Home Health New Algorithms Help Predict Diabetes Treatment Failure

New Algorithms Help Predict Diabetes Treatment Failure


Metformin is the recommended first line treatment for type 2 diabetes, and is often used to prevent progression of prediabetes to diabetes. Unfortunately, it will not work for over one third of patients who take it, a condition called “therapeutic failure.”

Historically there has been little way of knowing whether a patient will or will not respond to this drug, however, a recent study has identified a new way to predict the treatment failure of this widely used type 2 diabetes medication, giving the healthcare provider the opportunity to consider using metformin in combination with another medication or selecting a different medication entirely.

Researchers analyzed the treatment regimens and glycemic outcomes of 12,147 commercially insured and Medicare Advantage beneficiary adults and developed machine learning algorithms to predict which patients are not likely to achieve and maintain control of their blood glucose after one year of therapy. The most influential variables in the prediction were the patient’s baseline hemoglobin A1C level, starting metformin at a sub-therapeutic dose, and the presence of diabetes complications at the time of treatment initiation. By predicting metformin failure rates, physicians can adjust patient treatment strategies accordingly and achieve better outcomes.

An overwhelming majority of the machine learning model outperformed traditional models that used logistic regression. This research, through the use of these new methods in machine learning, highlights the potential for applying a type of artificial intelligence technology to problems in medicine.

Editor’s Note: Artificial intelligence (AI) is a branch of computer science which attempts to emulate human problem solving skills. Also called ‘cognitive computing,’ it includes concepts such as machine learning – including ‘deep learning,’ and natural language processing, which are especially relevant to health care.

Mayo Clinic recognizes the essence of AI as a set of techniques and technologies that serve to augment – not replace – human intelligence. Artificial intelligence and augmented human intelligence research and development activities at Mayo are overseen by the Mayo Clinic Robert D. and Patricia E. Kern Center for the Science of Health Care Delivery.