Ethical Artificial Intelligence Standards To Improve Patient Outcomes

As providers continue to target improved patient outcomes, more organizations are utilizing artificial intelligence, data analytics, and machine learning. However, organizations should follow ethical standards to ensure positive results when implementing new technology such as artificial intelligence.

To uphold ethical standards and continue improving their AI practices, Intermountain Healthcare created a Data Science and Artificial Intelligence Center of Excellence.

“We’ve been doing this work in data science and artificial intelligence for a very long time, and it became apparent that there was a need for consistency and standards. We also wanted to provide ethical oversight of some of the work that we’re doing,” Assistant Vice President of Analytics Services at Intermountain Healthcare Greg Nelson told HealthITAnalytics.

The new model brings together experts from multiple disciplines, including data analytics, applied mathematics and statistics, computer science, behavioral sciences, econometrics, computational linguistics, clinical informatics, and clinical specialists.

The collaborative effort produced large datasets that the artificial intelligence system could sort through accurately and efficiently.

“We don’t want to just think about the data that’s in front of us. One of the initiatives that we set out last year is we don’t have enough data analysts to look under every corner and every nook and cranny in our organization,” Nelson explained.

Intermountain developed a set of algorithms that can detect whether underlying data sets in their machine learning programs are biased and can identify patterns by sifting through large volumes of data.

“What we wanted to be able to do is use algorithms to help us identify and uncover where we might have blind spots,” Nelson said.

According to Intermountain, the center’s goal is to improve the quality and affordability of healthcare, identify care disparities, enhance the patient experience, and ensure ethical AI practices.

“We can synthesize the insights from our populations of data and see what interventions and what care pathways are working across populations of patients. AI really helps us sift through what’s working,” Nelson continued.

“We realize that all patients are unique, and each has different characteristics, life experiences, access to care, access to medicine, affordability, and so on. The basic hypothesis is if we can help our caregivers sift through that data in a way that is nuanced and personalized for the patient in front of them, that we can achieve better outcomes.”

To eliminate care disparities, it is critical to understand the external challenges an organization is facing. Social determinants of health greatly impact a patient’s access to care, further perpetuating health inequality.

“When we consider whole-person care, you cannot consider the whole person without understanding their access to food, access to greenways, access to grocery stores with fresh fruits and vegetables. Even socioeconomic factors that may limit the affordability clinic access and open availability to times to be able to meet their nuance needs,” Nelson explained.

“We believe strongly that we need to overlay some of our traditional clinical data with social determinants that help us create a better picture and design interventions that are appropriate and personalized for that patient.”

Over the summer, the World Health Organization released ethical standards and guidelines for implementing artificial intelligence into healthcare systems. According to Nelson, Intermountain found their practices to be in line with the recommendation from WHO.

“We were well down a path of establishing our own standards, and it just so happens they aligned well. It was so validating for the work that we had already put into our ethical standards,” Nelson said.