Machine learning-based binary classification of elevated HbA1c (≥6.5 %) for risk assessment.

Objective: One of the most important biomarkers for evaluating long-term glycemic management and estimating the risk of diabetes is glycated hemoglobin (HbA1c). Early risk assessment and intervention techniques can be improved by identifying important clinical and demographic variables. Through the integration of clinical indicators (lipid profiles, albumin, and liver enzymes) and demographic characteristics (age, gender), this study seeks to create a comprehensive HbA1c prediction model. Study design: To find the most important factors, logistic regression was used to evaluate 482 cases using a stepwise selection process. Results: With an area under the curve (AUC) of 0.797, the final model had strong predictive ability. Age (OR = 1.085, p < 0.001), glutamate pyruvate transaminase (GPT) (OR = 1.011, p = 0.0127), high-density lipoprotein (HDL) (OR = 0.969, p = 0.017), VitaminD3 (OR = 1.023, p = 0.014), and very low-density lipoprotein (VDL) (OR = 1.016, p = 0.018) were all significant predictors. Conclusions: The greatest predictor was age, which was positively correlated with elevated HbA1c levels, whereas HDL had a protective impact. The addition of VitaminD3, VDL, and GPT implies that indicators of liver and metabolic function have a major role in the variability of HbA1c. These results demonstrate how normal clinical and demographic data may be incorporated into predictive models for early diabetes risk assessment, enabling more individualized medical care and bettering patient outcomes.