A new AI tool developed by the University of Maryland offers dynamic, short-term risk predictions for diabetes complications, promising to enhance personalised patient care and early intervention strategies.
Researchers at the University of Maryland School of Medicine have developed an artificial intelligence tool that could help doctors predict which diabetes complications are most likely to emerge next, rather than simply warning patients about every possible risk at once. The system, called the Diabetes Complications Risk Calculator, or DCRC, was led by Rozalina G. McCoy and published in Nature Communications. It is designed to estimate short-term risk for multiple complications and revise those estimates as new information is added during routine care. [1]
The model was trained on health data from more than 400,000 adults newly diagnosed with diabetes in the US, using routinely collected clinical information such as age, diagnoses, medication use, laboratory results and kidney measures. According to the Nature Communications paper, the calculator can generate near-term predictions for nine complications, including cardiovascular disease, stroke, nephropathy and neuropathy, as well as severe high- and low-blood-sugar events. [2]
That matters because diabetes is not a single-condition disease. Persistently elevated glucose can damage both large and small blood vessels, raising the likelihood of heart attacks, strokes, kidney disease, eye disease, nerve damage and foot problems. Doctors already use risk scores in other areas of medicine, but the researchers argued that diabetes requires a more flexible approach because a patient’s most urgent threat can shift as treatments change and other conditions develop. [1][2]
The team tested the tool not only in the national claims dataset used to build it, but also in an independent Mayo Clinic population. In that external validation, the best-performing model showed strong discrimination, with areas under the receiver operating characteristic curve ranging from 0.77 to 0.85 across complications, according to the paper. The researchers said the results suggest the calculator may be useful for personalising risk discussions and supporting shared decision-making. [2][3]
The findings also highlight how quickly complications can appear after diagnosis. The study reported that around one in three patients developed at least one complication within a year, rising to more than 40% by two years. Risk was higher among older patients, those with high blood pressure, poorer kidney function and other medical conditions. Importantly, the calculator is designed to update as a patient’s health changes, making it more responsive than many traditional long-range prediction tools. [1][2]
Still, the researchers caution that the DCRC is a decision-support tool, not a replacement for clinical judgement. They note limits including the use of insured patient data, possible gaps in electronic health records and the risk that people who see doctors more often may have problems detected earlier. Before the system can be used routinely, it will need further testing in real-world care settings to show that its predictions actually improve treatment and outcomes. [1][2][3]
Disclaimer: This content is for informational purposes only and is not intended to be a substitute for professional medical judgment, advice, diagnosis, or treatment.
