Machine Learning Model Based Prediction of Risk Factors for Hypertension in Elderly HIV/AIDS Patients Using Electronic Medical Records

Objective: This study used electronic medical record information and machine learning methods to develop a model predicting the risk of hypertension in HIV/AIDS patients. Method: This study focused on HIV/AIDS patients admitted to the Henan Infectious Disease Hospital in China from September 2017 to October 2022. The researchers extracted 84 clinical features from electronic medical records. Feature selection was performed using mutual information, Recursive Feature Elimination, Lasso regression, and Principal Component Analysis. Five machine learning prediction models (eXtreme Gradient Boosting [XGBoost], Logistic Regression, Support Vector Machine [SVM], Back Propagation Neural Network [BPNN], and Decision Tree [DT]) were trained using these selected features. These models were tested using 20% of the data. Model performance was assessed using metrics including the area under the receiver operating characteristic curve (AUC), sensitivity, specificity, accuracy, precision, and F1 score. The optimal predictive model interpreted the feature values using Shapley Additive Explanations (SHAP). Result: After evaluating five machine learning models, the XGBoost prediction model demonstrated superior predictive capability. According to the SHAP interpretation of this model, CD4/CD3 (percentage of CD4 cells to CD3 cells), NE (Neutrophil Count), CRP (C-Reactive Protein), GLU (Glucose), PLT (Platelets), LDD-c (Low-Density Lipoprotein Cholesterol), ApoA1 (Apolipoprotein A1), TC (Total Cholesterol) were identified as the most crucial predictive features. Additionally, MCHC (Mean Corpuscular Hemoglobin Concentration), UCB (Unconjugated Bilirubin), TP (Total Protein), CK-MB (Creatine Kinase-MB), GT (Gamma-Glutamyl Transferase), PCV (Packed Cell Volume) have a significant impact on the occurrence of hypertension in HIV/AIDS patients. Conclusion: The predictive model is significant for the rational management and formulation of personalized treatment plans for hospitalized HIV/AIDS patients.