Objective To analyse the related factors influencing the occurrence of mild cognitive impairment (MCI) in elderly patients with type 2 diabetes mellitus (T2DM) and construct a predictive model.

Methods A retrospective collection of clinical data was conducted on elderly T2DM patients admitted to Shanxi Fenyang Hospital from October 2023 to October 2024. After excluding cases with missing values for independent and dependent variables, a final sample of 244 elderly T2DM patients was selected as the original dataset. Using simple random sampling and a 1∶1 ratio, the patientswere divided into modelling group (122 cases, with 67 cases of MCI and 55 cases without MCI) and validation group (122 cases, with 67 cases of MCI and 55 cases without MCI). Clinical data and laboratory test indicators were compared between different groups. Multivariate logistic regression analysis was employed to explore the risk factors for MCI in elderly T2DM patients and construct a predictive model. The receiver operating characteristic (ROC) curve was drawn to analyze the value of the model in predicting the occurrence of MCI in elderly T2DM patients.

Results In both the modelling and validation groups, the proportions of patients with T2DM duration ≥10 years, age >70 years, bone metabolism abnormalities, serum monocyte chemoattractant protein-1 (MCP-1) level ≥350 pg/mL, serum high-sensitivity C-reactive protein (hs-CRP) level ≥10 mg/L, and serum amyloid β-protein 1-42 (Aβ1-42) level >70 pg/mL were higher in the MCI subgroups than in the non-MCI subgroups, with statistically significant differences (P < 0.05). Multivariate logistic regression analysis revealed that T2DM duration ≥10 years, age >70 years, bone metabolism abnormalities, serum MCP-1 level ≥350 pg/mL, serum hs-CRP level ≥10 mg/L, and serum Aβ1-42 level >70 pg/mL were all independent risk factors for MCI in elderly T2DM patients (P < 0.05). ROC curve analysis showed that the area under the curve (AUC) of the constructed predictive model for predicting the occurrence of MCI in elderly T2DM patients was 0.853 (95%CI, 0.784 to 0.921), with a sensitivity of 86.57% and a specificity of 69.09%. The calibration curve results indicated that the predicted probabilities were close to the actual probabilities, suggesting that the predictive model had good discrimination, calibration, and predictive capabilities.

Conclusion The occurrence of MCI in elderly T2DM patients is closely related to T2DM duration, age, bone metabolism abnormalities, and serum levels of MCP-1, hs-CRP, and Aβ1-42. The constructed predictive model based on these factors has good predictive value. This study innovatively combines bone metabolism indicators with inflammatory markers, providing a reference for the early clinical identification of the risk of MCI occurrence in elderly T2DM patients.