologist, BJC HealthCare/Washington University, vice president, Digital Products and Innovation
• Sunny Lou, instructor in the Divisions of Cardiothoracic Anesthesiology, Clinical and Translational Research (DoCTR), and Institute for Informatics at the Washington University School of Medicine in St Louis
• Karen Blum, independent journalist and AHCJ core topic leader on health IT (moderator)

By Mara Silvers

Health care professionals and researchers are using artificial intelligence to smooth workflows, protect patient safety and reliably analyze complex data. But to be effective, experts say the new tools must be designed and used wisely.

Incorporating artificial intelligence in health care doesn’t mean researchers are being aided by robots as envisioned in fictionalized Hollywood films. Instead, panelists said computer programs can be trained to do specific tasks like reviewing medical imaging from CAT scans, X-rays and pathology units and flagging abnormal conditions. 

“All of these have really granular imaging elements that these algorithims can parse apart and extract data features to give an incredible degree of accuracy,” said Dr. Thomas Maddox, a cardiologist at Washington University School of Medicine in St. Louis and vice president of Digital Products and Innovation. 

Machine learning can also be used in high stress settings to alert health care providers about medical complications and risks following an operation. When used correctly, Dr. Joanna Abrahams said, those programs can serve as a backstop for accuracy and safety when physicians are dealing with stress and exhaustion.

“The idea is ... how can we use machine learning or artificial intelligence to prevent some of these post-operative complications and support the physicians to make better decisions?” Abraham said.

Some machine learning programs could be designed to help lower costs and reduce waste by pinpointing which materials and supplies are needed for an upcoming patient surgery. Dr. Sunny Lou highlighted research that found artificial intelligence could help avoid oversupplying for potential blood transfers. 

But as the medical field looks to expand uses of AI, experts cautioned that — as with any medical device — users should pay careful attention to how tools are created and their shortcomings. 

Dr. Yolonda Wilson highlighted how machine learning and other computer programs can be built on biased data or medical theories such as race-norming and risk repeating historic bias. That challenge is important to think about, she said, “as we write about AI [and] as we glorify AI.”

Mara Silvers covers health and human services at Montana Free Press.