The model addresses a longstanding challenge in hospital care. Low blood sugar, also called hypoglycemia, is a common and potentially life-threatening complication among hospitalized patients, including those receiving diabetes treatment, those who are fasting before procedures or those in critical care. However, there are no widely used tools for predicting which hospitalized patients may develop hypoglycemia.

“Today, most hospital care for hypoglycemia is reactive, and we respond after a patient’s blood sugar drops,” said Roma Gianchandani, MD, senior author of the study and vice chair of Quality and Innovation in the Department of Medicine and program director for Diabetes.  

The AI model developed by Cedars-Sinai investigators analyzes patterns in medications, lab results, meals and other data from patients’ electronic health records. It collects the information in four-hour intervals over a five-day period and uses it to predict whether a patient will develop hypoglycemia within the next 24 hours.

Researchers developed and tested the model using data from more than 143,000 adult hospital admissions across three Cedars-Sinai Health System hospitals between 2014 and 2025. Investigators also tested the tool using prospective hospital data to confirm their initial findings.

“The AI model is designed to alert patient care teams before a patient experiences low blood sugar and identify the key factors driving that risk,” said Amanda Momenzadeh, PharmD, lead author of the study and a project scientist in the Meyer Research Lab at Cedars-Sinai. “By offering actionable insights to care teams, it also aims to support hospital diabetes management programs.”

Researchers estimate the tool could help prevent about three to four cases of low blood sugar at a large hospital each day. Extrapolating across all hospital beds worldwide, the impact could be substantial.

“What’s exciting is that this isn’t just a theoretical model, but instead, it is built and validated to work prospectively in real time using data hospitals already collect,” said senior author of the study Jesse Meyer, PhD, assistant professor in the Department of Computational Biomedicine at Cedars-Sinai. “By identifying patients at risk earlier, we have an opportunity to reduce preventable complications and improve patient safety.”

If widely adopted, the model could lead to more proactive, data-driven care for hospitalized patients with diabetes and other conditions that affect blood sugar.

Additional Cedars-Sinai authors: Caleb Cranney, Dennis Chen, and Elizabeth Nguyen.

Funding: NIGMS R35GM142502, NIH National Center for Advancing Translational Science (NCATS), and UCLA CTSI Grant Number UL1TR001881

Disclosure: Jesse Meyer, Amanda Momenzadeh, and Caleb Cranney are listed as inventors on a patent application related to this AI tool.

Cedars-Sinai Health Sciences University is advancing groundbreaking research and educating future leaders in medicine, biomedical sciences and allied health sciences. Learn more about the university.

“Patients are not usually hospitalized because of diabetes, but because of another condition, so they are not admitted in one specific area of the hospital,” said Roma Gianchandani, MD, medical director of Diabetes Quality and vice chair of Quality and Innovation at Cedars-Sinai. “It’s important that specialized protocols and programs that prioritize diabetes care are available hospital-wide.”

Diabetes affects how the body turns food into energy, sometimes causing high blood sugar, also known as high blood glucose or hyperglycemia. Hospitalized patients who do not have diabetes also can develop high blood glucose—a condition called stress hyperglycemia—which can lead to poor outcomes.

Gianchandani, an endocrinologist, along with Cedars-Sinai Fellow Margaret Wei, MD, authored a paper published in Annals of Internal Medicine about hospital best practices for managing high blood glucose. Gianchandani leads the medical center’s interdisciplinary efforts to improve outcomes for hospitalized patients with diabetes or stress hyperglycemia.

The Cedars-Sinai Newsroom spoke with Gianchandani about these efforts.

What does inpatient diabetes management involve?

Hospitals should use diabetes experts to help patients reach their safest blood sugar levels, which is shown to shorten hospital stays and improve outcomes. There is no single best diabetes team format—institutions should do what works with their own processes—but the most effective teams use experts to manage a patient from admission to discharge and conduct post-discharge follow-ups.

At Cedars-Sinai, we have a heterogeneous team format that includes endocrinologists (faculty and private), endocrinology trainees, diabetes educators, pharmacists and information technologists with a focus on hospital diabetes management. We also are increasingly integrating diabetes technology into our medical ordering system to help non-diabetes practitioners better manage their patients. Our diabetes clinicians are familiar with the latest diabetes technologies and increasingly used devices like continuous glucose monitors.

