Habibpour M, Tanzanakis A, Salin E, Eskofier B (2026)
Publication Type: Conference contribution
Publication year: 2026
Publisher: Springer Science and Business Media Deutschland GmbH
Book Volume: 694 LNICST
Pages Range: 272-276
Conference Proceedings Title: Lecture Notes of the Institute for Computer Sciences, Social-Informatics and Telecommunications Engineering, LNICST
Event location: Eindhoven, NLD
ISBN: 9783032275844
DOI: 10.1007/978-3-032-27585-1_25
Accurate prediction of Hemoglobin A1c (HbA1c) levels is essential for managing Type 2 Diabetes Mellitus (T2DM) and enabling timely clinical intervention. While most studies treat HbA1c prediction as a classification task, we approach it from a regression perspective to provide finer-grained insights for personalized care. Using demographic, clinical, and laboratory data from hospitalized patients in the MIMIC-IV dataset, we developed and evaluated multiple supervised machine learning models and ensemble techniques with a focus on both predictive accuracy and interpretability. The inpatient nature of the data introduces complexity and heterogeneity, which we addressed through systematic preprocessing and thresholding strategies. Models were trained and evaluated with 5-fold cross-validation, with Random Forest and XGBoost achieving the best predictive performance. To enhance interpretability, we applied Shapley Additive Explanations (SHAP), which highlighted glucose and albumin as key predictors alongside other relevant clinical variables. These findings demonstrate the feasibility of numeric HbA1c prediction from real-world electronic health records (EHR) data, offering promising directions for personalized decision support in T2DM care.
APA:
Habibpour, M., Tanzanakis, A., Salin, E., & Eskofier, B. (2026). Predicting Numeric Hemoglobin A1c for Individualized Type 2 Diabetes Care Using Machine Learning: A MIMIC-IV Study. In Jun Hu, Karin Coninx, Bin Yu, Luigi Borzì, Maarten Houben (Eds.), Lecture Notes of the Institute for Computer Sciences, Social-Informatics and Telecommunications Engineering, LNICST (pp. 272-276). Eindhoven, NLD: Springer Science and Business Media Deutschland GmbH.
MLA:
Habibpour, Mahdis, et al. "Predicting Numeric Hemoglobin A1c for Individualized Type 2 Diabetes Care Using Machine Learning: A MIMIC-IV Study." Proceedings of the 19th EAI International Conference on Pervasive Computing Technologies for Healthcare, Pervasive Health 2025, Eindhoven, NLD Ed. Jun Hu, Karin Coninx, Bin Yu, Luigi Borzì, Maarten Houben, Springer Science and Business Media Deutschland GmbH, 2026. 272-276.
BibTeX: Download