by James Ralston, MD, MPH, a Kaiser Permanente physician and senior investigator at Kaiser Permanente Washington Health Research Institute
As a general internist, I have cared for several patients with diabetes who have had episodes of severe hypoglycemia — very low blood sugar — while taking insulin. Severe hypoglycemia can be a frightening and demoralizing experience for patients and their families.
I remember one patient who took years to rebuild the confidence and skills he needed to manage his insulin after experiencing severe hypoglycemia.
As food raises blood sugar, and medications and exercise lower it, people with diabetes may feel like they’re doing a kind of balancing act to maintain blood sugar levels that aren’t too high or low. Often, as with this patient, they may not realize when their blood sugar is dropping — and can decline quickly into severe hypoglycemia.
Hypoglycemia can lead to falling, car crashes, heart attacks, coma, and even death. And each year in the United States, around 100,000 hypoglycemia-related events result in emergency room visits. So I’ve long wished we could have a way to recognize which diabetes patients are at highest risk for low-blood-sugar emergencies like this.
Finally we have a tool that can help, and I’m delighted.
JAMA Internal Medicine just published a study led by my colleague, Andrew Karter, PhD, from Kaiser Permanente Division of Research that describes this tool. I was a coauthor on the paper along with Geoffrey Jackson, MHA, of KPWHRI and our other colleagues from Kaiser Permanente Northern California’s Division of Research, Yale, the University of Chicago, and the Center for Healthcare Organization and Implementation Research at Edith Nourse Rogers Memorial Veterans Hospital: Development and Validation of a Practical Tool to Identify Patients with Type 2 Diabetes at High Risk of Hypoglycemia-Related Utilization.
Dr. Karter led our team through developing and validating a risk stratification tool to identify patients with diabetes type 2 who are at the highest risk for having to visit an emergency room or hospital because of severe hypoglycemia.
This work started with collecting information from more than 200,000 patients at Kaiser Permanente Northern California. We used machine learning to develop a model to predict a patient’s risk of using an emergency department or hospital in the next year because of hypoglycemia.
We found and considered using 156 possible risk factors for hypoglycemia. But we ended up using only these six:
We validated the tool with electronic health record information routinely collected from more than 1.3 million members of the U.S. Veterans Health Administration and nearly 15,000 members of Kaiser Permanente Washington, all with type 2 diabetes.
Thanks to this tool, we can start working more closely with the small number of patients at highest risk — and, we hope, prevent future hypoglycemia and related emergency and hospital visits.
We are grateful to the U.S. Food and Drug Administration for funding the development of this tool under their Safe Use Initiative. The Centers for Medicare and Medicaid Services (CMS) is helping to disseminate the results. And several large health care systems and organizations — including CMS, the Mayo Clinic, and Kaiser Permanente — are now seeing how they can use the tool.
Learn more about Kaiser Permanente Washington Health Research Institute. Sign up for our free monthly newsletter.