What’s new about Cedars-Sinai’s approach to inpatient diabetes care?

Collectively, our teams have prioritized several key components that have been shown to improve outcomes. We have standardized care by creating nursing policies for monitoring blood sugar and administering insulin, as well as procedures for addressing low blood sugar.

We are at the forefront of technology integration. We are using a computerized tool to support nurses administering insulin through an IV—the nurse simply enters the patient’s blood sugar, and the tool calculates how much insulin to give without the nurse needing to make that complex calculation. We also have insulin pumps and glucose sensors integrated into the patient’s electronic medical record, so we can monitor patients’ blood sugar as needed after they go home. This would allow us in the future to follow up if the patient needed to adjust their insulin dosage.

I am proud to share that Cedars-Sinai also has a pharmacist surveillance program. A pharmacist reviews abnormally high and low blood sugar levels every morning and reaches out to patients’ physicians if medication orders need to be adjusted.

We also are evaluating predictive algorithms using artificial intelligence tools for managing diabetes. For example, monitoring patients over time, an algorithm could predict when their blood sugar might go too high or too low.

How does Cedars-Sinai educate diabetes patients in the hospital?

A hospital stay provides a unique opportunity for diabetes education. We have several educators in the hospital as well as outpatient educators. Our inpatient educators interact with patients during admission to address any gaps in care they may be experiencing, and they strive to ensure patients go home with the resources they need to maintain safe blood sugar levels.

Read more on the Cedars-Sinai Blog: How to Prevent Insulin Resistance and Keep Blood Sugar Levels Steady

The study results, published in Annals of Internal Medicine, could help researchers evaluate how the prescription patterns of GLP-1RAs have changed over time, particularly for the use for treating Type 2 diabetes versus obesity.

GLP-1RA medications are a newer class of medications that were originally developed to treat patients with Type 2 diabetes. However, given their promising results, in 2021 the U.S. Food and Drug Administration approved the medications for chronic weight management in adults with obesity or those who are overweight with at least one weight-related condition, such as high blood pressure, prediabetes or blood vessel disease.

Since then, the demand for these drugs has skyrocketed, with a marked surge in GLP-1RA prescriptions for weight loss management. However, the steady swell in demand has also had downstream effects leading to nationwide drug shortage and issues with accessibility.

“Essentially, after the medication was approved for obesity, GLP-1RA use took off so quickly that we lost control and vision of how fast people were picking up these medications, and the trends of use are uncertain,” said the study’s co-first author Ali Rezaie, MD, medical director of the Cedars-Sinai GI Motility Program and director of bioinformatics at the Medically Associated Science and Technology (MAST) program at Cedars-Sinai. “While GLP1-RAs offer several benefits, they are also associated with various common and uncommon side effects, necessitating careful monitoring of their prescription patterns.”   

To gain insight into how GLP-1RAs have been prescribed and consumed over the years, Cedars-Sinai investigators used a real-world nationwide database containing medical records of about 45 million individuals with at least one outpatient or inpatient visit in the U.S. from 2011 through 2023.

During this timeline, the investigators found approximately 1 million new GLP-1RA users and classified them based on whether they had diabetes, obesity or other related conditions when they were prescribed the medication.

The researchers noted that between 2011 and 2023, semaglutide (a GLP-1RA) users were disproportionately female, non-Hispanic white and with a body mass index of 30 or greater. During the same period, the data also revealed that the proportion of new users with Type 2 diabetes decreased. “On the contrary, new GLP-1RA prescriptions among those with obesity or relevant comorbid conditions without Type 2 diabetes doubled, with a notable uptick in medication use especially since 2020,” Rezaie said.  

This noticeable shift and rising trend in GLP-1RA prescriptions for those with obesity, scientists say, is crucial.

“This data suggests that more healthcare providers are seeing the benefits of these medications for treating obesity, which is a significant public health shift,” said Yee Hui Yeo, MD, co-first author of the study and a clinical fellow in the Karsh Division of Gastroenterology and Hepatology at Cedars-Sinai. “However, it also raises concerns about potential medication shortages and the need to ensure that patients with diabetes still have access to these treatments.”

