Artificial intelligence is becoming an increasingly important part of healthcare, helping doctors detect diseases earlier and make better treatment decisions.
Low blood sugar, also called hypoglycemia, is a common problem in hospitals. It happens when the level of glucose in the blood falls too low. Glucose is the body’s main source of energy, especially for the brain.
When blood sugar drops too far, people may feel shaky, sweaty, hungry, dizzy, confused, or weak. In severe cases, hypoglycemia can cause seizures, loss of consciousness, coma, dangerous heart rhythm problems, and even death if it is not treated quickly.
Hospital patients can develop low blood sugar for many reasons. Some take insulin or other medicines for diabetes. Others may need to stop eating before surgery or medical procedures.
Critically ill patients may also have changes in the way their bodies control blood sugar. Doctors and nurses usually discover hypoglycemia only after it happens by checking blood sugar levels. This means treatment is often reactive instead of preventive.
A new study from Cedars-Sinai Health Sciences University offers a promising solution. The research, published in npj Digital Medicine, describes an artificial intelligence system that can identify patients who are likely to develop low blood sugar up to 24 hours before it happens.
The research team developed an AI model known as a long short-term memory, or LSTM, model. This type of artificial intelligence is designed to recognize patterns that develop over time. Instead of looking at a single blood test, it studies many pieces of information together and learns how changes over several days may signal future risk.
The system examines information that hospitals already collect through electronic health records. These include medications, laboratory test results, meal records, blood sugar measurements, and other clinical information.
The AI reviews these data every four hours over a five-day period and then estimates whether a patient is likely to experience hypoglycemia within the next 24 hours.
To build the model, researchers analyzed data from more than 143,000 adult hospital admissions across three Cedars-Sinai hospitals between 2014 and 2025. After developing the model, they also tested it using new hospital data collected in real time to make sure it worked outside the original dataset.
According to senior author Dr. Roma Gianchandani, hospital teams currently respond only after blood sugar has already fallen. The new AI system could give doctors and nurses an early warning so they can adjust medications, provide food sooner, or monitor patients more closely before dangerous blood sugar levels develop.
Lead author Dr. Amanda Momenzadeh explained that the model does more than predict risk. It also identifies the main factors contributing to the prediction, helping healthcare teams understand why a patient is at risk and what actions may reduce that risk.
The researchers estimate that a large hospital could prevent three or four cases of hypoglycemia every day by using this tool.
While this number may sound small, preventing even a few severe episodes can protect patients from serious complications, shorten hospital stays, and reduce healthcare costs. If similar systems were adopted by hospitals around the world, the overall benefit could be substantial.
Senior author Dr. Jesse Meyer emphasized that this is not simply a computer experiment. The model was designed and tested to work with real hospital information as patients receive care, making it much more practical for everyday clinical use.
This study is one of the largest efforts to develop an AI system for predicting hospital-related hypoglycemia. Its strengths include the very large number of patients, testing across multiple hospitals, and prospective validation using real-time data.
However, the model still needs to be evaluated in hospitals outside the Cedars-Sinai health system to confirm that it performs equally well in different healthcare settings and patient populations. Future studies should also determine whether using the AI system actually reduces complications and improves patient outcomes in routine practice.
Even so, the findings suggest that artificial intelligence could help move hospital care from reacting to medical emergencies toward preventing them before they occur, making treatment safer for patients with diabetes and many other conditions.
If you care about diabetes, please read studies about Vitamin D and type 2 diabetes, and to people with diabetes, some fruits are better than others.
For more health information, please see recent studies that low calorie diets may help reverse diabetes, and 5 vitamins that may prevent complication in diabetes.
Source: Cedars-Sinai Health Sciences University.
