Using AI to Save Lives in Rural Alaska

Brian Rice, MD, uses machine learning to analyze medevac utilization in remote areas of Alaska where air transport is the lifeline for emergencies.

A mother in a small, rural Alaska village takes her child to the local community health aide. The child is having trouble breathing. The aide, who has limited medical training, must refer to a manual throughout the exam. The community is connected to the outside world only by airplane so the health aid must decide if the child’s condition warrants an airlift medevac to a larger facility hundreds of miles away.

Healthcare providers in Alaska do not have a shared, common classification to determine when to call for a medevac, despite the reliance on air transport in crises. Rice, associate professor of emergency medicine, hopes to offer providers a decision support tool to save lives.

Thanks to a five-year grant from the National Institutes of Health, Rice is leading a mixed methods study consisting of interviews and data analysis of health care records. Machine learning models will analyze appropriate medevac utilization in some of the most remote areas of the United States.

Rural Alaska is made up primarily of Alaska Native communities connected by air. 100% of the areas Rice is analyzing are not connected to the road system. When emergencies occur, the only way for patients to see a doctor is to fly hundreds of miles. A health aide, with limited information, must communicate with doctors at a distant hospital to determine if the patient needs a medevac.

The information about when and why a patient gets a medevac is scattered, according to Rice, and written in unstructured ways into clinical notes. Natural language processing is required to analyze what went into the decision. These data include clinical and non-clinical features such as time of day, experience levels of providers, and, most importantly, severe weather conditions that can make the difference between life and death in rural Alaska.

Rice is creating the model himself. He learned programming in his youth to construct computer versions of the Choose Your Own Adventure novels. More recently, Rice wrote programming to analyze free-text chief complaints presented at hospitals in Uganda. For his Alaska project, Rice is being mentored by Tina Hernandez-Boussard, PhD, Stanford professor of medicine (Biomedical Informatics).

Rice hopes the study will provide an understanding of what questions providers should ask undifferentiated patients. Therein lies a distinction between precision medicine and precision emergency medicine — while much of precision medicine focuses on therapeutics, precision emergency medicine places a greater emphasis on diagnostics.

According to Rice, “Finding the needle in a haystack is a lot different than sharpening the needle."

Updated Spring 2024