Type 2 diabetes predictors may identify worse long-term HbA1c
Click Here to Manage Email Alerts
Data from previous HbA1c levels, fasting plasma glucose, duration of type 2 diabetes and diabetes medication use may predict long-term glycemic profile, researchers reported in Clinical Epidemiology.
“Individualized prediction models based on [a] patient’s characteristics are needed to aid physicians and health care professionals in identification of the patients with poor treatment balance and tailoring treatment and monitoring according to the needs of the patients,” Piia Lavikainen, MD, from the School of Pharmacy at the University of Eastern Finland, Kuopio, and colleagues wrote. “In the present study... we applied artificial intelligence as a tool to examine individualized predictions by searching complex relationships from high-dimensional data.”
Researchers extracted data from electronic health records of primary and specialized care on 9,631 individuals with type 2 diabetes (mean age, 69.2 years; 47.2% women) from the North Karelia region in eastern Finland. Using these data, researchers assessed 6-year HbA1c trajectories for all of these individuals.
Overall, researchers categorized individuals into three HbA1c trajectories of glycemic control during the study period: “stable, adequate” for 86.5% of the cohort, “improving, but inadequate” for 7.3% and “fluctuating, inadequate” for 6.2%. The most important predictors for long-term treatment balance were prior glucose levels, type 2 diabetes duration, use of insulin alone, insulin use with oral diabetes medications and use of metformin alone.
Researchers observed a balanced accuracy of 85% for the prediction model and 91% for the receiving operating characteristic area under the curve. These results highlight high performance utilizing these predictors, according to researchers. In addition, the Shapley additive explanations values demonstrate possibilities to explain outcomes of machine learning methods at the population level and individual level.
According to the researchers, these findings are encouraging because individuals with type 2 diabetes with continuously inadequate treatment balance could have predictions with confidence that are based on individualized characteristics at any point in their disease course.
“Our findings suggest that heterogeneity in long-term treatment outcomes is predictable with patient’s unique risk factors,” the researchers wrote. “This, in turn, offers a useful tool to support treatment planning in the future. However, future studies are needed to obtain even more accurate and personalized predictions.”
Collapse
Read more about
Click Here to Manage Email Alerts