Predicting Short-term and Long-term HbA1c Response after Insulin Initiation in Patients with Type 2 Diabetes Mellitus using Machine Learning

AIM: To assess the potential of supervised machine learning techniques to identify clinical variables for predicting short-term and long-term glycated hemoglobin (HbA1c) response after insulin treatment initiation in patients with type 2 diabetes mellitus (T2DM). MATERIALS AND METHODS: We included patients with T2DM from the Groningen Initiative to ANalyze Type 2 diabetes Treatment (GIANTT) database who started insulin treatment between 2007-2013 with a minimum follow-up of 2 years. Short-term and long-term response were defined at 6 (± 2) and 24 (± 2) months after insulin initiation, respectively. Patients were defined as good responders if they had a decrease in HbA1c ≥ 5mmol/mol or reached the recommended level of HbA1c ≤ 53 mmol/mol. Twenty-four baseline clinical variables were used for the analysis and elastic net regularization technique was used for variables selection. The performance of three traditional machine learning algorithms was compared to predict short-term and long-term responses and the area under the receiver operator characteristic curve (AUC) was used to assess the performance of the prediction model. RESULTS: The elastic net regularization based generalized linear model, including baseline HbA1c and eGFR, correctly classified short-term and long-term HbA1c response after treatment initiation with an AUC (95% CI) = 0.80 (0.78 - 0.83) and 0.81 (0.79 - 0.84), respectively, and outperformed other machine learning algorithms. Using baseline HbA1c alone, an AUC = 0.71 (0.65 - 0.73) and 0.72 (0.66 - 0.75) was obtained for predicting short-term and long-term response, respectively. CONCLUSIONS: Machine-learning algorithm performed well in the prediction of an individual's short-term and long-term HbA1c response using baseline clinical variables. This article is protected by copyright. All rights reserved