Fig. 1
Overview of models under comparison. HbA1c is often translated to an estimated average glucose (eAG) using a population-based linear regression model. However, the eAG has a high variance with CGM-derived average glucose (CGM-AG). In contrast, novel personalized kinetic models use concurrent CGM and HbA1c data to compute a personalization factor (PF) that can later be used to compute a patient-specific predicted average glucose (pAG). We compare two kinetic models (models 1 and 2) proposed in prior studies to obtain pAGA and pAGA, G. Moreover, we introduce model 3, which employs a multiple linear regression model. This model incorporates demographic, clinical, and kinetic features to generate a pAGLR.