The researchers believe that the study's findings can play a crucial role in developing strategies to address the growing popularity and demand for GLP-1RAs.

“We hope our findings will inform healthcare policies and promote equitable access to these essential treatments,” Yeo said.

Additional Cedars-Sinai authors involved in the study are Srinivas Gaddam, MD, MPH; Ghada Mohamed, MD; Rabindra Watson, MD. Additional authors include Kevin Sheng-Kai Ma, DDS; Tina Yi-Jin Hsieh, MD; Xiaoqin Hu, MD; Gin Yi Lee, MD; Pin-Chia Huang, MSc.

Read more on the Cedars-Sinai Blog: Weight-Loss Drugs – What to Know Now

The findings, published in the peer-reviewed journal Diabetologia, also identified three therapeutic pathways that reversed these changes and partially restored wound-healing function to the cornea—a discovery that could ultimately inform new treatments for diabetes.

“We have found that diabetes induces more cellular changes than we were aware of previously,” said Alexander Ljubimov, PhD, director of the Eye Program at Cedars-Sinai’s Board of Governors Regenerative Medicine Institute and senior author of the paper. “The discovery does not affect gene sequence but entails specific DNA modifications altering gene expression—what are known as epigenetic alterations.”

More than 37 million people in the United States—11% of the population—have diabetes, a systemic disorder that can result in kidney disease, heart disease, amputation, stroke and nerve damage. Most diabetes drugs are designed to increase glucose tolerance or supply depleted insulin, but do not address molecular and cellular changes or their associated complications.

The new research also identifies for the first time an important role of Wnt-5a, a secreted signaling protein investigators found responsible for corneal wound healing and the function of stem cells—cells capable of differentiating into many cell types.

“Current treatments only address symptoms, so there is an urgent need to understand the molecular mechanisms of diabetes-related wound-healing problems,” said Ruchi Shah, PhD, a scientist in Ljubimov’s lab and the study’s first author. “Understanding of this novel epigenetically regulated wound-healing mechanism could lead to therapeutic treatments that could help patients avoid further long-term ocular health issues.”

Though much focus of diabetic eye disease is on the retina, up to 70% of diabetes patients suffer from problems of the cornea, the transparent, protective exterior surface of the eye. In advanced diabetes, corneal stem cells become dysfunctional, and the cornea heals more slowly and less completely following an injury or procedures such as cataract surgery and laser treatment for diabetic retinopathy.

To identify the epigenetic changes discovered in this study—changes not hard-wired into the genome from birth, but introduced later—Ljubimov and his team compared cells from the corneas from six diabetic patients with those of five healthy donors. They found that in diabetic corneas, the protein product of the WNT5A gene was repressed. Additionally, in diabetic samples, they found an increase in the microRNA that inhibits WNT5A.

The team of scientists then induced wounds to corneal cells in culture and corneal organ cultures, and tested three interventions designed to normalize Wnt-5a protein expression. They added the Wnt-5a protein directly; they introduced a DNA methylation inhibitor, originally approved to treat cancer; and they targeted microRNA levels with a novel gene therapy approach using a nanoscale compound. The team developed the compound, which uses synthetic molecules to block the microRNA, as a substitute for a viral gene therapy they found to be toxic to stem cells.

All three therapeutic methods, in the diabetic samples, stimulated stem cell marker production and improved tissue regeneration, accelerating wound healing.

“Novel therapies to reverse epigenetic effects could improve corneal function, and may also prove significant in other diabetic complications,” said Clive Svendsen, PhD, director of the Board of Governors Regenerative Medicine Institute and study co-author. “This work certainly helps move the field forward.”

Investigators will continue to analyze their data to better understand the mechanisms of WNT5A and other genes related to wound healing. They are also studying a combination therapy to target both microRNA and DNA methylation in hopes that it will more thoroughly normalize wound healing by increasing Wnt-5a protein.

“Our goal is to develop topical, sustained-release drugs for corneal wound healing,” said Ljubimov. “Drugs that are FDA [Food and Drug Administration] approved and could be easily applied may be one of the most promising approaches for effective future therapies.”

