Jian Ding ¹ Zheng Long ² Yiming Liu ³ Min Wang ^4*

• ¹ Shandong Public Health Clinical Center, Jinan, Shandong Province, China
• ² Xuanwu Hospital, Capital Medical University, Beijing, Beijing Municipality, China
• ³ Qilu Hospital, Shandong University, Jinan, Shandong Province, China
• ⁴ The Second Hospital of Shandong University, Jinan, China

Alzheimer's disease (AD) is a widespread progressive neurodegenerative disease often accompanied by a series of comorbidities. The presence of comorbidities increases the risk of death in AD patients. Our study employed Multivariate logistic regression and LASSO regression, Random Forest, Support Vector Machine (SVM), and eXtreme Gradient Boosting (XGBoost) modeling based on the MIMIC-V database to identify age-adjusted Charlson Comorbidity Index (aCCI) related feature factors in patients with AD. Subsequently, the regression model was used to verify the feature factors and the optimal model was selected to construct the nomogram. The results showed that Multivariate logistic regression, LASSO regression, and Random Forest showed excellent performance in classifying patients with aCCI. When regression models were constructed based on the feature factors derived from them, the model constructed from the LASSO regression feature factors has the best performance and the highest net benefit. Age, respiratory rate, Base Excess, Glucose, Red Cell Distribution Width (RDW), Alkaline Phosphatase, Whole Blood Potassium, Hematocrit, Phosphate, Creatinine, and Mean Corpuscular Hemoglobin (MCH) identified by LASSO regression are all closely related to comorbidities.In summary, the factors identified by LASSO regression could be used to predict patients' high aCCI probability and provide theoretical guidance for patients' personalized treatment.