Researchers at the MIT Sloan School say they have developed an algorithm that has the potential to improve the health of the 29 million Americans living with type 2 diabetes, a chronic, incurable illness that causes blood glucose levels to rise higher than normal.
“Our algorithm mines patient and drug data, finds what is most relevant to an individual patient based on his or her medical history and genes, and then provides a recommendation on whether a different treatment or drug would be more effective,” said Dimitris Bertsimas, Ph.D., who is the co-director of the Operations Research Center, a major analytics center at MIT. “The algorithm is applicable to other diseases—including cancer, Alzheimer's, and cardiovascular disease—and is a vivid illustration of how personalized medicine has the potential to transform patient care.”
The team's study (“Personalized Diabetes Management Using Electronic Medical Records“), which appears in a forthcoming issue of Diabetes Care, was conducted in partnership with Boston Medical Center. Dr. Bertsimas and his colleagues used a dataset that involved the electronic medical records (EMR) of about 11,000 patients who had three or more glucose level tests on record, a prescription for at least one blood glucose regulation drug, and no recorded diagnosis of type 1 diabetes. The researchers also had access to each patient’s demographic data, height, weight, body mass index, and prescription drug history. (All patient information was anonymous.)
Next, the team developed an algorithm to define precisely when each line of therapy ended and the next one began according to when the combination of drugs prescribed to the patients changed in the EMR data. All told, the algorithm considered 13 possible drug regimens.
Then the algorithm went to work, said Dr. Bertsimas. For each patient, the algorithm processed the menu of available treatment options, including the patient’s current treatment, as well as the treatment of his or her 30 “nearest neighbors” in terms of genomic and demographic similarity to predict potential effects. The algorithm assumed that the patient would inherit the average outcome of his or her nearest neighbors.
If the algorithm spotted potential for improvement, it proposed a change in treatment; if not, the algorithm suggested the patient remain on his or her existing regimen. In two-thirds of the patient sample, the algorithm did not propose a change.
The patients who received proposed new treatments saw dramatic results, continued Dr. Bertsimas. Using historical data from the database, the algorithm resulted in an average beneficial change in the hemoglobin of 0.44% at each doctor’s visit for which the system’s recommendation differed from standard of care.
“This is a meaningful and medically material improvement,” he noted. “By tailoring specific treatments to specific patients, we’re giving everyone the best possible opportunity for a healthier life.”
Based on the success of the initial study, Dr. Bertsimas and his team are working with doctors at Massachusetts General Hospital to organize a clinical trial.
After all, he said, while data sets and genomic sequencing are critical to the future of precision medicine, “human expertise” provides an important third element. “For it is the doctors—who have the education, skills, and relationships with patients—who make informed judgments about the potential courses of treatment. Even with the best data in the world, we need highly trained doctors to make smart medical decisions.”