Funding: This work was funded by National Institutes of Health grant numbers R01EY013431482, R01EY031377, R01EY025377, and R01CA206220; and the Cedars-Sinai Board of Governors Regenerative Medicine Institute.

Read more on the Cedars-Sinai Blog: Diabetes Dealt Her In and She’s Playing to Win

The study, published in the peer-reviewed journal Diabetes, found people with higher levels of a bacterium called Coprococcus tended to have higher insulin sensitivity, while those whose microbiomes had higher levels of the bacterium Flavonifractor tended to have lower insulin sensitivity. 

For years, investigators have sought to understand why people develop diabetes by studying the composition of the microbiome, which is a collection of microorganisms that include fungi, bacteria and viruses that live in the digestive tract. The microbiome is thought to be affected by medications and diet. Studies have also found that people who don’t process insulin properly have lower levels of a certain type of bacteria that produce a type of fatty acid called butyrate.

Mark Goodarzi, MD, PhD, the director of the Endocrine Genetics Laboratory at Cedars-Sinai, is leading an ongoing study that is following and observing people at risk for diabetes to learn whether those with lower levels of these bacteria develop the disease. 

 “The big question we're hoping to address is: Did the microbiome differences cause the diabetes, or did the diabetes cause the microbiome differences?” said Goodarzi, who is the senior author of the study and principal investigator of the multicenter study called Microbiome and Insulin Longitudinal Evaluation Study (MILES). 

Investigators involved in MILES have been collecting information from participating Black and non-Hispanic white adults between 40 and 80 years of age since 2018. An earlier cohort study from the MILES trial found that birth by cesarean section is associated with a higher risk for developing prediabetes and diabetes.

For the most recent study to come out of this ongoing trial, investigators analyzed data from 352 people without known diabetes who were recruited from the Wake Forest Baptist Health System in Winston-Salem, North Carolina.

Study participants were asked to attend three clinic visits and collect stool samples prior to the visits. Investigators analyzed data collected at the first visit. They conducted genetic sequencing on the stool samples, for example, to study the participants’ microbiomes, and specifically look for bacteria that earlier studies have found to be associated with insulin resistance. Each participant also filled out a diet questionnaire and took an oral glucose tolerance test, which was used to determine ability to process glucose.

Investigators found 28 people had oral glucose tolerance results that met the criteria for diabetes. They also found that 135 people had prediabetes, a condition in which a person’s blood-sugar levels are higher than normal but not high enough to meet the definition of diabetes. 

The research team analyzed associations between 36 butyrate-producing bacteria found in the stool samples and a person’s ability to maintain normal levels of insulin. They controlled for factors that could also contribute to a person’s diabetes risk, such as age, sex, body mass index and race. Coprococcus and related bacteria formed a network of bacteria with beneficial effects on insulin sensitivity. Despite being a producer of butyrate, Flavonifractor was associated with insulin resistance; prior work by others have found higher levels of Flavonifractor in the stool of people with diabetes. 

Investigators are continuing to study samples from patients who participated in this study to learn how insulin production and the composition of the microbiome change over time. They also plan to study how diet may affect the bacterial balance of the microbiome. 

Goodarzi emphasized, however, that it is too early to know how people can change their microbiome to reduce their diabetes risk.

“As far as the idea of taking probiotics, that would really be somewhat experimental,” said Goodarzi, who is also the Eris M. Field Chair in Diabetes Research at Cedars-Sinai. “We need more research to identify the specific bacteria that we need to be modulating to prevent or treat diabetes, but it's coming, probably in the next five to 10 years.”

Jinrui Cui, a biostatistician in the Goodarzi Laboratory at Cedars-Sinai, was the first author of the study.   

Funding: The study was funded by the National Institutes of Health (R01-DK109588), the National Institute of Diabetes and Digestive and Kidney Disease (P30-DK063491), the National Center for Advancing Translational Sciences (grants UL1TR001420, UL1TR001881). 

Follow Cedars-Sinai Academic Medicine on Twitter for more on the latest basic science and clinical research from Cedars-Sinai.